system
stringclasses
7 values
input
stringlengths
10
94.5k
output
stringlengths
14
55.5k
You are a helpful coding assistant specialized in converting HIP source code to AMD device assembly.
#include <hip/hip_runtime.h> #include "includes.h" __global__ void NormalizationExecutionKernel(unsigned char* src, float* dst, const int size, const float alpha, const float beta, const float bias) { int index = blockIdx.x * blockDim.x + threadIdx.x; if(index < size){ dst[index] = (float)(src[index] - alpha) / beta + bias; } }
.text .amdgcn_target "amdgcn-amd-amdhsa--gfx1100" .protected _Z28NormalizationExecutionKernelPhPfifff .globl _Z28NormalizationExecutionKernelPhPfifff .p2align 8 .type _Z28NormalizationExecutionKernelPhPfifff,@function _Z28NormalizationExecutionKernelPhPfifff: s_clause 0x1 s_load_b32 s2, s[0:1], 0x2c s_load_b32 s3, s[0:1], 0x10 s_waitcnt lgkmcnt(0) s_and_b32 s2, s2, 0xffff s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(VALU_DEP_1) v_mad_u64_u32 v[1:2], null, s15, s2, v[0:1] s_mov_b32 s2, exec_lo v_cmpx_gt_i32_e64 s3, v1 s_cbranch_execz .LBB0_2 s_clause 0x1 s_load_b128 s[4:7], s[0:1], 0x0 s_load_b64 s[2:3], s[0:1], 0x14 v_ashrrev_i32_e32 v2, 31, v1 s_load_b32 s0, s[0:1], 0x1c s_waitcnt lgkmcnt(0) v_add_co_u32 v3, vcc_lo, s4, v1 s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_3) | instid1(VALU_DEP_1) v_add_co_ci_u32_e32 v4, vcc_lo, s5, v2, vcc_lo global_load_u8 v0, v[3:4], off s_waitcnt vmcnt(0) v_cvt_f32_ubyte0_e32 v0, v0 v_subrev_f32_e32 v0, s2, v0 s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2) v_div_scale_f32 v3, null, s3, s3, v0 v_div_scale_f32 v6, vcc_lo, v0, s3, v0 v_rcp_f32_e32 v4, v3 s_waitcnt_depctr 0xfff v_fma_f32 v5, -v3, v4, 1.0 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_fmac_f32_e32 v4, v5, v4 v_mul_f32_e32 v5, v6, v4 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_fma_f32 v7, -v3, v5, v6 v_fmac_f32_e32 v5, v7, v4 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_fma_f32 v3, -v3, v5, v6 v_div_fmas_f32 v3, v3, v4, v5 s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2) v_div_fixup_f32 v3, v3, s3, v0 v_lshlrev_b64 v[0:1], 2, v[1:2] v_add_f32_e32 v2, s0, v3 s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3) v_add_co_u32 v0, vcc_lo, s6, v0 v_add_co_ci_u32_e32 v1, vcc_lo, s7, v1, vcc_lo global_store_b32 v[0:1], v2, off .LBB0_2: s_nop 0 s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) s_endpgm .section .rodata,"a",@progbits .p2align 6, 0x0 .amdhsa_kernel _Z28NormalizationExecutionKernelPhPfifff .amdhsa_group_segment_fixed_size 0 .amdhsa_private_segment_fixed_size 0 .amdhsa_kernarg_size 288 .amdhsa_user_sgpr_count 15 .amdhsa_user_sgpr_dispatch_ptr 0 .amdhsa_user_sgpr_queue_ptr 0 .amdhsa_user_sgpr_kernarg_segment_ptr 1 .amdhsa_user_sgpr_dispatch_id 0 .amdhsa_user_sgpr_private_segment_size 0 .amdhsa_wavefront_size32 1 .amdhsa_uses_dynamic_stack 0 .amdhsa_enable_private_segment 0 .amdhsa_system_sgpr_workgroup_id_x 1 .amdhsa_system_sgpr_workgroup_id_y 0 .amdhsa_system_sgpr_workgroup_id_z 0 .amdhsa_system_sgpr_workgroup_info 0 .amdhsa_system_vgpr_workitem_id 0 .amdhsa_next_free_vgpr 8 .amdhsa_next_free_sgpr 16 .amdhsa_float_round_mode_32 0 .amdhsa_float_round_mode_16_64 0 .amdhsa_float_denorm_mode_32 3 .amdhsa_float_denorm_mode_16_64 3 .amdhsa_dx10_clamp 1 .amdhsa_ieee_mode 1 .amdhsa_fp16_overflow 0 .amdhsa_workgroup_processor_mode 1 .amdhsa_memory_ordered 1 .amdhsa_forward_progress 0 .amdhsa_shared_vgpr_count 0 .amdhsa_exception_fp_ieee_invalid_op 0 .amdhsa_exception_fp_denorm_src 0 .amdhsa_exception_fp_ieee_div_zero 0 .amdhsa_exception_fp_ieee_overflow 0 .amdhsa_exception_fp_ieee_underflow 0 .amdhsa_exception_fp_ieee_inexact 0 .amdhsa_exception_int_div_zero 0 .end_amdhsa_kernel .text .Lfunc_end0: .size _Z28NormalizationExecutionKernelPhPfifff, .Lfunc_end0-_Z28NormalizationExecutionKernelPhPfifff .section .AMDGPU.csdata,"",@progbits .text .p2alignl 7, 3214868480 .fill 96, 4, 3214868480 .type __hip_cuid_,@object .section .bss,"aw",@nobits .globl __hip_cuid_ __hip_cuid_: .byte 0 .size __hip_cuid_, 1 .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym __hip_cuid_ .amdgpu_metadata --- amdhsa.kernels: - .args: - .address_space: global .offset: 0 .size: 8 .value_kind: global_buffer - .address_space: global .offset: 8 .size: 8 .value_kind: global_buffer - .offset: 16 .size: 4 .value_kind: by_value - .offset: 20 .size: 4 .value_kind: by_value - .offset: 24 .size: 4 .value_kind: by_value - .offset: 28 .size: 4 .value_kind: by_value - .offset: 32 .size: 4 .value_kind: hidden_block_count_x - .offset: 36 .size: 4 .value_kind: hidden_block_count_y - .offset: 40 .size: 4 .value_kind: hidden_block_count_z - .offset: 44 .size: 2 .value_kind: hidden_group_size_x - .offset: 46 .size: 2 .value_kind: hidden_group_size_y - .offset: 48 .size: 2 .value_kind: hidden_group_size_z - .offset: 50 .size: 2 .value_kind: hidden_remainder_x - .offset: 52 .size: 2 .value_kind: hidden_remainder_y - .offset: 54 .size: 2 .value_kind: hidden_remainder_z - .offset: 72 .size: 8 .value_kind: hidden_global_offset_x - .offset: 80 .size: 8 .value_kind: hidden_global_offset_y - .offset: 88 .size: 8 .value_kind: hidden_global_offset_z - .offset: 96 .size: 2 .value_kind: hidden_grid_dims .group_segment_fixed_size: 0 .kernarg_segment_align: 8 .kernarg_segment_size: 288 .language: OpenCL C .language_version: - 2 - 0 .max_flat_workgroup_size: 1024 .name: _Z28NormalizationExecutionKernelPhPfifff .private_segment_fixed_size: 0 .sgpr_count: 18 .sgpr_spill_count: 0 .symbol: _Z28NormalizationExecutionKernelPhPfifff.kd .uniform_work_group_size: 1 .uses_dynamic_stack: false .vgpr_count: 8 .vgpr_spill_count: 0 .wavefront_size: 32 .workgroup_processor_mode: 1 amdhsa.target: amdgcn-amd-amdhsa--gfx1100 amdhsa.version: - 1 - 2 ... .end_amdgpu_metadata
You are a helpful coding assistant specialized in converting HIP source code to AMD host assembly.
#include <hip/hip_runtime.h> #include "includes.h" __global__ void NormalizationExecutionKernel(unsigned char* src, float* dst, const int size, const float alpha, const float beta, const float bias) { int index = blockIdx.x * blockDim.x + threadIdx.x; if(index < size){ dst[index] = (float)(src[index] - alpha) / beta + bias; } }
.text .file "NormalizationExecutionKernel.hip" .globl _Z43__device_stub__NormalizationExecutionKernelPhPfifff # -- Begin function _Z43__device_stub__NormalizationExecutionKernelPhPfifff .p2align 4, 0x90 .type _Z43__device_stub__NormalizationExecutionKernelPhPfifff,@function _Z43__device_stub__NormalizationExecutionKernelPhPfifff: # @_Z43__device_stub__NormalizationExecutionKernelPhPfifff .cfi_startproc # %bb.0: subq $136, %rsp .cfi_def_cfa_offset 144 movq %rdi, 72(%rsp) movq %rsi, 64(%rsp) movl %edx, 12(%rsp) movss %xmm0, 8(%rsp) movss %xmm1, 4(%rsp) movss %xmm2, (%rsp) leaq 72(%rsp), %rax movq %rax, 80(%rsp) leaq 64(%rsp), %rax movq %rax, 88(%rsp) leaq 12(%rsp), %rax movq %rax, 96(%rsp) leaq 8(%rsp), %rax movq %rax, 104(%rsp) leaq 4(%rsp), %rax movq %rax, 112(%rsp) movq %rsp, %rax movq %rax, 120(%rsp) leaq 48(%rsp), %rdi leaq 32(%rsp), %rsi leaq 24(%rsp), %rdx leaq 16(%rsp), %rcx callq __hipPopCallConfiguration movq 48(%rsp), %rsi movl 56(%rsp), %edx movq 32(%rsp), %rcx movl 40(%rsp), %r8d leaq 80(%rsp), %r9 movl $_Z28NormalizationExecutionKernelPhPfifff, %edi pushq 16(%rsp) .cfi_adjust_cfa_offset 8 pushq 32(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $152, %rsp .cfi_adjust_cfa_offset -152 retq .Lfunc_end0: .size _Z43__device_stub__NormalizationExecutionKernelPhPfifff, .Lfunc_end0-_Z43__device_stub__NormalizationExecutionKernelPhPfifff .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_ctor .type __hip_module_ctor,@function __hip_module_ctor: # @__hip_module_ctor .cfi_startproc # %bb.0: subq $40, %rsp .cfi_def_cfa_offset 48 cmpq $0, __hip_gpubin_handle(%rip) jne .LBB1_2 # %bb.1: movl $__hip_fatbin_wrapper, %edi callq __hipRegisterFatBinary movq %rax, __hip_gpubin_handle(%rip) .LBB1_2: movq __hip_gpubin_handle(%rip), %rdi xorps %xmm0, %xmm0 movups %xmm0, 16(%rsp) movups %xmm0, (%rsp) movl $_Z28NormalizationExecutionKernelPhPfifff, %esi movl $.L__unnamed_1, %edx movl $.L__unnamed_1, %ecx movl $-1, %r8d xorl %r9d, %r9d callq __hipRegisterFunction movl $__hip_module_dtor, %edi addq $40, %rsp .cfi_def_cfa_offset 8 jmp atexit # TAILCALL .Lfunc_end1: .size __hip_module_ctor, .Lfunc_end1-__hip_module_ctor .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_dtor .type __hip_module_dtor,@function __hip_module_dtor: # @__hip_module_dtor .cfi_startproc # %bb.0: movq __hip_gpubin_handle(%rip), %rdi testq %rdi, %rdi je .LBB2_2 # %bb.1: pushq %rax .cfi_def_cfa_offset 16 callq __hipUnregisterFatBinary movq $0, __hip_gpubin_handle(%rip) addq $8, %rsp .cfi_def_cfa_offset 8 .LBB2_2: retq .Lfunc_end2: .size __hip_module_dtor, .Lfunc_end2-__hip_module_dtor .cfi_endproc # -- End function .type _Z28NormalizationExecutionKernelPhPfifff,@object # @_Z28NormalizationExecutionKernelPhPfifff .section .rodata,"a",@progbits .globl _Z28NormalizationExecutionKernelPhPfifff .p2align 3, 0x0 _Z28NormalizationExecutionKernelPhPfifff: .quad _Z43__device_stub__NormalizationExecutionKernelPhPfifff .size _Z28NormalizationExecutionKernelPhPfifff, 8 .type .L__unnamed_1,@object # @0 .section .rodata.str1.1,"aMS",@progbits,1 .L__unnamed_1: .asciz "_Z28NormalizationExecutionKernelPhPfifff" .size .L__unnamed_1, 41 .type __hip_fatbin_wrapper,@object # @__hip_fatbin_wrapper .section .hipFatBinSegment,"a",@progbits .p2align 3, 0x0 __hip_fatbin_wrapper: .long 1212764230 # 0x48495046 .long 1 # 0x1 .quad __hip_fatbin .quad 0 .size __hip_fatbin_wrapper, 24 .type __hip_gpubin_handle,@object # @__hip_gpubin_handle .local __hip_gpubin_handle .comm __hip_gpubin_handle,8,8 .section .init_array,"aw",@init_array .p2align 3, 0x0 .quad __hip_module_ctor .type __hip_cuid_,@object # @__hip_cuid_ .bss .globl __hip_cuid_ __hip_cuid_: .byte 0 # 0x0 .size __hip_cuid_, 1 .section ".linker-options","e",@llvm_linker_options .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym _Z43__device_stub__NormalizationExecutionKernelPhPfifff .addrsig_sym __hip_module_ctor .addrsig_sym __hip_module_dtor .addrsig_sym _Z28NormalizationExecutionKernelPhPfifff .addrsig_sym __hip_fatbin .addrsig_sym __hip_fatbin_wrapper .addrsig_sym __hip_cuid_
You are a helpful coding assistant specialized in converting from CUDA device assembly to AMD device assembly.
code for sm_80 Function : _Z28NormalizationExecutionKernelPhPfifff .headerflags @"EF_CUDA_TEXMODE_UNIFIED EF_CUDA_64BIT_ADDRESS EF_CUDA_SM80 EF_CUDA_VIRTUAL_SM(EF_CUDA_SM80)" /*0000*/ IMAD.MOV.U32 R1, RZ, RZ, c[0x0][0x28] ; /* 0x00000a00ff017624 */ /* 0x000fe400078e00ff */ /*0010*/ S2R R0, SR_CTAID.X ; /* 0x0000000000007919 */ /* 0x000e280000002500 */ /*0020*/ S2R R3, SR_TID.X ; /* 0x0000000000037919 */ /* 0x000e240000002100 */ /*0030*/ IMAD R0, R0, c[0x0][0x0], R3 ; /* 0x0000000000007a24 */ /* 0x001fca00078e0203 */ /*0040*/ ISETP.GE.AND P0, PT, R0, c[0x0][0x170], PT ; /* 0x00005c0000007a0c */ /* 0x000fda0003f06270 */ /*0050*/ @P0 EXIT ; /* 0x000000000000094d */ /* 0x000fea0003800000 */ /*0060*/ IADD3 R2, P0, R0, c[0x0][0x160], RZ ; /* 0x0000580000027a10 */ /* 0x000fe20007f1e0ff */ /*0070*/ ULDC.64 UR4, c[0x0][0x118] ; /* 0x0000460000047ab9 */ /* 0x000fc60000000a00 */ /*0080*/ LEA.HI.X.SX32 R3, R0, c[0x0][0x164], 0x1, P0 ; /* 0x0000590000037a11 */ /* 0x000fca00000f0eff */ /*0090*/ LDG.E.U8 R2, [R2.64] ; /* 0x0000000402027981 */ /* 0x000ea2000c1e1100 */ /*00a0*/ MUFU.RCP R5, c[0x0][0x178] ; /* 0x00005e0000057b08 */ /* 0x000e220000001000 */ /*00b0*/ IMAD.MOV.U32 R6, RZ, RZ, c[0x0][0x178] ; /* 0x00005e00ff067624 */ /* 0x000fe200078e00ff */ /*00c0*/ BSSY B0, 0x190 ; /* 0x000000c000007945 */ /* 0x000fe60003800000 */ /*00d0*/ FFMA R6, R5, -R6, 1 ; /* 0x3f80000005067423 */ /* 0x001fc80000000806 */ /*00e0*/ FFMA R5, R5, R6, R5 ; /* 0x0000000605057223 */ /* 0x000fe20000000005 */ /*00f0*/ I2F.U16 R4, R2 ; /* 0x0000000200047306 */ /* 0x004e240000101000 */ /*0100*/ FADD R4, R4, -c[0x0][0x174] ; /* 0x80005d0004047621 */ /* 0x001fcc0000000000 */ /*0110*/ FCHK P0, R4, c[0x0][0x178] ; /* 0x00005e0004007b02 */ /* 0x000e220000000000 */ /*0120*/ FFMA R6, R4, R5, RZ ; /* 0x0000000504067223 */ /* 0x000fc800000000ff */ /*0130*/ FFMA R7, R6, -c[0x0][0x178], R4 ; /* 0x80005e0006077a23 */ /* 0x000fc80000000004 */ /*0140*/ FFMA R5, R5, R7, R6 ; /* 0x0000000705057223 */ /* 0x000fe20000000006 */ /*0150*/ @!P0 BRA 0x180 ; /* 0x0000002000008947 */ /* 0x001fea0003800000 */ /*0160*/ MOV R2, 0x180 ; /* 0x0000018000027802 */ /* 0x000fe40000000f00 */ /*0170*/ CALL.REL.NOINC 0x1e0 ; /* 0x0000006000007944 */ /* 0x000fea0003c00000 */ /*0180*/ BSYNC B0 ; /* 0x0000000000007941 */ /* 0x000fea0003800000 */ /*0190*/ IMAD.MOV.U32 R3, RZ, RZ, 0x4 ; /* 0x00000004ff037424 */ /* 0x000fe400078e00ff */ /*01a0*/ FADD R5, R5, c[0x0][0x17c] ; /* 0x00005f0005057621 */ /* 0x000fe40000000000 */ /*01b0*/ IMAD.WIDE R2, R0, R3, c[0x0][0x168] ; /* 0x00005a0000027625 */ /* 0x000fca00078e0203 */ /*01c0*/ STG.E [R2.64], R5 ; /* 0x0000000502007986 */ /* 0x000fe2000c101904 */ /*01d0*/ EXIT ; /* 0x000000000000794d */ /* 0x000fea0003800000 */ /*01e0*/ IMAD.MOV.U32 R11, RZ, RZ, c[0x0][0x178] ; /* 0x00005e00ff0b7624 */ /* 0x000fe200078e00ff */ /*01f0*/ SHF.R.U32.HI R3, RZ, 0x17, R4.reuse ; /* 0x00000017ff037819 */ /* 0x100fe20000011604 */ /*0200*/ BSSY B1, 0x850 ; /* 0x0000064000017945 */ /* 0x000fe20003800000 */ /*0210*/ IMAD.MOV.U32 R6, RZ, RZ, R4 ; /* 0x000000ffff067224 */ /* 0x000fe400078e0004 */ /*0220*/ SHF.R.U32.HI R5, RZ, 0x17, R11 ; /* 0x00000017ff057819 */ /* 0x000fe2000001160b */ /*0230*/ IMAD.MOV.U32 R7, RZ, RZ, c[0x0][0x178] ; /* 0x00005e00ff077624 */ /* 0x000fe200078e00ff */ /*0240*/ LOP3.LUT R3, R3, 0xff, RZ, 0xc0, !PT ; /* 0x000000ff03037812 */ /* 0x000fe400078ec0ff */ /*0250*/ LOP3.LUT R5, R5, 0xff, RZ, 0xc0, !PT ; /* 0x000000ff05057812 */ /* 0x000fc400078ec0ff */ /*0260*/ IADD3 R9, R3, -0x1, RZ ; /* 0xffffffff03097810 */ /* 0x000fe40007ffe0ff */ /*0270*/ IADD3 R10, R5, -0x1, RZ ; /* 0xffffffff050a7810 */ /* 0x000fc80007ffe0ff */ /*0280*/ ISETP.GT.U32.AND P0, PT, R10, 0xfd, PT ; /* 0x000000fd0a00780c */ /* 0x000fc80003f04070 */ /*0290*/ ISETP.GT.U32.OR P0, PT, R9, 0xfd, P0 ; /* 0x000000fd0900780c */ /* 0x000fda0000704470 */ /*02a0*/ @!P0 IMAD.MOV.U32 R8, RZ, RZ, RZ ; /* 0x000000ffff088224 */ /* 0x000fe200078e00ff */ /*02b0*/ @!P0 BRA 0x430 ; /* 0x0000017000008947 */ /* 0x000fea0003800000 */ /*02c0*/ FSETP.GTU.FTZ.AND P1, PT, |R11|, +INF , PT ; /* 0x7f8000000b00780b */ /* 0x000fe40003f3c200 */ /*02d0*/ FSETP.GTU.FTZ.AND P0, PT, |R4|, +INF , PT ; /* 0x7f8000000400780b */ /* 0x000fc80003f1c200 */ /*02e0*/ PLOP3.LUT P0, PT, P0, P1, PT, 0xa8, 0x0 ; /* 0x000000000000781c */ /* 0x000fda0000703570 */ /*02f0*/ @P0 BRA 0x830 ; /* 0x0000053000000947 */ /* 0x000fea0003800000 */ /*0300*/ LOP3.LUT P0, RZ, R7, 0x7fffffff, R6, 0xc8, !PT ; /* 0x7fffffff07ff7812 */ /* 0x000fda000780c806 */ /*0310*/ @!P0 BRA 0x810 ; /* 0x000004f000008947 */ /* 0x000fea0003800000 */ /*0320*/ FSETP.NEU.FTZ.AND P2, PT, |R4|.reuse, +INF , PT ; /* 0x7f8000000400780b */ /* 0x040fe40003f5d200 */ /*0330*/ FSETP.NEU.FTZ.AND P1, PT, |R11|, +INF , PT ; /* 0x7f8000000b00780b */ /* 0x000fe40003f3d200 */ /*0340*/ FSETP.NEU.FTZ.AND P0, PT, |R4|, +INF , PT ; /* 0x7f8000000400780b */ /* 0x000fd60003f1d200 */ /*0350*/ @!P1 BRA !P2, 0x810 ; /* 0x000004b000009947 */ /* 0x000fea0005000000 */ /*0360*/ LOP3.LUT P2, RZ, R6, 0x7fffffff, RZ, 0xc0, !PT ; /* 0x7fffffff06ff7812 */ /* 0x000fc8000784c0ff */ /*0370*/ PLOP3.LUT P1, PT, P1, P2, PT, 0x2a, 0x0 ; /* 0x000000000000781c */ /* 0x000fda0000f24572 */ /*0380*/ @P1 BRA 0x7f0 ; /* 0x0000046000001947 */ /* 0x000fea0003800000 */ /*0390*/ LOP3.LUT P1, RZ, R7, 0x7fffffff, RZ, 0xc0, !PT ; /* 0x7fffffff07ff7812 */ /* 0x000fc8000782c0ff */ /*03a0*/ PLOP3.LUT P0, PT, P0, P1, PT, 0x2a, 0x0 ; /* 0x000000000000781c */ /* 0x000fda0000702572 */ /*03b0*/ @P0 BRA 0x7c0 ; /* 0x0000040000000947 */ /* 0x000fea0003800000 */ /*03c0*/ ISETP.GE.AND P0, PT, R9, RZ, PT ; /* 0x000000ff0900720c */ /* 0x000fe40003f06270 */ /*03d0*/ ISETP.GE.AND P1, PT, R10, RZ, PT ; /* 0x000000ff0a00720c */ /* 0x000fd60003f26270 */ /*03e0*/ @P0 IMAD.MOV.U32 R8, RZ, RZ, RZ ; /* 0x000000ffff080224 */ /* 0x000fe400078e00ff */ /*03f0*/ @!P0 IMAD.MOV.U32 R8, RZ, RZ, -0x40 ; /* 0xffffffc0ff088424 */ /* 0x000fe400078e00ff */ /*0400*/ @!P0 FFMA R6, R4, 1.84467440737095516160e+19, RZ ; /* 0x5f80000004068823 */ /* 0x000fe400000000ff */ /*0410*/ @!P1 FFMA R7, R11, 1.84467440737095516160e+19, RZ ; /* 0x5f8000000b079823 */ /* 0x000fe200000000ff */ /*0420*/ @!P1 IADD3 R8, R8, 0x40, RZ ; /* 0x0000004008089810 */ /* 0x000fe40007ffe0ff */ /*0430*/ LEA R4, R5, 0xc0800000, 0x17 ; /* 0xc080000005047811 */ /* 0x000fe200078eb8ff */ /*0440*/ BSSY B2, 0x7b0 ; /* 0x0000036000027945 */ /* 0x000fe80003800000 */ /*0450*/ IMAD.IADD R7, R7, 0x1, -R4 ; /* 0x0000000107077824 */ /* 0x000fe200078e0a04 */ /*0460*/ IADD3 R4, R3, -0x7f, RZ ; /* 0xffffff8103047810 */ /* 0x000fc60007ffe0ff */ /*0470*/ MUFU.RCP R9, R7 ; /* 0x0000000700097308 */ /* 0x000e220000001000 */ /*0480*/ FADD.FTZ R11, -R7, -RZ ; /* 0x800000ff070b7221 */ /* 0x000fe20000010100 */ /*0490*/ IADD3 R5, R4.reuse, 0x7f, -R5 ; /* 0x0000007f04057810 */ /* 0x040fe20007ffe805 */ /*04a0*/ IMAD R6, R4, -0x800000, R6 ; /* 0xff80000004067824 */ /* 0x000fc800078e0206 */ /*04b0*/ IMAD.IADD R5, R5, 0x1, R8 ; /* 0x0000000105057824 */ /* 0x000fe400078e0208 */ /*04c0*/ FFMA R10, R9, R11, 1 ; /* 0x3f800000090a7423 */ /* 0x001fc8000000000b */ /*04d0*/ FFMA R9, R9, R10, R9 ; /* 0x0000000a09097223 */ /* 0x000fc80000000009 */ /*04e0*/ FFMA R3, R6, R9, RZ ; /* 0x0000000906037223 */ /* 0x000fc800000000ff */ /*04f0*/ FFMA R10, R11, R3, R6 ; /* 0x000000030b0a7223 */ /* 0x000fc80000000006 */ /*0500*/ FFMA R10, R9, R10, R3 ; /* 0x0000000a090a7223 */ /* 0x000fc80000000003 */ /*0510*/ FFMA R11, R11, R10, R6 ; /* 0x0000000a0b0b7223 */ /* 0x000fc80000000006 */ /*0520*/ FFMA R3, R9, R11, R10 ; /* 0x0000000b09037223 */ /* 0x000fca000000000a */ /*0530*/ SHF.R.U32.HI R4, RZ, 0x17, R3 ; /* 0x00000017ff047819 */ /* 0x000fc80000011603 */ /*0540*/ LOP3.LUT R4, R4, 0xff, RZ, 0xc0, !PT ; /* 0x000000ff04047812 */ /* 0x000fca00078ec0ff */ /*0550*/ IMAD.IADD R8, R4, 0x1, R5 ; /* 0x0000000104087824 */ /* 0x000fca00078e0205 */ /*0560*/ IADD3 R4, R8, -0x1, RZ ; /* 0xffffffff08047810 */ /* 0x000fc80007ffe0ff */ /*0570*/ ISETP.GE.U32.AND P0, PT, R4, 0xfe, PT ; /* 0x000000fe0400780c */ /* 0x000fda0003f06070 */ /*0580*/ @!P0 BRA 0x790 ; /* 0x0000020000008947 */ /* 0x000fea0003800000 */ /*0590*/ ISETP.GT.AND P0, PT, R8, 0xfe, PT ; /* 0x000000fe0800780c */ /* 0x000fda0003f04270 */ /*05a0*/ @P0 BRA 0x760 ; /* 0x000001b000000947 */ /* 0x000fea0003800000 */ /*05b0*/ ISETP.GE.AND P0, PT, R8, 0x1, PT ; /* 0x000000010800780c */ /* 0x000fda0003f06270 */ /*05c0*/ @P0 BRA 0x7a0 ; /* 0x000001d000000947 */ /* 0x000fea0003800000 */ /*05d0*/ ISETP.GE.AND P0, PT, R8, -0x18, PT ; /* 0xffffffe80800780c */ /* 0x000fe40003f06270 */ /*05e0*/ LOP3.LUT R3, R3, 0x80000000, RZ, 0xc0, !PT ; /* 0x8000000003037812 */ /* 0x000fd600078ec0ff */ /*05f0*/ @!P0 BRA 0x7a0 ; /* 0x000001a000008947 */ /* 0x000fea0003800000 */ /*0600*/ FFMA.RZ R4, R9.reuse, R11.reuse, R10.reuse ; /* 0x0000000b09047223 */ /* 0x1c0fe2000000c00a */ /*0610*/ IADD3 R7, R8.reuse, 0x20, RZ ; /* 0x0000002008077810 */ /* 0x040fe20007ffe0ff */ /*0620*/ FFMA.RM R5, R9.reuse, R11.reuse, R10.reuse ; /* 0x0000000b09057223 */ /* 0x1c0fe2000000400a */ /*0630*/ ISETP.NE.AND P2, PT, R8, RZ, PT ; /* 0x000000ff0800720c */ /* 0x000fe40003f45270 */ /*0640*/ LOP3.LUT R6, R4, 0x7fffff, RZ, 0xc0, !PT ; /* 0x007fffff04067812 */ /* 0x000fe200078ec0ff */ /*0650*/ FFMA.RP R4, R9, R11, R10 ; /* 0x0000000b09047223 */ /* 0x000fe2000000800a */ /*0660*/ ISETP.NE.AND P1, PT, R8, RZ, PT ; /* 0x000000ff0800720c */ /* 0x000fe20003f25270 */ /*0670*/ IMAD.MOV R8, RZ, RZ, -R8 ; /* 0x000000ffff087224 */ /* 0x000fe200078e0a08 */ /*0680*/ LOP3.LUT R6, R6, 0x800000, RZ, 0xfc, !PT ; /* 0x0080000006067812 */ /* 0x000fe400078efcff */ /*0690*/ FSETP.NEU.FTZ.AND P0, PT, R4, R5, PT ; /* 0x000000050400720b */ /* 0x000fc40003f1d000 */ /*06a0*/ SHF.L.U32 R7, R6, R7, RZ ; /* 0x0000000706077219 */ /* 0x000fe400000006ff */ /*06b0*/ SEL R5, R8, RZ, P2 ; /* 0x000000ff08057207 */ /* 0x000fe40001000000 */ /*06c0*/ ISETP.NE.AND P1, PT, R7, RZ, P1 ; /* 0x000000ff0700720c */ /* 0x000fe40000f25270 */ /*06d0*/ SHF.R.U32.HI R5, RZ, R5, R6 ; /* 0x00000005ff057219 */ /* 0x000fe40000011606 */ /*06e0*/ PLOP3.LUT P0, PT, P0, P1, PT, 0xa8, 0x0 ; /* 0x000000000000781c */ /* 0x000fe40000703570 */ /*06f0*/ SHF.R.U32.HI R7, RZ, 0x1, R5 ; /* 0x00000001ff077819 */ /* 0x000fc40000011605 */ /*0700*/ SEL R4, RZ, 0x1, !P0 ; /* 0x00000001ff047807 */ /* 0x000fc80004000000 */ /*0710*/ LOP3.LUT R4, R4, 0x1, R7, 0xf8, !PT ; /* 0x0000000104047812 */ /* 0x000fc800078ef807 */ /*0720*/ LOP3.LUT R4, R4, R5, RZ, 0xc0, !PT ; /* 0x0000000504047212 */ /* 0x000fca00078ec0ff */ /*0730*/ IMAD.IADD R4, R7, 0x1, R4 ; /* 0x0000000107047824 */ /* 0x000fca00078e0204 */ /*0740*/ LOP3.LUT R3, R4, R3, RZ, 0xfc, !PT ; /* 0x0000000304037212 */ /* 0x000fe200078efcff */ /*0750*/ BRA 0x7a0 ; /* 0x0000004000007947 */ /* 0x000fea0003800000 */ /*0760*/ LOP3.LUT R3, R3, 0x80000000, RZ, 0xc0, !PT ; /* 0x8000000003037812 */ /* 0x000fc800078ec0ff */ /*0770*/ LOP3.LUT R3, R3, 0x7f800000, RZ, 0xfc, !PT ; /* 0x7f80000003037812 */ /* 0x000fe200078efcff */ /*0780*/ BRA 0x7a0 ; /* 0x0000001000007947 */ /* 0x000fea0003800000 */ /*0790*/ IMAD R3, R5, 0x800000, R3 ; /* 0x0080000005037824 */ /* 0x000fe400078e0203 */ /*07a0*/ BSYNC B2 ; /* 0x0000000000027941 */ /* 0x000fea0003800000 */ /*07b0*/ BRA 0x840 ; /* 0x0000008000007947 */ /* 0x000fea0003800000 */ /*07c0*/ LOP3.LUT R3, R7, 0x80000000, R6, 0x48, !PT ; /* 0x8000000007037812 */ /* 0x000fc800078e4806 */ /*07d0*/ LOP3.LUT R3, R3, 0x7f800000, RZ, 0xfc, !PT ; /* 0x7f80000003037812 */ /* 0x000fe200078efcff */ /*07e0*/ BRA 0x840 ; /* 0x0000005000007947 */ /* 0x000fea0003800000 */ /*07f0*/ LOP3.LUT R3, R7, 0x80000000, R6, 0x48, !PT ; /* 0x8000000007037812 */ /* 0x000fe200078e4806 */ /*0800*/ BRA 0x840 ; /* 0x0000003000007947 */ /* 0x000fea0003800000 */ /*0810*/ MUFU.RSQ R3, -QNAN ; /* 0xffc0000000037908 */ /* 0x000e220000001400 */ /*0820*/ BRA 0x840 ; /* 0x0000001000007947 */ /* 0x000fea0003800000 */ /*0830*/ FADD.FTZ R3, R4, c[0x0][0x178] ; /* 0x00005e0004037621 */ /* 0x000fe40000010000 */ /*0840*/ BSYNC B1 ; /* 0x0000000000017941 */ /* 0x000fea0003800000 */ /*0850*/ IMAD.MOV.U32 R5, RZ, RZ, R3 ; /* 0x000000ffff057224 */ /* 0x001fe400078e0003 */ /*0860*/ IMAD.MOV.U32 R3, RZ, RZ, 0x0 ; /* 0x00000000ff037424 */ /* 0x000fc800078e00ff */ /*0870*/ RET.REL.NODEC R2 0x0 ; /* 0xfffff78002007950 */ /* 0x000fea0003c3ffff */ /*0880*/ BRA 0x880; /* 0xfffffff000007947 */ /* 0x000fc0000383ffff */ /*0890*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*08a0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*08b0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*08c0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*08d0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*08e0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*08f0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0900*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0910*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0920*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0930*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0940*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0950*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0960*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0970*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ ..........
.text .amdgcn_target "amdgcn-amd-amdhsa--gfx1100" .protected _Z28NormalizationExecutionKernelPhPfifff .globl _Z28NormalizationExecutionKernelPhPfifff .p2align 8 .type _Z28NormalizationExecutionKernelPhPfifff,@function _Z28NormalizationExecutionKernelPhPfifff: s_clause 0x1 s_load_b32 s2, s[0:1], 0x2c s_load_b32 s3, s[0:1], 0x10 s_waitcnt lgkmcnt(0) s_and_b32 s2, s2, 0xffff s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(VALU_DEP_1) v_mad_u64_u32 v[1:2], null, s15, s2, v[0:1] s_mov_b32 s2, exec_lo v_cmpx_gt_i32_e64 s3, v1 s_cbranch_execz .LBB0_2 s_clause 0x1 s_load_b128 s[4:7], s[0:1], 0x0 s_load_b64 s[2:3], s[0:1], 0x14 v_ashrrev_i32_e32 v2, 31, v1 s_load_b32 s0, s[0:1], 0x1c s_waitcnt lgkmcnt(0) v_add_co_u32 v3, vcc_lo, s4, v1 s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_3) | instid1(VALU_DEP_1) v_add_co_ci_u32_e32 v4, vcc_lo, s5, v2, vcc_lo global_load_u8 v0, v[3:4], off s_waitcnt vmcnt(0) v_cvt_f32_ubyte0_e32 v0, v0 v_subrev_f32_e32 v0, s2, v0 s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2) v_div_scale_f32 v3, null, s3, s3, v0 v_div_scale_f32 v6, vcc_lo, v0, s3, v0 v_rcp_f32_e32 v4, v3 s_waitcnt_depctr 0xfff v_fma_f32 v5, -v3, v4, 1.0 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_fmac_f32_e32 v4, v5, v4 v_mul_f32_e32 v5, v6, v4 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_fma_f32 v7, -v3, v5, v6 v_fmac_f32_e32 v5, v7, v4 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_fma_f32 v3, -v3, v5, v6 v_div_fmas_f32 v3, v3, v4, v5 s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2) v_div_fixup_f32 v3, v3, s3, v0 v_lshlrev_b64 v[0:1], 2, v[1:2] v_add_f32_e32 v2, s0, v3 s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3) v_add_co_u32 v0, vcc_lo, s6, v0 v_add_co_ci_u32_e32 v1, vcc_lo, s7, v1, vcc_lo global_store_b32 v[0:1], v2, off .LBB0_2: s_nop 0 s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) s_endpgm .section .rodata,"a",@progbits .p2align 6, 0x0 .amdhsa_kernel _Z28NormalizationExecutionKernelPhPfifff .amdhsa_group_segment_fixed_size 0 .amdhsa_private_segment_fixed_size 0 .amdhsa_kernarg_size 288 .amdhsa_user_sgpr_count 15 .amdhsa_user_sgpr_dispatch_ptr 0 .amdhsa_user_sgpr_queue_ptr 0 .amdhsa_user_sgpr_kernarg_segment_ptr 1 .amdhsa_user_sgpr_dispatch_id 0 .amdhsa_user_sgpr_private_segment_size 0 .amdhsa_wavefront_size32 1 .amdhsa_uses_dynamic_stack 0 .amdhsa_enable_private_segment 0 .amdhsa_system_sgpr_workgroup_id_x 1 .amdhsa_system_sgpr_workgroup_id_y 0 .amdhsa_system_sgpr_workgroup_id_z 0 .amdhsa_system_sgpr_workgroup_info 0 .amdhsa_system_vgpr_workitem_id 0 .amdhsa_next_free_vgpr 8 .amdhsa_next_free_sgpr 16 .amdhsa_float_round_mode_32 0 .amdhsa_float_round_mode_16_64 0 .amdhsa_float_denorm_mode_32 3 .amdhsa_float_denorm_mode_16_64 3 .amdhsa_dx10_clamp 1 .amdhsa_ieee_mode 1 .amdhsa_fp16_overflow 0 .amdhsa_workgroup_processor_mode 1 .amdhsa_memory_ordered 1 .amdhsa_forward_progress 0 .amdhsa_shared_vgpr_count 0 .amdhsa_exception_fp_ieee_invalid_op 0 .amdhsa_exception_fp_denorm_src 0 .amdhsa_exception_fp_ieee_div_zero 0 .amdhsa_exception_fp_ieee_overflow 0 .amdhsa_exception_fp_ieee_underflow 0 .amdhsa_exception_fp_ieee_inexact 0 .amdhsa_exception_int_div_zero 0 .end_amdhsa_kernel .text .Lfunc_end0: .size _Z28NormalizationExecutionKernelPhPfifff, .Lfunc_end0-_Z28NormalizationExecutionKernelPhPfifff .section .AMDGPU.csdata,"",@progbits .text .p2alignl 7, 3214868480 .fill 96, 4, 3214868480 .type __hip_cuid_,@object .section .bss,"aw",@nobits .globl __hip_cuid_ __hip_cuid_: .byte 0 .size __hip_cuid_, 1 .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym __hip_cuid_ .amdgpu_metadata --- amdhsa.kernels: - .args: - .address_space: global .offset: 0 .size: 8 .value_kind: global_buffer - .address_space: global .offset: 8 .size: 8 .value_kind: global_buffer - .offset: 16 .size: 4 .value_kind: by_value - .offset: 20 .size: 4 .value_kind: by_value - .offset: 24 .size: 4 .value_kind: by_value - .offset: 28 .size: 4 .value_kind: by_value - .offset: 32 .size: 4 .value_kind: hidden_block_count_x - .offset: 36 .size: 4 .value_kind: hidden_block_count_y - .offset: 40 .size: 4 .value_kind: hidden_block_count_z - .offset: 44 .size: 2 .value_kind: hidden_group_size_x - .offset: 46 .size: 2 .value_kind: hidden_group_size_y - .offset: 48 .size: 2 .value_kind: hidden_group_size_z - .offset: 50 .size: 2 .value_kind: hidden_remainder_x - .offset: 52 .size: 2 .value_kind: hidden_remainder_y - .offset: 54 .size: 2 .value_kind: hidden_remainder_z - .offset: 72 .size: 8 .value_kind: hidden_global_offset_x - .offset: 80 .size: 8 .value_kind: hidden_global_offset_y - .offset: 88 .size: 8 .value_kind: hidden_global_offset_z - .offset: 96 .size: 2 .value_kind: hidden_grid_dims .group_segment_fixed_size: 0 .kernarg_segment_align: 8 .kernarg_segment_size: 288 .language: OpenCL C .language_version: - 2 - 0 .max_flat_workgroup_size: 1024 .name: _Z28NormalizationExecutionKernelPhPfifff .private_segment_fixed_size: 0 .sgpr_count: 18 .sgpr_spill_count: 0 .symbol: _Z28NormalizationExecutionKernelPhPfifff.kd .uniform_work_group_size: 1 .uses_dynamic_stack: false .vgpr_count: 8 .vgpr_spill_count: 0 .wavefront_size: 32 .workgroup_processor_mode: 1 amdhsa.target: amdgcn-amd-amdhsa--gfx1100 amdhsa.version: - 1 - 2 ... .end_amdgpu_metadata
You are a helpful coding assistant specialized in converting from CUDA host assembly to AMD host assembly.
.file "tmpxft_001510f0_00000000-6_NormalizationExecutionKernel.cudafe1.cpp" .text #APP #NO_APP .type _ZL26__cudaUnregisterBinaryUtilv, @function _ZL26__cudaUnregisterBinaryUtilv: .LFB2029: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaUnregisterFatBinary@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2029: .size _ZL26__cudaUnregisterBinaryUtilv, .-_ZL26__cudaUnregisterBinaryUtilv .globl _Z54__device_stub__Z28NormalizationExecutionKernelPhPfifffPhPfifff .type _Z54__device_stub__Z28NormalizationExecutionKernelPhPfifffPhPfifff, @function _Z54__device_stub__Z28NormalizationExecutionKernelPhPfifffPhPfifff: .LFB2051: .cfi_startproc endbr64 subq $168, %rsp .cfi_def_cfa_offset 176 movq %rdi, 24(%rsp) movq %rsi, 16(%rsp) movl %edx, 12(%rsp) movss %xmm0, 8(%rsp) movss %xmm1, 4(%rsp) movss %xmm2, (%rsp) movq %fs:40, %rax movq %rax, 152(%rsp) xorl %eax, %eax leaq 24(%rsp), %rax movq %rax, 96(%rsp) leaq 16(%rsp), %rax movq %rax, 104(%rsp) leaq 12(%rsp), %rax movq %rax, 112(%rsp) leaq 8(%rsp), %rax movq %rax, 120(%rsp) leaq 4(%rsp), %rax movq %rax, 128(%rsp) movq %rsp, %rax movq %rax, 136(%rsp) movl $1, 48(%rsp) movl $1, 52(%rsp) movl $1, 56(%rsp) movl $1, 60(%rsp) movl $1, 64(%rsp) movl $1, 68(%rsp) leaq 40(%rsp), %rcx leaq 32(%rsp), %rdx leaq 60(%rsp), %rsi leaq 48(%rsp), %rdi call __cudaPopCallConfiguration@PLT testl %eax, %eax je .L7 .L3: movq 152(%rsp), %rax subq %fs:40, %rax jne .L8 addq $168, %rsp .cfi_remember_state .cfi_def_cfa_offset 8 ret .L7: .cfi_restore_state pushq 40(%rsp) .cfi_def_cfa_offset 184 pushq 40(%rsp) .cfi_def_cfa_offset 192 leaq 112(%rsp), %r9 movq 76(%rsp), %rcx movl 84(%rsp), %r8d movq 64(%rsp), %rsi movl 72(%rsp), %edx leaq _Z28NormalizationExecutionKernelPhPfifff(%rip), %rdi call cudaLaunchKernel@PLT addq $16, %rsp .cfi_def_cfa_offset 176 jmp .L3 .L8: call __stack_chk_fail@PLT .cfi_endproc .LFE2051: .size _Z54__device_stub__Z28NormalizationExecutionKernelPhPfifffPhPfifff, .-_Z54__device_stub__Z28NormalizationExecutionKernelPhPfifffPhPfifff .globl _Z28NormalizationExecutionKernelPhPfifff .type _Z28NormalizationExecutionKernelPhPfifff, @function _Z28NormalizationExecutionKernelPhPfifff: .LFB2052: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 call _Z54__device_stub__Z28NormalizationExecutionKernelPhPfifffPhPfifff addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2052: .size _Z28NormalizationExecutionKernelPhPfifff, .-_Z28NormalizationExecutionKernelPhPfifff .section .rodata.str1.8,"aMS",@progbits,1 .align 8 .LC0: .string "_Z28NormalizationExecutionKernelPhPfifff" .text .type _ZL24__sti____cudaRegisterAllv, @function _ZL24__sti____cudaRegisterAllv: .LFB2054: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 leaq _ZL15__fatDeviceText(%rip), %rdi call __cudaRegisterFatBinary@PLT movq %rax, %rdi movq %rax, _ZL20__cudaFatCubinHandle(%rip) pushq $0 .cfi_def_cfa_offset 24 pushq $0 .cfi_def_cfa_offset 32 pushq $0 .cfi_def_cfa_offset 40 pushq $0 .cfi_def_cfa_offset 48 movl $0, %r9d movl $-1, %r8d leaq .LC0(%rip), %rdx movq %rdx, %rcx leaq _Z28NormalizationExecutionKernelPhPfifff(%rip), %rsi call __cudaRegisterFunction@PLT addq $32, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaRegisterFatBinaryEnd@PLT leaq _ZL26__cudaUnregisterBinaryUtilv(%rip), %rdi call atexit@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2054: .size _ZL24__sti____cudaRegisterAllv, .-_ZL24__sti____cudaRegisterAllv .section .init_array,"aw" .align 8 .quad _ZL24__sti____cudaRegisterAllv .section .nvFatBinSegment,"aw" .align 8 .type _ZL15__fatDeviceText, @object .size _ZL15__fatDeviceText, 24 _ZL15__fatDeviceText: .long 1180844977 .long 1 .quad fatbinData .quad 0 .local _ZL20__cudaFatCubinHandle .comm _ZL20__cudaFatCubinHandle,8,8 .ident "GCC: (Ubuntu 13.3.0-6ubuntu2~24.04) 13.3.0" .section .note.GNU-stack,"",@progbits .section .note.gnu.property,"a" .align 8 .long 1f - 0f .long 4f - 1f .long 5 0: .string "GNU" 1: .align 8 .long 0xc0000002 .long 3f - 2f 2: .long 0x3 3: .align 8 4:
.text .file "NormalizationExecutionKernel.hip" .globl _Z43__device_stub__NormalizationExecutionKernelPhPfifff # -- Begin function _Z43__device_stub__NormalizationExecutionKernelPhPfifff .p2align 4, 0x90 .type _Z43__device_stub__NormalizationExecutionKernelPhPfifff,@function _Z43__device_stub__NormalizationExecutionKernelPhPfifff: # @_Z43__device_stub__NormalizationExecutionKernelPhPfifff .cfi_startproc # %bb.0: subq $136, %rsp .cfi_def_cfa_offset 144 movq %rdi, 72(%rsp) movq %rsi, 64(%rsp) movl %edx, 12(%rsp) movss %xmm0, 8(%rsp) movss %xmm1, 4(%rsp) movss %xmm2, (%rsp) leaq 72(%rsp), %rax movq %rax, 80(%rsp) leaq 64(%rsp), %rax movq %rax, 88(%rsp) leaq 12(%rsp), %rax movq %rax, 96(%rsp) leaq 8(%rsp), %rax movq %rax, 104(%rsp) leaq 4(%rsp), %rax movq %rax, 112(%rsp) movq %rsp, %rax movq %rax, 120(%rsp) leaq 48(%rsp), %rdi leaq 32(%rsp), %rsi leaq 24(%rsp), %rdx leaq 16(%rsp), %rcx callq __hipPopCallConfiguration movq 48(%rsp), %rsi movl 56(%rsp), %edx movq 32(%rsp), %rcx movl 40(%rsp), %r8d leaq 80(%rsp), %r9 movl $_Z28NormalizationExecutionKernelPhPfifff, %edi pushq 16(%rsp) .cfi_adjust_cfa_offset 8 pushq 32(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $152, %rsp .cfi_adjust_cfa_offset -152 retq .Lfunc_end0: .size _Z43__device_stub__NormalizationExecutionKernelPhPfifff, .Lfunc_end0-_Z43__device_stub__NormalizationExecutionKernelPhPfifff .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_ctor .type __hip_module_ctor,@function __hip_module_ctor: # @__hip_module_ctor .cfi_startproc # %bb.0: subq $40, %rsp .cfi_def_cfa_offset 48 cmpq $0, __hip_gpubin_handle(%rip) jne .LBB1_2 # %bb.1: movl $__hip_fatbin_wrapper, %edi callq __hipRegisterFatBinary movq %rax, __hip_gpubin_handle(%rip) .LBB1_2: movq __hip_gpubin_handle(%rip), %rdi xorps %xmm0, %xmm0 movups %xmm0, 16(%rsp) movups %xmm0, (%rsp) movl $_Z28NormalizationExecutionKernelPhPfifff, %esi movl $.L__unnamed_1, %edx movl $.L__unnamed_1, %ecx movl $-1, %r8d xorl %r9d, %r9d callq __hipRegisterFunction movl $__hip_module_dtor, %edi addq $40, %rsp .cfi_def_cfa_offset 8 jmp atexit # TAILCALL .Lfunc_end1: .size __hip_module_ctor, .Lfunc_end1-__hip_module_ctor .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_dtor .type __hip_module_dtor,@function __hip_module_dtor: # @__hip_module_dtor .cfi_startproc # %bb.0: movq __hip_gpubin_handle(%rip), %rdi testq %rdi, %rdi je .LBB2_2 # %bb.1: pushq %rax .cfi_def_cfa_offset 16 callq __hipUnregisterFatBinary movq $0, __hip_gpubin_handle(%rip) addq $8, %rsp .cfi_def_cfa_offset 8 .LBB2_2: retq .Lfunc_end2: .size __hip_module_dtor, .Lfunc_end2-__hip_module_dtor .cfi_endproc # -- End function .type _Z28NormalizationExecutionKernelPhPfifff,@object # @_Z28NormalizationExecutionKernelPhPfifff .section .rodata,"a",@progbits .globl _Z28NormalizationExecutionKernelPhPfifff .p2align 3, 0x0 _Z28NormalizationExecutionKernelPhPfifff: .quad _Z43__device_stub__NormalizationExecutionKernelPhPfifff .size _Z28NormalizationExecutionKernelPhPfifff, 8 .type .L__unnamed_1,@object # @0 .section .rodata.str1.1,"aMS",@progbits,1 .L__unnamed_1: .asciz "_Z28NormalizationExecutionKernelPhPfifff" .size .L__unnamed_1, 41 .type __hip_fatbin_wrapper,@object # @__hip_fatbin_wrapper .section .hipFatBinSegment,"a",@progbits .p2align 3, 0x0 __hip_fatbin_wrapper: .long 1212764230 # 0x48495046 .long 1 # 0x1 .quad __hip_fatbin .quad 0 .size __hip_fatbin_wrapper, 24 .type __hip_gpubin_handle,@object # @__hip_gpubin_handle .local __hip_gpubin_handle .comm __hip_gpubin_handle,8,8 .section .init_array,"aw",@init_array .p2align 3, 0x0 .quad __hip_module_ctor .type __hip_cuid_,@object # @__hip_cuid_ .bss .globl __hip_cuid_ __hip_cuid_: .byte 0 # 0x0 .size __hip_cuid_, 1 .section ".linker-options","e",@llvm_linker_options .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym _Z43__device_stub__NormalizationExecutionKernelPhPfifff .addrsig_sym __hip_module_ctor .addrsig_sym __hip_module_dtor .addrsig_sym _Z28NormalizationExecutionKernelPhPfifff .addrsig_sym __hip_fatbin .addrsig_sym __hip_fatbin_wrapper .addrsig_sym __hip_cuid_
You are a helpful coding assistant specialized in converting CUDA source code to CUDA device assembly.
#include <stdio.h> #include "sys/time.h" #define GPU_ID 0 // #define USE_SINGLE_PRECISION /* Comment this line using "!" if you want to use double precision. */ #ifdef USE_SINGLE_PRECISION #define DAT float #define PRECIS 4 #else #define DAT double #define PRECIS 8 #endif #define zeros(A,nx,ny,nz) DAT *A##_d,*A##_h; A##_h = (DAT*)malloc((nx)*(ny)*(nz)*sizeof(DAT)); \ for(i=0; i < (nx)*(ny)*(nz); i++){ A##_h[i]=(DAT)0.0; } \ cudaMalloc(&A##_d ,(nx)*(ny)*(nz)*sizeof(DAT)); \ cudaMemcpy( A##_d,A##_h,(nx)*(ny)*(nz)*sizeof(DAT),cudaMemcpyHostToDevice); #define free_all(A) free(A##_h);cudaFree(A##_d); #define BLOCK_X 32 #define BLOCK_Y 16 #define BLOCK_Z 2 #define GRID_X 32 #define GRID_Y 64 #define GRID_Z 128 const int nx = GRID_X*BLOCK_X; const int ny = GRID_Y*BLOCK_Y; const int nz = GRID_Z*BLOCK_Z; const int nt = 100; // Timer double timer_start = 0; double cpu_sec(){ struct timeval tp; gettimeofday(&tp,NULL); return tp.tv_sec+1e-6*tp.tv_usec; } void tic(){ timer_start = cpu_sec(); } double toc(){ return cpu_sec()-timer_start; } void tim(const char *what, double n){ double s=toc(); printf("%s: %8.3f seconds",what,s);if(n>0)printf(", %8.3f GB/s", n/s); printf("\n"); } void timPrint(const char *what, double n, int nx, int ny, int nz){ double s=toc(); printf("%s: %8.3f seconds",what,s);if(n>0)printf(", %8.3f GB/s", n/s); printf("\n"); FILE*fid; fid=fopen("PERF_memcpy.dat","a"); fprintf(fid,"nx=%d ny=%d nz=%d GBs=%1.4f time_s=%1.4f \n", nx, ny, nz, n/s, s); fclose(fid); } void clean_cuda(){ cudaError_t ce = cudaGetLastError(); if(ce != cudaSuccess){ printf("ERROR launching GPU C-CUDA program: %s\n", cudaGetErrorString(ce)); cudaDeviceReset();} } __global__ void memcopy(DAT*A, DAT*B, DAT*C, const int nx,const int ny,const int nz){ int ix = blockIdx.x*blockDim.x + threadIdx.x; // thread ID, dimension x int iy = blockIdx.y*blockDim.y + threadIdx.y; // thread ID, dimension x int iz = blockIdx.z*blockDim.z + threadIdx.z; // thread ID, dimension x if (iz<nz && iy<ny && ix<nx) A[ix + iy*nx + iz*nx*ny] = B[ix + iy*nx + iz*nx*ny] + C[ix + iy*nx + iz*nx*ny]; } ////////// main ////////// int main(){ size_t i, it, N=nx*ny*nz, mem=3*N*sizeof(DAT); dim3 grid, block; block.x = BLOCK_X; block.y = BLOCK_Y; block.z = BLOCK_Z; grid.x = GRID_X; grid.y = GRID_Y; grid.z = GRID_Z; int gpu_id=-1; gpu_id=GPU_ID; cudaSetDevice(gpu_id); cudaGetDevice(&gpu_id); cudaDeviceReset(); cudaDeviceSetCacheConfig(cudaFuncCachePreferL1); // set L1 to prefered printf("Process uses GPU with id %d.\n",gpu_id); printf("%dx%dx%d, %1.3f GB, %d iterations.\n", nx,ny,nz, mem/1024./1024./1024., nt); printf("launching (%dx%dx%d) grid of (%dx%dx%d) blocks.\n", grid.x, grid.y, grid.z, block.x, block.y, block.z); // initializations zeros(A, nx,ny,nz); zeros(B, nx,ny,nz); zeros(C, nx,ny,nz); // time loop for(it=0; it<nt; it++){ if (it==10){ tic(); } memcopy<<<grid, block>>>(A_d,B_d,C_d,nx,ny,nz); cudaDeviceSynchronize(); } // tim("Performance", mem*(nt-3)*2/1024./1024./1024.); timPrint("Performance", mem*(nt-10)/1024./1024./1024.,nx,ny,nz); free_all(A); free_all(B); free_all(C); clean_cuda(); return 0; }
code for sm_80 Function : _Z7memcopyPdS_S_iii .headerflags @"EF_CUDA_TEXMODE_UNIFIED EF_CUDA_64BIT_ADDRESS EF_CUDA_SM80 EF_CUDA_VIRTUAL_SM(EF_CUDA_SM80)" /*0000*/ MOV R1, c[0x0][0x28] ; /* 0x00000a0000017a02 */ /* 0x000fe40000000f00 */ /*0010*/ S2R R3, SR_CTAID.Z ; /* 0x0000000000037919 */ /* 0x000e280000002700 */ /*0020*/ S2R R4, SR_TID.Z ; /* 0x0000000000047919 */ /* 0x000e280000002300 */ /*0030*/ S2R R2, SR_CTAID.Y ; /* 0x0000000000027919 */ /* 0x000e680000002600 */ /*0040*/ S2R R7, SR_TID.Y ; /* 0x0000000000077919 */ /* 0x000e680000002200 */ /*0050*/ S2R R0, SR_CTAID.X ; /* 0x0000000000007919 */ /* 0x000ea80000002500 */ /*0060*/ S2R R5, SR_TID.X ; /* 0x0000000000057919 */ /* 0x000ea20000002100 */ /*0070*/ IMAD R3, R3, c[0x0][0x8], R4 ; /* 0x0000020003037a24 */ /* 0x001fca00078e0204 */ /*0080*/ ISETP.GE.AND P0, PT, R3, c[0x0][0x180], PT ; /* 0x0000600003007a0c */ /* 0x000fe20003f06270 */ /*0090*/ IMAD R2, R2, c[0x0][0x4], R7 ; /* 0x0000010002027a24 */ /* 0x002fca00078e0207 */ /*00a0*/ ISETP.GE.OR P0, PT, R2, c[0x0][0x17c], P0 ; /* 0x00005f0002007a0c */ /* 0x000fe20000706670 */ /*00b0*/ IMAD R0, R0, c[0x0][0x0], R5 ; /* 0x0000000000007a24 */ /* 0x004fca00078e0205 */ /*00c0*/ ISETP.GE.OR P0, PT, R0, c[0x0][0x178], P0 ; /* 0x00005e0000007a0c */ /* 0x000fda0000706670 */ /*00d0*/ @P0 EXIT ; /* 0x000000000000094d */ /* 0x000fea0003800000 */ /*00e0*/ HFMA2.MMA R9, -RZ, RZ, 0, 4.76837158203125e-07 ; /* 0x00000008ff097435 */ /* 0x000fe200000001ff */ /*00f0*/ IMAD R3, R3, c[0x0][0x17c], R2 ; /* 0x00005f0003037a24 */ /* 0x000fe200078e0202 */ /*0100*/ ULDC.64 UR4, c[0x0][0x118] ; /* 0x0000460000047ab9 */ /* 0x000fc60000000a00 */ /*0110*/ IMAD R0, R3, c[0x0][0x178], R0 ; /* 0x00005e0003007a24 */ /* 0x000fca00078e0200 */ /*0120*/ IMAD.WIDE R4, R0, R9, c[0x0][0x170] ; /* 0x00005c0000047625 */ /* 0x000fc800078e0209 */ /*0130*/ IMAD.WIDE R2, R0.reuse, R9.reuse, c[0x0][0x168] ; /* 0x00005a0000027625 */ /* 0x0c0fe400078e0209 */ /*0140*/ LDG.E.64 R4, [R4.64] ; /* 0x0000000404047981 */ /* 0x000ea8000c1e1b00 */ /*0150*/ LDG.E.64 R2, [R2.64] ; /* 0x0000000402027981 */ /* 0x000ea2000c1e1b00 */ /*0160*/ IMAD.WIDE R8, R0, R9, c[0x0][0x160] ; /* 0x0000580000087625 */ /* 0x000fe200078e0209 */ /*0170*/ DADD R6, R4, R2 ; /* 0x0000000004067229 */ /* 0x004e0e0000000002 */ /*0180*/ STG.E.64 [R8.64], R6 ; /* 0x0000000608007986 */ /* 0x001fe2000c101b04 */ /*0190*/ EXIT ; /* 0x000000000000794d */ /* 0x000fea0003800000 */ /*01a0*/ BRA 0x1a0; /* 0xfffffff000007947 */ /* 0x000fc0000383ffff */ /*01b0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*01c0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*01d0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*01e0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*01f0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0200*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0210*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0220*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0230*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0240*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0250*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0260*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0270*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ ..........
You are a helpful coding assistant specialized in converting CUDA source code to CUDA host assembly.
#include <stdio.h> #include "sys/time.h" #define GPU_ID 0 // #define USE_SINGLE_PRECISION /* Comment this line using "!" if you want to use double precision. */ #ifdef USE_SINGLE_PRECISION #define DAT float #define PRECIS 4 #else #define DAT double #define PRECIS 8 #endif #define zeros(A,nx,ny,nz) DAT *A##_d,*A##_h; A##_h = (DAT*)malloc((nx)*(ny)*(nz)*sizeof(DAT)); \ for(i=0; i < (nx)*(ny)*(nz); i++){ A##_h[i]=(DAT)0.0; } \ cudaMalloc(&A##_d ,(nx)*(ny)*(nz)*sizeof(DAT)); \ cudaMemcpy( A##_d,A##_h,(nx)*(ny)*(nz)*sizeof(DAT),cudaMemcpyHostToDevice); #define free_all(A) free(A##_h);cudaFree(A##_d); #define BLOCK_X 32 #define BLOCK_Y 16 #define BLOCK_Z 2 #define GRID_X 32 #define GRID_Y 64 #define GRID_Z 128 const int nx = GRID_X*BLOCK_X; const int ny = GRID_Y*BLOCK_Y; const int nz = GRID_Z*BLOCK_Z; const int nt = 100; // Timer double timer_start = 0; double cpu_sec(){ struct timeval tp; gettimeofday(&tp,NULL); return tp.tv_sec+1e-6*tp.tv_usec; } void tic(){ timer_start = cpu_sec(); } double toc(){ return cpu_sec()-timer_start; } void tim(const char *what, double n){ double s=toc(); printf("%s: %8.3f seconds",what,s);if(n>0)printf(", %8.3f GB/s", n/s); printf("\n"); } void timPrint(const char *what, double n, int nx, int ny, int nz){ double s=toc(); printf("%s: %8.3f seconds",what,s);if(n>0)printf(", %8.3f GB/s", n/s); printf("\n"); FILE*fid; fid=fopen("PERF_memcpy.dat","a"); fprintf(fid,"nx=%d ny=%d nz=%d GBs=%1.4f time_s=%1.4f \n", nx, ny, nz, n/s, s); fclose(fid); } void clean_cuda(){ cudaError_t ce = cudaGetLastError(); if(ce != cudaSuccess){ printf("ERROR launching GPU C-CUDA program: %s\n", cudaGetErrorString(ce)); cudaDeviceReset();} } __global__ void memcopy(DAT*A, DAT*B, DAT*C, const int nx,const int ny,const int nz){ int ix = blockIdx.x*blockDim.x + threadIdx.x; // thread ID, dimension x int iy = blockIdx.y*blockDim.y + threadIdx.y; // thread ID, dimension x int iz = blockIdx.z*blockDim.z + threadIdx.z; // thread ID, dimension x if (iz<nz && iy<ny && ix<nx) A[ix + iy*nx + iz*nx*ny] = B[ix + iy*nx + iz*nx*ny] + C[ix + iy*nx + iz*nx*ny]; } ////////// main ////////// int main(){ size_t i, it, N=nx*ny*nz, mem=3*N*sizeof(DAT); dim3 grid, block; block.x = BLOCK_X; block.y = BLOCK_Y; block.z = BLOCK_Z; grid.x = GRID_X; grid.y = GRID_Y; grid.z = GRID_Z; int gpu_id=-1; gpu_id=GPU_ID; cudaSetDevice(gpu_id); cudaGetDevice(&gpu_id); cudaDeviceReset(); cudaDeviceSetCacheConfig(cudaFuncCachePreferL1); // set L1 to prefered printf("Process uses GPU with id %d.\n",gpu_id); printf("%dx%dx%d, %1.3f GB, %d iterations.\n", nx,ny,nz, mem/1024./1024./1024., nt); printf("launching (%dx%dx%d) grid of (%dx%dx%d) blocks.\n", grid.x, grid.y, grid.z, block.x, block.y, block.z); // initializations zeros(A, nx,ny,nz); zeros(B, nx,ny,nz); zeros(C, nx,ny,nz); // time loop for(it=0; it<nt; it++){ if (it==10){ tic(); } memcopy<<<grid, block>>>(A_d,B_d,C_d,nx,ny,nz); cudaDeviceSynchronize(); } // tim("Performance", mem*(nt-3)*2/1024./1024./1024.); timPrint("Performance", mem*(nt-10)/1024./1024./1024.,nx,ny,nz); free_all(A); free_all(B); free_all(C); clean_cuda(); return 0; }
.file "tmpxft_0001762b_00000000-6_memcopy.cudafe1.cpp" .text #APP #NO_APP .type _ZL26__cudaUnregisterBinaryUtilv, @function _ZL26__cudaUnregisterBinaryUtilv: .LFB2066: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaUnregisterFatBinary@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2066: .size _ZL26__cudaUnregisterBinaryUtilv, .-_ZL26__cudaUnregisterBinaryUtilv .globl _Z7cpu_secv .type _Z7cpu_secv, @function _Z7cpu_secv: .LFB2057: .cfi_startproc endbr64 subq $40, %rsp .cfi_def_cfa_offset 48 movq %fs:40, %rax movq %rax, 24(%rsp) xorl %eax, %eax movq %rsp, %rdi movl $0, %esi call gettimeofday@PLT pxor %xmm0, %xmm0 cvtsi2sdq 8(%rsp), %xmm0 mulsd .LC0(%rip), %xmm0 pxor %xmm1, %xmm1 cvtsi2sdq (%rsp), %xmm1 addsd %xmm1, %xmm0 movq 24(%rsp), %rax subq %fs:40, %rax jne .L6 addq $40, %rsp .cfi_remember_state .cfi_def_cfa_offset 8 ret .L6: .cfi_restore_state call __stack_chk_fail@PLT .cfi_endproc .LFE2057: .size _Z7cpu_secv, .-_Z7cpu_secv .globl _Z3ticv .type _Z3ticv, @function _Z3ticv: .LFB2058: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 call _Z7cpu_secv movsd %xmm0, timer_start(%rip) addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2058: .size _Z3ticv, .-_Z3ticv .globl _Z3tocv .type _Z3tocv, @function _Z3tocv: .LFB2059: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 call _Z7cpu_secv subsd timer_start(%rip), %xmm0 addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2059: .size _Z3tocv, .-_Z3tocv .section .rodata.str1.1,"aMS",@progbits,1 .LC1: .string "%s: %8.3f seconds" .LC3: .string ", %8.3f GB/s" .LC4: .string "\n" .text .globl _Z3timPKcd .type _Z3timPKcd, @function _Z3timPKcd: .LFB2060: .cfi_startproc endbr64 pushq %rbx .cfi_def_cfa_offset 16 .cfi_offset 3, -16 subq $16, %rsp .cfi_def_cfa_offset 32 movq %rdi, %rbx movsd %xmm0, (%rsp) call _Z3tocv movsd %xmm0, 8(%rsp) movq %rbx, %rdx leaq .LC1(%rip), %rsi movl $2, %edi movl $1, %eax call __printf_chk@PLT pxor %xmm0, %xmm0 movsd (%rsp), %xmm1 comisd %xmm0, %xmm1 ja .L16 .L12: leaq .LC4(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT addq $16, %rsp .cfi_remember_state .cfi_def_cfa_offset 16 popq %rbx .cfi_def_cfa_offset 8 ret .L16: .cfi_restore_state divsd 8(%rsp), %xmm1 movapd %xmm1, %xmm0 leaq .LC3(%rip), %rsi movl $2, %edi movl $1, %eax call __printf_chk@PLT jmp .L12 .cfi_endproc .LFE2060: .size _Z3timPKcd, .-_Z3timPKcd .section .rodata.str1.1 .LC5: .string "a" .LC6: .string "PERF_memcpy.dat" .section .rodata.str1.8,"aMS",@progbits,1 .align 8 .LC7: .string "nx=%d ny=%d nz=%d GBs=%1.4f time_s=%1.4f \n" .text .globl _Z8timPrintPKcdiii .type _Z8timPrintPKcdiii, @function _Z8timPrintPKcdiii: .LFB2061: .cfi_startproc endbr64 pushq %r13 .cfi_def_cfa_offset 16 .cfi_offset 13, -16 pushq %r12 .cfi_def_cfa_offset 24 .cfi_offset 12, -24 pushq %rbp .cfi_def_cfa_offset 32 .cfi_offset 6, -32 pushq %rbx .cfi_def_cfa_offset 40 .cfi_offset 3, -40 subq $24, %rsp .cfi_def_cfa_offset 64 movq %rdi, %rbx movsd %xmm0, (%rsp) movl %esi, %ebp movl %edx, %r12d movl %ecx, %r13d call _Z3tocv movsd %xmm0, 8(%rsp) movq %rbx, %rdx leaq .LC1(%rip), %rsi movl $2, %edi movl $1, %eax call __printf_chk@PLT pxor %xmm0, %xmm0 movsd (%rsp), %xmm2 comisd %xmm0, %xmm2 ja .L22 .L18: leaq .LC4(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT leaq .LC5(%rip), %rsi leaq .LC6(%rip), %rdi call fopen@PLT movq %rax, %rbx movsd (%rsp), %xmm0 movsd 8(%rsp), %xmm1 divsd %xmm1, %xmm0 movl %r13d, %r9d movl %r12d, %r8d movl %ebp, %ecx leaq .LC7(%rip), %rdx movl $2, %esi movq %rax, %rdi movl $2, %eax call __fprintf_chk@PLT movq %rbx, %rdi call fclose@PLT addq $24, %rsp .cfi_remember_state .cfi_def_cfa_offset 40 popq %rbx .cfi_def_cfa_offset 32 popq %rbp .cfi_def_cfa_offset 24 popq %r12 .cfi_def_cfa_offset 16 popq %r13 .cfi_def_cfa_offset 8 ret .L22: .cfi_restore_state divsd 8(%rsp), %xmm2 movapd %xmm2, %xmm0 leaq .LC3(%rip), %rsi movl $2, %edi movl $1, %eax call __printf_chk@PLT jmp .L18 .cfi_endproc .LFE2061: .size _Z8timPrintPKcdiii, .-_Z8timPrintPKcdiii .section .rodata.str1.8 .align 8 .LC8: .string "ERROR launching GPU C-CUDA program: %s\n" .text .globl _Z10clean_cudav .type _Z10clean_cudav, @function _Z10clean_cudav: .LFB2062: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 call cudaGetLastError@PLT testl %eax, %eax jne .L26 .L23: addq $8, %rsp .cfi_remember_state .cfi_def_cfa_offset 8 ret .L26: .cfi_restore_state movl %eax, %edi call cudaGetErrorString@PLT movq %rax, %rdx leaq .LC8(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT call cudaDeviceReset@PLT jmp .L23 .cfi_endproc .LFE2062: .size _Z10clean_cudav, .-_Z10clean_cudav .globl _Z33__device_stub__Z7memcopyPdS_S_iiiPdS_S_iii .type _Z33__device_stub__Z7memcopyPdS_S_iiiPdS_S_iii, @function _Z33__device_stub__Z7memcopyPdS_S_iiiPdS_S_iii: .LFB2088: .cfi_startproc endbr64 subq $184, %rsp .cfi_def_cfa_offset 192 movq %rdi, 40(%rsp) movq %rsi, 32(%rsp) movq %rdx, 24(%rsp) movl %ecx, 20(%rsp) movl %r8d, 16(%rsp) movl %r9d, 12(%rsp) movq %fs:40, %rax movq %rax, 168(%rsp) xorl %eax, %eax leaq 40(%rsp), %rax movq %rax, 112(%rsp) leaq 32(%rsp), %rax movq %rax, 120(%rsp) leaq 24(%rsp), %rax movq %rax, 128(%rsp) leaq 20(%rsp), %rax movq %rax, 136(%rsp) leaq 16(%rsp), %rax movq %rax, 144(%rsp) leaq 12(%rsp), %rax movq %rax, 152(%rsp) movl $1, 64(%rsp) movl $1, 68(%rsp) movl $1, 72(%rsp) movl $1, 76(%rsp) movl $1, 80(%rsp) movl $1, 84(%rsp) leaq 56(%rsp), %rcx leaq 48(%rsp), %rdx leaq 76(%rsp), %rsi leaq 64(%rsp), %rdi call __cudaPopCallConfiguration@PLT testl %eax, %eax je .L31 .L27: movq 168(%rsp), %rax subq %fs:40, %rax jne .L32 addq $184, %rsp .cfi_remember_state .cfi_def_cfa_offset 8 ret .L31: .cfi_restore_state pushq 56(%rsp) .cfi_def_cfa_offset 200 pushq 56(%rsp) .cfi_def_cfa_offset 208 leaq 128(%rsp), %r9 movq 92(%rsp), %rcx movl 100(%rsp), %r8d movq 80(%rsp), %rsi movl 88(%rsp), %edx leaq _Z7memcopyPdS_S_iii(%rip), %rdi call cudaLaunchKernel@PLT addq $16, %rsp .cfi_def_cfa_offset 192 jmp .L27 .L32: call __stack_chk_fail@PLT .cfi_endproc .LFE2088: .size _Z33__device_stub__Z7memcopyPdS_S_iiiPdS_S_iii, .-_Z33__device_stub__Z7memcopyPdS_S_iiiPdS_S_iii .globl _Z7memcopyPdS_S_iii .type _Z7memcopyPdS_S_iii, @function _Z7memcopyPdS_S_iii: .LFB2089: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 call _Z33__device_stub__Z7memcopyPdS_S_iiiPdS_S_iii addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2089: .size _Z7memcopyPdS_S_iii, .-_Z7memcopyPdS_S_iii .section .rodata.str1.1 .LC9: .string "Process uses GPU with id %d.\n" .section .rodata.str1.8 .align 8 .LC11: .string "%dx%dx%d, %1.3f GB, %d iterations.\n" .align 8 .LC12: .string "launching (%dx%dx%d) grid of (%dx%dx%d) blocks.\n" .section .rodata.str1.1 .LC14: .string "Performance" .text .globl main .type main, @function main: .LFB2063: .cfi_startproc endbr64 pushq %r13 .cfi_def_cfa_offset 16 .cfi_offset 13, -16 pushq %r12 .cfi_def_cfa_offset 24 .cfi_offset 12, -24 pushq %rbp .cfi_def_cfa_offset 32 .cfi_offset 6, -32 pushq %rbx .cfi_def_cfa_offset 40 .cfi_offset 3, -40 subq $72, %rsp .cfi_def_cfa_offset 112 movq %fs:40, %rax movq %rax, 56(%rsp) xorl %eax, %eax movl $0, 4(%rsp) movl $0, %edi call cudaSetDevice@PLT leaq 4(%rsp), %rdi call cudaGetDevice@PLT call cudaDeviceReset@PLT movl $2, %edi call cudaDeviceSetCacheConfig@PLT movl 4(%rsp), %edx leaq .LC9(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT movl $100, %r9d movsd .LC10(%rip), %xmm0 movl $256, %r8d movl $1024, %ecx movl $1024, %edx leaq .LC11(%rip), %rsi movl $2, %edi movl $1, %eax call __printf_chk@PLT pushq $2 .cfi_def_cfa_offset 120 pushq $16 .cfi_def_cfa_offset 128 movl $32, %r9d movl $128, %r8d movl $64, %ecx movl $32, %edx leaq .LC12(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT movl $2147483648, %ebx movq %rbx, %rdi call malloc@PLT movq %rax, %rbp leaq (%rax,%rbx), %rdx addq $16, %rsp .cfi_def_cfa_offset 112 .L36: movq $0x000000000, (%rax) addq $8, %rax cmpq %rdx, %rax jne .L36 leaq 8(%rsp), %rdi movl $2147483648, %ebx movq %rbx, %rsi call cudaMalloc@PLT movl $1, %ecx movq %rbx, %rdx movq %rbp, %rsi movq 8(%rsp), %rdi call cudaMemcpy@PLT movq %rbx, %rdi call malloc@PLT movq %rax, %r12 leaq (%rax,%rbx), %rdx .L37: movq $0x000000000, (%rax) addq $8, %rax cmpq %rdx, %rax jne .L37 leaq 16(%rsp), %rdi movl $2147483648, %ebx movq %rbx, %rsi call cudaMalloc@PLT movl $1, %ecx movq %rbx, %rdx movq %r12, %rsi movq 16(%rsp), %rdi call cudaMemcpy@PLT movq %rbx, %rdi call malloc@PLT movq %rax, %r13 leaq (%rax,%rbx), %rdx .L38: movq $0x000000000, (%rax) addq $8, %rax cmpq %rdx, %rax jne .L38 leaq 24(%rsp), %rdi movl $2147483648, %ebx movq %rbx, %rsi call cudaMalloc@PLT movl $1, %ecx movq %rbx, %rdx movq %r13, %rsi movq 24(%rsp), %rdi call cudaMemcpy@PLT movl $0, %ebx jmp .L44 .L53: call _Z3ticv movl $32, 32(%rsp) movl $64, 36(%rsp) movl $32, 44(%rsp) movl $16, 48(%rsp) movl $0, %r9d movl $0, %r8d movq 44(%rsp), %rdx movl $2, %ecx movq 32(%rsp), %rdi movl $128, %esi call __cudaPushCallConfiguration@PLT testl %eax, %eax je .L41 call cudaDeviceSynchronize@PLT addq $1, %rbx jmp .L44 .L41: movl $256, %r9d movl $1024, %r8d movl $1024, %ecx movq 24(%rsp), %rdx movq 16(%rsp), %rsi movq 8(%rsp), %rdi call _Z33__device_stub__Z7memcopyPdS_S_iiiPdS_S_iii .L42: call cudaDeviceSynchronize@PLT addq $1, %rbx cmpq $100, %rbx je .L52 .L44: cmpq $10, %rbx je .L53 movl $32, 32(%rsp) movl $64, 36(%rsp) movl $32, 44(%rsp) movl $16, 48(%rsp) movl $0, %r9d movl $0, %r8d movq 44(%rsp), %rdx movl $2, %ecx movq 32(%rsp), %rdi movl $128, %esi call __cudaPushCallConfiguration@PLT testl %eax, %eax jne .L42 jmp .L41 .L52: movl $256, %ecx movl $1024, %edx movl $1024, %esi movsd .LC13(%rip), %xmm0 leaq .LC14(%rip), %rdi call _Z8timPrintPKcdiii movq %rbp, %rdi call free@PLT movq 8(%rsp), %rdi call cudaFree@PLT movq %r12, %rdi call free@PLT movq 16(%rsp), %rdi call cudaFree@PLT movq %r13, %rdi call free@PLT movq 24(%rsp), %rdi call cudaFree@PLT call _Z10clean_cudav movq 56(%rsp), %rax subq %fs:40, %rax jne .L54 movl $0, %eax addq $72, %rsp .cfi_remember_state .cfi_def_cfa_offset 40 popq %rbx .cfi_def_cfa_offset 32 popq %rbp .cfi_def_cfa_offset 24 popq %r12 .cfi_def_cfa_offset 16 popq %r13 .cfi_def_cfa_offset 8 ret .L54: .cfi_restore_state call __stack_chk_fail@PLT .cfi_endproc .LFE2063: .size main, .-main .section .rodata.str1.1 .LC15: .string "_Z7memcopyPdS_S_iii" .text .type _ZL24__sti____cudaRegisterAllv, @function _ZL24__sti____cudaRegisterAllv: .LFB2091: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 leaq _ZL15__fatDeviceText(%rip), %rdi call __cudaRegisterFatBinary@PLT movq %rax, %rdi movq %rax, _ZL20__cudaFatCubinHandle(%rip) pushq $0 .cfi_def_cfa_offset 24 pushq $0 .cfi_def_cfa_offset 32 pushq $0 .cfi_def_cfa_offset 40 pushq $0 .cfi_def_cfa_offset 48 movl $0, %r9d movl $-1, %r8d leaq .LC15(%rip), %rdx movq %rdx, %rcx leaq _Z7memcopyPdS_S_iii(%rip), %rsi call __cudaRegisterFunction@PLT addq $32, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaRegisterFatBinaryEnd@PLT leaq _ZL26__cudaUnregisterBinaryUtilv(%rip), %rdi call atexit@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2091: .size _ZL24__sti____cudaRegisterAllv, .-_ZL24__sti____cudaRegisterAllv .section .init_array,"aw" .align 8 .quad _ZL24__sti____cudaRegisterAllv .section .nvFatBinSegment,"aw" .align 8 .type _ZL15__fatDeviceText, @object .size _ZL15__fatDeviceText, 24 _ZL15__fatDeviceText: .long 1180844977 .long 1 .quad fatbinData .quad 0 .local _ZL20__cudaFatCubinHandle .comm _ZL20__cudaFatCubinHandle,8,8 .globl timer_start .bss .align 8 .type timer_start, @object .size timer_start, 8 timer_start: .zero 8 .section .rodata.cst8,"aM",@progbits,8 .align 8 .LC0: .long -1598689907 .long 1051772663 .align 8 .LC10: .long 0 .long 1075314688 .align 8 .LC13: .long 0 .long 1082187776 .ident "GCC: (Ubuntu 13.3.0-6ubuntu2~24.04) 13.3.0" .section .note.GNU-stack,"",@progbits .section .note.gnu.property,"a" .align 8 .long 1f - 0f .long 4f - 1f .long 5 0: .string "GNU" 1: .align 8 .long 0xc0000002 .long 3f - 2f 2: .long 0x3 3: .align 8 4:
You are a helpful coding assistant specialized in converting CUDA source code to HIP/AMD source code.
#include <stdio.h> #include "sys/time.h" #define GPU_ID 0 // #define USE_SINGLE_PRECISION /* Comment this line using "!" if you want to use double precision. */ #ifdef USE_SINGLE_PRECISION #define DAT float #define PRECIS 4 #else #define DAT double #define PRECIS 8 #endif #define zeros(A,nx,ny,nz) DAT *A##_d,*A##_h; A##_h = (DAT*)malloc((nx)*(ny)*(nz)*sizeof(DAT)); \ for(i=0; i < (nx)*(ny)*(nz); i++){ A##_h[i]=(DAT)0.0; } \ cudaMalloc(&A##_d ,(nx)*(ny)*(nz)*sizeof(DAT)); \ cudaMemcpy( A##_d,A##_h,(nx)*(ny)*(nz)*sizeof(DAT),cudaMemcpyHostToDevice); #define free_all(A) free(A##_h);cudaFree(A##_d); #define BLOCK_X 32 #define BLOCK_Y 16 #define BLOCK_Z 2 #define GRID_X 32 #define GRID_Y 64 #define GRID_Z 128 const int nx = GRID_X*BLOCK_X; const int ny = GRID_Y*BLOCK_Y; const int nz = GRID_Z*BLOCK_Z; const int nt = 100; // Timer double timer_start = 0; double cpu_sec(){ struct timeval tp; gettimeofday(&tp,NULL); return tp.tv_sec+1e-6*tp.tv_usec; } void tic(){ timer_start = cpu_sec(); } double toc(){ return cpu_sec()-timer_start; } void tim(const char *what, double n){ double s=toc(); printf("%s: %8.3f seconds",what,s);if(n>0)printf(", %8.3f GB/s", n/s); printf("\n"); } void timPrint(const char *what, double n, int nx, int ny, int nz){ double s=toc(); printf("%s: %8.3f seconds",what,s);if(n>0)printf(", %8.3f GB/s", n/s); printf("\n"); FILE*fid; fid=fopen("PERF_memcpy.dat","a"); fprintf(fid,"nx=%d ny=%d nz=%d GBs=%1.4f time_s=%1.4f \n", nx, ny, nz, n/s, s); fclose(fid); } void clean_cuda(){ cudaError_t ce = cudaGetLastError(); if(ce != cudaSuccess){ printf("ERROR launching GPU C-CUDA program: %s\n", cudaGetErrorString(ce)); cudaDeviceReset();} } __global__ void memcopy(DAT*A, DAT*B, DAT*C, const int nx,const int ny,const int nz){ int ix = blockIdx.x*blockDim.x + threadIdx.x; // thread ID, dimension x int iy = blockIdx.y*blockDim.y + threadIdx.y; // thread ID, dimension x int iz = blockIdx.z*blockDim.z + threadIdx.z; // thread ID, dimension x if (iz<nz && iy<ny && ix<nx) A[ix + iy*nx + iz*nx*ny] = B[ix + iy*nx + iz*nx*ny] + C[ix + iy*nx + iz*nx*ny]; } ////////// main ////////// int main(){ size_t i, it, N=nx*ny*nz, mem=3*N*sizeof(DAT); dim3 grid, block; block.x = BLOCK_X; block.y = BLOCK_Y; block.z = BLOCK_Z; grid.x = GRID_X; grid.y = GRID_Y; grid.z = GRID_Z; int gpu_id=-1; gpu_id=GPU_ID; cudaSetDevice(gpu_id); cudaGetDevice(&gpu_id); cudaDeviceReset(); cudaDeviceSetCacheConfig(cudaFuncCachePreferL1); // set L1 to prefered printf("Process uses GPU with id %d.\n",gpu_id); printf("%dx%dx%d, %1.3f GB, %d iterations.\n", nx,ny,nz, mem/1024./1024./1024., nt); printf("launching (%dx%dx%d) grid of (%dx%dx%d) blocks.\n", grid.x, grid.y, grid.z, block.x, block.y, block.z); // initializations zeros(A, nx,ny,nz); zeros(B, nx,ny,nz); zeros(C, nx,ny,nz); // time loop for(it=0; it<nt; it++){ if (it==10){ tic(); } memcopy<<<grid, block>>>(A_d,B_d,C_d,nx,ny,nz); cudaDeviceSynchronize(); } // tim("Performance", mem*(nt-3)*2/1024./1024./1024.); timPrint("Performance", mem*(nt-10)/1024./1024./1024.,nx,ny,nz); free_all(A); free_all(B); free_all(C); clean_cuda(); return 0; }
#include <hip/hip_runtime.h> #include <stdio.h> #include "sys/time.h" #define GPU_ID 0 // #define USE_SINGLE_PRECISION /* Comment this line using "!" if you want to use double precision. */ #ifdef USE_SINGLE_PRECISION #define DAT float #define PRECIS 4 #else #define DAT double #define PRECIS 8 #endif #define zeros(A,nx,ny,nz) DAT *A##_d,*A##_h; A##_h = (DAT*)malloc((nx)*(ny)*(nz)*sizeof(DAT)); \ for(i=0; i < (nx)*(ny)*(nz); i++){ A##_h[i]=(DAT)0.0; } \ hipMalloc(&A##_d ,(nx)*(ny)*(nz)*sizeof(DAT)); \ hipMemcpy( A##_d,A##_h,(nx)*(ny)*(nz)*sizeof(DAT),hipMemcpyHostToDevice); #define free_all(A) free(A##_h);hipFree(A##_d); #define BLOCK_X 32 #define BLOCK_Y 16 #define BLOCK_Z 2 #define GRID_X 32 #define GRID_Y 64 #define GRID_Z 128 const int nx = GRID_X*BLOCK_X; const int ny = GRID_Y*BLOCK_Y; const int nz = GRID_Z*BLOCK_Z; const int nt = 100; // Timer double timer_start = 0; double cpu_sec(){ struct timeval tp; gettimeofday(&tp,NULL); return tp.tv_sec+1e-6*tp.tv_usec; } void tic(){ timer_start = cpu_sec(); } double toc(){ return cpu_sec()-timer_start; } void tim(const char *what, double n){ double s=toc(); printf("%s: %8.3f seconds",what,s);if(n>0)printf(", %8.3f GB/s", n/s); printf("\n"); } void timPrint(const char *what, double n, int nx, int ny, int nz){ double s=toc(); printf("%s: %8.3f seconds",what,s);if(n>0)printf(", %8.3f GB/s", n/s); printf("\n"); FILE*fid; fid=fopen("PERF_memcpy.dat","a"); fprintf(fid,"nx=%d ny=%d nz=%d GBs=%1.4f time_s=%1.4f \n", nx, ny, nz, n/s, s); fclose(fid); } void clean_cuda(){ hipError_t ce = hipGetLastError(); if(ce != hipSuccess){ printf("ERROR launching GPU C-CUDA program: %s\n", hipGetErrorString(ce)); hipDeviceReset();} } __global__ void memcopy(DAT*A, DAT*B, DAT*C, const int nx,const int ny,const int nz){ int ix = blockIdx.x*blockDim.x + threadIdx.x; // thread ID, dimension x int iy = blockIdx.y*blockDim.y + threadIdx.y; // thread ID, dimension x int iz = blockIdx.z*blockDim.z + threadIdx.z; // thread ID, dimension x if (iz<nz && iy<ny && ix<nx) A[ix + iy*nx + iz*nx*ny] = B[ix + iy*nx + iz*nx*ny] + C[ix + iy*nx + iz*nx*ny]; } ////////// main ////////// int main(){ size_t i, it, N=nx*ny*nz, mem=3*N*sizeof(DAT); dim3 grid, block; block.x = BLOCK_X; block.y = BLOCK_Y; block.z = BLOCK_Z; grid.x = GRID_X; grid.y = GRID_Y; grid.z = GRID_Z; int gpu_id=-1; gpu_id=GPU_ID; hipSetDevice(gpu_id); hipGetDevice(&gpu_id); hipDeviceReset(); hipDeviceSetCacheConfig(hipFuncCachePreferL1); // set L1 to prefered printf("Process uses GPU with id %d.\n",gpu_id); printf("%dx%dx%d, %1.3f GB, %d iterations.\n", nx,ny,nz, mem/1024./1024./1024., nt); printf("launching (%dx%dx%d) grid of (%dx%dx%d) blocks.\n", grid.x, grid.y, grid.z, block.x, block.y, block.z); // initializations zeros(A, nx,ny,nz); zeros(B, nx,ny,nz); zeros(C, nx,ny,nz); // time loop for(it=0; it<nt; it++){ if (it==10){ tic(); } memcopy<<<grid, block>>>(A_d,B_d,C_d,nx,ny,nz); hipDeviceSynchronize(); } // tim("Performance", mem*(nt-3)*2/1024./1024./1024.); timPrint("Performance", mem*(nt-10)/1024./1024./1024.,nx,ny,nz); free_all(A); free_all(B); free_all(C); clean_cuda(); return 0; }
You are a helpful coding assistant specialized in converting HIP source code to AMD device assembly.
#include <hip/hip_runtime.h> #include <stdio.h> #include "sys/time.h" #define GPU_ID 0 // #define USE_SINGLE_PRECISION /* Comment this line using "!" if you want to use double precision. */ #ifdef USE_SINGLE_PRECISION #define DAT float #define PRECIS 4 #else #define DAT double #define PRECIS 8 #endif #define zeros(A,nx,ny,nz) DAT *A##_d,*A##_h; A##_h = (DAT*)malloc((nx)*(ny)*(nz)*sizeof(DAT)); \ for(i=0; i < (nx)*(ny)*(nz); i++){ A##_h[i]=(DAT)0.0; } \ hipMalloc(&A##_d ,(nx)*(ny)*(nz)*sizeof(DAT)); \ hipMemcpy( A##_d,A##_h,(nx)*(ny)*(nz)*sizeof(DAT),hipMemcpyHostToDevice); #define free_all(A) free(A##_h);hipFree(A##_d); #define BLOCK_X 32 #define BLOCK_Y 16 #define BLOCK_Z 2 #define GRID_X 32 #define GRID_Y 64 #define GRID_Z 128 const int nx = GRID_X*BLOCK_X; const int ny = GRID_Y*BLOCK_Y; const int nz = GRID_Z*BLOCK_Z; const int nt = 100; // Timer double timer_start = 0; double cpu_sec(){ struct timeval tp; gettimeofday(&tp,NULL); return tp.tv_sec+1e-6*tp.tv_usec; } void tic(){ timer_start = cpu_sec(); } double toc(){ return cpu_sec()-timer_start; } void tim(const char *what, double n){ double s=toc(); printf("%s: %8.3f seconds",what,s);if(n>0)printf(", %8.3f GB/s", n/s); printf("\n"); } void timPrint(const char *what, double n, int nx, int ny, int nz){ double s=toc(); printf("%s: %8.3f seconds",what,s);if(n>0)printf(", %8.3f GB/s", n/s); printf("\n"); FILE*fid; fid=fopen("PERF_memcpy.dat","a"); fprintf(fid,"nx=%d ny=%d nz=%d GBs=%1.4f time_s=%1.4f \n", nx, ny, nz, n/s, s); fclose(fid); } void clean_cuda(){ hipError_t ce = hipGetLastError(); if(ce != hipSuccess){ printf("ERROR launching GPU C-CUDA program: %s\n", hipGetErrorString(ce)); hipDeviceReset();} } __global__ void memcopy(DAT*A, DAT*B, DAT*C, const int nx,const int ny,const int nz){ int ix = blockIdx.x*blockDim.x + threadIdx.x; // thread ID, dimension x int iy = blockIdx.y*blockDim.y + threadIdx.y; // thread ID, dimension x int iz = blockIdx.z*blockDim.z + threadIdx.z; // thread ID, dimension x if (iz<nz && iy<ny && ix<nx) A[ix + iy*nx + iz*nx*ny] = B[ix + iy*nx + iz*nx*ny] + C[ix + iy*nx + iz*nx*ny]; } ////////// main ////////// int main(){ size_t i, it, N=nx*ny*nz, mem=3*N*sizeof(DAT); dim3 grid, block; block.x = BLOCK_X; block.y = BLOCK_Y; block.z = BLOCK_Z; grid.x = GRID_X; grid.y = GRID_Y; grid.z = GRID_Z; int gpu_id=-1; gpu_id=GPU_ID; hipSetDevice(gpu_id); hipGetDevice(&gpu_id); hipDeviceReset(); hipDeviceSetCacheConfig(hipFuncCachePreferL1); // set L1 to prefered printf("Process uses GPU with id %d.\n",gpu_id); printf("%dx%dx%d, %1.3f GB, %d iterations.\n", nx,ny,nz, mem/1024./1024./1024., nt); printf("launching (%dx%dx%d) grid of (%dx%dx%d) blocks.\n", grid.x, grid.y, grid.z, block.x, block.y, block.z); // initializations zeros(A, nx,ny,nz); zeros(B, nx,ny,nz); zeros(C, nx,ny,nz); // time loop for(it=0; it<nt; it++){ if (it==10){ tic(); } memcopy<<<grid, block>>>(A_d,B_d,C_d,nx,ny,nz); hipDeviceSynchronize(); } // tim("Performance", mem*(nt-3)*2/1024./1024./1024.); timPrint("Performance", mem*(nt-10)/1024./1024./1024.,nx,ny,nz); free_all(A); free_all(B); free_all(C); clean_cuda(); return 0; }
.text .amdgcn_target "amdgcn-amd-amdhsa--gfx1100" .protected _Z7memcopyPdS_S_iii .globl _Z7memcopyPdS_S_iii .p2align 8 .type _Z7memcopyPdS_S_iii,@function _Z7memcopyPdS_S_iii: s_clause 0x2 s_load_b64 s[2:3], s[0:1], 0x34 s_load_b32 s6, s[0:1], 0x20 s_load_b64 s[4:5], s[0:1], 0x18 v_and_b32_e32 v5, 0x3ff, v0 v_bfe_u32 v1, v0, 10, 10 v_bfe_u32 v0, v0, 20, 10 s_waitcnt lgkmcnt(0) s_and_b32 s3, s3, 0xffff s_and_b32 s7, s2, 0xffff s_delay_alu instid0(VALU_DEP_1) v_mad_u64_u32 v[2:3], null, s15, s3, v[0:1] s_lshr_b32 s2, s2, 16 s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1) v_mad_u64_u32 v[3:4], null, s14, s2, v[1:2] v_mad_u64_u32 v[0:1], null, s13, s7, v[5:6] v_cmp_gt_i32_e32 vcc_lo, s6, v2 s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3) v_cmp_gt_i32_e64 s2, s5, v3 v_cmp_gt_i32_e64 s3, s4, v0 s_delay_alu instid0(VALU_DEP_2) s_and_b32 s2, s2, vcc_lo s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1) s_and_b32 s2, s3, s2 s_delay_alu instid0(SALU_CYCLE_1) s_and_saveexec_b32 s3, s2 s_cbranch_execz .LBB0_2 v_mad_u64_u32 v[4:5], null, v2, s5, v[3:4] s_clause 0x1 s_load_b128 s[8:11], s[0:1], 0x0 s_load_b64 s[0:1], s[0:1], 0x10 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_mad_u64_u32 v[1:2], null, v4, s4, v[0:1] v_ashrrev_i32_e32 v2, 31, v1 s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1) v_lshlrev_b64 v[0:1], 3, v[1:2] s_waitcnt lgkmcnt(0) v_add_co_u32 v2, vcc_lo, s10, v0 s_delay_alu instid0(VALU_DEP_2) v_add_co_ci_u32_e32 v3, vcc_lo, s11, v1, vcc_lo v_add_co_u32 v4, vcc_lo, s0, v0 v_add_co_ci_u32_e32 v5, vcc_lo, s1, v1, vcc_lo v_add_co_u32 v0, vcc_lo, s8, v0 global_load_b64 v[2:3], v[2:3], off global_load_b64 v[4:5], v[4:5], off v_add_co_ci_u32_e32 v1, vcc_lo, s9, v1, vcc_lo s_waitcnt vmcnt(0) v_add_f64 v[2:3], v[2:3], v[4:5] global_store_b64 v[0:1], v[2:3], off .LBB0_2: s_nop 0 s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) s_endpgm .section .rodata,"a",@progbits .p2align 6, 0x0 .amdhsa_kernel _Z7memcopyPdS_S_iii .amdhsa_group_segment_fixed_size 0 .amdhsa_private_segment_fixed_size 0 .amdhsa_kernarg_size 296 .amdhsa_user_sgpr_count 13 .amdhsa_user_sgpr_dispatch_ptr 0 .amdhsa_user_sgpr_queue_ptr 0 .amdhsa_user_sgpr_kernarg_segment_ptr 1 .amdhsa_user_sgpr_dispatch_id 0 .amdhsa_user_sgpr_private_segment_size 0 .amdhsa_wavefront_size32 1 .amdhsa_uses_dynamic_stack 0 .amdhsa_enable_private_segment 0 .amdhsa_system_sgpr_workgroup_id_x 1 .amdhsa_system_sgpr_workgroup_id_y 1 .amdhsa_system_sgpr_workgroup_id_z 1 .amdhsa_system_sgpr_workgroup_info 0 .amdhsa_system_vgpr_workitem_id 2 .amdhsa_next_free_vgpr 7 .amdhsa_next_free_sgpr 16 .amdhsa_float_round_mode_32 0 .amdhsa_float_round_mode_16_64 0 .amdhsa_float_denorm_mode_32 3 .amdhsa_float_denorm_mode_16_64 3 .amdhsa_dx10_clamp 1 .amdhsa_ieee_mode 1 .amdhsa_fp16_overflow 0 .amdhsa_workgroup_processor_mode 1 .amdhsa_memory_ordered 1 .amdhsa_forward_progress 0 .amdhsa_shared_vgpr_count 0 .amdhsa_exception_fp_ieee_invalid_op 0 .amdhsa_exception_fp_denorm_src 0 .amdhsa_exception_fp_ieee_div_zero 0 .amdhsa_exception_fp_ieee_overflow 0 .amdhsa_exception_fp_ieee_underflow 0 .amdhsa_exception_fp_ieee_inexact 0 .amdhsa_exception_int_div_zero 0 .end_amdhsa_kernel .text .Lfunc_end0: .size _Z7memcopyPdS_S_iii, .Lfunc_end0-_Z7memcopyPdS_S_iii .section .AMDGPU.csdata,"",@progbits .text .p2alignl 7, 3214868480 .fill 96, 4, 3214868480 .type __hip_cuid_,@object .section .bss,"aw",@nobits .globl __hip_cuid_ __hip_cuid_: .byte 0 .size __hip_cuid_, 1 .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym __hip_cuid_ .amdgpu_metadata --- amdhsa.kernels: - .args: - .address_space: global .offset: 0 .size: 8 .value_kind: global_buffer - .address_space: global .offset: 8 .size: 8 .value_kind: global_buffer - .address_space: global .offset: 16 .size: 8 .value_kind: global_buffer - .offset: 24 .size: 4 .value_kind: by_value - .offset: 28 .size: 4 .value_kind: by_value - .offset: 32 .size: 4 .value_kind: by_value - .offset: 40 .size: 4 .value_kind: hidden_block_count_x - .offset: 44 .size: 4 .value_kind: hidden_block_count_y - .offset: 48 .size: 4 .value_kind: hidden_block_count_z - .offset: 52 .size: 2 .value_kind: hidden_group_size_x - .offset: 54 .size: 2 .value_kind: hidden_group_size_y - .offset: 56 .size: 2 .value_kind: hidden_group_size_z - .offset: 58 .size: 2 .value_kind: hidden_remainder_x - .offset: 60 .size: 2 .value_kind: hidden_remainder_y - .offset: 62 .size: 2 .value_kind: hidden_remainder_z - .offset: 80 .size: 8 .value_kind: hidden_global_offset_x - .offset: 88 .size: 8 .value_kind: hidden_global_offset_y - .offset: 96 .size: 8 .value_kind: hidden_global_offset_z - .offset: 104 .size: 2 .value_kind: hidden_grid_dims .group_segment_fixed_size: 0 .kernarg_segment_align: 8 .kernarg_segment_size: 296 .language: OpenCL C .language_version: - 2 - 0 .max_flat_workgroup_size: 1024 .name: _Z7memcopyPdS_S_iii .private_segment_fixed_size: 0 .sgpr_count: 18 .sgpr_spill_count: 0 .symbol: _Z7memcopyPdS_S_iii.kd .uniform_work_group_size: 1 .uses_dynamic_stack: false .vgpr_count: 7 .vgpr_spill_count: 0 .wavefront_size: 32 .workgroup_processor_mode: 1 amdhsa.target: amdgcn-amd-amdhsa--gfx1100 amdhsa.version: - 1 - 2 ... .end_amdgpu_metadata
You are a helpful coding assistant specialized in converting HIP source code to AMD host assembly.
#include <hip/hip_runtime.h> #include <stdio.h> #include "sys/time.h" #define GPU_ID 0 // #define USE_SINGLE_PRECISION /* Comment this line using "!" if you want to use double precision. */ #ifdef USE_SINGLE_PRECISION #define DAT float #define PRECIS 4 #else #define DAT double #define PRECIS 8 #endif #define zeros(A,nx,ny,nz) DAT *A##_d,*A##_h; A##_h = (DAT*)malloc((nx)*(ny)*(nz)*sizeof(DAT)); \ for(i=0; i < (nx)*(ny)*(nz); i++){ A##_h[i]=(DAT)0.0; } \ hipMalloc(&A##_d ,(nx)*(ny)*(nz)*sizeof(DAT)); \ hipMemcpy( A##_d,A##_h,(nx)*(ny)*(nz)*sizeof(DAT),hipMemcpyHostToDevice); #define free_all(A) free(A##_h);hipFree(A##_d); #define BLOCK_X 32 #define BLOCK_Y 16 #define BLOCK_Z 2 #define GRID_X 32 #define GRID_Y 64 #define GRID_Z 128 const int nx = GRID_X*BLOCK_X; const int ny = GRID_Y*BLOCK_Y; const int nz = GRID_Z*BLOCK_Z; const int nt = 100; // Timer double timer_start = 0; double cpu_sec(){ struct timeval tp; gettimeofday(&tp,NULL); return tp.tv_sec+1e-6*tp.tv_usec; } void tic(){ timer_start = cpu_sec(); } double toc(){ return cpu_sec()-timer_start; } void tim(const char *what, double n){ double s=toc(); printf("%s: %8.3f seconds",what,s);if(n>0)printf(", %8.3f GB/s", n/s); printf("\n"); } void timPrint(const char *what, double n, int nx, int ny, int nz){ double s=toc(); printf("%s: %8.3f seconds",what,s);if(n>0)printf(", %8.3f GB/s", n/s); printf("\n"); FILE*fid; fid=fopen("PERF_memcpy.dat","a"); fprintf(fid,"nx=%d ny=%d nz=%d GBs=%1.4f time_s=%1.4f \n", nx, ny, nz, n/s, s); fclose(fid); } void clean_cuda(){ hipError_t ce = hipGetLastError(); if(ce != hipSuccess){ printf("ERROR launching GPU C-CUDA program: %s\n", hipGetErrorString(ce)); hipDeviceReset();} } __global__ void memcopy(DAT*A, DAT*B, DAT*C, const int nx,const int ny,const int nz){ int ix = blockIdx.x*blockDim.x + threadIdx.x; // thread ID, dimension x int iy = blockIdx.y*blockDim.y + threadIdx.y; // thread ID, dimension x int iz = blockIdx.z*blockDim.z + threadIdx.z; // thread ID, dimension x if (iz<nz && iy<ny && ix<nx) A[ix + iy*nx + iz*nx*ny] = B[ix + iy*nx + iz*nx*ny] + C[ix + iy*nx + iz*nx*ny]; } ////////// main ////////// int main(){ size_t i, it, N=nx*ny*nz, mem=3*N*sizeof(DAT); dim3 grid, block; block.x = BLOCK_X; block.y = BLOCK_Y; block.z = BLOCK_Z; grid.x = GRID_X; grid.y = GRID_Y; grid.z = GRID_Z; int gpu_id=-1; gpu_id=GPU_ID; hipSetDevice(gpu_id); hipGetDevice(&gpu_id); hipDeviceReset(); hipDeviceSetCacheConfig(hipFuncCachePreferL1); // set L1 to prefered printf("Process uses GPU with id %d.\n",gpu_id); printf("%dx%dx%d, %1.3f GB, %d iterations.\n", nx,ny,nz, mem/1024./1024./1024., nt); printf("launching (%dx%dx%d) grid of (%dx%dx%d) blocks.\n", grid.x, grid.y, grid.z, block.x, block.y, block.z); // initializations zeros(A, nx,ny,nz); zeros(B, nx,ny,nz); zeros(C, nx,ny,nz); // time loop for(it=0; it<nt; it++){ if (it==10){ tic(); } memcopy<<<grid, block>>>(A_d,B_d,C_d,nx,ny,nz); hipDeviceSynchronize(); } // tim("Performance", mem*(nt-3)*2/1024./1024./1024.); timPrint("Performance", mem*(nt-10)/1024./1024./1024.,nx,ny,nz); free_all(A); free_all(B); free_all(C); clean_cuda(); return 0; }
.text .file "memcopy.hip" .section .rodata.cst8,"aM",@progbits,8 .p2align 3, 0x0 # -- Begin function _Z7cpu_secv .LCPI0_0: .quad 0x3eb0c6f7a0b5ed8d # double 9.9999999999999995E-7 .text .globl _Z7cpu_secv .p2align 4, 0x90 .type _Z7cpu_secv,@function _Z7cpu_secv: # @_Z7cpu_secv .cfi_startproc # %bb.0: subq $24, %rsp .cfi_def_cfa_offset 32 leaq 8(%rsp), %rdi xorl %esi, %esi callq gettimeofday cvtsi2sdq 8(%rsp), %xmm1 cvtsi2sdq 16(%rsp), %xmm0 mulsd .LCPI0_0(%rip), %xmm0 addsd %xmm1, %xmm0 addq $24, %rsp .cfi_def_cfa_offset 8 retq .Lfunc_end0: .size _Z7cpu_secv, .Lfunc_end0-_Z7cpu_secv .cfi_endproc # -- End function .section .rodata.cst8,"aM",@progbits,8 .p2align 3, 0x0 # -- Begin function _Z3ticv .LCPI1_0: .quad 0x3eb0c6f7a0b5ed8d # double 9.9999999999999995E-7 .text .globl _Z3ticv .p2align 4, 0x90 .type _Z3ticv,@function _Z3ticv: # @_Z3ticv .cfi_startproc # %bb.0: subq $24, %rsp .cfi_def_cfa_offset 32 leaq 8(%rsp), %rdi xorl %esi, %esi callq gettimeofday cvtsi2sdq 8(%rsp), %xmm0 cvtsi2sdq 16(%rsp), %xmm1 mulsd .LCPI1_0(%rip), %xmm1 addsd %xmm0, %xmm1 movsd %xmm1, timer_start(%rip) addq $24, %rsp .cfi_def_cfa_offset 8 retq .Lfunc_end1: .size _Z3ticv, .Lfunc_end1-_Z3ticv .cfi_endproc # -- End function .section .rodata.cst8,"aM",@progbits,8 .p2align 3, 0x0 # -- Begin function _Z3tocv .LCPI2_0: .quad 0x3eb0c6f7a0b5ed8d # double 9.9999999999999995E-7 .text .globl _Z3tocv .p2align 4, 0x90 .type _Z3tocv,@function _Z3tocv: # @_Z3tocv .cfi_startproc # %bb.0: subq $24, %rsp .cfi_def_cfa_offset 32 leaq 8(%rsp), %rdi xorl %esi, %esi callq gettimeofday cvtsi2sdq 8(%rsp), %xmm1 cvtsi2sdq 16(%rsp), %xmm0 mulsd .LCPI2_0(%rip), %xmm0 addsd %xmm1, %xmm0 subsd timer_start(%rip), %xmm0 addq $24, %rsp .cfi_def_cfa_offset 8 retq .Lfunc_end2: .size _Z3tocv, .Lfunc_end2-_Z3tocv .cfi_endproc # -- End function .section .rodata.cst8,"aM",@progbits,8 .p2align 3, 0x0 # -- Begin function _Z3timPKcd .LCPI3_0: .quad 0x3eb0c6f7a0b5ed8d # double 9.9999999999999995E-7 .text .globl _Z3timPKcd .p2align 4, 0x90 .type _Z3timPKcd,@function _Z3timPKcd: # @_Z3timPKcd .cfi_startproc # %bb.0: pushq %rbx .cfi_def_cfa_offset 16 subq $32, %rsp .cfi_def_cfa_offset 48 .cfi_offset %rbx, -16 movsd %xmm0, 8(%rsp) # 8-byte Spill movq %rdi, %rbx leaq 16(%rsp), %rdi xorl %esi, %esi callq gettimeofday cvtsi2sdq 16(%rsp), %xmm1 xorps %xmm0, %xmm0 cvtsi2sdq 24(%rsp), %xmm0 mulsd .LCPI3_0(%rip), %xmm0 addsd %xmm1, %xmm0 subsd timer_start(%rip), %xmm0 movl $.L.str, %edi movq %rbx, %rsi movsd %xmm0, (%rsp) # 8-byte Spill movb $1, %al callq printf xorpd %xmm0, %xmm0 movsd 8(%rsp), %xmm1 # 8-byte Reload # xmm1 = mem[0],zero ucomisd %xmm0, %xmm1 jbe .LBB3_2 # %bb.1: movapd %xmm1, %xmm0 divsd (%rsp), %xmm0 # 8-byte Folded Reload movl $.L.str.1, %edi movb $1, %al callq printf .LBB3_2: movl $10, %edi addq $32, %rsp .cfi_def_cfa_offset 16 popq %rbx .cfi_def_cfa_offset 8 jmp putchar@PLT # TAILCALL .Lfunc_end3: .size _Z3timPKcd, .Lfunc_end3-_Z3timPKcd .cfi_endproc # -- End function .section .rodata.cst8,"aM",@progbits,8 .p2align 3, 0x0 # -- Begin function _Z8timPrintPKcdiii .LCPI4_0: .quad 0x3eb0c6f7a0b5ed8d # double 9.9999999999999995E-7 .text .globl _Z8timPrintPKcdiii .p2align 4, 0x90 .type _Z8timPrintPKcdiii,@function _Z8timPrintPKcdiii: # @_Z8timPrintPKcdiii .cfi_startproc # %bb.0: pushq %rbp .cfi_def_cfa_offset 16 pushq %r15 .cfi_def_cfa_offset 24 pushq %r14 .cfi_def_cfa_offset 32 pushq %rbx .cfi_def_cfa_offset 40 subq $40, %rsp .cfi_def_cfa_offset 80 .cfi_offset %rbx, -40 .cfi_offset %r14, -32 .cfi_offset %r15, -24 .cfi_offset %rbp, -16 movl %ecx, %ebx movl %edx, %ebp movl %esi, %r14d movsd %xmm0, 8(%rsp) # 8-byte Spill movq %rdi, %r15 leaq 24(%rsp), %rdi xorl %esi, %esi callq gettimeofday cvtsi2sdq 24(%rsp), %xmm1 xorps %xmm0, %xmm0 cvtsi2sdq 32(%rsp), %xmm0 mulsd .LCPI4_0(%rip), %xmm0 addsd %xmm1, %xmm0 subsd timer_start(%rip), %xmm0 movsd %xmm0, 16(%rsp) # 8-byte Spill movl $.L.str, %edi movq %r15, %rsi movb $1, %al callq printf xorpd %xmm0, %xmm0 movsd 8(%rsp), %xmm1 # 8-byte Reload # xmm1 = mem[0],zero ucomisd %xmm0, %xmm1 divsd 16(%rsp), %xmm1 # 8-byte Folded Reload movsd %xmm1, 8(%rsp) # 8-byte Spill jbe .LBB4_2 # %bb.1: movl $.L.str.1, %edi movsd 8(%rsp), %xmm0 # 8-byte Reload # xmm0 = mem[0],zero movb $1, %al callq printf .LBB4_2: movl $10, %edi callq putchar@PLT movl $.L.str.3, %edi movl $.L.str.4, %esi callq fopen movq %rax, %r15 movl $.L.str.5, %esi movq %rax, %rdi movl %r14d, %edx movl %ebp, %ecx movl %ebx, %r8d movsd 8(%rsp), %xmm0 # 8-byte Reload # xmm0 = mem[0],zero movsd 16(%rsp), %xmm1 # 8-byte Reload # xmm1 = mem[0],zero movb $2, %al callq fprintf movq %r15, %rdi addq $40, %rsp .cfi_def_cfa_offset 40 popq %rbx .cfi_def_cfa_offset 32 popq %r14 .cfi_def_cfa_offset 24 popq %r15 .cfi_def_cfa_offset 16 popq %rbp .cfi_def_cfa_offset 8 jmp fclose # TAILCALL .Lfunc_end4: .size _Z8timPrintPKcdiii, .Lfunc_end4-_Z8timPrintPKcdiii .cfi_endproc # -- End function .globl _Z10clean_cudav # -- Begin function _Z10clean_cudav .p2align 4, 0x90 .type _Z10clean_cudav,@function _Z10clean_cudav: # @_Z10clean_cudav .cfi_startproc # %bb.0: pushq %rax .cfi_def_cfa_offset 16 callq hipGetLastError testl %eax, %eax je .LBB5_1 # %bb.2: movl %eax, %edi callq hipGetErrorString movl $.L.str.6, %edi movq %rax, %rsi xorl %eax, %eax callq printf popq %rax .cfi_def_cfa_offset 8 jmp hipDeviceReset # TAILCALL .LBB5_1: .cfi_def_cfa_offset 16 popq %rax .cfi_def_cfa_offset 8 retq .Lfunc_end5: .size _Z10clean_cudav, .Lfunc_end5-_Z10clean_cudav .cfi_endproc # -- End function .globl _Z22__device_stub__memcopyPdS_S_iii # -- Begin function _Z22__device_stub__memcopyPdS_S_iii .p2align 4, 0x90 .type _Z22__device_stub__memcopyPdS_S_iii,@function _Z22__device_stub__memcopyPdS_S_iii: # @_Z22__device_stub__memcopyPdS_S_iii .cfi_startproc # %bb.0: subq $152, %rsp .cfi_def_cfa_offset 160 movq %rdi, 88(%rsp) movq %rsi, 80(%rsp) movq %rdx, 72(%rsp) movl %ecx, 20(%rsp) movl %r8d, 16(%rsp) movl %r9d, 12(%rsp) leaq 88(%rsp), %rax movq %rax, 96(%rsp) leaq 80(%rsp), %rax movq %rax, 104(%rsp) leaq 72(%rsp), %rax movq %rax, 112(%rsp) leaq 20(%rsp), %rax movq %rax, 120(%rsp) leaq 16(%rsp), %rax movq %rax, 128(%rsp) leaq 12(%rsp), %rax movq %rax, 136(%rsp) leaq 56(%rsp), %rdi leaq 40(%rsp), %rsi leaq 32(%rsp), %rdx leaq 24(%rsp), %rcx callq __hipPopCallConfiguration movq 56(%rsp), %rsi movl 64(%rsp), %edx movq 40(%rsp), %rcx movl 48(%rsp), %r8d leaq 96(%rsp), %r9 movl $_Z7memcopyPdS_S_iii, %edi pushq 24(%rsp) .cfi_adjust_cfa_offset 8 pushq 40(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $168, %rsp .cfi_adjust_cfa_offset -168 retq .Lfunc_end6: .size _Z22__device_stub__memcopyPdS_S_iii, .Lfunc_end6-_Z22__device_stub__memcopyPdS_S_iii .cfi_endproc # -- End function .section .rodata.cst8,"aM",@progbits,8 .p2align 3, 0x0 # -- Begin function main .LCPI7_0: .quad 0x4018000000000000 # double 6 .LCPI7_1: .quad 0x3eb0c6f7a0b5ed8d # double 9.9999999999999995E-7 .LCPI7_2: .quad 0x4080e00000000000 # double 540 .text .globl main .p2align 4, 0x90 .type main,@function main: # @main .cfi_startproc # %bb.0: pushq %rbp .cfi_def_cfa_offset 16 pushq %r15 .cfi_def_cfa_offset 24 pushq %r14 .cfi_def_cfa_offset 32 pushq %r13 .cfi_def_cfa_offset 40 pushq %r12 .cfi_def_cfa_offset 48 pushq %rbx .cfi_def_cfa_offset 56 subq $216, %rsp .cfi_def_cfa_offset 272 .cfi_offset %rbx, -56 .cfi_offset %r12, -48 .cfi_offset %r13, -40 .cfi_offset %r14, -32 .cfi_offset %r15, -24 .cfi_offset %rbp, -16 movl $0, 12(%rsp) xorl %edi, %edi callq hipSetDevice leaq 12(%rsp), %rdi callq hipGetDevice callq hipDeviceReset movl $2, %edi callq hipDeviceSetCacheConfig movl 12(%rsp), %esi movl $.L.str.7, %edi xorl %eax, %eax callq printf movsd .LCPI7_0(%rip), %xmm0 # xmm0 = mem[0],zero movl $.L.str.8, %edi movl $1024, %esi # imm = 0x400 movl $1024, %edx # imm = 0x400 movl $256, %ecx # imm = 0x100 movl $100, %r8d movb $1, %al callq printf subq $8, %rsp .cfi_adjust_cfa_offset 8 movl $.L.str.9, %edi movl $32, %esi movl $64, %edx movl $128, %ecx movl $32, %r8d movl $16, %r9d xorl %eax, %eax pushq $2 .cfi_adjust_cfa_offset 8 callq printf addq $16, %rsp .cfi_adjust_cfa_offset -16 movl $2147483648, %edi # imm = 0x80000000 callq malloc movq %rax, %rbx movl $2147483648, %edx # imm = 0x80000000 movq %rax, %rdi xorl %esi, %esi callq memset@PLT leaq 32(%rsp), %rdi movl $2147483648, %esi # imm = 0x80000000 callq hipMalloc movq 32(%rsp), %rdi movl $2147483648, %edx # imm = 0x80000000 movq %rbx, 136(%rsp) # 8-byte Spill movq %rbx, %rsi movl $1, %ecx callq hipMemcpy movl $2147483648, %edi # imm = 0x80000000 callq malloc movq %rax, %rbx movl $2147483648, %edx # imm = 0x80000000 movq %rax, %rdi xorl %esi, %esi callq memset@PLT leaq 24(%rsp), %rdi movl $2147483648, %esi # imm = 0x80000000 callq hipMalloc movq 24(%rsp), %rdi movl $2147483648, %edx # imm = 0x80000000 movq %rbx, 128(%rsp) # 8-byte Spill movq %rbx, %rsi movl $1, %ecx callq hipMemcpy movl $2147483648, %edi # imm = 0x80000000 callq malloc movq %rax, %rbx movl $2147483648, %edx # imm = 0x80000000 movq %rax, %rdi xorl %esi, %esi callq memset@PLT leaq 16(%rsp), %rdi movl $2147483648, %esi # imm = 0x80000000 callq hipMalloc movq 16(%rsp), %rdi movl $2147483648, %edx # imm = 0x80000000 movq %rbx, 120(%rsp) # 8-byte Spill movq %rbx, %rsi movl $1, %ecx callq hipMemcpy movl $100, %r14d leaq 64(%rsp), %r12 movabsq $274877906976, %r13 # imm = 0x4000000020 movabsq $68719476768, %rbp # imm = 0x1000000020 leaq 152(%rsp), %r15 leaq 144(%rsp), %rbx jmp .LBB7_1 .p2align 4, 0x90 .LBB7_3: # in Loop: Header=BB7_1 Depth=1 movq %r13, %rdi movl $128, %esi movq %rbp, %rdx movl $2, %ecx xorl %r8d, %r8d xorl %r9d, %r9d callq __hipPushCallConfiguration testl %eax, %eax je .LBB7_4 .LBB7_5: # in Loop: Header=BB7_1 Depth=1 callq hipDeviceSynchronize decq %r14 je .LBB7_6 .LBB7_1: # =>This Inner Loop Header: Depth=1 cmpq $90, %r14 jne .LBB7_3 # %bb.2: # in Loop: Header=BB7_1 Depth=1 movq %r12, %rdi xorl %esi, %esi callq gettimeofday xorps %xmm0, %xmm0 cvtsi2sdq 64(%rsp), %xmm0 xorps %xmm1, %xmm1 cvtsi2sdq 72(%rsp), %xmm1 mulsd .LCPI7_1(%rip), %xmm1 addsd %xmm0, %xmm1 movsd %xmm1, timer_start(%rip) jmp .LBB7_3 .p2align 4, 0x90 .LBB7_4: # in Loop: Header=BB7_1 Depth=1 movq 32(%rsp), %rax movq 24(%rsp), %rcx movq 16(%rsp), %rdx movq %rax, 208(%rsp) movq %rcx, 200(%rsp) movq %rdx, 192(%rsp) movl $1024, 52(%rsp) # imm = 0x400 movl $1024, 48(%rsp) # imm = 0x400 movl $256, 44(%rsp) # imm = 0x100 leaq 208(%rsp), %rax movq %rax, 64(%rsp) leaq 200(%rsp), %rax movq %rax, 72(%rsp) leaq 192(%rsp), %rax movq %rax, 80(%rsp) leaq 52(%rsp), %rax movq %rax, 88(%rsp) leaq 48(%rsp), %rax movq %rax, 96(%rsp) leaq 44(%rsp), %rax movq %rax, 104(%rsp) leaq 176(%rsp), %rdi leaq 160(%rsp), %rsi movq %r15, %rdx movq %rbx, %rcx callq __hipPopCallConfiguration movq 176(%rsp), %rsi movl 184(%rsp), %edx movq 160(%rsp), %rcx movl 168(%rsp), %r8d movl $_Z7memcopyPdS_S_iii, %edi movq %r12, %r9 pushq 144(%rsp) .cfi_adjust_cfa_offset 8 pushq 160(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $16, %rsp .cfi_adjust_cfa_offset -16 jmp .LBB7_5 .LBB7_6: leaq 64(%rsp), %rdi xorl %esi, %esi callq gettimeofday xorps %xmm1, %xmm1 cvtsi2sdq 64(%rsp), %xmm1 xorps %xmm0, %xmm0 cvtsi2sdq 72(%rsp), %xmm0 mulsd .LCPI7_1(%rip), %xmm0 addsd %xmm1, %xmm0 subsd timer_start(%rip), %xmm0 movsd %xmm0, 56(%rsp) # 8-byte Spill movl $.L.str, %edi movl $.L.str.10, %esi movb $1, %al callq printf movsd .LCPI7_2(%rip), %xmm0 # xmm0 = mem[0],zero divsd 56(%rsp), %xmm0 # 8-byte Folded Reload movsd %xmm0, 112(%rsp) # 8-byte Spill movl $.L.str.1, %edi movb $1, %al callq printf movl $10, %edi callq putchar@PLT movl $.L.str.3, %edi movl $.L.str.4, %esi callq fopen movq %rax, %rbx movl $.L.str.5, %esi movq %rax, %rdi movl $1024, %edx # imm = 0x400 movl $1024, %ecx # imm = 0x400 movl $256, %r8d # imm = 0x100 movsd 112(%rsp), %xmm0 # 8-byte Reload # xmm0 = mem[0],zero movsd 56(%rsp), %xmm1 # 8-byte Reload # xmm1 = mem[0],zero movb $2, %al callq fprintf movq %rbx, %rdi callq fclose movq 136(%rsp), %rdi # 8-byte Reload callq free movq 32(%rsp), %rdi callq hipFree movq 128(%rsp), %rdi # 8-byte Reload callq free movq 24(%rsp), %rdi callq hipFree movq 120(%rsp), %rdi # 8-byte Reload callq free movq 16(%rsp), %rdi callq hipFree callq hipGetLastError testl %eax, %eax je .LBB7_8 # %bb.7: movl %eax, %edi callq hipGetErrorString movl $.L.str.6, %edi movq %rax, %rsi xorl %eax, %eax callq printf callq hipDeviceReset .LBB7_8: # %_Z10clean_cudav.exit xorl %eax, %eax addq $216, %rsp .cfi_def_cfa_offset 56 popq %rbx .cfi_def_cfa_offset 48 popq %r12 .cfi_def_cfa_offset 40 popq %r13 .cfi_def_cfa_offset 32 popq %r14 .cfi_def_cfa_offset 24 popq %r15 .cfi_def_cfa_offset 16 popq %rbp .cfi_def_cfa_offset 8 retq .Lfunc_end7: .size main, .Lfunc_end7-main .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_ctor .type __hip_module_ctor,@function __hip_module_ctor: # @__hip_module_ctor .cfi_startproc # %bb.0: subq $40, %rsp .cfi_def_cfa_offset 48 cmpq $0, __hip_gpubin_handle(%rip) jne .LBB8_2 # %bb.1: movl $__hip_fatbin_wrapper, %edi callq __hipRegisterFatBinary movq %rax, __hip_gpubin_handle(%rip) .LBB8_2: movq __hip_gpubin_handle(%rip), %rdi xorps %xmm0, %xmm0 movups %xmm0, 16(%rsp) movups %xmm0, (%rsp) movl $_Z7memcopyPdS_S_iii, %esi movl $.L__unnamed_1, %edx movl $.L__unnamed_1, %ecx movl $-1, %r8d xorl %r9d, %r9d callq __hipRegisterFunction movl $__hip_module_dtor, %edi addq $40, %rsp .cfi_def_cfa_offset 8 jmp atexit # TAILCALL .Lfunc_end8: .size __hip_module_ctor, .Lfunc_end8-__hip_module_ctor .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_dtor .type __hip_module_dtor,@function __hip_module_dtor: # @__hip_module_dtor .cfi_startproc # %bb.0: movq __hip_gpubin_handle(%rip), %rdi testq %rdi, %rdi je .LBB9_2 # %bb.1: pushq %rax .cfi_def_cfa_offset 16 callq __hipUnregisterFatBinary movq $0, __hip_gpubin_handle(%rip) addq $8, %rsp .cfi_def_cfa_offset 8 .LBB9_2: retq .Lfunc_end9: .size __hip_module_dtor, .Lfunc_end9-__hip_module_dtor .cfi_endproc # -- End function .type timer_start,@object # @timer_start .bss .globl timer_start .p2align 3, 0x0 timer_start: .quad 0x0000000000000000 # double 0 .size timer_start, 8 .type .L.str,@object # @.str .section .rodata.str1.1,"aMS",@progbits,1 .L.str: .asciz "%s: %8.3f seconds" .size .L.str, 18 .type .L.str.1,@object # @.str.1 .L.str.1: .asciz ", %8.3f GB/s" .size .L.str.1, 13 .type .L.str.3,@object # @.str.3 .L.str.3: .asciz "PERF_memcpy.dat" .size .L.str.3, 16 .type .L.str.4,@object # @.str.4 .L.str.4: .asciz "a" .size .L.str.4, 2 .type .L.str.5,@object # @.str.5 .L.str.5: .asciz "nx=%d ny=%d nz=%d GBs=%1.4f time_s=%1.4f \n" .size .L.str.5, 44 .type .L.str.6,@object # @.str.6 .L.str.6: .asciz "ERROR launching GPU C-CUDA program: %s\n" .size .L.str.6, 40 .type _Z7memcopyPdS_S_iii,@object # @_Z7memcopyPdS_S_iii .section .rodata,"a",@progbits .globl _Z7memcopyPdS_S_iii .p2align 3, 0x0 _Z7memcopyPdS_S_iii: .quad _Z22__device_stub__memcopyPdS_S_iii .size _Z7memcopyPdS_S_iii, 8 .type .L.str.7,@object # @.str.7 .section .rodata.str1.1,"aMS",@progbits,1 .L.str.7: .asciz "Process uses GPU with id %d.\n" .size .L.str.7, 30 .type .L.str.8,@object # @.str.8 .L.str.8: .asciz "%dx%dx%d, %1.3f GB, %d iterations.\n" .size .L.str.8, 36 .type .L.str.9,@object # @.str.9 .L.str.9: .asciz "launching (%dx%dx%d) grid of (%dx%dx%d) blocks.\n" .size .L.str.9, 49 .type .L.str.10,@object # @.str.10 .L.str.10: .asciz "Performance" .size .L.str.10, 12 .type .L__unnamed_1,@object # @0 .L__unnamed_1: .asciz "_Z7memcopyPdS_S_iii" .size .L__unnamed_1, 20 .type __hip_fatbin_wrapper,@object # @__hip_fatbin_wrapper .section .hipFatBinSegment,"a",@progbits .p2align 3, 0x0 __hip_fatbin_wrapper: .long 1212764230 # 0x48495046 .long 1 # 0x1 .quad __hip_fatbin .quad 0 .size __hip_fatbin_wrapper, 24 .type __hip_gpubin_handle,@object # @__hip_gpubin_handle .local __hip_gpubin_handle .comm __hip_gpubin_handle,8,8 .section .init_array,"aw",@init_array .p2align 3, 0x0 .quad __hip_module_ctor .type __hip_cuid_,@object # @__hip_cuid_ .bss .globl __hip_cuid_ __hip_cuid_: .byte 0 # 0x0 .size __hip_cuid_, 1 .section ".linker-options","e",@llvm_linker_options .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym _Z22__device_stub__memcopyPdS_S_iii .addrsig_sym __hip_module_ctor .addrsig_sym __hip_module_dtor .addrsig_sym _Z7memcopyPdS_S_iii .addrsig_sym __hip_fatbin .addrsig_sym __hip_fatbin_wrapper .addrsig_sym __hip_cuid_
You are a helpful coding assistant specialized in converting from CUDA device assembly to AMD device assembly.
code for sm_80 Function : _Z7memcopyPdS_S_iii .headerflags @"EF_CUDA_TEXMODE_UNIFIED EF_CUDA_64BIT_ADDRESS EF_CUDA_SM80 EF_CUDA_VIRTUAL_SM(EF_CUDA_SM80)" /*0000*/ MOV R1, c[0x0][0x28] ; /* 0x00000a0000017a02 */ /* 0x000fe40000000f00 */ /*0010*/ S2R R3, SR_CTAID.Z ; /* 0x0000000000037919 */ /* 0x000e280000002700 */ /*0020*/ S2R R4, SR_TID.Z ; /* 0x0000000000047919 */ /* 0x000e280000002300 */ /*0030*/ S2R R2, SR_CTAID.Y ; /* 0x0000000000027919 */ /* 0x000e680000002600 */ /*0040*/ S2R R7, SR_TID.Y ; /* 0x0000000000077919 */ /* 0x000e680000002200 */ /*0050*/ S2R R0, SR_CTAID.X ; /* 0x0000000000007919 */ /* 0x000ea80000002500 */ /*0060*/ S2R R5, SR_TID.X ; /* 0x0000000000057919 */ /* 0x000ea20000002100 */ /*0070*/ IMAD R3, R3, c[0x0][0x8], R4 ; /* 0x0000020003037a24 */ /* 0x001fca00078e0204 */ /*0080*/ ISETP.GE.AND P0, PT, R3, c[0x0][0x180], PT ; /* 0x0000600003007a0c */ /* 0x000fe20003f06270 */ /*0090*/ IMAD R2, R2, c[0x0][0x4], R7 ; /* 0x0000010002027a24 */ /* 0x002fca00078e0207 */ /*00a0*/ ISETP.GE.OR P0, PT, R2, c[0x0][0x17c], P0 ; /* 0x00005f0002007a0c */ /* 0x000fe20000706670 */ /*00b0*/ IMAD R0, R0, c[0x0][0x0], R5 ; /* 0x0000000000007a24 */ /* 0x004fca00078e0205 */ /*00c0*/ ISETP.GE.OR P0, PT, R0, c[0x0][0x178], P0 ; /* 0x00005e0000007a0c */ /* 0x000fda0000706670 */ /*00d0*/ @P0 EXIT ; /* 0x000000000000094d */ /* 0x000fea0003800000 */ /*00e0*/ HFMA2.MMA R9, -RZ, RZ, 0, 4.76837158203125e-07 ; /* 0x00000008ff097435 */ /* 0x000fe200000001ff */ /*00f0*/ IMAD R3, R3, c[0x0][0x17c], R2 ; /* 0x00005f0003037a24 */ /* 0x000fe200078e0202 */ /*0100*/ ULDC.64 UR4, c[0x0][0x118] ; /* 0x0000460000047ab9 */ /* 0x000fc60000000a00 */ /*0110*/ IMAD R0, R3, c[0x0][0x178], R0 ; /* 0x00005e0003007a24 */ /* 0x000fca00078e0200 */ /*0120*/ IMAD.WIDE R4, R0, R9, c[0x0][0x170] ; /* 0x00005c0000047625 */ /* 0x000fc800078e0209 */ /*0130*/ IMAD.WIDE R2, R0.reuse, R9.reuse, c[0x0][0x168] ; /* 0x00005a0000027625 */ /* 0x0c0fe400078e0209 */ /*0140*/ LDG.E.64 R4, [R4.64] ; /* 0x0000000404047981 */ /* 0x000ea8000c1e1b00 */ /*0150*/ LDG.E.64 R2, [R2.64] ; /* 0x0000000402027981 */ /* 0x000ea2000c1e1b00 */ /*0160*/ IMAD.WIDE R8, R0, R9, c[0x0][0x160] ; /* 0x0000580000087625 */ /* 0x000fe200078e0209 */ /*0170*/ DADD R6, R4, R2 ; /* 0x0000000004067229 */ /* 0x004e0e0000000002 */ /*0180*/ STG.E.64 [R8.64], R6 ; /* 0x0000000608007986 */ /* 0x001fe2000c101b04 */ /*0190*/ EXIT ; /* 0x000000000000794d */ /* 0x000fea0003800000 */ /*01a0*/ BRA 0x1a0; /* 0xfffffff000007947 */ /* 0x000fc0000383ffff */ /*01b0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*01c0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*01d0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*01e0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*01f0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0200*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0210*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0220*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0230*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0240*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0250*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0260*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0270*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ ..........
.text .amdgcn_target "amdgcn-amd-amdhsa--gfx1100" .protected _Z7memcopyPdS_S_iii .globl _Z7memcopyPdS_S_iii .p2align 8 .type _Z7memcopyPdS_S_iii,@function _Z7memcopyPdS_S_iii: s_clause 0x2 s_load_b64 s[2:3], s[0:1], 0x34 s_load_b32 s6, s[0:1], 0x20 s_load_b64 s[4:5], s[0:1], 0x18 v_and_b32_e32 v5, 0x3ff, v0 v_bfe_u32 v1, v0, 10, 10 v_bfe_u32 v0, v0, 20, 10 s_waitcnt lgkmcnt(0) s_and_b32 s3, s3, 0xffff s_and_b32 s7, s2, 0xffff s_delay_alu instid0(VALU_DEP_1) v_mad_u64_u32 v[2:3], null, s15, s3, v[0:1] s_lshr_b32 s2, s2, 16 s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1) v_mad_u64_u32 v[3:4], null, s14, s2, v[1:2] v_mad_u64_u32 v[0:1], null, s13, s7, v[5:6] v_cmp_gt_i32_e32 vcc_lo, s6, v2 s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3) v_cmp_gt_i32_e64 s2, s5, v3 v_cmp_gt_i32_e64 s3, s4, v0 s_delay_alu instid0(VALU_DEP_2) s_and_b32 s2, s2, vcc_lo s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1) s_and_b32 s2, s3, s2 s_delay_alu instid0(SALU_CYCLE_1) s_and_saveexec_b32 s3, s2 s_cbranch_execz .LBB0_2 v_mad_u64_u32 v[4:5], null, v2, s5, v[3:4] s_clause 0x1 s_load_b128 s[8:11], s[0:1], 0x0 s_load_b64 s[0:1], s[0:1], 0x10 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_mad_u64_u32 v[1:2], null, v4, s4, v[0:1] v_ashrrev_i32_e32 v2, 31, v1 s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1) v_lshlrev_b64 v[0:1], 3, v[1:2] s_waitcnt lgkmcnt(0) v_add_co_u32 v2, vcc_lo, s10, v0 s_delay_alu instid0(VALU_DEP_2) v_add_co_ci_u32_e32 v3, vcc_lo, s11, v1, vcc_lo v_add_co_u32 v4, vcc_lo, s0, v0 v_add_co_ci_u32_e32 v5, vcc_lo, s1, v1, vcc_lo v_add_co_u32 v0, vcc_lo, s8, v0 global_load_b64 v[2:3], v[2:3], off global_load_b64 v[4:5], v[4:5], off v_add_co_ci_u32_e32 v1, vcc_lo, s9, v1, vcc_lo s_waitcnt vmcnt(0) v_add_f64 v[2:3], v[2:3], v[4:5] global_store_b64 v[0:1], v[2:3], off .LBB0_2: s_nop 0 s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) s_endpgm .section .rodata,"a",@progbits .p2align 6, 0x0 .amdhsa_kernel _Z7memcopyPdS_S_iii .amdhsa_group_segment_fixed_size 0 .amdhsa_private_segment_fixed_size 0 .amdhsa_kernarg_size 296 .amdhsa_user_sgpr_count 13 .amdhsa_user_sgpr_dispatch_ptr 0 .amdhsa_user_sgpr_queue_ptr 0 .amdhsa_user_sgpr_kernarg_segment_ptr 1 .amdhsa_user_sgpr_dispatch_id 0 .amdhsa_user_sgpr_private_segment_size 0 .amdhsa_wavefront_size32 1 .amdhsa_uses_dynamic_stack 0 .amdhsa_enable_private_segment 0 .amdhsa_system_sgpr_workgroup_id_x 1 .amdhsa_system_sgpr_workgroup_id_y 1 .amdhsa_system_sgpr_workgroup_id_z 1 .amdhsa_system_sgpr_workgroup_info 0 .amdhsa_system_vgpr_workitem_id 2 .amdhsa_next_free_vgpr 7 .amdhsa_next_free_sgpr 16 .amdhsa_float_round_mode_32 0 .amdhsa_float_round_mode_16_64 0 .amdhsa_float_denorm_mode_32 3 .amdhsa_float_denorm_mode_16_64 3 .amdhsa_dx10_clamp 1 .amdhsa_ieee_mode 1 .amdhsa_fp16_overflow 0 .amdhsa_workgroup_processor_mode 1 .amdhsa_memory_ordered 1 .amdhsa_forward_progress 0 .amdhsa_shared_vgpr_count 0 .amdhsa_exception_fp_ieee_invalid_op 0 .amdhsa_exception_fp_denorm_src 0 .amdhsa_exception_fp_ieee_div_zero 0 .amdhsa_exception_fp_ieee_overflow 0 .amdhsa_exception_fp_ieee_underflow 0 .amdhsa_exception_fp_ieee_inexact 0 .amdhsa_exception_int_div_zero 0 .end_amdhsa_kernel .text .Lfunc_end0: .size _Z7memcopyPdS_S_iii, .Lfunc_end0-_Z7memcopyPdS_S_iii .section .AMDGPU.csdata,"",@progbits .text .p2alignl 7, 3214868480 .fill 96, 4, 3214868480 .type __hip_cuid_,@object .section .bss,"aw",@nobits .globl __hip_cuid_ __hip_cuid_: .byte 0 .size __hip_cuid_, 1 .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym __hip_cuid_ .amdgpu_metadata --- amdhsa.kernels: - .args: - .address_space: global .offset: 0 .size: 8 .value_kind: global_buffer - .address_space: global .offset: 8 .size: 8 .value_kind: global_buffer - .address_space: global .offset: 16 .size: 8 .value_kind: global_buffer - .offset: 24 .size: 4 .value_kind: by_value - .offset: 28 .size: 4 .value_kind: by_value - .offset: 32 .size: 4 .value_kind: by_value - .offset: 40 .size: 4 .value_kind: hidden_block_count_x - .offset: 44 .size: 4 .value_kind: hidden_block_count_y - .offset: 48 .size: 4 .value_kind: hidden_block_count_z - .offset: 52 .size: 2 .value_kind: hidden_group_size_x - .offset: 54 .size: 2 .value_kind: hidden_group_size_y - .offset: 56 .size: 2 .value_kind: hidden_group_size_z - .offset: 58 .size: 2 .value_kind: hidden_remainder_x - .offset: 60 .size: 2 .value_kind: hidden_remainder_y - .offset: 62 .size: 2 .value_kind: hidden_remainder_z - .offset: 80 .size: 8 .value_kind: hidden_global_offset_x - .offset: 88 .size: 8 .value_kind: hidden_global_offset_y - .offset: 96 .size: 8 .value_kind: hidden_global_offset_z - .offset: 104 .size: 2 .value_kind: hidden_grid_dims .group_segment_fixed_size: 0 .kernarg_segment_align: 8 .kernarg_segment_size: 296 .language: OpenCL C .language_version: - 2 - 0 .max_flat_workgroup_size: 1024 .name: _Z7memcopyPdS_S_iii .private_segment_fixed_size: 0 .sgpr_count: 18 .sgpr_spill_count: 0 .symbol: _Z7memcopyPdS_S_iii.kd .uniform_work_group_size: 1 .uses_dynamic_stack: false .vgpr_count: 7 .vgpr_spill_count: 0 .wavefront_size: 32 .workgroup_processor_mode: 1 amdhsa.target: amdgcn-amd-amdhsa--gfx1100 amdhsa.version: - 1 - 2 ... .end_amdgpu_metadata
You are a helpful coding assistant specialized in converting from CUDA host assembly to AMD host assembly.
.file "tmpxft_0001762b_00000000-6_memcopy.cudafe1.cpp" .text #APP #NO_APP .type _ZL26__cudaUnregisterBinaryUtilv, @function _ZL26__cudaUnregisterBinaryUtilv: .LFB2066: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaUnregisterFatBinary@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2066: .size _ZL26__cudaUnregisterBinaryUtilv, .-_ZL26__cudaUnregisterBinaryUtilv .globl _Z7cpu_secv .type _Z7cpu_secv, @function _Z7cpu_secv: .LFB2057: .cfi_startproc endbr64 subq $40, %rsp .cfi_def_cfa_offset 48 movq %fs:40, %rax movq %rax, 24(%rsp) xorl %eax, %eax movq %rsp, %rdi movl $0, %esi call gettimeofday@PLT pxor %xmm0, %xmm0 cvtsi2sdq 8(%rsp), %xmm0 mulsd .LC0(%rip), %xmm0 pxor %xmm1, %xmm1 cvtsi2sdq (%rsp), %xmm1 addsd %xmm1, %xmm0 movq 24(%rsp), %rax subq %fs:40, %rax jne .L6 addq $40, %rsp .cfi_remember_state .cfi_def_cfa_offset 8 ret .L6: .cfi_restore_state call __stack_chk_fail@PLT .cfi_endproc .LFE2057: .size _Z7cpu_secv, .-_Z7cpu_secv .globl _Z3ticv .type _Z3ticv, @function _Z3ticv: .LFB2058: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 call _Z7cpu_secv movsd %xmm0, timer_start(%rip) addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2058: .size _Z3ticv, .-_Z3ticv .globl _Z3tocv .type _Z3tocv, @function _Z3tocv: .LFB2059: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 call _Z7cpu_secv subsd timer_start(%rip), %xmm0 addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2059: .size _Z3tocv, .-_Z3tocv .section .rodata.str1.1,"aMS",@progbits,1 .LC1: .string "%s: %8.3f seconds" .LC3: .string ", %8.3f GB/s" .LC4: .string "\n" .text .globl _Z3timPKcd .type _Z3timPKcd, @function _Z3timPKcd: .LFB2060: .cfi_startproc endbr64 pushq %rbx .cfi_def_cfa_offset 16 .cfi_offset 3, -16 subq $16, %rsp .cfi_def_cfa_offset 32 movq %rdi, %rbx movsd %xmm0, (%rsp) call _Z3tocv movsd %xmm0, 8(%rsp) movq %rbx, %rdx leaq .LC1(%rip), %rsi movl $2, %edi movl $1, %eax call __printf_chk@PLT pxor %xmm0, %xmm0 movsd (%rsp), %xmm1 comisd %xmm0, %xmm1 ja .L16 .L12: leaq .LC4(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT addq $16, %rsp .cfi_remember_state .cfi_def_cfa_offset 16 popq %rbx .cfi_def_cfa_offset 8 ret .L16: .cfi_restore_state divsd 8(%rsp), %xmm1 movapd %xmm1, %xmm0 leaq .LC3(%rip), %rsi movl $2, %edi movl $1, %eax call __printf_chk@PLT jmp .L12 .cfi_endproc .LFE2060: .size _Z3timPKcd, .-_Z3timPKcd .section .rodata.str1.1 .LC5: .string "a" .LC6: .string "PERF_memcpy.dat" .section .rodata.str1.8,"aMS",@progbits,1 .align 8 .LC7: .string "nx=%d ny=%d nz=%d GBs=%1.4f time_s=%1.4f \n" .text .globl _Z8timPrintPKcdiii .type _Z8timPrintPKcdiii, @function _Z8timPrintPKcdiii: .LFB2061: .cfi_startproc endbr64 pushq %r13 .cfi_def_cfa_offset 16 .cfi_offset 13, -16 pushq %r12 .cfi_def_cfa_offset 24 .cfi_offset 12, -24 pushq %rbp .cfi_def_cfa_offset 32 .cfi_offset 6, -32 pushq %rbx .cfi_def_cfa_offset 40 .cfi_offset 3, -40 subq $24, %rsp .cfi_def_cfa_offset 64 movq %rdi, %rbx movsd %xmm0, (%rsp) movl %esi, %ebp movl %edx, %r12d movl %ecx, %r13d call _Z3tocv movsd %xmm0, 8(%rsp) movq %rbx, %rdx leaq .LC1(%rip), %rsi movl $2, %edi movl $1, %eax call __printf_chk@PLT pxor %xmm0, %xmm0 movsd (%rsp), %xmm2 comisd %xmm0, %xmm2 ja .L22 .L18: leaq .LC4(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT leaq .LC5(%rip), %rsi leaq .LC6(%rip), %rdi call fopen@PLT movq %rax, %rbx movsd (%rsp), %xmm0 movsd 8(%rsp), %xmm1 divsd %xmm1, %xmm0 movl %r13d, %r9d movl %r12d, %r8d movl %ebp, %ecx leaq .LC7(%rip), %rdx movl $2, %esi movq %rax, %rdi movl $2, %eax call __fprintf_chk@PLT movq %rbx, %rdi call fclose@PLT addq $24, %rsp .cfi_remember_state .cfi_def_cfa_offset 40 popq %rbx .cfi_def_cfa_offset 32 popq %rbp .cfi_def_cfa_offset 24 popq %r12 .cfi_def_cfa_offset 16 popq %r13 .cfi_def_cfa_offset 8 ret .L22: .cfi_restore_state divsd 8(%rsp), %xmm2 movapd %xmm2, %xmm0 leaq .LC3(%rip), %rsi movl $2, %edi movl $1, %eax call __printf_chk@PLT jmp .L18 .cfi_endproc .LFE2061: .size _Z8timPrintPKcdiii, .-_Z8timPrintPKcdiii .section .rodata.str1.8 .align 8 .LC8: .string "ERROR launching GPU C-CUDA program: %s\n" .text .globl _Z10clean_cudav .type _Z10clean_cudav, @function _Z10clean_cudav: .LFB2062: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 call cudaGetLastError@PLT testl %eax, %eax jne .L26 .L23: addq $8, %rsp .cfi_remember_state .cfi_def_cfa_offset 8 ret .L26: .cfi_restore_state movl %eax, %edi call cudaGetErrorString@PLT movq %rax, %rdx leaq .LC8(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT call cudaDeviceReset@PLT jmp .L23 .cfi_endproc .LFE2062: .size _Z10clean_cudav, .-_Z10clean_cudav .globl _Z33__device_stub__Z7memcopyPdS_S_iiiPdS_S_iii .type _Z33__device_stub__Z7memcopyPdS_S_iiiPdS_S_iii, @function _Z33__device_stub__Z7memcopyPdS_S_iiiPdS_S_iii: .LFB2088: .cfi_startproc endbr64 subq $184, %rsp .cfi_def_cfa_offset 192 movq %rdi, 40(%rsp) movq %rsi, 32(%rsp) movq %rdx, 24(%rsp) movl %ecx, 20(%rsp) movl %r8d, 16(%rsp) movl %r9d, 12(%rsp) movq %fs:40, %rax movq %rax, 168(%rsp) xorl %eax, %eax leaq 40(%rsp), %rax movq %rax, 112(%rsp) leaq 32(%rsp), %rax movq %rax, 120(%rsp) leaq 24(%rsp), %rax movq %rax, 128(%rsp) leaq 20(%rsp), %rax movq %rax, 136(%rsp) leaq 16(%rsp), %rax movq %rax, 144(%rsp) leaq 12(%rsp), %rax movq %rax, 152(%rsp) movl $1, 64(%rsp) movl $1, 68(%rsp) movl $1, 72(%rsp) movl $1, 76(%rsp) movl $1, 80(%rsp) movl $1, 84(%rsp) leaq 56(%rsp), %rcx leaq 48(%rsp), %rdx leaq 76(%rsp), %rsi leaq 64(%rsp), %rdi call __cudaPopCallConfiguration@PLT testl %eax, %eax je .L31 .L27: movq 168(%rsp), %rax subq %fs:40, %rax jne .L32 addq $184, %rsp .cfi_remember_state .cfi_def_cfa_offset 8 ret .L31: .cfi_restore_state pushq 56(%rsp) .cfi_def_cfa_offset 200 pushq 56(%rsp) .cfi_def_cfa_offset 208 leaq 128(%rsp), %r9 movq 92(%rsp), %rcx movl 100(%rsp), %r8d movq 80(%rsp), %rsi movl 88(%rsp), %edx leaq _Z7memcopyPdS_S_iii(%rip), %rdi call cudaLaunchKernel@PLT addq $16, %rsp .cfi_def_cfa_offset 192 jmp .L27 .L32: call __stack_chk_fail@PLT .cfi_endproc .LFE2088: .size _Z33__device_stub__Z7memcopyPdS_S_iiiPdS_S_iii, .-_Z33__device_stub__Z7memcopyPdS_S_iiiPdS_S_iii .globl _Z7memcopyPdS_S_iii .type _Z7memcopyPdS_S_iii, @function _Z7memcopyPdS_S_iii: .LFB2089: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 call _Z33__device_stub__Z7memcopyPdS_S_iiiPdS_S_iii addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2089: .size _Z7memcopyPdS_S_iii, .-_Z7memcopyPdS_S_iii .section .rodata.str1.1 .LC9: .string "Process uses GPU with id %d.\n" .section .rodata.str1.8 .align 8 .LC11: .string "%dx%dx%d, %1.3f GB, %d iterations.\n" .align 8 .LC12: .string "launching (%dx%dx%d) grid of (%dx%dx%d) blocks.\n" .section .rodata.str1.1 .LC14: .string "Performance" .text .globl main .type main, @function main: .LFB2063: .cfi_startproc endbr64 pushq %r13 .cfi_def_cfa_offset 16 .cfi_offset 13, -16 pushq %r12 .cfi_def_cfa_offset 24 .cfi_offset 12, -24 pushq %rbp .cfi_def_cfa_offset 32 .cfi_offset 6, -32 pushq %rbx .cfi_def_cfa_offset 40 .cfi_offset 3, -40 subq $72, %rsp .cfi_def_cfa_offset 112 movq %fs:40, %rax movq %rax, 56(%rsp) xorl %eax, %eax movl $0, 4(%rsp) movl $0, %edi call cudaSetDevice@PLT leaq 4(%rsp), %rdi call cudaGetDevice@PLT call cudaDeviceReset@PLT movl $2, %edi call cudaDeviceSetCacheConfig@PLT movl 4(%rsp), %edx leaq .LC9(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT movl $100, %r9d movsd .LC10(%rip), %xmm0 movl $256, %r8d movl $1024, %ecx movl $1024, %edx leaq .LC11(%rip), %rsi movl $2, %edi movl $1, %eax call __printf_chk@PLT pushq $2 .cfi_def_cfa_offset 120 pushq $16 .cfi_def_cfa_offset 128 movl $32, %r9d movl $128, %r8d movl $64, %ecx movl $32, %edx leaq .LC12(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT movl $2147483648, %ebx movq %rbx, %rdi call malloc@PLT movq %rax, %rbp leaq (%rax,%rbx), %rdx addq $16, %rsp .cfi_def_cfa_offset 112 .L36: movq $0x000000000, (%rax) addq $8, %rax cmpq %rdx, %rax jne .L36 leaq 8(%rsp), %rdi movl $2147483648, %ebx movq %rbx, %rsi call cudaMalloc@PLT movl $1, %ecx movq %rbx, %rdx movq %rbp, %rsi movq 8(%rsp), %rdi call cudaMemcpy@PLT movq %rbx, %rdi call malloc@PLT movq %rax, %r12 leaq (%rax,%rbx), %rdx .L37: movq $0x000000000, (%rax) addq $8, %rax cmpq %rdx, %rax jne .L37 leaq 16(%rsp), %rdi movl $2147483648, %ebx movq %rbx, %rsi call cudaMalloc@PLT movl $1, %ecx movq %rbx, %rdx movq %r12, %rsi movq 16(%rsp), %rdi call cudaMemcpy@PLT movq %rbx, %rdi call malloc@PLT movq %rax, %r13 leaq (%rax,%rbx), %rdx .L38: movq $0x000000000, (%rax) addq $8, %rax cmpq %rdx, %rax jne .L38 leaq 24(%rsp), %rdi movl $2147483648, %ebx movq %rbx, %rsi call cudaMalloc@PLT movl $1, %ecx movq %rbx, %rdx movq %r13, %rsi movq 24(%rsp), %rdi call cudaMemcpy@PLT movl $0, %ebx jmp .L44 .L53: call _Z3ticv movl $32, 32(%rsp) movl $64, 36(%rsp) movl $32, 44(%rsp) movl $16, 48(%rsp) movl $0, %r9d movl $0, %r8d movq 44(%rsp), %rdx movl $2, %ecx movq 32(%rsp), %rdi movl $128, %esi call __cudaPushCallConfiguration@PLT testl %eax, %eax je .L41 call cudaDeviceSynchronize@PLT addq $1, %rbx jmp .L44 .L41: movl $256, %r9d movl $1024, %r8d movl $1024, %ecx movq 24(%rsp), %rdx movq 16(%rsp), %rsi movq 8(%rsp), %rdi call _Z33__device_stub__Z7memcopyPdS_S_iiiPdS_S_iii .L42: call cudaDeviceSynchronize@PLT addq $1, %rbx cmpq $100, %rbx je .L52 .L44: cmpq $10, %rbx je .L53 movl $32, 32(%rsp) movl $64, 36(%rsp) movl $32, 44(%rsp) movl $16, 48(%rsp) movl $0, %r9d movl $0, %r8d movq 44(%rsp), %rdx movl $2, %ecx movq 32(%rsp), %rdi movl $128, %esi call __cudaPushCallConfiguration@PLT testl %eax, %eax jne .L42 jmp .L41 .L52: movl $256, %ecx movl $1024, %edx movl $1024, %esi movsd .LC13(%rip), %xmm0 leaq .LC14(%rip), %rdi call _Z8timPrintPKcdiii movq %rbp, %rdi call free@PLT movq 8(%rsp), %rdi call cudaFree@PLT movq %r12, %rdi call free@PLT movq 16(%rsp), %rdi call cudaFree@PLT movq %r13, %rdi call free@PLT movq 24(%rsp), %rdi call cudaFree@PLT call _Z10clean_cudav movq 56(%rsp), %rax subq %fs:40, %rax jne .L54 movl $0, %eax addq $72, %rsp .cfi_remember_state .cfi_def_cfa_offset 40 popq %rbx .cfi_def_cfa_offset 32 popq %rbp .cfi_def_cfa_offset 24 popq %r12 .cfi_def_cfa_offset 16 popq %r13 .cfi_def_cfa_offset 8 ret .L54: .cfi_restore_state call __stack_chk_fail@PLT .cfi_endproc .LFE2063: .size main, .-main .section .rodata.str1.1 .LC15: .string "_Z7memcopyPdS_S_iii" .text .type _ZL24__sti____cudaRegisterAllv, @function _ZL24__sti____cudaRegisterAllv: .LFB2091: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 leaq _ZL15__fatDeviceText(%rip), %rdi call __cudaRegisterFatBinary@PLT movq %rax, %rdi movq %rax, _ZL20__cudaFatCubinHandle(%rip) pushq $0 .cfi_def_cfa_offset 24 pushq $0 .cfi_def_cfa_offset 32 pushq $0 .cfi_def_cfa_offset 40 pushq $0 .cfi_def_cfa_offset 48 movl $0, %r9d movl $-1, %r8d leaq .LC15(%rip), %rdx movq %rdx, %rcx leaq _Z7memcopyPdS_S_iii(%rip), %rsi call __cudaRegisterFunction@PLT addq $32, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaRegisterFatBinaryEnd@PLT leaq _ZL26__cudaUnregisterBinaryUtilv(%rip), %rdi call atexit@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2091: .size _ZL24__sti____cudaRegisterAllv, .-_ZL24__sti____cudaRegisterAllv .section .init_array,"aw" .align 8 .quad _ZL24__sti____cudaRegisterAllv .section .nvFatBinSegment,"aw" .align 8 .type _ZL15__fatDeviceText, @object .size _ZL15__fatDeviceText, 24 _ZL15__fatDeviceText: .long 1180844977 .long 1 .quad fatbinData .quad 0 .local _ZL20__cudaFatCubinHandle .comm _ZL20__cudaFatCubinHandle,8,8 .globl timer_start .bss .align 8 .type timer_start, @object .size timer_start, 8 timer_start: .zero 8 .section .rodata.cst8,"aM",@progbits,8 .align 8 .LC0: .long -1598689907 .long 1051772663 .align 8 .LC10: .long 0 .long 1075314688 .align 8 .LC13: .long 0 .long 1082187776 .ident "GCC: (Ubuntu 13.3.0-6ubuntu2~24.04) 13.3.0" .section .note.GNU-stack,"",@progbits .section .note.gnu.property,"a" .align 8 .long 1f - 0f .long 4f - 1f .long 5 0: .string "GNU" 1: .align 8 .long 0xc0000002 .long 3f - 2f 2: .long 0x3 3: .align 8 4:
.text .file "memcopy.hip" .section .rodata.cst8,"aM",@progbits,8 .p2align 3, 0x0 # -- Begin function _Z7cpu_secv .LCPI0_0: .quad 0x3eb0c6f7a0b5ed8d # double 9.9999999999999995E-7 .text .globl _Z7cpu_secv .p2align 4, 0x90 .type _Z7cpu_secv,@function _Z7cpu_secv: # @_Z7cpu_secv .cfi_startproc # %bb.0: subq $24, %rsp .cfi_def_cfa_offset 32 leaq 8(%rsp), %rdi xorl %esi, %esi callq gettimeofday cvtsi2sdq 8(%rsp), %xmm1 cvtsi2sdq 16(%rsp), %xmm0 mulsd .LCPI0_0(%rip), %xmm0 addsd %xmm1, %xmm0 addq $24, %rsp .cfi_def_cfa_offset 8 retq .Lfunc_end0: .size _Z7cpu_secv, .Lfunc_end0-_Z7cpu_secv .cfi_endproc # -- End function .section .rodata.cst8,"aM",@progbits,8 .p2align 3, 0x0 # -- Begin function _Z3ticv .LCPI1_0: .quad 0x3eb0c6f7a0b5ed8d # double 9.9999999999999995E-7 .text .globl _Z3ticv .p2align 4, 0x90 .type _Z3ticv,@function _Z3ticv: # @_Z3ticv .cfi_startproc # %bb.0: subq $24, %rsp .cfi_def_cfa_offset 32 leaq 8(%rsp), %rdi xorl %esi, %esi callq gettimeofday cvtsi2sdq 8(%rsp), %xmm0 cvtsi2sdq 16(%rsp), %xmm1 mulsd .LCPI1_0(%rip), %xmm1 addsd %xmm0, %xmm1 movsd %xmm1, timer_start(%rip) addq $24, %rsp .cfi_def_cfa_offset 8 retq .Lfunc_end1: .size _Z3ticv, .Lfunc_end1-_Z3ticv .cfi_endproc # -- End function .section .rodata.cst8,"aM",@progbits,8 .p2align 3, 0x0 # -- Begin function _Z3tocv .LCPI2_0: .quad 0x3eb0c6f7a0b5ed8d # double 9.9999999999999995E-7 .text .globl _Z3tocv .p2align 4, 0x90 .type _Z3tocv,@function _Z3tocv: # @_Z3tocv .cfi_startproc # %bb.0: subq $24, %rsp .cfi_def_cfa_offset 32 leaq 8(%rsp), %rdi xorl %esi, %esi callq gettimeofday cvtsi2sdq 8(%rsp), %xmm1 cvtsi2sdq 16(%rsp), %xmm0 mulsd .LCPI2_0(%rip), %xmm0 addsd %xmm1, %xmm0 subsd timer_start(%rip), %xmm0 addq $24, %rsp .cfi_def_cfa_offset 8 retq .Lfunc_end2: .size _Z3tocv, .Lfunc_end2-_Z3tocv .cfi_endproc # -- End function .section .rodata.cst8,"aM",@progbits,8 .p2align 3, 0x0 # -- Begin function _Z3timPKcd .LCPI3_0: .quad 0x3eb0c6f7a0b5ed8d # double 9.9999999999999995E-7 .text .globl _Z3timPKcd .p2align 4, 0x90 .type _Z3timPKcd,@function _Z3timPKcd: # @_Z3timPKcd .cfi_startproc # %bb.0: pushq %rbx .cfi_def_cfa_offset 16 subq $32, %rsp .cfi_def_cfa_offset 48 .cfi_offset %rbx, -16 movsd %xmm0, 8(%rsp) # 8-byte Spill movq %rdi, %rbx leaq 16(%rsp), %rdi xorl %esi, %esi callq gettimeofday cvtsi2sdq 16(%rsp), %xmm1 xorps %xmm0, %xmm0 cvtsi2sdq 24(%rsp), %xmm0 mulsd .LCPI3_0(%rip), %xmm0 addsd %xmm1, %xmm0 subsd timer_start(%rip), %xmm0 movl $.L.str, %edi movq %rbx, %rsi movsd %xmm0, (%rsp) # 8-byte Spill movb $1, %al callq printf xorpd %xmm0, %xmm0 movsd 8(%rsp), %xmm1 # 8-byte Reload # xmm1 = mem[0],zero ucomisd %xmm0, %xmm1 jbe .LBB3_2 # %bb.1: movapd %xmm1, %xmm0 divsd (%rsp), %xmm0 # 8-byte Folded Reload movl $.L.str.1, %edi movb $1, %al callq printf .LBB3_2: movl $10, %edi addq $32, %rsp .cfi_def_cfa_offset 16 popq %rbx .cfi_def_cfa_offset 8 jmp putchar@PLT # TAILCALL .Lfunc_end3: .size _Z3timPKcd, .Lfunc_end3-_Z3timPKcd .cfi_endproc # -- End function .section .rodata.cst8,"aM",@progbits,8 .p2align 3, 0x0 # -- Begin function _Z8timPrintPKcdiii .LCPI4_0: .quad 0x3eb0c6f7a0b5ed8d # double 9.9999999999999995E-7 .text .globl _Z8timPrintPKcdiii .p2align 4, 0x90 .type _Z8timPrintPKcdiii,@function _Z8timPrintPKcdiii: # @_Z8timPrintPKcdiii .cfi_startproc # %bb.0: pushq %rbp .cfi_def_cfa_offset 16 pushq %r15 .cfi_def_cfa_offset 24 pushq %r14 .cfi_def_cfa_offset 32 pushq %rbx .cfi_def_cfa_offset 40 subq $40, %rsp .cfi_def_cfa_offset 80 .cfi_offset %rbx, -40 .cfi_offset %r14, -32 .cfi_offset %r15, -24 .cfi_offset %rbp, -16 movl %ecx, %ebx movl %edx, %ebp movl %esi, %r14d movsd %xmm0, 8(%rsp) # 8-byte Spill movq %rdi, %r15 leaq 24(%rsp), %rdi xorl %esi, %esi callq gettimeofday cvtsi2sdq 24(%rsp), %xmm1 xorps %xmm0, %xmm0 cvtsi2sdq 32(%rsp), %xmm0 mulsd .LCPI4_0(%rip), %xmm0 addsd %xmm1, %xmm0 subsd timer_start(%rip), %xmm0 movsd %xmm0, 16(%rsp) # 8-byte Spill movl $.L.str, %edi movq %r15, %rsi movb $1, %al callq printf xorpd %xmm0, %xmm0 movsd 8(%rsp), %xmm1 # 8-byte Reload # xmm1 = mem[0],zero ucomisd %xmm0, %xmm1 divsd 16(%rsp), %xmm1 # 8-byte Folded Reload movsd %xmm1, 8(%rsp) # 8-byte Spill jbe .LBB4_2 # %bb.1: movl $.L.str.1, %edi movsd 8(%rsp), %xmm0 # 8-byte Reload # xmm0 = mem[0],zero movb $1, %al callq printf .LBB4_2: movl $10, %edi callq putchar@PLT movl $.L.str.3, %edi movl $.L.str.4, %esi callq fopen movq %rax, %r15 movl $.L.str.5, %esi movq %rax, %rdi movl %r14d, %edx movl %ebp, %ecx movl %ebx, %r8d movsd 8(%rsp), %xmm0 # 8-byte Reload # xmm0 = mem[0],zero movsd 16(%rsp), %xmm1 # 8-byte Reload # xmm1 = mem[0],zero movb $2, %al callq fprintf movq %r15, %rdi addq $40, %rsp .cfi_def_cfa_offset 40 popq %rbx .cfi_def_cfa_offset 32 popq %r14 .cfi_def_cfa_offset 24 popq %r15 .cfi_def_cfa_offset 16 popq %rbp .cfi_def_cfa_offset 8 jmp fclose # TAILCALL .Lfunc_end4: .size _Z8timPrintPKcdiii, .Lfunc_end4-_Z8timPrintPKcdiii .cfi_endproc # -- End function .globl _Z10clean_cudav # -- Begin function _Z10clean_cudav .p2align 4, 0x90 .type _Z10clean_cudav,@function _Z10clean_cudav: # @_Z10clean_cudav .cfi_startproc # %bb.0: pushq %rax .cfi_def_cfa_offset 16 callq hipGetLastError testl %eax, %eax je .LBB5_1 # %bb.2: movl %eax, %edi callq hipGetErrorString movl $.L.str.6, %edi movq %rax, %rsi xorl %eax, %eax callq printf popq %rax .cfi_def_cfa_offset 8 jmp hipDeviceReset # TAILCALL .LBB5_1: .cfi_def_cfa_offset 16 popq %rax .cfi_def_cfa_offset 8 retq .Lfunc_end5: .size _Z10clean_cudav, .Lfunc_end5-_Z10clean_cudav .cfi_endproc # -- End function .globl _Z22__device_stub__memcopyPdS_S_iii # -- Begin function _Z22__device_stub__memcopyPdS_S_iii .p2align 4, 0x90 .type _Z22__device_stub__memcopyPdS_S_iii,@function _Z22__device_stub__memcopyPdS_S_iii: # @_Z22__device_stub__memcopyPdS_S_iii .cfi_startproc # %bb.0: subq $152, %rsp .cfi_def_cfa_offset 160 movq %rdi, 88(%rsp) movq %rsi, 80(%rsp) movq %rdx, 72(%rsp) movl %ecx, 20(%rsp) movl %r8d, 16(%rsp) movl %r9d, 12(%rsp) leaq 88(%rsp), %rax movq %rax, 96(%rsp) leaq 80(%rsp), %rax movq %rax, 104(%rsp) leaq 72(%rsp), %rax movq %rax, 112(%rsp) leaq 20(%rsp), %rax movq %rax, 120(%rsp) leaq 16(%rsp), %rax movq %rax, 128(%rsp) leaq 12(%rsp), %rax movq %rax, 136(%rsp) leaq 56(%rsp), %rdi leaq 40(%rsp), %rsi leaq 32(%rsp), %rdx leaq 24(%rsp), %rcx callq __hipPopCallConfiguration movq 56(%rsp), %rsi movl 64(%rsp), %edx movq 40(%rsp), %rcx movl 48(%rsp), %r8d leaq 96(%rsp), %r9 movl $_Z7memcopyPdS_S_iii, %edi pushq 24(%rsp) .cfi_adjust_cfa_offset 8 pushq 40(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $168, %rsp .cfi_adjust_cfa_offset -168 retq .Lfunc_end6: .size _Z22__device_stub__memcopyPdS_S_iii, .Lfunc_end6-_Z22__device_stub__memcopyPdS_S_iii .cfi_endproc # -- End function .section .rodata.cst8,"aM",@progbits,8 .p2align 3, 0x0 # -- Begin function main .LCPI7_0: .quad 0x4018000000000000 # double 6 .LCPI7_1: .quad 0x3eb0c6f7a0b5ed8d # double 9.9999999999999995E-7 .LCPI7_2: .quad 0x4080e00000000000 # double 540 .text .globl main .p2align 4, 0x90 .type main,@function main: # @main .cfi_startproc # %bb.0: pushq %rbp .cfi_def_cfa_offset 16 pushq %r15 .cfi_def_cfa_offset 24 pushq %r14 .cfi_def_cfa_offset 32 pushq %r13 .cfi_def_cfa_offset 40 pushq %r12 .cfi_def_cfa_offset 48 pushq %rbx .cfi_def_cfa_offset 56 subq $216, %rsp .cfi_def_cfa_offset 272 .cfi_offset %rbx, -56 .cfi_offset %r12, -48 .cfi_offset %r13, -40 .cfi_offset %r14, -32 .cfi_offset %r15, -24 .cfi_offset %rbp, -16 movl $0, 12(%rsp) xorl %edi, %edi callq hipSetDevice leaq 12(%rsp), %rdi callq hipGetDevice callq hipDeviceReset movl $2, %edi callq hipDeviceSetCacheConfig movl 12(%rsp), %esi movl $.L.str.7, %edi xorl %eax, %eax callq printf movsd .LCPI7_0(%rip), %xmm0 # xmm0 = mem[0],zero movl $.L.str.8, %edi movl $1024, %esi # imm = 0x400 movl $1024, %edx # imm = 0x400 movl $256, %ecx # imm = 0x100 movl $100, %r8d movb $1, %al callq printf subq $8, %rsp .cfi_adjust_cfa_offset 8 movl $.L.str.9, %edi movl $32, %esi movl $64, %edx movl $128, %ecx movl $32, %r8d movl $16, %r9d xorl %eax, %eax pushq $2 .cfi_adjust_cfa_offset 8 callq printf addq $16, %rsp .cfi_adjust_cfa_offset -16 movl $2147483648, %edi # imm = 0x80000000 callq malloc movq %rax, %rbx movl $2147483648, %edx # imm = 0x80000000 movq %rax, %rdi xorl %esi, %esi callq memset@PLT leaq 32(%rsp), %rdi movl $2147483648, %esi # imm = 0x80000000 callq hipMalloc movq 32(%rsp), %rdi movl $2147483648, %edx # imm = 0x80000000 movq %rbx, 136(%rsp) # 8-byte Spill movq %rbx, %rsi movl $1, %ecx callq hipMemcpy movl $2147483648, %edi # imm = 0x80000000 callq malloc movq %rax, %rbx movl $2147483648, %edx # imm = 0x80000000 movq %rax, %rdi xorl %esi, %esi callq memset@PLT leaq 24(%rsp), %rdi movl $2147483648, %esi # imm = 0x80000000 callq hipMalloc movq 24(%rsp), %rdi movl $2147483648, %edx # imm = 0x80000000 movq %rbx, 128(%rsp) # 8-byte Spill movq %rbx, %rsi movl $1, %ecx callq hipMemcpy movl $2147483648, %edi # imm = 0x80000000 callq malloc movq %rax, %rbx movl $2147483648, %edx # imm = 0x80000000 movq %rax, %rdi xorl %esi, %esi callq memset@PLT leaq 16(%rsp), %rdi movl $2147483648, %esi # imm = 0x80000000 callq hipMalloc movq 16(%rsp), %rdi movl $2147483648, %edx # imm = 0x80000000 movq %rbx, 120(%rsp) # 8-byte Spill movq %rbx, %rsi movl $1, %ecx callq hipMemcpy movl $100, %r14d leaq 64(%rsp), %r12 movabsq $274877906976, %r13 # imm = 0x4000000020 movabsq $68719476768, %rbp # imm = 0x1000000020 leaq 152(%rsp), %r15 leaq 144(%rsp), %rbx jmp .LBB7_1 .p2align 4, 0x90 .LBB7_3: # in Loop: Header=BB7_1 Depth=1 movq %r13, %rdi movl $128, %esi movq %rbp, %rdx movl $2, %ecx xorl %r8d, %r8d xorl %r9d, %r9d callq __hipPushCallConfiguration testl %eax, %eax je .LBB7_4 .LBB7_5: # in Loop: Header=BB7_1 Depth=1 callq hipDeviceSynchronize decq %r14 je .LBB7_6 .LBB7_1: # =>This Inner Loop Header: Depth=1 cmpq $90, %r14 jne .LBB7_3 # %bb.2: # in Loop: Header=BB7_1 Depth=1 movq %r12, %rdi xorl %esi, %esi callq gettimeofday xorps %xmm0, %xmm0 cvtsi2sdq 64(%rsp), %xmm0 xorps %xmm1, %xmm1 cvtsi2sdq 72(%rsp), %xmm1 mulsd .LCPI7_1(%rip), %xmm1 addsd %xmm0, %xmm1 movsd %xmm1, timer_start(%rip) jmp .LBB7_3 .p2align 4, 0x90 .LBB7_4: # in Loop: Header=BB7_1 Depth=1 movq 32(%rsp), %rax movq 24(%rsp), %rcx movq 16(%rsp), %rdx movq %rax, 208(%rsp) movq %rcx, 200(%rsp) movq %rdx, 192(%rsp) movl $1024, 52(%rsp) # imm = 0x400 movl $1024, 48(%rsp) # imm = 0x400 movl $256, 44(%rsp) # imm = 0x100 leaq 208(%rsp), %rax movq %rax, 64(%rsp) leaq 200(%rsp), %rax movq %rax, 72(%rsp) leaq 192(%rsp), %rax movq %rax, 80(%rsp) leaq 52(%rsp), %rax movq %rax, 88(%rsp) leaq 48(%rsp), %rax movq %rax, 96(%rsp) leaq 44(%rsp), %rax movq %rax, 104(%rsp) leaq 176(%rsp), %rdi leaq 160(%rsp), %rsi movq %r15, %rdx movq %rbx, %rcx callq __hipPopCallConfiguration movq 176(%rsp), %rsi movl 184(%rsp), %edx movq 160(%rsp), %rcx movl 168(%rsp), %r8d movl $_Z7memcopyPdS_S_iii, %edi movq %r12, %r9 pushq 144(%rsp) .cfi_adjust_cfa_offset 8 pushq 160(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $16, %rsp .cfi_adjust_cfa_offset -16 jmp .LBB7_5 .LBB7_6: leaq 64(%rsp), %rdi xorl %esi, %esi callq gettimeofday xorps %xmm1, %xmm1 cvtsi2sdq 64(%rsp), %xmm1 xorps %xmm0, %xmm0 cvtsi2sdq 72(%rsp), %xmm0 mulsd .LCPI7_1(%rip), %xmm0 addsd %xmm1, %xmm0 subsd timer_start(%rip), %xmm0 movsd %xmm0, 56(%rsp) # 8-byte Spill movl $.L.str, %edi movl $.L.str.10, %esi movb $1, %al callq printf movsd .LCPI7_2(%rip), %xmm0 # xmm0 = mem[0],zero divsd 56(%rsp), %xmm0 # 8-byte Folded Reload movsd %xmm0, 112(%rsp) # 8-byte Spill movl $.L.str.1, %edi movb $1, %al callq printf movl $10, %edi callq putchar@PLT movl $.L.str.3, %edi movl $.L.str.4, %esi callq fopen movq %rax, %rbx movl $.L.str.5, %esi movq %rax, %rdi movl $1024, %edx # imm = 0x400 movl $1024, %ecx # imm = 0x400 movl $256, %r8d # imm = 0x100 movsd 112(%rsp), %xmm0 # 8-byte Reload # xmm0 = mem[0],zero movsd 56(%rsp), %xmm1 # 8-byte Reload # xmm1 = mem[0],zero movb $2, %al callq fprintf movq %rbx, %rdi callq fclose movq 136(%rsp), %rdi # 8-byte Reload callq free movq 32(%rsp), %rdi callq hipFree movq 128(%rsp), %rdi # 8-byte Reload callq free movq 24(%rsp), %rdi callq hipFree movq 120(%rsp), %rdi # 8-byte Reload callq free movq 16(%rsp), %rdi callq hipFree callq hipGetLastError testl %eax, %eax je .LBB7_8 # %bb.7: movl %eax, %edi callq hipGetErrorString movl $.L.str.6, %edi movq %rax, %rsi xorl %eax, %eax callq printf callq hipDeviceReset .LBB7_8: # %_Z10clean_cudav.exit xorl %eax, %eax addq $216, %rsp .cfi_def_cfa_offset 56 popq %rbx .cfi_def_cfa_offset 48 popq %r12 .cfi_def_cfa_offset 40 popq %r13 .cfi_def_cfa_offset 32 popq %r14 .cfi_def_cfa_offset 24 popq %r15 .cfi_def_cfa_offset 16 popq %rbp .cfi_def_cfa_offset 8 retq .Lfunc_end7: .size main, .Lfunc_end7-main .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_ctor .type __hip_module_ctor,@function __hip_module_ctor: # @__hip_module_ctor .cfi_startproc # %bb.0: subq $40, %rsp .cfi_def_cfa_offset 48 cmpq $0, __hip_gpubin_handle(%rip) jne .LBB8_2 # %bb.1: movl $__hip_fatbin_wrapper, %edi callq __hipRegisterFatBinary movq %rax, __hip_gpubin_handle(%rip) .LBB8_2: movq __hip_gpubin_handle(%rip), %rdi xorps %xmm0, %xmm0 movups %xmm0, 16(%rsp) movups %xmm0, (%rsp) movl $_Z7memcopyPdS_S_iii, %esi movl $.L__unnamed_1, %edx movl $.L__unnamed_1, %ecx movl $-1, %r8d xorl %r9d, %r9d callq __hipRegisterFunction movl $__hip_module_dtor, %edi addq $40, %rsp .cfi_def_cfa_offset 8 jmp atexit # TAILCALL .Lfunc_end8: .size __hip_module_ctor, .Lfunc_end8-__hip_module_ctor .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_dtor .type __hip_module_dtor,@function __hip_module_dtor: # @__hip_module_dtor .cfi_startproc # %bb.0: movq __hip_gpubin_handle(%rip), %rdi testq %rdi, %rdi je .LBB9_2 # %bb.1: pushq %rax .cfi_def_cfa_offset 16 callq __hipUnregisterFatBinary movq $0, __hip_gpubin_handle(%rip) addq $8, %rsp .cfi_def_cfa_offset 8 .LBB9_2: retq .Lfunc_end9: .size __hip_module_dtor, .Lfunc_end9-__hip_module_dtor .cfi_endproc # -- End function .type timer_start,@object # @timer_start .bss .globl timer_start .p2align 3, 0x0 timer_start: .quad 0x0000000000000000 # double 0 .size timer_start, 8 .type .L.str,@object # @.str .section .rodata.str1.1,"aMS",@progbits,1 .L.str: .asciz "%s: %8.3f seconds" .size .L.str, 18 .type .L.str.1,@object # @.str.1 .L.str.1: .asciz ", %8.3f GB/s" .size .L.str.1, 13 .type .L.str.3,@object # @.str.3 .L.str.3: .asciz "PERF_memcpy.dat" .size .L.str.3, 16 .type .L.str.4,@object # @.str.4 .L.str.4: .asciz "a" .size .L.str.4, 2 .type .L.str.5,@object # @.str.5 .L.str.5: .asciz "nx=%d ny=%d nz=%d GBs=%1.4f time_s=%1.4f \n" .size .L.str.5, 44 .type .L.str.6,@object # @.str.6 .L.str.6: .asciz "ERROR launching GPU C-CUDA program: %s\n" .size .L.str.6, 40 .type _Z7memcopyPdS_S_iii,@object # @_Z7memcopyPdS_S_iii .section .rodata,"a",@progbits .globl _Z7memcopyPdS_S_iii .p2align 3, 0x0 _Z7memcopyPdS_S_iii: .quad _Z22__device_stub__memcopyPdS_S_iii .size _Z7memcopyPdS_S_iii, 8 .type .L.str.7,@object # @.str.7 .section .rodata.str1.1,"aMS",@progbits,1 .L.str.7: .asciz "Process uses GPU with id %d.\n" .size .L.str.7, 30 .type .L.str.8,@object # @.str.8 .L.str.8: .asciz "%dx%dx%d, %1.3f GB, %d iterations.\n" .size .L.str.8, 36 .type .L.str.9,@object # @.str.9 .L.str.9: .asciz "launching (%dx%dx%d) grid of (%dx%dx%d) blocks.\n" .size .L.str.9, 49 .type .L.str.10,@object # @.str.10 .L.str.10: .asciz "Performance" .size .L.str.10, 12 .type .L__unnamed_1,@object # @0 .L__unnamed_1: .asciz "_Z7memcopyPdS_S_iii" .size .L__unnamed_1, 20 .type __hip_fatbin_wrapper,@object # @__hip_fatbin_wrapper .section .hipFatBinSegment,"a",@progbits .p2align 3, 0x0 __hip_fatbin_wrapper: .long 1212764230 # 0x48495046 .long 1 # 0x1 .quad __hip_fatbin .quad 0 .size __hip_fatbin_wrapper, 24 .type __hip_gpubin_handle,@object # @__hip_gpubin_handle .local __hip_gpubin_handle .comm __hip_gpubin_handle,8,8 .section .init_array,"aw",@init_array .p2align 3, 0x0 .quad __hip_module_ctor .type __hip_cuid_,@object # @__hip_cuid_ .bss .globl __hip_cuid_ __hip_cuid_: .byte 0 # 0x0 .size __hip_cuid_, 1 .section ".linker-options","e",@llvm_linker_options .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym _Z22__device_stub__memcopyPdS_S_iii .addrsig_sym __hip_module_ctor .addrsig_sym __hip_module_dtor .addrsig_sym _Z7memcopyPdS_S_iii .addrsig_sym __hip_fatbin .addrsig_sym __hip_fatbin_wrapper .addrsig_sym __hip_cuid_
You are a helpful coding assistant specialized in converting CUDA source code to HIP/AMD source code.
/* Daniel Willen, 2019 * * Solve the transient heat conduction problem with homogeneous Dirichlet * boundary conditions: * * u(x={0,L}) = u(y={0,L}) = 0 * * and initial condition: * * u(x,y,0) = sin(x) * sin(y) * * on the domain 0 <= x,y <= L, with L = pi. * * This program solves the above problem on a single GPU with the Jacobi method. * */ #include <stdlib.h> #include <stdio.h> #include <string.h> #include <math.h> #include <cuda.h> #include <thrust/device_ptr.h> #include <thrust/reduce.h> #define PI 3.14159265358979323846 #define MAX_THREADS_DIM 16 // Note that this depends on the hardware /* Note on the structure of this file: * - Cuda device constant memory declarations are at the top * - Functions definitions are in the middle. Functions include: * - - parse_cmdline: Read command-line arguments for domain size * - - jacobi_solver: Advance the soln to the next time step using Jacobi * - - check_error: Calculate the error b/t the numeric and analytic solns * - The `main' function is at the bottom * * Note that it is good practice to use header files and break functions out * into separate files. This has not been done here for simplicity. */ /*** Auxiliary Functions ***/ /* Read the command line inputs */ // - argv[0] is the program name // - argv[1] is the first input (number of points) int parse_cmdline(int argc, char *argv[]) { int nx; if (argc == 2) { nx = atoi(argv[1]); // Number of grid points if (nx < MAX_THREADS_DIM) { printf("Expecting a number of grid cells in one dimension to be at least %d\n", MAX_THREADS_DIM); exit(EXIT_FAILURE); } printf("Grid is %d by %d\n\n", nx, nx); } else { printf("Input error. Run like: \n\n"); printf(" $ ./parallel.c n\n\n"); printf(" where n is the number of grid cells in one dimension\n"); exit(EXIT_FAILURE); } return nx; } /******************************************************************************* * Step IV: Launch the GPU kernel to advance to the next time step with the * * Jacobi method here. * ******************************************************************************/ __global__ void computeNextJacobiStep(int nx, int ny, double pref, double* _u, double* _u_new) { int ti = blockDim.x * blockIdx.x + threadIdx.x; int tj = blockDim.y * blockIdx.y + threadIdx.y; if (ti < (nx-1) && ti > 0 && tj < (ny-1) && tj > 0) { double leftTerm = _u[tj*nx + ti]; double rightTerm = pref * ( _u[tj*nx + (ti+1)] + _u[tj*nx + (ti-1)] + _u[(tj+1)*nx + ti] + _u[(tj-1)*nx + ti] - 4*_u[tj*nx + ti] ); _u_new[tj*nx + ti] = leftTerm + rightTerm; } } /****************************************************************************** * Step V: Launch the GPU kernel to calculate the error at each grid point * * here. * *****************************************************************************/ __global__ void computeJacobiError(int nx, int ny, double dx, double dy, double D, double t, double* _u, double* _error) { int ti = blockDim.x * blockIdx.x + threadIdx.x; int tj = blockDim.y * blockIdx.y + threadIdx.y; if (ti < (nx-1) && ti > 0 && tj < (ny-1) && tj > 0) { double discretizedValue = _u[tj*nx + ti]; double analyticalValue = sin(dx * ti)*sin(dy * tj)*exp(-2*D*t); _error[tj*nx + ti] = abs(discretizedValue - analyticalValue); } } /*** Main Function ***/ int main(int argc, char *argv[]) { /* Variable declaration */ double Lx = PI; // Domain length in x-direction double Ly = PI; // Domain length in y-direction double D = 1.; // Diffusion constant int nx, ny; // Grid points (grid cells + 1) double dx, dy; // Grid spacing double dt; // Time step size double sim_time; // Length of sim time, arbitrary for simplicity double pref; // Pre-factor in the Jacobi method double error = 0.; // Mean percent-difference at each grid point error = error; // To prevent compiler warning /* Parse command-line for problem size */ nx = parse_cmdline(argc, argv); ny = nx; // Assume a square grid /* Initialize variables */ dx = Lx / (nx - 1); // Cell width in x-direction dy = Ly / (ny - 1); // Cell width in y-direction dt = 0.25*dx*dy/D; // Limited by diffusive stability sim_time = 0.5*Lx*Ly/D; // Arbitrary simulation length pref = D*dt/(dx*dx); // Jacobi pre-factor /***************************************************************************** * Step I: Declare, allocate, and initialize memory for the field variable * * u on the CPU. * ****************************************************************************/ double* u = (double*) malloc(nx*ny * sizeof(double)); for (int j = 0; j < ny; ++j) { for (int i = 0; i < nx; ++i) { u[j*nx + i] = sin(i * dx) * sin(j * dy); } } /***************************************************************************** * Step II: Declare and allocate GPU memory for _u, _u_new, and _error. Copy * * the initial condition to the GPU. * ****************************************************************************/ double *_u, *_u_new, *_error; cudaMalloc(&_u, nx*ny * sizeof(double)); cudaMemcpy(_u, u, nx*ny * sizeof(double), cudaMemcpyHostToDevice); cudaMalloc(&_u_new, nx*ny * sizeof(double)); cudaMalloc(&_error, nx*ny * sizeof(double)); // Set the new soln and error to 0 cudaMemset(_u_new, 0., nx*ny * sizeof(double)); cudaMemset(_error, 0., nx*ny * sizeof(double)); // Create thrust pointers to device memory for error calculation thrust::device_ptr<double> t_error(_error); /***************************************************************************** * Step III: Set up the kernel execution configuration for the domain based * * on the input domain size and the MAX_THREADS_DIM variable. * ****************************************************************************/ int tx = MAX_THREADS_DIM; int ty = MAX_THREADS_DIM; int bx = (int) ceil((double) nx / tx); int by = (int) ceil((double) ny / ty); dim3 dimBlocks(tx, ty); dim3 numBlocks(bx, by); /***************************/ /* Main Time-Stepping Loop */ /***************************/ for (double time = 0.; time <= sim_time; time += dt) { /*************************************************************************** * Step IV: Launch the GPU kernel to advance to the next time step with * * the Jacobi method here. * **************************************************************************/ computeNextJacobiStep<<<numBlocks, dimBlocks>>>(nx, ny, pref, _u, _u_new); cudaDeviceSynchronize(); /*************************************************************************** * Step V: Launch the GPU kernel to calculate the error at each grid point * * here. * **************************************************************************/ computeJacobiError<<<numBlocks, dimBlocks>>>(nx, ny, dx, dy, D, time, _u, _error); cudaDeviceSynchronize(); // Use thrust to do a parallel reduction on the error error = thrust::reduce(t_error, t_error + nx*ny, 0., thrust::plus<double>()); printf("Error at t* = %.5lf is %e\n", time*D/(Lx*Lx), error/(nx*ny)); // Copy new soln to old. This also blocks to ensure computations are finished. cudaMemcpy(_u, _u_new, nx*ny * sizeof(double), cudaMemcpyDeviceToDevice); } /***************************************************************************** * Step VI: Copy the memory back to the CPU. * ****************************************************************************/ cudaMemcpy(u, _u, nx*ny * sizeof(double), cudaMemcpyDeviceToHost); /***************************************************************************** * Step I and Step II: Free the memory that you declared and allocated * * earlier in the program. * ****************************************************************************/ free(u); cudaFree(_u); cudaFree(_u_new); cudaFree(_error); return EXIT_SUCCESS; }
/* Daniel Willen, 2019 * * Solve the transient heat conduction problem with homogeneous Dirichlet * boundary conditions: * * u(x={0,L}) = u(y={0,L}) = 0 * * and initial condition: * * u(x,y,0) = sin(x) * sin(y) * * on the domain 0 <= x,y <= L, with L = pi. * * This program solves the above problem on a single GPU with the Jacobi method. * */ #include <stdlib.h> #include <stdio.h> #include <string.h> #include <math.h> #include <hip/hip_runtime.h> #include <thrust/device_ptr.h> #include <thrust/reduce.h> #define PI 3.14159265358979323846 #define MAX_THREADS_DIM 16 // Note that this depends on the hardware /* Note on the structure of this file: * - Cuda device constant memory declarations are at the top * - Functions definitions are in the middle. Functions include: * - - parse_cmdline: Read command-line arguments for domain size * - - jacobi_solver: Advance the soln to the next time step using Jacobi * - - check_error: Calculate the error b/t the numeric and analytic solns * - The `main' function is at the bottom * * Note that it is good practice to use header files and break functions out * into separate files. This has not been done here for simplicity. */ /*** Auxiliary Functions ***/ /* Read the command line inputs */ // - argv[0] is the program name // - argv[1] is the first input (number of points) int parse_cmdline(int argc, char *argv[]) { int nx; if (argc == 2) { nx = atoi(argv[1]); // Number of grid points if (nx < MAX_THREADS_DIM) { printf("Expecting a number of grid cells in one dimension to be at least %d\n", MAX_THREADS_DIM); exit(EXIT_FAILURE); } printf("Grid is %d by %d\n\n", nx, nx); } else { printf("Input error. Run like: \n\n"); printf(" $ ./parallel.c n\n\n"); printf(" where n is the number of grid cells in one dimension\n"); exit(EXIT_FAILURE); } return nx; } /******************************************************************************* * Step IV: Launch the GPU kernel to advance to the next time step with the * * Jacobi method here. * ******************************************************************************/ __global__ void computeNextJacobiStep(int nx, int ny, double pref, double* _u, double* _u_new) { int ti = blockDim.x * blockIdx.x + threadIdx.x; int tj = blockDim.y * blockIdx.y + threadIdx.y; if (ti < (nx-1) && ti > 0 && tj < (ny-1) && tj > 0) { double leftTerm = _u[tj*nx + ti]; double rightTerm = pref * ( _u[tj*nx + (ti+1)] + _u[tj*nx + (ti-1)] + _u[(tj+1)*nx + ti] + _u[(tj-1)*nx + ti] - 4*_u[tj*nx + ti] ); _u_new[tj*nx + ti] = leftTerm + rightTerm; } } /****************************************************************************** * Step V: Launch the GPU kernel to calculate the error at each grid point * * here. * *****************************************************************************/ __global__ void computeJacobiError(int nx, int ny, double dx, double dy, double D, double t, double* _u, double* _error) { int ti = blockDim.x * blockIdx.x + threadIdx.x; int tj = blockDim.y * blockIdx.y + threadIdx.y; if (ti < (nx-1) && ti > 0 && tj < (ny-1) && tj > 0) { double discretizedValue = _u[tj*nx + ti]; double analyticalValue = sin(dx * ti)*sin(dy * tj)*exp(-2*D*t); _error[tj*nx + ti] = abs(discretizedValue - analyticalValue); } } /*** Main Function ***/ int main(int argc, char *argv[]) { /* Variable declaration */ double Lx = PI; // Domain length in x-direction double Ly = PI; // Domain length in y-direction double D = 1.; // Diffusion constant int nx, ny; // Grid points (grid cells + 1) double dx, dy; // Grid spacing double dt; // Time step size double sim_time; // Length of sim time, arbitrary for simplicity double pref; // Pre-factor in the Jacobi method double error = 0.; // Mean percent-difference at each grid point error = error; // To prevent compiler warning /* Parse command-line for problem size */ nx = parse_cmdline(argc, argv); ny = nx; // Assume a square grid /* Initialize variables */ dx = Lx / (nx - 1); // Cell width in x-direction dy = Ly / (ny - 1); // Cell width in y-direction dt = 0.25*dx*dy/D; // Limited by diffusive stability sim_time = 0.5*Lx*Ly/D; // Arbitrary simulation length pref = D*dt/(dx*dx); // Jacobi pre-factor /***************************************************************************** * Step I: Declare, allocate, and initialize memory for the field variable * * u on the CPU. * ****************************************************************************/ double* u = (double*) malloc(nx*ny * sizeof(double)); for (int j = 0; j < ny; ++j) { for (int i = 0; i < nx; ++i) { u[j*nx + i] = sin(i * dx) * sin(j * dy); } } /***************************************************************************** * Step II: Declare and allocate GPU memory for _u, _u_new, and _error. Copy * * the initial condition to the GPU. * ****************************************************************************/ double *_u, *_u_new, *_error; hipMalloc(&_u, nx*ny * sizeof(double)); hipMemcpy(_u, u, nx*ny * sizeof(double), hipMemcpyHostToDevice); hipMalloc(&_u_new, nx*ny * sizeof(double)); hipMalloc(&_error, nx*ny * sizeof(double)); // Set the new soln and error to 0 hipMemset(_u_new, 0., nx*ny * sizeof(double)); hipMemset(_error, 0., nx*ny * sizeof(double)); // Create thrust pointers to device memory for error calculation thrust::device_ptr<double> t_error(_error); /***************************************************************************** * Step III: Set up the kernel execution configuration for the domain based * * on the input domain size and the MAX_THREADS_DIM variable. * ****************************************************************************/ int tx = MAX_THREADS_DIM; int ty = MAX_THREADS_DIM; int bx = (int) ceil((double) nx / tx); int by = (int) ceil((double) ny / ty); dim3 dimBlocks(tx, ty); dim3 numBlocks(bx, by); /***************************/ /* Main Time-Stepping Loop */ /***************************/ for (double time = 0.; time <= sim_time; time += dt) { /*************************************************************************** * Step IV: Launch the GPU kernel to advance to the next time step with * * the Jacobi method here. * **************************************************************************/ computeNextJacobiStep<<<numBlocks, dimBlocks>>>(nx, ny, pref, _u, _u_new); hipDeviceSynchronize(); /*************************************************************************** * Step V: Launch the GPU kernel to calculate the error at each grid point * * here. * **************************************************************************/ computeJacobiError<<<numBlocks, dimBlocks>>>(nx, ny, dx, dy, D, time, _u, _error); hipDeviceSynchronize(); // Use thrust to do a parallel reduction on the error error = thrust::reduce(t_error, t_error + nx*ny, 0., thrust::plus<double>()); printf("Error at t* = %.5lf is %e\n", time*D/(Lx*Lx), error/(nx*ny)); // Copy new soln to old. This also blocks to ensure computations are finished. hipMemcpy(_u, _u_new, nx*ny * sizeof(double), hipMemcpyDeviceToDevice); } /***************************************************************************** * Step VI: Copy the memory back to the CPU. * ****************************************************************************/ hipMemcpy(u, _u, nx*ny * sizeof(double), hipMemcpyDeviceToHost); /***************************************************************************** * Step I and Step II: Free the memory that you declared and allocated * * earlier in the program. * ****************************************************************************/ free(u); hipFree(_u); hipFree(_u_new); hipFree(_error); return EXIT_SUCCESS; }
You are a helpful coding assistant specialized in converting CUDA source code to CUDA device assembly.
// cudaDCA.cu // //This file contains the recursive DCA function, and the function that is used to invoke DCA and //interperate the results. //Included Files #include <iostream> //Function Prototypes // Functions found in this file void RecDCA(double Zs[], int n, int i, double AF[], int cut_off,double Xs[]); // Functions found in Init_setup.cu void CudaInitialize(double m[], double l[], double I[], double x[], int n, double Zs[]); // Functions found in Assemble_setup.cu void cudaAssemble(double Zs[],double Xs[], int num, double nZs[], double nXs[], int odd, int newlen, int data); // Functions found in Disassemble_setup.cu void cudaDisassemble(double OldAF[], double Zs[], double Xs[],double nZs[], double nXs[], int odd, int morelen, int lesslen, double AF[], int data); // Functions found in Assemble.cu void Assemble(double Zs[], double Xs[],double nZs[], double nXs[], int len, int odd, int n); // Functions found in Disassemble.cu void Disassemble(double lessZs[], double lessXs[],double moreZs[], double moreXs[], double oldAs[] ,double newAs[], int num, int odd); // Functions found in SolveBCs.cu void solve_BCs(double Zs[], double Xs[], double AF[]); void printa(double A[], int n); void printm(double A[6][6]); void Update_Properties(double bodyZetas[],double nZetas[], int n, double state[], double m[], double l[], double II[]); //DCAhelp: // Function that prepares the list of bodies for DCA and finds the final state vector // state is the state of the system at that timestep // bs is a list of bodies used for initialization // js is a list of joints // n is the number of bodies // Y is the array where the final velocities and accelerations are stored void DCAhelp(double state[], double m[], double l[], double I[],int n, double Y[],int cut_off, double bZs[]) { //Create the list that will hold all acceleration and force values for all bodies double *AF = (double*) malloc(sizeof(double)*n*4*6); //double A[6][n*2]; //Create the matrix where only the accelerations will be stored double *Zs=(double*)malloc(sizeof(double)*n*6*26); double *Xs=(double*)malloc(sizeof(double)*n*5*5); Update_Properties(bZs,Zs,n,state,m,l,I); for(int r =0; r<6; r++) //CudaInitialize(m,l,I, state, n, Zs); //Initialize the bodies, finding all zeta values //Pass the list of bodies to DCA and return the accelerations //and forces of both handles of every body in the list //RecDCA(Zs, n, 0, AF, cut_off,Xs); Y[n]=AF[8*n]; //For a pendulum, the fist acceleration value is in A[2][0] for(int i = n+1, j=2; i<n*2; i++, j+=2) //Loop through the acceleration matrix { Y[i]=AF[2*4*n+2*j]-AF[2*4*n+2*(j-1)]; //Find and save all generalized accelerations } for(int i = 0; i<n; i++) //Loop through the state vector { Y[i]=state[i+n]; //Save the velocities } //Free memory free(AF); free(Zs); free(Xs); } //RecDCA: // Function used to solve for the velocty and acceleration of the list of bodies at // the current timestep. This is a recursive function that continues to call itself // until there is a single body left. Once at this point the accelerations and forces // are found using the boundary conditions of a pendulum. These values are then returned // to the previous level of recursion which then finds the new accelerations and forces // for the disassembled bodies. This continues until all bodies are disassembled, ultimately // returning the forces and accelerations at both handles of every body in the system. These // results are intererated by DCAhelp (above) to obtain the actual generalized accelerations. // bodies is the list of bodies // n is the number of bodies // i is the level of recursion // AF is the array in which the accelerations and forces at the handles of the bodies // will be stored. void RecDCA(double Zs[], int n, int i, double AF[], int cut_off,double Xs[],int gpu, int data) { if (n==1) //If there is only 1 body { } else //If there is more than 1 body { int newlen; //New number of bodies after assembly int odd = 0; //Flag to keep track of the parity of the length of the list of if(n % 2 == 0) //If there is an even number of bodies { newlen = (int) (n/2); //The new number of bodies will be half the original number } else //If there is an odd number of bodies { newlen = (int)((n+1)/2); //The new number of bodies will be half the original number //rounded down, plus 1 odd = 1; //odd is set to 1 because there are an odd number of bodies } double *nZs=(double*)malloc(sizeof(double)*newlen*26*6); double *nXs=(double*) malloc(sizeof(double)*(newlen)*5*5); double *AFo=(double*)malloc(sizeof(double)*6*newlen*4); //Call the DCA function again to return the accelerations and forces of the new bodies RecDCA(nZs,newlen,i+1 ,AF,cut_off,nXs,gpu, data); if(gpu) { cudaDisassemble(AFo, Zs,Xs , nZs,nXs, odd, n, newlen, AF,data); } else { Disassemble(nZs,nXs,Zs,Xs,AFo, AF, newlen,odd); } //Free memory free(nZs); free(nXs); } }
code for sm_80
You are a helpful coding assistant specialized in converting CUDA source code to CUDA host assembly.
// cudaDCA.cu // //This file contains the recursive DCA function, and the function that is used to invoke DCA and //interperate the results. //Included Files #include <iostream> //Function Prototypes // Functions found in this file void RecDCA(double Zs[], int n, int i, double AF[], int cut_off,double Xs[]); // Functions found in Init_setup.cu void CudaInitialize(double m[], double l[], double I[], double x[], int n, double Zs[]); // Functions found in Assemble_setup.cu void cudaAssemble(double Zs[],double Xs[], int num, double nZs[], double nXs[], int odd, int newlen, int data); // Functions found in Disassemble_setup.cu void cudaDisassemble(double OldAF[], double Zs[], double Xs[],double nZs[], double nXs[], int odd, int morelen, int lesslen, double AF[], int data); // Functions found in Assemble.cu void Assemble(double Zs[], double Xs[],double nZs[], double nXs[], int len, int odd, int n); // Functions found in Disassemble.cu void Disassemble(double lessZs[], double lessXs[],double moreZs[], double moreXs[], double oldAs[] ,double newAs[], int num, int odd); // Functions found in SolveBCs.cu void solve_BCs(double Zs[], double Xs[], double AF[]); void printa(double A[], int n); void printm(double A[6][6]); void Update_Properties(double bodyZetas[],double nZetas[], int n, double state[], double m[], double l[], double II[]); //DCAhelp: // Function that prepares the list of bodies for DCA and finds the final state vector // state is the state of the system at that timestep // bs is a list of bodies used for initialization // js is a list of joints // n is the number of bodies // Y is the array where the final velocities and accelerations are stored void DCAhelp(double state[], double m[], double l[], double I[],int n, double Y[],int cut_off, double bZs[]) { //Create the list that will hold all acceleration and force values for all bodies double *AF = (double*) malloc(sizeof(double)*n*4*6); //double A[6][n*2]; //Create the matrix where only the accelerations will be stored double *Zs=(double*)malloc(sizeof(double)*n*6*26); double *Xs=(double*)malloc(sizeof(double)*n*5*5); Update_Properties(bZs,Zs,n,state,m,l,I); for(int r =0; r<6; r++) //CudaInitialize(m,l,I, state, n, Zs); //Initialize the bodies, finding all zeta values //Pass the list of bodies to DCA and return the accelerations //and forces of both handles of every body in the list //RecDCA(Zs, n, 0, AF, cut_off,Xs); Y[n]=AF[8*n]; //For a pendulum, the fist acceleration value is in A[2][0] for(int i = n+1, j=2; i<n*2; i++, j+=2) //Loop through the acceleration matrix { Y[i]=AF[2*4*n+2*j]-AF[2*4*n+2*(j-1)]; //Find and save all generalized accelerations } for(int i = 0; i<n; i++) //Loop through the state vector { Y[i]=state[i+n]; //Save the velocities } //Free memory free(AF); free(Zs); free(Xs); } //RecDCA: // Function used to solve for the velocty and acceleration of the list of bodies at // the current timestep. This is a recursive function that continues to call itself // until there is a single body left. Once at this point the accelerations and forces // are found using the boundary conditions of a pendulum. These values are then returned // to the previous level of recursion which then finds the new accelerations and forces // for the disassembled bodies. This continues until all bodies are disassembled, ultimately // returning the forces and accelerations at both handles of every body in the system. These // results are intererated by DCAhelp (above) to obtain the actual generalized accelerations. // bodies is the list of bodies // n is the number of bodies // i is the level of recursion // AF is the array in which the accelerations and forces at the handles of the bodies // will be stored. void RecDCA(double Zs[], int n, int i, double AF[], int cut_off,double Xs[],int gpu, int data) { if (n==1) //If there is only 1 body { } else //If there is more than 1 body { int newlen; //New number of bodies after assembly int odd = 0; //Flag to keep track of the parity of the length of the list of if(n % 2 == 0) //If there is an even number of bodies { newlen = (int) (n/2); //The new number of bodies will be half the original number } else //If there is an odd number of bodies { newlen = (int)((n+1)/2); //The new number of bodies will be half the original number //rounded down, plus 1 odd = 1; //odd is set to 1 because there are an odd number of bodies } double *nZs=(double*)malloc(sizeof(double)*newlen*26*6); double *nXs=(double*) malloc(sizeof(double)*(newlen)*5*5); double *AFo=(double*)malloc(sizeof(double)*6*newlen*4); //Call the DCA function again to return the accelerations and forces of the new bodies RecDCA(nZs,newlen,i+1 ,AF,cut_off,nXs,gpu, data); if(gpu) { cudaDisassemble(AFo, Zs,Xs , nZs,nXs, odd, n, newlen, AF,data); } else { Disassemble(nZs,nXs,Zs,Xs,AFo, AF, newlen,odd); } //Free memory free(nZs); free(nXs); } }
.file "tmpxft_00064b10_00000000-6_cudaDCA.cudafe1.cpp" .text #APP .globl _ZSt21ios_base_library_initv #NO_APP .type _ZL26__cudaUnregisterBinaryUtilv, @function _ZL26__cudaUnregisterBinaryUtilv: .LFB3673: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaUnregisterFatBinary@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE3673: .size _ZL26__cudaUnregisterBinaryUtilv, .-_ZL26__cudaUnregisterBinaryUtilv .globl _Z7DCAhelpPdS_S_S_iS_iS_ .type _Z7DCAhelpPdS_S_S_iS_iS_, @function _Z7DCAhelpPdS_S_S_iS_iS_: .LFB3669: .cfi_startproc endbr64 pushq %r15 .cfi_def_cfa_offset 16 .cfi_offset 15, -16 pushq %r14 .cfi_def_cfa_offset 24 .cfi_offset 14, -24 pushq %r13 .cfi_def_cfa_offset 32 .cfi_offset 13, -32 pushq %r12 .cfi_def_cfa_offset 40 .cfi_offset 12, -40 pushq %rbp .cfi_def_cfa_offset 48 .cfi_offset 6, -48 pushq %rbx .cfi_def_cfa_offset 56 .cfi_offset 3, -56 subq $40, %rsp .cfi_def_cfa_offset 96 movq %rdi, %rbx movq %rsi, 8(%rsp) movq %rdx, 16(%rsp) movq %rcx, 24(%rsp) movl %r8d, %r12d movq %r9, %rbp movslq %r8d, %r14 leaq (%r14,%r14,2), %rdi salq $6, %rdi call malloc@PLT movq %rax, %r13 imulq $1248, %r14, %rdi call malloc@PLT movq %rax, %r15 subq $8, %rsp .cfi_def_cfa_offset 104 pushq 32(%rsp) .cfi_def_cfa_offset 112 movq 32(%rsp), %r9 movq 24(%rsp), %r8 movq %rbx, %rcx movl %r12d, %edx movq %rax, %rsi movq 120(%rsp), %rdi call _Z17Update_PropertiesPdS_iS_S_S_S_@PLT leal 0(,%r12,8), %ecx movslq %ecx, %rcx movsd 0(%r13,%rcx,8), %xmm0 movsd %xmm0, 0(%rbp,%rcx) leal 1(%r12), %edx leal (%r12,%r12), %eax addq $16, %rsp .cfi_def_cfa_offset 96 cmpl %eax, %edx jge .L4 leal 4(,%r12,8), %eax cltq leaq 0(%r13,%rax,8), %rdx leaq 8(%rbp,%rcx), %rax leal -2(%r12), %esi addq %r14, %rsi leaq 16(%rbp,%rsi,8), %rsi .L5: movsd (%rdx), %xmm0 subsd -16(%rdx), %xmm0 movsd %xmm0, (%rax) addq $32, %rdx addq $8, %rax cmpq %rsi, %rax jne .L5 .L4: testl %r12d, %r12d jle .L6 addq %rcx, %rbx movl $0, %eax .L7: movsd (%rbx,%rax), %xmm0 movsd %xmm0, 0(%rbp,%rax) addq $8, %rax cmpq %rax, %rcx jne .L7 .L6: movq %r13, %rdi call free@PLT movq %r15, %rdi call free@PLT addq $40, %rsp .cfi_def_cfa_offset 56 popq %rbx .cfi_def_cfa_offset 48 popq %rbp .cfi_def_cfa_offset 40 popq %r12 .cfi_def_cfa_offset 32 popq %r13 .cfi_def_cfa_offset 24 popq %r14 .cfi_def_cfa_offset 16 popq %r15 .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE3669: .size _Z7DCAhelpPdS_S_S_iS_iS_, .-_Z7DCAhelpPdS_S_S_iS_iS_ .globl _Z6RecDCAPdiiS_iS_ii .type _Z6RecDCAPdiiS_iS_ii, @function _Z6RecDCAPdiiS_iS_ii: .LFB3670: .cfi_startproc endbr64 pushq %r15 .cfi_def_cfa_offset 16 .cfi_offset 15, -16 pushq %r14 .cfi_def_cfa_offset 24 .cfi_offset 14, -24 pushq %r13 .cfi_def_cfa_offset 32 .cfi_offset 13, -32 pushq %r12 .cfi_def_cfa_offset 40 .cfi_offset 12, -40 pushq %rbp .cfi_def_cfa_offset 48 .cfi_offset 6, -48 pushq %rbx .cfi_def_cfa_offset 56 .cfi_offset 3, -56 subq $40, %rsp .cfi_def_cfa_offset 96 movq %rdi, (%rsp) movl %edx, 12(%rsp) movl %r8d, 24(%rsp) movq %r9, 16(%rsp) cmpl $1, %esi je .L11 movl %esi, %ebx movq %rcx, %r12 movl %esi, %eax andl $1, %eax movl %eax, 28(%rsp) jne .L13 movl %esi, %ebp shrl $31, %ebp addl %esi, %ebp sarl %ebp .L14: movslq %ebp, %r14 imulq $1248, %r14, %rdi call malloc@PLT movq %rax, %r13 leaq (%r14,%r14,4), %rax leaq (%rax,%rax,4), %rdi salq $3, %rdi call malloc@PLT movq %rax, %r15 leaq (%r14,%r14,2), %rdi salq $6, %rdi call malloc@PLT movq %rax, %r14 movl 12(%rsp), %edx addl $1, %edx movl 104(%rsp), %eax pushq %rax .cfi_def_cfa_offset 104 movl 104(%rsp), %eax pushq %rax .cfi_def_cfa_offset 112 movq %r15, %r9 movl 40(%rsp), %r8d movq %r12, %rcx movl %ebp, %esi movq %r13, %rdi call _Z6RecDCAPdiiS_iS_ii addq $16, %rsp .cfi_def_cfa_offset 96 cmpl $0, 96(%rsp) je .L15 movl 104(%rsp), %eax pushq %rax .cfi_def_cfa_offset 104 pushq %r12 .cfi_def_cfa_offset 112 pushq %rbp .cfi_def_cfa_offset 120 pushq %rbx .cfi_def_cfa_offset 128 movl 60(%rsp), %r9d movq %r15, %r8 movq %r13, %rcx movq 48(%rsp), %rdx movq 32(%rsp), %rsi movq %r14, %rdi call _Z15cudaDisassemblePdS_S_S_S_iiiS_i@PLT addq $32, %rsp .cfi_def_cfa_offset 96 .L16: movq %r13, %rdi call free@PLT movq %r15, %rdi call free@PLT .L11: addq $40, %rsp .cfi_remember_state .cfi_def_cfa_offset 56 popq %rbx .cfi_def_cfa_offset 48 popq %rbp .cfi_def_cfa_offset 40 popq %r12 .cfi_def_cfa_offset 32 popq %r13 .cfi_def_cfa_offset 24 popq %r14 .cfi_def_cfa_offset 16 popq %r15 .cfi_def_cfa_offset 8 ret .L13: .cfi_restore_state leal 1(%rsi), %eax movl %eax, %ebp shrl $31, %ebp addl %eax, %ebp sarl %ebp jmp .L14 .L15: movl 28(%rsp), %eax pushq %rax .cfi_def_cfa_offset 104 pushq %rbp .cfi_def_cfa_offset 112 movq %r12, %r9 movq %r14, %r8 movq 32(%rsp), %rcx movq 16(%rsp), %rdx movq %r15, %rsi movq %r13, %rdi call _Z11DisassemblePdS_S_S_S_S_ii@PLT addq $16, %rsp .cfi_def_cfa_offset 96 jmp .L16 .cfi_endproc .LFE3670: .size _Z6RecDCAPdiiS_iS_ii, .-_Z6RecDCAPdiiS_iS_ii .type _ZL24__sti____cudaRegisterAllv, @function _ZL24__sti____cudaRegisterAllv: .LFB3696: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 leaq _ZL15__fatDeviceText(%rip), %rdi call __cudaRegisterFatBinary@PLT movq %rax, %rdi movq %rax, _ZL20__cudaFatCubinHandle(%rip) call __cudaRegisterFatBinaryEnd@PLT leaq _ZL26__cudaUnregisterBinaryUtilv(%rip), %rdi call atexit@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE3696: .size _ZL24__sti____cudaRegisterAllv, .-_ZL24__sti____cudaRegisterAllv .section .init_array,"aw" .align 8 .quad _ZL24__sti____cudaRegisterAllv .section .nvFatBinSegment,"aw" .align 8 .type _ZL15__fatDeviceText, @object .size _ZL15__fatDeviceText, 24 _ZL15__fatDeviceText: .long 1180844977 .long 1 .quad fatbinData .quad 0 .local _ZL20__cudaFatCubinHandle .comm _ZL20__cudaFatCubinHandle,8,8 .ident "GCC: (Ubuntu 13.3.0-6ubuntu2~24.04) 13.3.0" .section .note.GNU-stack,"",@progbits .section .note.gnu.property,"a" .align 8 .long 1f - 0f .long 4f - 1f .long 5 0: .string "GNU" 1: .align 8 .long 0xc0000002 .long 3f - 2f 2: .long 0x3 3: .align 8 4:
You are a helpful coding assistant specialized in converting CUDA source code to HIP/AMD source code.
// cudaDCA.cu // //This file contains the recursive DCA function, and the function that is used to invoke DCA and //interperate the results. //Included Files #include <iostream> //Function Prototypes // Functions found in this file void RecDCA(double Zs[], int n, int i, double AF[], int cut_off,double Xs[]); // Functions found in Init_setup.cu void CudaInitialize(double m[], double l[], double I[], double x[], int n, double Zs[]); // Functions found in Assemble_setup.cu void cudaAssemble(double Zs[],double Xs[], int num, double nZs[], double nXs[], int odd, int newlen, int data); // Functions found in Disassemble_setup.cu void cudaDisassemble(double OldAF[], double Zs[], double Xs[],double nZs[], double nXs[], int odd, int morelen, int lesslen, double AF[], int data); // Functions found in Assemble.cu void Assemble(double Zs[], double Xs[],double nZs[], double nXs[], int len, int odd, int n); // Functions found in Disassemble.cu void Disassemble(double lessZs[], double lessXs[],double moreZs[], double moreXs[], double oldAs[] ,double newAs[], int num, int odd); // Functions found in SolveBCs.cu void solve_BCs(double Zs[], double Xs[], double AF[]); void printa(double A[], int n); void printm(double A[6][6]); void Update_Properties(double bodyZetas[],double nZetas[], int n, double state[], double m[], double l[], double II[]); //DCAhelp: // Function that prepares the list of bodies for DCA and finds the final state vector // state is the state of the system at that timestep // bs is a list of bodies used for initialization // js is a list of joints // n is the number of bodies // Y is the array where the final velocities and accelerations are stored void DCAhelp(double state[], double m[], double l[], double I[],int n, double Y[],int cut_off, double bZs[]) { //Create the list that will hold all acceleration and force values for all bodies double *AF = (double*) malloc(sizeof(double)*n*4*6); //double A[6][n*2]; //Create the matrix where only the accelerations will be stored double *Zs=(double*)malloc(sizeof(double)*n*6*26); double *Xs=(double*)malloc(sizeof(double)*n*5*5); Update_Properties(bZs,Zs,n,state,m,l,I); for(int r =0; r<6; r++) //CudaInitialize(m,l,I, state, n, Zs); //Initialize the bodies, finding all zeta values //Pass the list of bodies to DCA and return the accelerations //and forces of both handles of every body in the list //RecDCA(Zs, n, 0, AF, cut_off,Xs); Y[n]=AF[8*n]; //For a pendulum, the fist acceleration value is in A[2][0] for(int i = n+1, j=2; i<n*2; i++, j+=2) //Loop through the acceleration matrix { Y[i]=AF[2*4*n+2*j]-AF[2*4*n+2*(j-1)]; //Find and save all generalized accelerations } for(int i = 0; i<n; i++) //Loop through the state vector { Y[i]=state[i+n]; //Save the velocities } //Free memory free(AF); free(Zs); free(Xs); } //RecDCA: // Function used to solve for the velocty and acceleration of the list of bodies at // the current timestep. This is a recursive function that continues to call itself // until there is a single body left. Once at this point the accelerations and forces // are found using the boundary conditions of a pendulum. These values are then returned // to the previous level of recursion which then finds the new accelerations and forces // for the disassembled bodies. This continues until all bodies are disassembled, ultimately // returning the forces and accelerations at both handles of every body in the system. These // results are intererated by DCAhelp (above) to obtain the actual generalized accelerations. // bodies is the list of bodies // n is the number of bodies // i is the level of recursion // AF is the array in which the accelerations and forces at the handles of the bodies // will be stored. void RecDCA(double Zs[], int n, int i, double AF[], int cut_off,double Xs[],int gpu, int data) { if (n==1) //If there is only 1 body { } else //If there is more than 1 body { int newlen; //New number of bodies after assembly int odd = 0; //Flag to keep track of the parity of the length of the list of if(n % 2 == 0) //If there is an even number of bodies { newlen = (int) (n/2); //The new number of bodies will be half the original number } else //If there is an odd number of bodies { newlen = (int)((n+1)/2); //The new number of bodies will be half the original number //rounded down, plus 1 odd = 1; //odd is set to 1 because there are an odd number of bodies } double *nZs=(double*)malloc(sizeof(double)*newlen*26*6); double *nXs=(double*) malloc(sizeof(double)*(newlen)*5*5); double *AFo=(double*)malloc(sizeof(double)*6*newlen*4); //Call the DCA function again to return the accelerations and forces of the new bodies RecDCA(nZs,newlen,i+1 ,AF,cut_off,nXs,gpu, data); if(gpu) { cudaDisassemble(AFo, Zs,Xs , nZs,nXs, odd, n, newlen, AF,data); } else { Disassemble(nZs,nXs,Zs,Xs,AFo, AF, newlen,odd); } //Free memory free(nZs); free(nXs); } }
// cudaDCA.cu // //This file contains the recursive DCA function, and the function that is used to invoke DCA and //interperate the results. //Included Files #include <hip/hip_runtime.h> #include <iostream> //Function Prototypes // Functions found in this file void RecDCA(double Zs[], int n, int i, double AF[], int cut_off,double Xs[]); // Functions found in Init_setup.cu void CudaInitialize(double m[], double l[], double I[], double x[], int n, double Zs[]); // Functions found in Assemble_setup.cu void cudaAssemble(double Zs[],double Xs[], int num, double nZs[], double nXs[], int odd, int newlen, int data); // Functions found in Disassemble_setup.cu void cudaDisassemble(double OldAF[], double Zs[], double Xs[],double nZs[], double nXs[], int odd, int morelen, int lesslen, double AF[], int data); // Functions found in Assemble.cu void Assemble(double Zs[], double Xs[],double nZs[], double nXs[], int len, int odd, int n); // Functions found in Disassemble.cu void Disassemble(double lessZs[], double lessXs[],double moreZs[], double moreXs[], double oldAs[] ,double newAs[], int num, int odd); // Functions found in SolveBCs.cu void solve_BCs(double Zs[], double Xs[], double AF[]); void printa(double A[], int n); void printm(double A[6][6]); void Update_Properties(double bodyZetas[],double nZetas[], int n, double state[], double m[], double l[], double II[]); //DCAhelp: // Function that prepares the list of bodies for DCA and finds the final state vector // state is the state of the system at that timestep // bs is a list of bodies used for initialization // js is a list of joints // n is the number of bodies // Y is the array where the final velocities and accelerations are stored void DCAhelp(double state[], double m[], double l[], double I[],int n, double Y[],int cut_off, double bZs[]) { //Create the list that will hold all acceleration and force values for all bodies double *AF = (double*) malloc(sizeof(double)*n*4*6); //double A[6][n*2]; //Create the matrix where only the accelerations will be stored double *Zs=(double*)malloc(sizeof(double)*n*6*26); double *Xs=(double*)malloc(sizeof(double)*n*5*5); Update_Properties(bZs,Zs,n,state,m,l,I); for(int r =0; r<6; r++) //CudaInitialize(m,l,I, state, n, Zs); //Initialize the bodies, finding all zeta values //Pass the list of bodies to DCA and return the accelerations //and forces of both handles of every body in the list //RecDCA(Zs, n, 0, AF, cut_off,Xs); Y[n]=AF[8*n]; //For a pendulum, the fist acceleration value is in A[2][0] for(int i = n+1, j=2; i<n*2; i++, j+=2) //Loop through the acceleration matrix { Y[i]=AF[2*4*n+2*j]-AF[2*4*n+2*(j-1)]; //Find and save all generalized accelerations } for(int i = 0; i<n; i++) //Loop through the state vector { Y[i]=state[i+n]; //Save the velocities } //Free memory free(AF); free(Zs); free(Xs); } //RecDCA: // Function used to solve for the velocty and acceleration of the list of bodies at // the current timestep. This is a recursive function that continues to call itself // until there is a single body left. Once at this point the accelerations and forces // are found using the boundary conditions of a pendulum. These values are then returned // to the previous level of recursion which then finds the new accelerations and forces // for the disassembled bodies. This continues until all bodies are disassembled, ultimately // returning the forces and accelerations at both handles of every body in the system. These // results are intererated by DCAhelp (above) to obtain the actual generalized accelerations. // bodies is the list of bodies // n is the number of bodies // i is the level of recursion // AF is the array in which the accelerations and forces at the handles of the bodies // will be stored. void RecDCA(double Zs[], int n, int i, double AF[], int cut_off,double Xs[],int gpu, int data) { if (n==1) //If there is only 1 body { } else //If there is more than 1 body { int newlen; //New number of bodies after assembly int odd = 0; //Flag to keep track of the parity of the length of the list of if(n % 2 == 0) //If there is an even number of bodies { newlen = (int) (n/2); //The new number of bodies will be half the original number } else //If there is an odd number of bodies { newlen = (int)((n+1)/2); //The new number of bodies will be half the original number //rounded down, plus 1 odd = 1; //odd is set to 1 because there are an odd number of bodies } double *nZs=(double*)malloc(sizeof(double)*newlen*26*6); double *nXs=(double*) malloc(sizeof(double)*(newlen)*5*5); double *AFo=(double*)malloc(sizeof(double)*6*newlen*4); //Call the DCA function again to return the accelerations and forces of the new bodies RecDCA(nZs,newlen,i+1 ,AF,cut_off,nXs,gpu, data); if(gpu) { cudaDisassemble(AFo, Zs,Xs , nZs,nXs, odd, n, newlen, AF,data); } else { Disassemble(nZs,nXs,Zs,Xs,AFo, AF, newlen,odd); } //Free memory free(nZs); free(nXs); } }
You are a helpful coding assistant specialized in converting HIP source code to AMD device assembly.
// cudaDCA.cu // //This file contains the recursive DCA function, and the function that is used to invoke DCA and //interperate the results. //Included Files #include <hip/hip_runtime.h> #include <iostream> //Function Prototypes // Functions found in this file void RecDCA(double Zs[], int n, int i, double AF[], int cut_off,double Xs[]); // Functions found in Init_setup.cu void CudaInitialize(double m[], double l[], double I[], double x[], int n, double Zs[]); // Functions found in Assemble_setup.cu void cudaAssemble(double Zs[],double Xs[], int num, double nZs[], double nXs[], int odd, int newlen, int data); // Functions found in Disassemble_setup.cu void cudaDisassemble(double OldAF[], double Zs[], double Xs[],double nZs[], double nXs[], int odd, int morelen, int lesslen, double AF[], int data); // Functions found in Assemble.cu void Assemble(double Zs[], double Xs[],double nZs[], double nXs[], int len, int odd, int n); // Functions found in Disassemble.cu void Disassemble(double lessZs[], double lessXs[],double moreZs[], double moreXs[], double oldAs[] ,double newAs[], int num, int odd); // Functions found in SolveBCs.cu void solve_BCs(double Zs[], double Xs[], double AF[]); void printa(double A[], int n); void printm(double A[6][6]); void Update_Properties(double bodyZetas[],double nZetas[], int n, double state[], double m[], double l[], double II[]); //DCAhelp: // Function that prepares the list of bodies for DCA and finds the final state vector // state is the state of the system at that timestep // bs is a list of bodies used for initialization // js is a list of joints // n is the number of bodies // Y is the array where the final velocities and accelerations are stored void DCAhelp(double state[], double m[], double l[], double I[],int n, double Y[],int cut_off, double bZs[]) { //Create the list that will hold all acceleration and force values for all bodies double *AF = (double*) malloc(sizeof(double)*n*4*6); //double A[6][n*2]; //Create the matrix where only the accelerations will be stored double *Zs=(double*)malloc(sizeof(double)*n*6*26); double *Xs=(double*)malloc(sizeof(double)*n*5*5); Update_Properties(bZs,Zs,n,state,m,l,I); for(int r =0; r<6; r++) //CudaInitialize(m,l,I, state, n, Zs); //Initialize the bodies, finding all zeta values //Pass the list of bodies to DCA and return the accelerations //and forces of both handles of every body in the list //RecDCA(Zs, n, 0, AF, cut_off,Xs); Y[n]=AF[8*n]; //For a pendulum, the fist acceleration value is in A[2][0] for(int i = n+1, j=2; i<n*2; i++, j+=2) //Loop through the acceleration matrix { Y[i]=AF[2*4*n+2*j]-AF[2*4*n+2*(j-1)]; //Find and save all generalized accelerations } for(int i = 0; i<n; i++) //Loop through the state vector { Y[i]=state[i+n]; //Save the velocities } //Free memory free(AF); free(Zs); free(Xs); } //RecDCA: // Function used to solve for the velocty and acceleration of the list of bodies at // the current timestep. This is a recursive function that continues to call itself // until there is a single body left. Once at this point the accelerations and forces // are found using the boundary conditions of a pendulum. These values are then returned // to the previous level of recursion which then finds the new accelerations and forces // for the disassembled bodies. This continues until all bodies are disassembled, ultimately // returning the forces and accelerations at both handles of every body in the system. These // results are intererated by DCAhelp (above) to obtain the actual generalized accelerations. // bodies is the list of bodies // n is the number of bodies // i is the level of recursion // AF is the array in which the accelerations and forces at the handles of the bodies // will be stored. void RecDCA(double Zs[], int n, int i, double AF[], int cut_off,double Xs[],int gpu, int data) { if (n==1) //If there is only 1 body { } else //If there is more than 1 body { int newlen; //New number of bodies after assembly int odd = 0; //Flag to keep track of the parity of the length of the list of if(n % 2 == 0) //If there is an even number of bodies { newlen = (int) (n/2); //The new number of bodies will be half the original number } else //If there is an odd number of bodies { newlen = (int)((n+1)/2); //The new number of bodies will be half the original number //rounded down, plus 1 odd = 1; //odd is set to 1 because there are an odd number of bodies } double *nZs=(double*)malloc(sizeof(double)*newlen*26*6); double *nXs=(double*) malloc(sizeof(double)*(newlen)*5*5); double *AFo=(double*)malloc(sizeof(double)*6*newlen*4); //Call the DCA function again to return the accelerations and forces of the new bodies RecDCA(nZs,newlen,i+1 ,AF,cut_off,nXs,gpu, data); if(gpu) { cudaDisassemble(AFo, Zs,Xs , nZs,nXs, odd, n, newlen, AF,data); } else { Disassemble(nZs,nXs,Zs,Xs,AFo, AF, newlen,odd); } //Free memory free(nZs); free(nXs); } }
.text .p2alignl 7, 3214868480 .fill 96, 4, 3214868480 .type __hip_cuid_,@object .section .bss,"aw",@nobits .globl __hip_cuid_ __hip_cuid_: .byte 0 .size __hip_cuid_, 1 .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym __hip_cuid_ .amdgcn_target "amdgcn-amd-amdhsa--gfx1100" .amdgpu_metadata --- amdhsa.kernels: [] amdhsa.target: amdgcn-amd-amdhsa--gfx1100 amdhsa.version: - 1 - 2 ... .end_amdgpu_metadata
You are a helpful coding assistant specialized in converting HIP source code to AMD host assembly.
// cudaDCA.cu // //This file contains the recursive DCA function, and the function that is used to invoke DCA and //interperate the results. //Included Files #include <hip/hip_runtime.h> #include <iostream> //Function Prototypes // Functions found in this file void RecDCA(double Zs[], int n, int i, double AF[], int cut_off,double Xs[]); // Functions found in Init_setup.cu void CudaInitialize(double m[], double l[], double I[], double x[], int n, double Zs[]); // Functions found in Assemble_setup.cu void cudaAssemble(double Zs[],double Xs[], int num, double nZs[], double nXs[], int odd, int newlen, int data); // Functions found in Disassemble_setup.cu void cudaDisassemble(double OldAF[], double Zs[], double Xs[],double nZs[], double nXs[], int odd, int morelen, int lesslen, double AF[], int data); // Functions found in Assemble.cu void Assemble(double Zs[], double Xs[],double nZs[], double nXs[], int len, int odd, int n); // Functions found in Disassemble.cu void Disassemble(double lessZs[], double lessXs[],double moreZs[], double moreXs[], double oldAs[] ,double newAs[], int num, int odd); // Functions found in SolveBCs.cu void solve_BCs(double Zs[], double Xs[], double AF[]); void printa(double A[], int n); void printm(double A[6][6]); void Update_Properties(double bodyZetas[],double nZetas[], int n, double state[], double m[], double l[], double II[]); //DCAhelp: // Function that prepares the list of bodies for DCA and finds the final state vector // state is the state of the system at that timestep // bs is a list of bodies used for initialization // js is a list of joints // n is the number of bodies // Y is the array where the final velocities and accelerations are stored void DCAhelp(double state[], double m[], double l[], double I[],int n, double Y[],int cut_off, double bZs[]) { //Create the list that will hold all acceleration and force values for all bodies double *AF = (double*) malloc(sizeof(double)*n*4*6); //double A[6][n*2]; //Create the matrix where only the accelerations will be stored double *Zs=(double*)malloc(sizeof(double)*n*6*26); double *Xs=(double*)malloc(sizeof(double)*n*5*5); Update_Properties(bZs,Zs,n,state,m,l,I); for(int r =0; r<6; r++) //CudaInitialize(m,l,I, state, n, Zs); //Initialize the bodies, finding all zeta values //Pass the list of bodies to DCA and return the accelerations //and forces of both handles of every body in the list //RecDCA(Zs, n, 0, AF, cut_off,Xs); Y[n]=AF[8*n]; //For a pendulum, the fist acceleration value is in A[2][0] for(int i = n+1, j=2; i<n*2; i++, j+=2) //Loop through the acceleration matrix { Y[i]=AF[2*4*n+2*j]-AF[2*4*n+2*(j-1)]; //Find and save all generalized accelerations } for(int i = 0; i<n; i++) //Loop through the state vector { Y[i]=state[i+n]; //Save the velocities } //Free memory free(AF); free(Zs); free(Xs); } //RecDCA: // Function used to solve for the velocty and acceleration of the list of bodies at // the current timestep. This is a recursive function that continues to call itself // until there is a single body left. Once at this point the accelerations and forces // are found using the boundary conditions of a pendulum. These values are then returned // to the previous level of recursion which then finds the new accelerations and forces // for the disassembled bodies. This continues until all bodies are disassembled, ultimately // returning the forces and accelerations at both handles of every body in the system. These // results are intererated by DCAhelp (above) to obtain the actual generalized accelerations. // bodies is the list of bodies // n is the number of bodies // i is the level of recursion // AF is the array in which the accelerations and forces at the handles of the bodies // will be stored. void RecDCA(double Zs[], int n, int i, double AF[], int cut_off,double Xs[],int gpu, int data) { if (n==1) //If there is only 1 body { } else //If there is more than 1 body { int newlen; //New number of bodies after assembly int odd = 0; //Flag to keep track of the parity of the length of the list of if(n % 2 == 0) //If there is an even number of bodies { newlen = (int) (n/2); //The new number of bodies will be half the original number } else //If there is an odd number of bodies { newlen = (int)((n+1)/2); //The new number of bodies will be half the original number //rounded down, plus 1 odd = 1; //odd is set to 1 because there are an odd number of bodies } double *nZs=(double*)malloc(sizeof(double)*newlen*26*6); double *nXs=(double*) malloc(sizeof(double)*(newlen)*5*5); double *AFo=(double*)malloc(sizeof(double)*6*newlen*4); //Call the DCA function again to return the accelerations and forces of the new bodies RecDCA(nZs,newlen,i+1 ,AF,cut_off,nXs,gpu, data); if(gpu) { cudaDisassemble(AFo, Zs,Xs , nZs,nXs, odd, n, newlen, AF,data); } else { Disassemble(nZs,nXs,Zs,Xs,AFo, AF, newlen,odd); } //Free memory free(nZs); free(nXs); } }
.text .file "cudaDCA.hip" # Start of file scope inline assembly .globl _ZSt21ios_base_library_initv # End of file scope inline assembly .globl _Z7DCAhelpPdS_S_S_iS_iS_ # -- Begin function _Z7DCAhelpPdS_S_S_iS_iS_ .p2align 4, 0x90 .type _Z7DCAhelpPdS_S_S_iS_iS_,@function _Z7DCAhelpPdS_S_S_iS_iS_: # @_Z7DCAhelpPdS_S_S_iS_iS_ .cfi_startproc # %bb.0: # %.preheader43 pushq %rbp .cfi_def_cfa_offset 16 pushq %r15 .cfi_def_cfa_offset 24 pushq %r14 .cfi_def_cfa_offset 32 pushq %r13 .cfi_def_cfa_offset 40 pushq %r12 .cfi_def_cfa_offset 48 pushq %rbx .cfi_def_cfa_offset 56 subq $40, %rsp .cfi_def_cfa_offset 96 .cfi_offset %rbx, -56 .cfi_offset %r12, -48 .cfi_offset %r13, -40 .cfi_offset %r14, -32 .cfi_offset %r15, -24 .cfi_offset %rbp, -16 movq %r9, %rbx movl %r8d, %r14d movq %rcx, 16(%rsp) # 8-byte Spill movq %rdx, 32(%rsp) # 8-byte Spill movq %rsi, 24(%rsp) # 8-byte Spill movq %rdi, %r13 movq 104(%rsp), %r15 movslq %r8d, %rbp movq %rbp, %rax shlq $6, %rax leaq (%rax,%rax,2), %rdi callq malloc movq %rax, %r12 imulq $1248, %rbp, %rdi # imm = 0x4E0 callq malloc movq 16(%rsp), %rcx # 8-byte Reload movq %rcx, (%rsp) movq %r15, %rdi movq %rax, %r15 movq %rax, %rsi movl %ebp, %edx movq %r13, %rcx movq 24(%rsp), %r8 # 8-byte Reload movq 32(%rsp), %r9 # 8-byte Reload callq _Z17Update_PropertiesPdS_iS_S_S_S_ leal (,%rbp,8), %eax cltq movsd (%r12,%rax,8), %xmm0 # xmm0 = mem[0],zero movsd %xmm0, (%rbx,%rbp,8) leal (,%rbp,2), %edx leal 1(%rbp), %ecx cmpl %edx, %ecx jge .LBB0_3 # %bb.1: # %.lr.ph movslq %ecx, %rcx leaq (%r12,%rax,8), %rax addq $32, %rax leaq (%rbx,%rcx,8), %rcx leal -1(%r14), %edx xorl %esi, %esi .p2align 4, 0x90 .LBB0_2: # =>This Inner Loop Header: Depth=1 movsd (%rax), %xmm0 # xmm0 = mem[0],zero subsd -16(%rax), %xmm0 movsd %xmm0, (%rcx,%rsi,8) addq $32, %rax incq %rsi cmpl %esi, %edx jne .LBB0_2 .LBB0_3: # %.preheader testl %r14d, %r14d jle .LBB0_6 # %bb.4: # %.lr.ph48.preheader movl %r14d, %eax leaq (,%rax,8), %rcx addq %r13, %rcx xorl %edx, %edx .p2align 4, 0x90 .LBB0_5: # %.lr.ph48 # =>This Inner Loop Header: Depth=1 movsd (%rcx,%rdx,8), %xmm0 # xmm0 = mem[0],zero movsd %xmm0, (%rbx,%rdx,8) incq %rdx cmpq %rdx, %rax jne .LBB0_5 .LBB0_6: # %._crit_edge movq %r12, %rdi callq free movq %r15, %rdi addq $40, %rsp .cfi_def_cfa_offset 56 popq %rbx .cfi_def_cfa_offset 48 popq %r12 .cfi_def_cfa_offset 40 popq %r13 .cfi_def_cfa_offset 32 popq %r14 .cfi_def_cfa_offset 24 popq %r15 .cfi_def_cfa_offset 16 popq %rbp .cfi_def_cfa_offset 8 jmp free # TAILCALL .Lfunc_end0: .size _Z7DCAhelpPdS_S_S_iS_iS_, .Lfunc_end0-_Z7DCAhelpPdS_S_S_iS_iS_ .cfi_endproc # -- End function .globl _Z6RecDCAPdiiS_iS_ii # -- Begin function _Z6RecDCAPdiiS_iS_ii .p2align 4, 0x90 .type _Z6RecDCAPdiiS_iS_ii,@function _Z6RecDCAPdiiS_iS_ii: # @_Z6RecDCAPdiiS_iS_ii .cfi_startproc # %bb.0: cmpl $1, %esi jne .LBB1_1 # %bb.5: retq .LBB1_1: pushq %rbp .cfi_def_cfa_offset 16 pushq %r15 .cfi_def_cfa_offset 24 pushq %r14 .cfi_def_cfa_offset 32 pushq %r13 .cfi_def_cfa_offset 40 pushq %r12 .cfi_def_cfa_offset 48 pushq %rbx .cfi_def_cfa_offset 56 subq $72, %rsp .cfi_def_cfa_offset 128 .cfi_offset %rbx, -56 .cfi_offset %r12, -48 .cfi_offset %r13, -40 .cfi_offset %r14, -32 .cfi_offset %r15, -24 .cfi_offset %rbp, -16 movl %edx, %ebx movl %esi, %r15d movq %rdi, 16(%rsp) # 8-byte Spill movq %r9, 24(%rsp) # 8-byte Spill movl 136(%rsp), %eax movq %rax, 40(%rsp) # 8-byte Spill movl 128(%rsp), %eax movq %rax, 56(%rsp) # 8-byte Spill movl %esi, %ebp andl $1, %ebp leal (%r15,%rbp), %r12d sarl %r12d movq %rcx, 64(%rsp) # 8-byte Spill movslq %r12d, %r14 imulq $1248, %r14, %rdi # imm = 0x4E0 movl %r8d, 4(%rsp) # 4-byte Spill callq malloc movq %rax, %r13 imulq $200, %r14, %rdi callq malloc movq %rax, 48(%rsp) # 8-byte Spill movq %r14, %rax shlq $6, %rax leaq (%rax,%rax,2), %rdi callq malloc movq %rax, 8(%rsp) # 8-byte Spill incl %ebx movq %r13, 32(%rsp) # 8-byte Spill movq %r13, %rdi movq 48(%rsp), %r13 # 8-byte Reload movl %r14d, %esi movq 64(%rsp), %r14 # 8-byte Reload movl %ebx, %edx movq %r14, %rcx movl 4(%rsp), %r8d # 4-byte Reload movq %r13, %r9 pushq 40(%rsp) # 8-byte Folded Reload .cfi_adjust_cfa_offset 8 movq 64(%rsp), %rbx # 8-byte Reload pushq %rbx .cfi_adjust_cfa_offset 8 callq _Z6RecDCAPdiiS_iS_ii addq $16, %rsp .cfi_adjust_cfa_offset -16 testl %ebx, %ebx je .LBB1_3 # %bb.2: movq 8(%rsp), %rdi # 8-byte Reload movq 16(%rsp), %rsi # 8-byte Reload movq 24(%rsp), %rdx # 8-byte Reload movq 32(%rsp), %rbx # 8-byte Reload movq %rbx, %rcx movq %r13, %r8 movl %ebp, %r9d pushq 40(%rsp) # 8-byte Folded Reload .cfi_adjust_cfa_offset 8 pushq %r14 .cfi_adjust_cfa_offset 8 pushq %r12 .cfi_adjust_cfa_offset 8 pushq %r15 .cfi_adjust_cfa_offset 8 callq _Z15cudaDisassemblePdS_S_S_S_iiiS_i addq $32, %rsp .cfi_adjust_cfa_offset -32 jmp .LBB1_4 .LBB1_3: movq 32(%rsp), %rbx # 8-byte Reload movq %rbx, %rdi movq %r13, %rsi movq 16(%rsp), %rdx # 8-byte Reload movq 24(%rsp), %rcx # 8-byte Reload movq 8(%rsp), %r8 # 8-byte Reload movq %r14, %r9 pushq %rbp .cfi_adjust_cfa_offset 8 pushq %r12 .cfi_adjust_cfa_offset 8 callq _Z11DisassemblePdS_S_S_S_S_ii addq $16, %rsp .cfi_adjust_cfa_offset -16 .LBB1_4: movq %rbx, %rdi callq free movq %r13, %rdi addq $72, %rsp .cfi_def_cfa_offset 56 popq %rbx .cfi_def_cfa_offset 48 popq %r12 .cfi_def_cfa_offset 40 popq %r13 .cfi_def_cfa_offset 32 popq %r14 .cfi_def_cfa_offset 24 popq %r15 .cfi_def_cfa_offset 16 popq %rbp .cfi_def_cfa_offset 8 jmp free # TAILCALL .Lfunc_end1: .size _Z6RecDCAPdiiS_iS_ii, .Lfunc_end1-_Z6RecDCAPdiiS_iS_ii .cfi_endproc # -- End function .type __hip_cuid_,@object # @__hip_cuid_ .bss .globl __hip_cuid_ __hip_cuid_: .byte 0 # 0x0 .size __hip_cuid_, 1 .section ".linker-options","e",@llvm_linker_options .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym __hip_cuid_
You are a helpful coding assistant specialized in converting from CUDA device assembly to AMD device assembly.
code for sm_80
.text .p2alignl 7, 3214868480 .fill 96, 4, 3214868480 .type __hip_cuid_,@object .section .bss,"aw",@nobits .globl __hip_cuid_ __hip_cuid_: .byte 0 .size __hip_cuid_, 1 .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym __hip_cuid_ .amdgcn_target "amdgcn-amd-amdhsa--gfx1100" .amdgpu_metadata --- amdhsa.kernels: [] amdhsa.target: amdgcn-amd-amdhsa--gfx1100 amdhsa.version: - 1 - 2 ... .end_amdgpu_metadata
You are a helpful coding assistant specialized in converting from CUDA host assembly to AMD host assembly.
.file "tmpxft_00064b10_00000000-6_cudaDCA.cudafe1.cpp" .text #APP .globl _ZSt21ios_base_library_initv #NO_APP .type _ZL26__cudaUnregisterBinaryUtilv, @function _ZL26__cudaUnregisterBinaryUtilv: .LFB3673: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaUnregisterFatBinary@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE3673: .size _ZL26__cudaUnregisterBinaryUtilv, .-_ZL26__cudaUnregisterBinaryUtilv .globl _Z7DCAhelpPdS_S_S_iS_iS_ .type _Z7DCAhelpPdS_S_S_iS_iS_, @function _Z7DCAhelpPdS_S_S_iS_iS_: .LFB3669: .cfi_startproc endbr64 pushq %r15 .cfi_def_cfa_offset 16 .cfi_offset 15, -16 pushq %r14 .cfi_def_cfa_offset 24 .cfi_offset 14, -24 pushq %r13 .cfi_def_cfa_offset 32 .cfi_offset 13, -32 pushq %r12 .cfi_def_cfa_offset 40 .cfi_offset 12, -40 pushq %rbp .cfi_def_cfa_offset 48 .cfi_offset 6, -48 pushq %rbx .cfi_def_cfa_offset 56 .cfi_offset 3, -56 subq $40, %rsp .cfi_def_cfa_offset 96 movq %rdi, %rbx movq %rsi, 8(%rsp) movq %rdx, 16(%rsp) movq %rcx, 24(%rsp) movl %r8d, %r12d movq %r9, %rbp movslq %r8d, %r14 leaq (%r14,%r14,2), %rdi salq $6, %rdi call malloc@PLT movq %rax, %r13 imulq $1248, %r14, %rdi call malloc@PLT movq %rax, %r15 subq $8, %rsp .cfi_def_cfa_offset 104 pushq 32(%rsp) .cfi_def_cfa_offset 112 movq 32(%rsp), %r9 movq 24(%rsp), %r8 movq %rbx, %rcx movl %r12d, %edx movq %rax, %rsi movq 120(%rsp), %rdi call _Z17Update_PropertiesPdS_iS_S_S_S_@PLT leal 0(,%r12,8), %ecx movslq %ecx, %rcx movsd 0(%r13,%rcx,8), %xmm0 movsd %xmm0, 0(%rbp,%rcx) leal 1(%r12), %edx leal (%r12,%r12), %eax addq $16, %rsp .cfi_def_cfa_offset 96 cmpl %eax, %edx jge .L4 leal 4(,%r12,8), %eax cltq leaq 0(%r13,%rax,8), %rdx leaq 8(%rbp,%rcx), %rax leal -2(%r12), %esi addq %r14, %rsi leaq 16(%rbp,%rsi,8), %rsi .L5: movsd (%rdx), %xmm0 subsd -16(%rdx), %xmm0 movsd %xmm0, (%rax) addq $32, %rdx addq $8, %rax cmpq %rsi, %rax jne .L5 .L4: testl %r12d, %r12d jle .L6 addq %rcx, %rbx movl $0, %eax .L7: movsd (%rbx,%rax), %xmm0 movsd %xmm0, 0(%rbp,%rax) addq $8, %rax cmpq %rax, %rcx jne .L7 .L6: movq %r13, %rdi call free@PLT movq %r15, %rdi call free@PLT addq $40, %rsp .cfi_def_cfa_offset 56 popq %rbx .cfi_def_cfa_offset 48 popq %rbp .cfi_def_cfa_offset 40 popq %r12 .cfi_def_cfa_offset 32 popq %r13 .cfi_def_cfa_offset 24 popq %r14 .cfi_def_cfa_offset 16 popq %r15 .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE3669: .size _Z7DCAhelpPdS_S_S_iS_iS_, .-_Z7DCAhelpPdS_S_S_iS_iS_ .globl _Z6RecDCAPdiiS_iS_ii .type _Z6RecDCAPdiiS_iS_ii, @function _Z6RecDCAPdiiS_iS_ii: .LFB3670: .cfi_startproc endbr64 pushq %r15 .cfi_def_cfa_offset 16 .cfi_offset 15, -16 pushq %r14 .cfi_def_cfa_offset 24 .cfi_offset 14, -24 pushq %r13 .cfi_def_cfa_offset 32 .cfi_offset 13, -32 pushq %r12 .cfi_def_cfa_offset 40 .cfi_offset 12, -40 pushq %rbp .cfi_def_cfa_offset 48 .cfi_offset 6, -48 pushq %rbx .cfi_def_cfa_offset 56 .cfi_offset 3, -56 subq $40, %rsp .cfi_def_cfa_offset 96 movq %rdi, (%rsp) movl %edx, 12(%rsp) movl %r8d, 24(%rsp) movq %r9, 16(%rsp) cmpl $1, %esi je .L11 movl %esi, %ebx movq %rcx, %r12 movl %esi, %eax andl $1, %eax movl %eax, 28(%rsp) jne .L13 movl %esi, %ebp shrl $31, %ebp addl %esi, %ebp sarl %ebp .L14: movslq %ebp, %r14 imulq $1248, %r14, %rdi call malloc@PLT movq %rax, %r13 leaq (%r14,%r14,4), %rax leaq (%rax,%rax,4), %rdi salq $3, %rdi call malloc@PLT movq %rax, %r15 leaq (%r14,%r14,2), %rdi salq $6, %rdi call malloc@PLT movq %rax, %r14 movl 12(%rsp), %edx addl $1, %edx movl 104(%rsp), %eax pushq %rax .cfi_def_cfa_offset 104 movl 104(%rsp), %eax pushq %rax .cfi_def_cfa_offset 112 movq %r15, %r9 movl 40(%rsp), %r8d movq %r12, %rcx movl %ebp, %esi movq %r13, %rdi call _Z6RecDCAPdiiS_iS_ii addq $16, %rsp .cfi_def_cfa_offset 96 cmpl $0, 96(%rsp) je .L15 movl 104(%rsp), %eax pushq %rax .cfi_def_cfa_offset 104 pushq %r12 .cfi_def_cfa_offset 112 pushq %rbp .cfi_def_cfa_offset 120 pushq %rbx .cfi_def_cfa_offset 128 movl 60(%rsp), %r9d movq %r15, %r8 movq %r13, %rcx movq 48(%rsp), %rdx movq 32(%rsp), %rsi movq %r14, %rdi call _Z15cudaDisassemblePdS_S_S_S_iiiS_i@PLT addq $32, %rsp .cfi_def_cfa_offset 96 .L16: movq %r13, %rdi call free@PLT movq %r15, %rdi call free@PLT .L11: addq $40, %rsp .cfi_remember_state .cfi_def_cfa_offset 56 popq %rbx .cfi_def_cfa_offset 48 popq %rbp .cfi_def_cfa_offset 40 popq %r12 .cfi_def_cfa_offset 32 popq %r13 .cfi_def_cfa_offset 24 popq %r14 .cfi_def_cfa_offset 16 popq %r15 .cfi_def_cfa_offset 8 ret .L13: .cfi_restore_state leal 1(%rsi), %eax movl %eax, %ebp shrl $31, %ebp addl %eax, %ebp sarl %ebp jmp .L14 .L15: movl 28(%rsp), %eax pushq %rax .cfi_def_cfa_offset 104 pushq %rbp .cfi_def_cfa_offset 112 movq %r12, %r9 movq %r14, %r8 movq 32(%rsp), %rcx movq 16(%rsp), %rdx movq %r15, %rsi movq %r13, %rdi call _Z11DisassemblePdS_S_S_S_S_ii@PLT addq $16, %rsp .cfi_def_cfa_offset 96 jmp .L16 .cfi_endproc .LFE3670: .size _Z6RecDCAPdiiS_iS_ii, .-_Z6RecDCAPdiiS_iS_ii .type _ZL24__sti____cudaRegisterAllv, @function _ZL24__sti____cudaRegisterAllv: .LFB3696: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 leaq _ZL15__fatDeviceText(%rip), %rdi call __cudaRegisterFatBinary@PLT movq %rax, %rdi movq %rax, _ZL20__cudaFatCubinHandle(%rip) call __cudaRegisterFatBinaryEnd@PLT leaq _ZL26__cudaUnregisterBinaryUtilv(%rip), %rdi call atexit@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE3696: .size _ZL24__sti____cudaRegisterAllv, .-_ZL24__sti____cudaRegisterAllv .section .init_array,"aw" .align 8 .quad _ZL24__sti____cudaRegisterAllv .section .nvFatBinSegment,"aw" .align 8 .type _ZL15__fatDeviceText, @object .size _ZL15__fatDeviceText, 24 _ZL15__fatDeviceText: .long 1180844977 .long 1 .quad fatbinData .quad 0 .local _ZL20__cudaFatCubinHandle .comm _ZL20__cudaFatCubinHandle,8,8 .ident "GCC: (Ubuntu 13.3.0-6ubuntu2~24.04) 13.3.0" .section .note.GNU-stack,"",@progbits .section .note.gnu.property,"a" .align 8 .long 1f - 0f .long 4f - 1f .long 5 0: .string "GNU" 1: .align 8 .long 0xc0000002 .long 3f - 2f 2: .long 0x3 3: .align 8 4:
.text .file "cudaDCA.hip" # Start of file scope inline assembly .globl _ZSt21ios_base_library_initv # End of file scope inline assembly .globl _Z7DCAhelpPdS_S_S_iS_iS_ # -- Begin function _Z7DCAhelpPdS_S_S_iS_iS_ .p2align 4, 0x90 .type _Z7DCAhelpPdS_S_S_iS_iS_,@function _Z7DCAhelpPdS_S_S_iS_iS_: # @_Z7DCAhelpPdS_S_S_iS_iS_ .cfi_startproc # %bb.0: # %.preheader43 pushq %rbp .cfi_def_cfa_offset 16 pushq %r15 .cfi_def_cfa_offset 24 pushq %r14 .cfi_def_cfa_offset 32 pushq %r13 .cfi_def_cfa_offset 40 pushq %r12 .cfi_def_cfa_offset 48 pushq %rbx .cfi_def_cfa_offset 56 subq $40, %rsp .cfi_def_cfa_offset 96 .cfi_offset %rbx, -56 .cfi_offset %r12, -48 .cfi_offset %r13, -40 .cfi_offset %r14, -32 .cfi_offset %r15, -24 .cfi_offset %rbp, -16 movq %r9, %rbx movl %r8d, %r14d movq %rcx, 16(%rsp) # 8-byte Spill movq %rdx, 32(%rsp) # 8-byte Spill movq %rsi, 24(%rsp) # 8-byte Spill movq %rdi, %r13 movq 104(%rsp), %r15 movslq %r8d, %rbp movq %rbp, %rax shlq $6, %rax leaq (%rax,%rax,2), %rdi callq malloc movq %rax, %r12 imulq $1248, %rbp, %rdi # imm = 0x4E0 callq malloc movq 16(%rsp), %rcx # 8-byte Reload movq %rcx, (%rsp) movq %r15, %rdi movq %rax, %r15 movq %rax, %rsi movl %ebp, %edx movq %r13, %rcx movq 24(%rsp), %r8 # 8-byte Reload movq 32(%rsp), %r9 # 8-byte Reload callq _Z17Update_PropertiesPdS_iS_S_S_S_ leal (,%rbp,8), %eax cltq movsd (%r12,%rax,8), %xmm0 # xmm0 = mem[0],zero movsd %xmm0, (%rbx,%rbp,8) leal (,%rbp,2), %edx leal 1(%rbp), %ecx cmpl %edx, %ecx jge .LBB0_3 # %bb.1: # %.lr.ph movslq %ecx, %rcx leaq (%r12,%rax,8), %rax addq $32, %rax leaq (%rbx,%rcx,8), %rcx leal -1(%r14), %edx xorl %esi, %esi .p2align 4, 0x90 .LBB0_2: # =>This Inner Loop Header: Depth=1 movsd (%rax), %xmm0 # xmm0 = mem[0],zero subsd -16(%rax), %xmm0 movsd %xmm0, (%rcx,%rsi,8) addq $32, %rax incq %rsi cmpl %esi, %edx jne .LBB0_2 .LBB0_3: # %.preheader testl %r14d, %r14d jle .LBB0_6 # %bb.4: # %.lr.ph48.preheader movl %r14d, %eax leaq (,%rax,8), %rcx addq %r13, %rcx xorl %edx, %edx .p2align 4, 0x90 .LBB0_5: # %.lr.ph48 # =>This Inner Loop Header: Depth=1 movsd (%rcx,%rdx,8), %xmm0 # xmm0 = mem[0],zero movsd %xmm0, (%rbx,%rdx,8) incq %rdx cmpq %rdx, %rax jne .LBB0_5 .LBB0_6: # %._crit_edge movq %r12, %rdi callq free movq %r15, %rdi addq $40, %rsp .cfi_def_cfa_offset 56 popq %rbx .cfi_def_cfa_offset 48 popq %r12 .cfi_def_cfa_offset 40 popq %r13 .cfi_def_cfa_offset 32 popq %r14 .cfi_def_cfa_offset 24 popq %r15 .cfi_def_cfa_offset 16 popq %rbp .cfi_def_cfa_offset 8 jmp free # TAILCALL .Lfunc_end0: .size _Z7DCAhelpPdS_S_S_iS_iS_, .Lfunc_end0-_Z7DCAhelpPdS_S_S_iS_iS_ .cfi_endproc # -- End function .globl _Z6RecDCAPdiiS_iS_ii # -- Begin function _Z6RecDCAPdiiS_iS_ii .p2align 4, 0x90 .type _Z6RecDCAPdiiS_iS_ii,@function _Z6RecDCAPdiiS_iS_ii: # @_Z6RecDCAPdiiS_iS_ii .cfi_startproc # %bb.0: cmpl $1, %esi jne .LBB1_1 # %bb.5: retq .LBB1_1: pushq %rbp .cfi_def_cfa_offset 16 pushq %r15 .cfi_def_cfa_offset 24 pushq %r14 .cfi_def_cfa_offset 32 pushq %r13 .cfi_def_cfa_offset 40 pushq %r12 .cfi_def_cfa_offset 48 pushq %rbx .cfi_def_cfa_offset 56 subq $72, %rsp .cfi_def_cfa_offset 128 .cfi_offset %rbx, -56 .cfi_offset %r12, -48 .cfi_offset %r13, -40 .cfi_offset %r14, -32 .cfi_offset %r15, -24 .cfi_offset %rbp, -16 movl %edx, %ebx movl %esi, %r15d movq %rdi, 16(%rsp) # 8-byte Spill movq %r9, 24(%rsp) # 8-byte Spill movl 136(%rsp), %eax movq %rax, 40(%rsp) # 8-byte Spill movl 128(%rsp), %eax movq %rax, 56(%rsp) # 8-byte Spill movl %esi, %ebp andl $1, %ebp leal (%r15,%rbp), %r12d sarl %r12d movq %rcx, 64(%rsp) # 8-byte Spill movslq %r12d, %r14 imulq $1248, %r14, %rdi # imm = 0x4E0 movl %r8d, 4(%rsp) # 4-byte Spill callq malloc movq %rax, %r13 imulq $200, %r14, %rdi callq malloc movq %rax, 48(%rsp) # 8-byte Spill movq %r14, %rax shlq $6, %rax leaq (%rax,%rax,2), %rdi callq malloc movq %rax, 8(%rsp) # 8-byte Spill incl %ebx movq %r13, 32(%rsp) # 8-byte Spill movq %r13, %rdi movq 48(%rsp), %r13 # 8-byte Reload movl %r14d, %esi movq 64(%rsp), %r14 # 8-byte Reload movl %ebx, %edx movq %r14, %rcx movl 4(%rsp), %r8d # 4-byte Reload movq %r13, %r9 pushq 40(%rsp) # 8-byte Folded Reload .cfi_adjust_cfa_offset 8 movq 64(%rsp), %rbx # 8-byte Reload pushq %rbx .cfi_adjust_cfa_offset 8 callq _Z6RecDCAPdiiS_iS_ii addq $16, %rsp .cfi_adjust_cfa_offset -16 testl %ebx, %ebx je .LBB1_3 # %bb.2: movq 8(%rsp), %rdi # 8-byte Reload movq 16(%rsp), %rsi # 8-byte Reload movq 24(%rsp), %rdx # 8-byte Reload movq 32(%rsp), %rbx # 8-byte Reload movq %rbx, %rcx movq %r13, %r8 movl %ebp, %r9d pushq 40(%rsp) # 8-byte Folded Reload .cfi_adjust_cfa_offset 8 pushq %r14 .cfi_adjust_cfa_offset 8 pushq %r12 .cfi_adjust_cfa_offset 8 pushq %r15 .cfi_adjust_cfa_offset 8 callq _Z15cudaDisassemblePdS_S_S_S_iiiS_i addq $32, %rsp .cfi_adjust_cfa_offset -32 jmp .LBB1_4 .LBB1_3: movq 32(%rsp), %rbx # 8-byte Reload movq %rbx, %rdi movq %r13, %rsi movq 16(%rsp), %rdx # 8-byte Reload movq 24(%rsp), %rcx # 8-byte Reload movq 8(%rsp), %r8 # 8-byte Reload movq %r14, %r9 pushq %rbp .cfi_adjust_cfa_offset 8 pushq %r12 .cfi_adjust_cfa_offset 8 callq _Z11DisassemblePdS_S_S_S_S_ii addq $16, %rsp .cfi_adjust_cfa_offset -16 .LBB1_4: movq %rbx, %rdi callq free movq %r13, %rdi addq $72, %rsp .cfi_def_cfa_offset 56 popq %rbx .cfi_def_cfa_offset 48 popq %r12 .cfi_def_cfa_offset 40 popq %r13 .cfi_def_cfa_offset 32 popq %r14 .cfi_def_cfa_offset 24 popq %r15 .cfi_def_cfa_offset 16 popq %rbp .cfi_def_cfa_offset 8 jmp free # TAILCALL .Lfunc_end1: .size _Z6RecDCAPdiiS_iS_ii, .Lfunc_end1-_Z6RecDCAPdiiS_iS_ii .cfi_endproc # -- End function .type __hip_cuid_,@object # @__hip_cuid_ .bss .globl __hip_cuid_ __hip_cuid_: .byte 0 # 0x0 .size __hip_cuid_, 1 .section ".linker-options","e",@llvm_linker_options .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym __hip_cuid_
You are a helpful coding assistant specialized in converting CUDA source code to CUDA device assembly.
/* #include <cfloat> #include <vector> #include "caffe/layer.hpp" #include "caffe/util/math_functions.hpp" #include "caffe/vision_layers.hpp" namespace caffe { template <typename Dtype> void SuperCategoryFMLayer<Dtype>::Forward_gpu(const vector<Blob<Dtype>*>& bottom, const vector<Blob<Dtype>*>& top) { caffe_copy(bottom[0]->count(), bottom[0]->gpu_data(), top[depth_-1]->mutable_gpu_data()); for(int i = 0; i < depth_-1; ++i) caffe_gpu_set(top[i]->count(), (Dtype)0., top[i]->mutable_gpu_data()); for(int m = 0; m < M_; ++m) { for( int i = depth_-2; i >= 0; --i ) { Blob<Dtype> * tops = top[i]; Blob<Dtype> * bottoms = top[i+1]; int base_idx = base_index_per_level_[i]; for(int j = 0; j < node_num_per_level_[i]; ++j) { Tree * node = serialized_tree_[base_idx + j]; const std::vector<shared_ptr<Tree> >* children = node->GetChildren(); Dtype * top_data = &tops->mutable_gpu_data()[tops->offset(m,node->GetLabel())]; for(std::vector<shared_ptr<Tree> >::const_iterator it = children->begin(); it != children->end(); ++it) { int offset = bottoms->offset(m,(*it)->GetLabel()); const Dtype * bottom_data = &bottoms->gpu_data()[offset]; caffe_gpu_axpy(H_*W_,(Dtype)(1.),bottom_data,top_data); } caffe_gpu_scal(H_*W_,(Dtype)(1./children->size()),top_data); } } } } template <typename Dtype> void SuperCategoryFMLayer<Dtype>::Backward_gpu(const vector<Blob<Dtype>*>& top, const vector<bool>& propagate_down, const vector<Blob<Dtype>*>& bottom) { if( propagate_down[0] == false ) return; for(int m = 0; m < M_; ++m) { for( int i = 0; i < depth_-1; ++i ) { Blob<Dtype> * tops = top[i]; Blob<Dtype> * bottoms = top[i+1]; int base_idx = base_index_per_level_[i]; for(int j = 0; j < node_num_per_level_[i]; ++j) { Tree * node = serialized_tree_[base_idx + j]; const std::vector<shared_ptr<Tree> >* children = node->GetChildren(); const Dtype * top_diff = &tops->gpu_diff()[tops->offset(m,node->GetLabel())]; for(std::vector<shared_ptr<Tree> >::const_iterator it = children->begin(); it != children->end(); ++it) { int offset = bottoms->offset(m,(*it)->GetLabel()); Dtype * bottom_diff = &bottoms->mutable_gpu_diff()[offset]; caffe_gpu_axpy(H_*W_,(Dtype)(1./children->size()),top_diff,bottom_diff); } } } } caffe_copy(bottom[0]->count(), top[depth_-1]->gpu_diff(), bottom[0]->mutable_gpu_diff()); } INSTANTIATE_LAYER_GPU_FUNCS(SuperCategoryFMLayer); } // namespace caffe */
code for sm_80
You are a helpful coding assistant specialized in converting CUDA source code to CUDA host assembly.
/* #include <cfloat> #include <vector> #include "caffe/layer.hpp" #include "caffe/util/math_functions.hpp" #include "caffe/vision_layers.hpp" namespace caffe { template <typename Dtype> void SuperCategoryFMLayer<Dtype>::Forward_gpu(const vector<Blob<Dtype>*>& bottom, const vector<Blob<Dtype>*>& top) { caffe_copy(bottom[0]->count(), bottom[0]->gpu_data(), top[depth_-1]->mutable_gpu_data()); for(int i = 0; i < depth_-1; ++i) caffe_gpu_set(top[i]->count(), (Dtype)0., top[i]->mutable_gpu_data()); for(int m = 0; m < M_; ++m) { for( int i = depth_-2; i >= 0; --i ) { Blob<Dtype> * tops = top[i]; Blob<Dtype> * bottoms = top[i+1]; int base_idx = base_index_per_level_[i]; for(int j = 0; j < node_num_per_level_[i]; ++j) { Tree * node = serialized_tree_[base_idx + j]; const std::vector<shared_ptr<Tree> >* children = node->GetChildren(); Dtype * top_data = &tops->mutable_gpu_data()[tops->offset(m,node->GetLabel())]; for(std::vector<shared_ptr<Tree> >::const_iterator it = children->begin(); it != children->end(); ++it) { int offset = bottoms->offset(m,(*it)->GetLabel()); const Dtype * bottom_data = &bottoms->gpu_data()[offset]; caffe_gpu_axpy(H_*W_,(Dtype)(1.),bottom_data,top_data); } caffe_gpu_scal(H_*W_,(Dtype)(1./children->size()),top_data); } } } } template <typename Dtype> void SuperCategoryFMLayer<Dtype>::Backward_gpu(const vector<Blob<Dtype>*>& top, const vector<bool>& propagate_down, const vector<Blob<Dtype>*>& bottom) { if( propagate_down[0] == false ) return; for(int m = 0; m < M_; ++m) { for( int i = 0; i < depth_-1; ++i ) { Blob<Dtype> * tops = top[i]; Blob<Dtype> * bottoms = top[i+1]; int base_idx = base_index_per_level_[i]; for(int j = 0; j < node_num_per_level_[i]; ++j) { Tree * node = serialized_tree_[base_idx + j]; const std::vector<shared_ptr<Tree> >* children = node->GetChildren(); const Dtype * top_diff = &tops->gpu_diff()[tops->offset(m,node->GetLabel())]; for(std::vector<shared_ptr<Tree> >::const_iterator it = children->begin(); it != children->end(); ++it) { int offset = bottoms->offset(m,(*it)->GetLabel()); Dtype * bottom_diff = &bottoms->mutable_gpu_diff()[offset]; caffe_gpu_axpy(H_*W_,(Dtype)(1./children->size()),top_diff,bottom_diff); } } } } caffe_copy(bottom[0]->count(), top[depth_-1]->gpu_diff(), bottom[0]->mutable_gpu_diff()); } INSTANTIATE_LAYER_GPU_FUNCS(SuperCategoryFMLayer); } // namespace caffe */
.file "tmpxft_001588f5_00000000-6_super_category_fm_layer.cudafe1.cpp" .text #APP #NO_APP .type _ZL26__cudaUnregisterBinaryUtilv, @function _ZL26__cudaUnregisterBinaryUtilv: .LFB2029: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaUnregisterFatBinary@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2029: .size _ZL26__cudaUnregisterBinaryUtilv, .-_ZL26__cudaUnregisterBinaryUtilv .type _ZL24__sti____cudaRegisterAllv, @function _ZL24__sti____cudaRegisterAllv: .LFB2052: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 leaq _ZL15__fatDeviceText(%rip), %rdi call __cudaRegisterFatBinary@PLT movq %rax, %rdi movq %rax, _ZL20__cudaFatCubinHandle(%rip) call __cudaRegisterFatBinaryEnd@PLT leaq _ZL26__cudaUnregisterBinaryUtilv(%rip), %rdi call atexit@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2052: .size _ZL24__sti____cudaRegisterAllv, .-_ZL24__sti____cudaRegisterAllv .section .init_array,"aw" .align 8 .quad _ZL24__sti____cudaRegisterAllv .section .nvFatBinSegment,"aw" .align 8 .type _ZL15__fatDeviceText, @object .size _ZL15__fatDeviceText, 24 _ZL15__fatDeviceText: .long 1180844977 .long 1 .quad fatbinData .quad 0 .local _ZL20__cudaFatCubinHandle .comm _ZL20__cudaFatCubinHandle,8,8 .ident "GCC: (Ubuntu 13.3.0-6ubuntu2~24.04) 13.3.0" .section .note.GNU-stack,"",@progbits .section .note.gnu.property,"a" .align 8 .long 1f - 0f .long 4f - 1f .long 5 0: .string "GNU" 1: .align 8 .long 0xc0000002 .long 3f - 2f 2: .long 0x3 3: .align 8 4:
You are a helpful coding assistant specialized in converting CUDA source code to HIP/AMD source code.
/* #include <cfloat> #include <vector> #include "caffe/layer.hpp" #include "caffe/util/math_functions.hpp" #include "caffe/vision_layers.hpp" namespace caffe { template <typename Dtype> void SuperCategoryFMLayer<Dtype>::Forward_gpu(const vector<Blob<Dtype>*>& bottom, const vector<Blob<Dtype>*>& top) { caffe_copy(bottom[0]->count(), bottom[0]->gpu_data(), top[depth_-1]->mutable_gpu_data()); for(int i = 0; i < depth_-1; ++i) caffe_gpu_set(top[i]->count(), (Dtype)0., top[i]->mutable_gpu_data()); for(int m = 0; m < M_; ++m) { for( int i = depth_-2; i >= 0; --i ) { Blob<Dtype> * tops = top[i]; Blob<Dtype> * bottoms = top[i+1]; int base_idx = base_index_per_level_[i]; for(int j = 0; j < node_num_per_level_[i]; ++j) { Tree * node = serialized_tree_[base_idx + j]; const std::vector<shared_ptr<Tree> >* children = node->GetChildren(); Dtype * top_data = &tops->mutable_gpu_data()[tops->offset(m,node->GetLabel())]; for(std::vector<shared_ptr<Tree> >::const_iterator it = children->begin(); it != children->end(); ++it) { int offset = bottoms->offset(m,(*it)->GetLabel()); const Dtype * bottom_data = &bottoms->gpu_data()[offset]; caffe_gpu_axpy(H_*W_,(Dtype)(1.),bottom_data,top_data); } caffe_gpu_scal(H_*W_,(Dtype)(1./children->size()),top_data); } } } } template <typename Dtype> void SuperCategoryFMLayer<Dtype>::Backward_gpu(const vector<Blob<Dtype>*>& top, const vector<bool>& propagate_down, const vector<Blob<Dtype>*>& bottom) { if( propagate_down[0] == false ) return; for(int m = 0; m < M_; ++m) { for( int i = 0; i < depth_-1; ++i ) { Blob<Dtype> * tops = top[i]; Blob<Dtype> * bottoms = top[i+1]; int base_idx = base_index_per_level_[i]; for(int j = 0; j < node_num_per_level_[i]; ++j) { Tree * node = serialized_tree_[base_idx + j]; const std::vector<shared_ptr<Tree> >* children = node->GetChildren(); const Dtype * top_diff = &tops->gpu_diff()[tops->offset(m,node->GetLabel())]; for(std::vector<shared_ptr<Tree> >::const_iterator it = children->begin(); it != children->end(); ++it) { int offset = bottoms->offset(m,(*it)->GetLabel()); Dtype * bottom_diff = &bottoms->mutable_gpu_diff()[offset]; caffe_gpu_axpy(H_*W_,(Dtype)(1./children->size()),top_diff,bottom_diff); } } } } caffe_copy(bottom[0]->count(), top[depth_-1]->gpu_diff(), bottom[0]->mutable_gpu_diff()); } INSTANTIATE_LAYER_GPU_FUNCS(SuperCategoryFMLayer); } // namespace caffe */
#include <hip/hip_runtime.h> /* #include <cfloat> #include <vector> #include "caffe/layer.hpp" #include "caffe/util/math_functions.hpp" #include "caffe/vision_layers.hpp" namespace caffe { template <typename Dtype> void SuperCategoryFMLayer<Dtype>::Forward_gpu(const vector<Blob<Dtype>*>& bottom, const vector<Blob<Dtype>*>& top) { caffe_copy(bottom[0]->count(), bottom[0]->gpu_data(), top[depth_-1]->mutable_gpu_data()); for(int i = 0; i < depth_-1; ++i) caffe_gpu_set(top[i]->count(), (Dtype)0., top[i]->mutable_gpu_data()); for(int m = 0; m < M_; ++m) { for( int i = depth_-2; i >= 0; --i ) { Blob<Dtype> * tops = top[i]; Blob<Dtype> * bottoms = top[i+1]; int base_idx = base_index_per_level_[i]; for(int j = 0; j < node_num_per_level_[i]; ++j) { Tree * node = serialized_tree_[base_idx + j]; const std::vector<shared_ptr<Tree> >* children = node->GetChildren(); Dtype * top_data = &tops->mutable_gpu_data()[tops->offset(m,node->GetLabel())]; for(std::vector<shared_ptr<Tree> >::const_iterator it = children->begin(); it != children->end(); ++it) { int offset = bottoms->offset(m,(*it)->GetLabel()); const Dtype * bottom_data = &bottoms->gpu_data()[offset]; caffe_gpu_axpy(H_*W_,(Dtype)(1.),bottom_data,top_data); } caffe_gpu_scal(H_*W_,(Dtype)(1./children->size()),top_data); } } } } template <typename Dtype> void SuperCategoryFMLayer<Dtype>::Backward_gpu(const vector<Blob<Dtype>*>& top, const vector<bool>& propagate_down, const vector<Blob<Dtype>*>& bottom) { if( propagate_down[0] == false ) return; for(int m = 0; m < M_; ++m) { for( int i = 0; i < depth_-1; ++i ) { Blob<Dtype> * tops = top[i]; Blob<Dtype> * bottoms = top[i+1]; int base_idx = base_index_per_level_[i]; for(int j = 0; j < node_num_per_level_[i]; ++j) { Tree * node = serialized_tree_[base_idx + j]; const std::vector<shared_ptr<Tree> >* children = node->GetChildren(); const Dtype * top_diff = &tops->gpu_diff()[tops->offset(m,node->GetLabel())]; for(std::vector<shared_ptr<Tree> >::const_iterator it = children->begin(); it != children->end(); ++it) { int offset = bottoms->offset(m,(*it)->GetLabel()); Dtype * bottom_diff = &bottoms->mutable_gpu_diff()[offset]; caffe_gpu_axpy(H_*W_,(Dtype)(1./children->size()),top_diff,bottom_diff); } } } } caffe_copy(bottom[0]->count(), top[depth_-1]->gpu_diff(), bottom[0]->mutable_gpu_diff()); } INSTANTIATE_LAYER_GPU_FUNCS(SuperCategoryFMLayer); } // namespace caffe */
You are a helpful coding assistant specialized in converting HIP source code to AMD device assembly.
#include <hip/hip_runtime.h> /* #include <cfloat> #include <vector> #include "caffe/layer.hpp" #include "caffe/util/math_functions.hpp" #include "caffe/vision_layers.hpp" namespace caffe { template <typename Dtype> void SuperCategoryFMLayer<Dtype>::Forward_gpu(const vector<Blob<Dtype>*>& bottom, const vector<Blob<Dtype>*>& top) { caffe_copy(bottom[0]->count(), bottom[0]->gpu_data(), top[depth_-1]->mutable_gpu_data()); for(int i = 0; i < depth_-1; ++i) caffe_gpu_set(top[i]->count(), (Dtype)0., top[i]->mutable_gpu_data()); for(int m = 0; m < M_; ++m) { for( int i = depth_-2; i >= 0; --i ) { Blob<Dtype> * tops = top[i]; Blob<Dtype> * bottoms = top[i+1]; int base_idx = base_index_per_level_[i]; for(int j = 0; j < node_num_per_level_[i]; ++j) { Tree * node = serialized_tree_[base_idx + j]; const std::vector<shared_ptr<Tree> >* children = node->GetChildren(); Dtype * top_data = &tops->mutable_gpu_data()[tops->offset(m,node->GetLabel())]; for(std::vector<shared_ptr<Tree> >::const_iterator it = children->begin(); it != children->end(); ++it) { int offset = bottoms->offset(m,(*it)->GetLabel()); const Dtype * bottom_data = &bottoms->gpu_data()[offset]; caffe_gpu_axpy(H_*W_,(Dtype)(1.),bottom_data,top_data); } caffe_gpu_scal(H_*W_,(Dtype)(1./children->size()),top_data); } } } } template <typename Dtype> void SuperCategoryFMLayer<Dtype>::Backward_gpu(const vector<Blob<Dtype>*>& top, const vector<bool>& propagate_down, const vector<Blob<Dtype>*>& bottom) { if( propagate_down[0] == false ) return; for(int m = 0; m < M_; ++m) { for( int i = 0; i < depth_-1; ++i ) { Blob<Dtype> * tops = top[i]; Blob<Dtype> * bottoms = top[i+1]; int base_idx = base_index_per_level_[i]; for(int j = 0; j < node_num_per_level_[i]; ++j) { Tree * node = serialized_tree_[base_idx + j]; const std::vector<shared_ptr<Tree> >* children = node->GetChildren(); const Dtype * top_diff = &tops->gpu_diff()[tops->offset(m,node->GetLabel())]; for(std::vector<shared_ptr<Tree> >::const_iterator it = children->begin(); it != children->end(); ++it) { int offset = bottoms->offset(m,(*it)->GetLabel()); Dtype * bottom_diff = &bottoms->mutable_gpu_diff()[offset]; caffe_gpu_axpy(H_*W_,(Dtype)(1./children->size()),top_diff,bottom_diff); } } } } caffe_copy(bottom[0]->count(), top[depth_-1]->gpu_diff(), bottom[0]->mutable_gpu_diff()); } INSTANTIATE_LAYER_GPU_FUNCS(SuperCategoryFMLayer); } // namespace caffe */
.text .p2alignl 7, 3214868480 .fill 96, 4, 3214868480 .type __hip_cuid_,@object .section .bss,"aw",@nobits .globl __hip_cuid_ __hip_cuid_: .byte 0 .size __hip_cuid_, 1 .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym __hip_cuid_ .amdgcn_target "amdgcn-amd-amdhsa--gfx1100" .amdgpu_metadata --- amdhsa.kernels: [] amdhsa.target: amdgcn-amd-amdhsa--gfx1100 amdhsa.version: - 1 - 2 ... .end_amdgpu_metadata
You are a helpful coding assistant specialized in converting HIP source code to AMD host assembly.
#include <hip/hip_runtime.h> /* #include <cfloat> #include <vector> #include "caffe/layer.hpp" #include "caffe/util/math_functions.hpp" #include "caffe/vision_layers.hpp" namespace caffe { template <typename Dtype> void SuperCategoryFMLayer<Dtype>::Forward_gpu(const vector<Blob<Dtype>*>& bottom, const vector<Blob<Dtype>*>& top) { caffe_copy(bottom[0]->count(), bottom[0]->gpu_data(), top[depth_-1]->mutable_gpu_data()); for(int i = 0; i < depth_-1; ++i) caffe_gpu_set(top[i]->count(), (Dtype)0., top[i]->mutable_gpu_data()); for(int m = 0; m < M_; ++m) { for( int i = depth_-2; i >= 0; --i ) { Blob<Dtype> * tops = top[i]; Blob<Dtype> * bottoms = top[i+1]; int base_idx = base_index_per_level_[i]; for(int j = 0; j < node_num_per_level_[i]; ++j) { Tree * node = serialized_tree_[base_idx + j]; const std::vector<shared_ptr<Tree> >* children = node->GetChildren(); Dtype * top_data = &tops->mutable_gpu_data()[tops->offset(m,node->GetLabel())]; for(std::vector<shared_ptr<Tree> >::const_iterator it = children->begin(); it != children->end(); ++it) { int offset = bottoms->offset(m,(*it)->GetLabel()); const Dtype * bottom_data = &bottoms->gpu_data()[offset]; caffe_gpu_axpy(H_*W_,(Dtype)(1.),bottom_data,top_data); } caffe_gpu_scal(H_*W_,(Dtype)(1./children->size()),top_data); } } } } template <typename Dtype> void SuperCategoryFMLayer<Dtype>::Backward_gpu(const vector<Blob<Dtype>*>& top, const vector<bool>& propagate_down, const vector<Blob<Dtype>*>& bottom) { if( propagate_down[0] == false ) return; for(int m = 0; m < M_; ++m) { for( int i = 0; i < depth_-1; ++i ) { Blob<Dtype> * tops = top[i]; Blob<Dtype> * bottoms = top[i+1]; int base_idx = base_index_per_level_[i]; for(int j = 0; j < node_num_per_level_[i]; ++j) { Tree * node = serialized_tree_[base_idx + j]; const std::vector<shared_ptr<Tree> >* children = node->GetChildren(); const Dtype * top_diff = &tops->gpu_diff()[tops->offset(m,node->GetLabel())]; for(std::vector<shared_ptr<Tree> >::const_iterator it = children->begin(); it != children->end(); ++it) { int offset = bottoms->offset(m,(*it)->GetLabel()); Dtype * bottom_diff = &bottoms->mutable_gpu_diff()[offset]; caffe_gpu_axpy(H_*W_,(Dtype)(1./children->size()),top_diff,bottom_diff); } } } } caffe_copy(bottom[0]->count(), top[depth_-1]->gpu_diff(), bottom[0]->mutable_gpu_diff()); } INSTANTIATE_LAYER_GPU_FUNCS(SuperCategoryFMLayer); } // namespace caffe */
.text .file "super_category_fm_layer.hip" .type __hip_cuid_,@object # @__hip_cuid_ .bss .globl __hip_cuid_ __hip_cuid_: .byte 0 # 0x0 .size __hip_cuid_, 1 .section ".linker-options","e",@llvm_linker_options .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym __hip_cuid_
You are a helpful coding assistant specialized in converting from CUDA device assembly to AMD device assembly.
code for sm_80
.text .p2alignl 7, 3214868480 .fill 96, 4, 3214868480 .type __hip_cuid_,@object .section .bss,"aw",@nobits .globl __hip_cuid_ __hip_cuid_: .byte 0 .size __hip_cuid_, 1 .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym __hip_cuid_ .amdgcn_target "amdgcn-amd-amdhsa--gfx1100" .amdgpu_metadata --- amdhsa.kernels: [] amdhsa.target: amdgcn-amd-amdhsa--gfx1100 amdhsa.version: - 1 - 2 ... .end_amdgpu_metadata
You are a helpful coding assistant specialized in converting from CUDA host assembly to AMD host assembly.
.file "tmpxft_001588f5_00000000-6_super_category_fm_layer.cudafe1.cpp" .text #APP #NO_APP .type _ZL26__cudaUnregisterBinaryUtilv, @function _ZL26__cudaUnregisterBinaryUtilv: .LFB2029: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaUnregisterFatBinary@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2029: .size _ZL26__cudaUnregisterBinaryUtilv, .-_ZL26__cudaUnregisterBinaryUtilv .type _ZL24__sti____cudaRegisterAllv, @function _ZL24__sti____cudaRegisterAllv: .LFB2052: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 leaq _ZL15__fatDeviceText(%rip), %rdi call __cudaRegisterFatBinary@PLT movq %rax, %rdi movq %rax, _ZL20__cudaFatCubinHandle(%rip) call __cudaRegisterFatBinaryEnd@PLT leaq _ZL26__cudaUnregisterBinaryUtilv(%rip), %rdi call atexit@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2052: .size _ZL24__sti____cudaRegisterAllv, .-_ZL24__sti____cudaRegisterAllv .section .init_array,"aw" .align 8 .quad _ZL24__sti____cudaRegisterAllv .section .nvFatBinSegment,"aw" .align 8 .type _ZL15__fatDeviceText, @object .size _ZL15__fatDeviceText, 24 _ZL15__fatDeviceText: .long 1180844977 .long 1 .quad fatbinData .quad 0 .local _ZL20__cudaFatCubinHandle .comm _ZL20__cudaFatCubinHandle,8,8 .ident "GCC: (Ubuntu 13.3.0-6ubuntu2~24.04) 13.3.0" .section .note.GNU-stack,"",@progbits .section .note.gnu.property,"a" .align 8 .long 1f - 0f .long 4f - 1f .long 5 0: .string "GNU" 1: .align 8 .long 0xc0000002 .long 3f - 2f 2: .long 0x3 3: .align 8 4:
.text .file "super_category_fm_layer.hip" .type __hip_cuid_,@object # @__hip_cuid_ .bss .globl __hip_cuid_ __hip_cuid_: .byte 0 # 0x0 .size __hip_cuid_, 1 .section ".linker-options","e",@llvm_linker_options .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym __hip_cuid_
You are a helpful coding assistant specialized in converting CUDA source code to CUDA device assembly.
//The max possible number of roots #define DEV_MAX_ROOTS_NUM 3 //Number of threads per block. #define NT 1024 //The array index in global memory. __device__ int devArrayIndex; //The answer array stored roots in global memory __device__ unsigned long long int devAnswer[DEV_MAX_ROOTS_NUM]; /** * Device kernel to find (a) cube root(s) for a given integer c with a modular n. * <P> * Called with a one-dimensional grid of one-dimensional blocks. * * @param c The mod cube (input). * @param n The module (input). * * @author Junan Zhao * @version 26-Nov-2018 */ extern "C" __global__ void modCubeRoot( int c, int n) { //Determine number of threads and this thread's m (test number). unsigned long long m = blockIdx.x*NT + threadIdx.x; unsigned long long size = gridDim.x*NT; for(; m<n; m+=size) //use loop to cover all range of n if n is a pretty large integer cannot covered by one round { unsigned long long temp = m*m; temp = temp%n; temp = temp*m; temp = temp%n; if(c==(int)temp) //once found a root { int oldIndex = atomicAdd(&devArrayIndex,1); devAnswer[oldIndex] = m; } } }
code for sm_80 Function : modCubeRoot .headerflags @"EF_CUDA_TEXMODE_UNIFIED EF_CUDA_64BIT_ADDRESS EF_CUDA_SM80 EF_CUDA_VIRTUAL_SM(EF_CUDA_SM80)" /*0000*/ IMAD.MOV.U32 R1, RZ, RZ, c[0x0][0x28] ; /* 0x00000a00ff017624 */ /* 0x000fe400078e00ff */ /*0010*/ S2R R0, SR_CTAID.X ; /* 0x0000000000007919 */ /* 0x000e220000002500 */ /*0020*/ ULDC UR4, c[0x0][0x164] ; /* 0x0000590000047ab9 */ /* 0x000fe40000000800 */ /*0030*/ USHF.R.S32.HI UR4, URZ, 0x1f, UR4 ; /* 0x0000001f3f047899 */ /* 0x000fe20008011404 */ /*0040*/ S2R R3, SR_TID.X ; /* 0x0000000000037919 */ /* 0x000e240000002100 */ /*0050*/ IMAD R0, R0, 0x400, R3 ; /* 0x0000040000007824 */ /* 0x001fca00078e0203 */ /*0060*/ ISETP.GE.U32.AND P0, PT, R0, c[0x0][0x164], PT ; /* 0x0000590000007a0c */ /* 0x000fc80003f06070 */ /*0070*/ ISETP.GE.U32.AND.EX P0, PT, RZ, UR4, PT, P0 ; /* 0x00000004ff007c0c */ /* 0x000fda000bf06100 */ /*0080*/ @P0 EXIT ; /* 0x000000000000094d */ /* 0x000fea0003800000 */ /*0090*/ HFMA2.MMA R3, -RZ, RZ, 0, 0 ; /* 0x00000000ff037435 */ /* 0x000fe200000001ff */ /*00a0*/ IMAD.MOV.U32 R5, RZ, RZ, c[0x0][0xc] ; /* 0x00000300ff057624 */ /* 0x000fe200078e00ff */ /*00b0*/ ULDC.64 UR6, c[0x0][0x118] ; /* 0x0000460000067ab9 */ /* 0x000fd00000000a00 */ /*00c0*/ IMAD R2, R3, R0.reuse, RZ ; /* 0x0000000003027224 */ /* 0x080fe200078e02ff */ /*00d0*/ YIELD ; /* 0x0000000000007946 */ /* 0x000fe20003800000 */ /*00e0*/ IMAD.WIDE.U32 R6, R0.reuse, R0, RZ ; /* 0x0000000000067225 */ /* 0x040fe200078e00ff */ /*00f0*/ BSSY B0, 0x300 ; /* 0x0000020000007945 */ /* 0x000fe60003800000 */ /*0100*/ IMAD R9, R0, R3, R2 ; /* 0x0000000300097224 */ /* 0x000fc800078e0202 */ /*0110*/ IMAD.IADD R8, R7, 0x1, R9 ; /* 0x0000000107087824 */ /* 0x000fca00078e0209 */ /*0120*/ LOP3.LUT R2, R8, UR4, RZ, 0xfc, !PT ; /* 0x0000000408027c12 */ /* 0x000fc8000f8efcff */ /*0130*/ ISETP.NE.U32.AND P0, PT, R2, RZ, PT ; /* 0x000000ff0200720c */ /* 0x000fda0003f05070 */ /*0140*/ @!P0 BRA 0x1c0 ; /* 0x0000007000008947 */ /* 0x000fea0003800000 */ /*0150*/ MOV R2, R6 ; /* 0x0000000600027202 */ /* 0x000fe20000000f00 */ /*0160*/ IMAD.MOV.U32 R6, RZ, RZ, c[0x0][0x164] ; /* 0x00005900ff067624 */ /* 0x000fe200078e00ff */ /*0170*/ MOV R4, 0x1a0 ; /* 0x000001a000047802 */ /* 0x000fe20000000f00 */ /*0180*/ IMAD.U32 R7, RZ, RZ, UR4 ; /* 0x00000004ff077e24 */ /* 0x000fe4000f8e00ff */ /*0190*/ CALL.REL.NOINC 0x6b0 ; /* 0x0000051000007944 */ /* 0x000fea0003c00000 */ /*01a0*/ MOV R8, R2 ; /* 0x0000000200087202 */ /* 0x000fe20000000f00 */ /*01b0*/ BRA 0x2f0 ; /* 0x0000013000007947 */ /* 0x000fea0003800000 */ /*01c0*/ I2F.U32.RP R2, c[0x0][0x164] ; /* 0x0000590000027b06 */ /* 0x000e220000209000 */ /*01d0*/ ISETP.NE.U32.AND P1, PT, RZ, c[0x0][0x164], PT ; /* 0x00005900ff007a0c */ /* 0x000fce0003f25070 */ /*01e0*/ MUFU.RCP R2, R2 ; /* 0x0000000200027308 */ /* 0x001e240000001000 */ /*01f0*/ IADD3 R8, R2, 0xffffffe, RZ ; /* 0x0ffffffe02087810 */ /* 0x001fcc0007ffe0ff */ /*0200*/ F2I.FTZ.U32.TRUNC.NTZ R9, R8 ; /* 0x0000000800097305 */ /* 0x000064000021f000 */ /*0210*/ IMAD.MOV.U32 R8, RZ, RZ, RZ ; /* 0x000000ffff087224 */ /* 0x001fe400078e00ff */ /*0220*/ IMAD.MOV R7, RZ, RZ, -R9 ; /* 0x000000ffff077224 */ /* 0x002fc800078e0a09 */ /*0230*/ IMAD R7, R7, c[0x0][0x164], RZ ; /* 0x0000590007077a24 */ /* 0x000fc800078e02ff */ /*0240*/ IMAD.HI.U32 R9, R9, R7, R8 ; /* 0x0000000709097227 */ /* 0x000fcc00078e0008 */ /*0250*/ IMAD.HI.U32 R9, R9, R6, RZ ; /* 0x0000000609097227 */ /* 0x000fca00078e00ff */ /*0260*/ IADD3 R9, -R9, RZ, RZ ; /* 0x000000ff09097210 */ /* 0x000fca0007ffe1ff */ /*0270*/ IMAD R6, R9, c[0x0][0x164], R6 ; /* 0x0000590009067a24 */ /* 0x000fe400078e0206 */ /*0280*/ IMAD.MOV.U32 R9, RZ, RZ, RZ ; /* 0x000000ffff097224 */ /* 0x000fc600078e00ff */ /*0290*/ ISETP.GE.U32.AND P0, PT, R6, c[0x0][0x164], PT ; /* 0x0000590006007a0c */ /* 0x000fda0003f06070 */ /*02a0*/ @P0 IADD3 R6, R6, -c[0x0][0x164], RZ ; /* 0x8000590006060a10 */ /* 0x000fc80007ffe0ff */ /*02b0*/ ISETP.GE.U32.AND P0, PT, R6, c[0x0][0x164], PT ; /* 0x0000590006007a0c */ /* 0x000fda0003f06070 */ /*02c0*/ @P0 IADD3 R6, R6, -c[0x0][0x164], RZ ; /* 0x8000590006060a10 */ /* 0x000fe40007ffe0ff */ /*02d0*/ @!P1 LOP3.LUT R6, RZ, c[0x0][0x164], RZ, 0x33, !PT ; /* 0x00005900ff069a12 */ /* 0x000fca00078e33ff */ /*02e0*/ IMAD.MOV.U32 R8, RZ, RZ, R6 ; /* 0x000000ffff087224 */ /* 0x000fe400078e0006 */ /*02f0*/ BSYNC B0 ; /* 0x0000000000007941 */ /* 0x000fea0003800000 */ /*0300*/ IMAD R2, R9, R0.reuse, RZ ; /* 0x0000000009027224 */ /* 0x080fe200078e02ff */ /*0310*/ BSSY B0, 0x500 ; /* 0x000001e000007945 */ /* 0x000fe20003800000 */ /*0320*/ IMAD.WIDE.U32 R6, R8, R0, RZ ; /* 0x0000000008067225 */ /* 0x000fc800078e00ff */ /*0330*/ IMAD R9, R3, R8, R2 ; /* 0x0000000803097224 */ /* 0x000fca00078e0202 */ /*0340*/ IADD3 R8, R7, R9, RZ ; /* 0x0000000907087210 */ /* 0x000fc80007ffe0ff */ /*0350*/ LOP3.LUT R2, R8, UR4, RZ, 0xfc, !PT ; /* 0x0000000408027c12 */ /* 0x000fc8000f8efcff */ /*0360*/ ISETP.NE.U32.AND P0, PT, R2, RZ, PT ; /* 0x000000ff0200720c */ /* 0x000fda0003f05070 */ /*0370*/ @!P0 BRA 0x3e0 ; /* 0x0000006000008947 */ /* 0x000fea0003800000 */ /*0380*/ IMAD.MOV.U32 R2, RZ, RZ, R6 ; /* 0x000000ffff027224 */ /* 0x000fe200078e0006 */ /*0390*/ MOV R7, UR4 ; /* 0x0000000400077c02 */ /* 0x000fe20008000f00 */ /*03a0*/ IMAD.MOV.U32 R6, RZ, RZ, c[0x0][0x164] ; /* 0x00005900ff067624 */ /* 0x000fe200078e00ff */ /*03b0*/ MOV R4, 0x3d0 ; /* 0x000003d000047802 */ /* 0x000fe40000000f00 */ /*03c0*/ CALL.REL.NOINC 0x6b0 ; /* 0x000002e000007944 */ /* 0x000fea0003c00000 */ /*03d0*/ BRA 0x4f0 ; /* 0x0000011000007947 */ /* 0x000fea0003800000 */ /*03e0*/ I2F.U32.RP R4, c[0x0][0x164] ; /* 0x0000590000047b06 */ /* 0x000e220000209000 */ /*03f0*/ ISETP.NE.U32.AND P1, PT, RZ, c[0x0][0x164], PT ; /* 0x00005900ff007a0c */ /* 0x000fce0003f25070 */ /*0400*/ MUFU.RCP R4, R4 ; /* 0x0000000400047308 */ /* 0x001e240000001000 */ /*0410*/ IADD3 R8, R4, 0xffffffe, RZ ; /* 0x0ffffffe04087810 */ /* 0x001fcc0007ffe0ff */ /*0420*/ F2I.FTZ.U32.TRUNC.NTZ R9, R8 ; /* 0x0000000800097305 */ /* 0x000064000021f000 */ /*0430*/ IMAD.MOV.U32 R8, RZ, RZ, RZ ; /* 0x000000ffff087224 */ /* 0x001fe400078e00ff */ /*0440*/ IMAD.MOV R7, RZ, RZ, -R9 ; /* 0x000000ffff077224 */ /* 0x002fc800078e0a09 */ /*0450*/ IMAD R7, R7, c[0x0][0x164], RZ ; /* 0x0000590007077a24 */ /* 0x000fc800078e02ff */ /*0460*/ IMAD.HI.U32 R7, R9, R7, R8 ; /* 0x0000000709077227 */ /* 0x000fcc00078e0008 */ /*0470*/ IMAD.HI.U32 R2, R7, R6, RZ ; /* 0x0000000607027227 */ /* 0x000fca00078e00ff */ /*0480*/ IADD3 R7, -R2, RZ, RZ ; /* 0x000000ff02077210 */ /* 0x000fca0007ffe1ff */ /*0490*/ IMAD R2, R7, c[0x0][0x164], R6 ; /* 0x0000590007027a24 */ /* 0x000fca00078e0206 */ /*04a0*/ ISETP.GE.U32.AND P0, PT, R2, c[0x0][0x164], PT ; /* 0x0000590002007a0c */ /* 0x000fda0003f06070 */ /*04b0*/ @P0 IADD3 R2, R2, -c[0x0][0x164], RZ ; /* 0x8000590002020a10 */ /* 0x000fc80007ffe0ff */ /*04c0*/ ISETP.GE.U32.AND P0, PT, R2, c[0x0][0x164], PT ; /* 0x0000590002007a0c */ /* 0x000fda0003f06070 */ /*04d0*/ @P0 IADD3 R2, R2, -c[0x0][0x164], RZ ; /* 0x8000590002020a10 */ /* 0x000fe40007ffe0ff */ /*04e0*/ @!P1 LOP3.LUT R2, RZ, c[0x0][0x164], RZ, 0x33, !PT ; /* 0x00005900ff029a12 */ /* 0x000fe400078e33ff */ /*04f0*/ BSYNC B0 ; /* 0x0000000000007941 */ /* 0x000fea0003800000 */ /*0500*/ ISETP.NE.AND P0, PT, R2, c[0x0][0x160], PT ; /* 0x0000580002007a0c */ /* 0x000fe20003f05270 */ /*0510*/ BSSY B0, 0x650 ; /* 0x0000013000007945 */ /* 0x000fd80003800000 */ /*0520*/ @P0 BRA 0x640 ; /* 0x0000011000000947 */ /* 0x000fea0003800000 */ /*0530*/ S2R R7, SR_LANEID ; /* 0x0000000000077919 */ /* 0x000e220000000000 */ /*0540*/ VOTEU.ANY UR5, UPT, PT ; /* 0x0000000000057886 */ /* 0x000fe200038e0100 */ /*0550*/ IMAD.MOV.U32 R6, RZ, RZ, c[0x4][0x0] ; /* 0x01000000ff067624 */ /* 0x000fe200078e00ff */ /*0560*/ FLO.U32 R2, UR5 ; /* 0x0000000500027d00 */ /* 0x000e3000080e0000 */ /*0570*/ POPC R11, UR5 ; /* 0x00000005000b7d09 */ /* 0x000e620008000000 */ /*0580*/ ISETP.EQ.U32.AND P0, PT, R2, R7, PT ; /* 0x000000070200720c */ /* 0x001fe20003f02070 */ /*0590*/ IMAD.MOV.U32 R7, RZ, RZ, c[0x4][0x4] ; /* 0x01000100ff077624 */ /* 0x000fd800078e00ff */ /*05a0*/ @P0 ATOMG.E.ADD.STRONG.GPU PT, R7, [R6.64], R11 ; /* 0x0000000b060709a8 */ /* 0x002ea200081ee1c6 */ /*05b0*/ HFMA2.MMA R13, -RZ, RZ, 0, 4.76837158203125e-07 ; /* 0x00000008ff0d7435 */ /* 0x000fc600000001ff */ /*05c0*/ S2R R4, SR_LTMASK ; /* 0x0000000000047919 */ /* 0x000e240000003900 */ /*05d0*/ LOP3.LUT R4, R4, UR5, RZ, 0xc0, !PT ; /* 0x0000000504047c12 */ /* 0x001fc8000f8ec0ff */ /*05e0*/ POPC R9, R4 ; /* 0x0000000400097309 */ /* 0x000e220000000000 */ /*05f0*/ SHFL.IDX PT, R8, R7, R2, 0x1f ; /* 0x00001f0207087589 */ /* 0x00422400000e0000 */ /*0600*/ IMAD.MOV.U32 R2, RZ, RZ, R0 ; /* 0x000000ffff027224 */ /* 0x002fe400078e0000 */ /*0610*/ IMAD.IADD R8, R8, 0x1, R9 ; /* 0x0000000108087824 */ /* 0x001fc800078e0209 */ /*0620*/ IMAD.WIDE R8, R8, R13, c[0x4][0x8] ; /* 0x0100020008087625 */ /* 0x000fca00078e020d */ /*0630*/ STG.E.64 [R8.64], R2 ; /* 0x0000000208007986 */ /* 0x0001e4000c101b06 */ /*0640*/ BSYNC B0 ; /* 0x0000000000007941 */ /* 0x000fea0003800000 */ /*0650*/ LEA R0, P0, R5, R0, 0xa ; /* 0x0000000005007211 */ /* 0x000fc800078050ff */ /*0660*/ IADD3.X R3, RZ, R3, RZ, P0, !PT ; /* 0x00000003ff037210 */ /* 0x001fe400007fe4ff */ /*0670*/ ISETP.GE.U32.AND P0, PT, R0, c[0x0][0x164], PT ; /* 0x0000590000007a0c */ /* 0x000fc80003f06070 */ /*0680*/ ISETP.GE.U32.AND.EX P0, PT, R3, UR4, PT, P0 ; /* 0x0000000403007c0c */ /* 0x000fda000bf06100 */ /*0690*/ @!P0 BRA 0xc0 ; /* 0xfffffa2000008947 */ /* 0x000fea000383ffff */ /*06a0*/ EXIT ; /* 0x000000000000794d */ /* 0x000fea0003800000 */ /*06b0*/ I2F.U64.RP R9, R6 ; /* 0x0000000600097312 */ /* 0x000e300000309000 */ /*06c0*/ MUFU.RCP R9, R9 ; /* 0x0000000900097308 */ /* 0x001e240000001000 */ /*06d0*/ IADD3 R10, R9, 0x1ffffffe, RZ ; /* 0x1ffffffe090a7810 */ /* 0x001fcc0007ffe0ff */ /*06e0*/ F2I.U64.TRUNC R10, R10 ; /* 0x0000000a000a7311 */ /* 0x000e24000020d800 */ /*06f0*/ IMAD.WIDE.U32 R12, R10, R6, RZ ; /* 0x000000060a0c7225 */ /* 0x001fc800078e00ff */ /*0700*/ IMAD R13, R10, R7, R13 ; /* 0x000000070a0d7224 */ /* 0x000fe200078e020d */ /*0710*/ IADD3 R15, P0, RZ, -R12, RZ ; /* 0x8000000cff0f7210 */ /* 0x000fc60007f1e0ff */ /*0720*/ IMAD R13, R11, R6, R13 ; /* 0x000000060b0d7224 */ /* 0x000fe400078e020d */ /*0730*/ IMAD.HI.U32 R12, R10, R15, RZ ; /* 0x0000000f0a0c7227 */ /* 0x000fc800078e00ff */ /*0740*/ IMAD.X R17, RZ, RZ, ~R13, P0 ; /* 0x000000ffff117224 */ /* 0x000fe400000e0e0d */ /*0750*/ IMAD.MOV.U32 R13, RZ, RZ, R10 ; /* 0x000000ffff0d7224 */ /* 0x000fe400078e000a */ /*0760*/ IMAD R19, R11, R17.reuse, RZ ; /* 0x000000110b137224 */ /* 0x080fe400078e02ff */ /*0770*/ IMAD.WIDE.U32 R12, P0, R10, R17, R12 ; /* 0x000000110a0c7225 */ /* 0x000fc8000780000c */ /*0780*/ IMAD.HI.U32 R9, R11, R17, RZ ; /* 0x000000110b097227 */ /* 0x000fc800078e00ff */ /*0790*/ IMAD.HI.U32 R12, P1, R11, R15, R12 ; /* 0x0000000f0b0c7227 */ /* 0x000fe2000782000c */ /*07a0*/ IADD3.X R9, R9, R11, RZ, P0, !PT ; /* 0x0000000b09097210 */ /* 0x000fc800007fe4ff */ /*07b0*/ IADD3 R13, P2, R19, R12, RZ ; /* 0x0000000c130d7210 */ /* 0x000fc80007f5e0ff */ /*07c0*/ IADD3.X R9, RZ, RZ, R9, P2, P1 ; /* 0x000000ffff097210 */ /* 0x000fe200017e2409 */ /*07d0*/ IMAD.WIDE.U32 R10, R13, R6, RZ ; /* 0x000000060d0a7225 */ /* 0x000fc800078e00ff */ /*07e0*/ IMAD R11, R13, R7, R11 ; /* 0x000000070d0b7224 */ /* 0x000fe200078e020b */ /*07f0*/ IADD3 R15, P0, RZ, -R10, RZ ; /* 0x8000000aff0f7210 */ /* 0x000fc60007f1e0ff */ /*0800*/ IMAD R11, R9, R6, R11 ; /* 0x00000006090b7224 */ /* 0x000fe400078e020b */ /*0810*/ IMAD.HI.U32 R12, R13, R15, RZ ; /* 0x0000000f0d0c7227 */ /* 0x000fc800078e00ff */ /*0820*/ IMAD.X R10, RZ, RZ, ~R11, P0 ; /* 0x000000ffff0a7224 */ /* 0x000fe400000e0e0b */ /*0830*/ IMAD.MOV.U32 R11, RZ, RZ, RZ ; /* 0x000000ffff0b7224 */ /* 0x000fe400078e00ff */ /*0840*/ IMAD.WIDE.U32 R12, P0, R13, R10, R12 ; /* 0x0000000a0d0c7225 */ /* 0x000fc8000780000c */ /*0850*/ IMAD R14, R9.reuse, R10, RZ ; /* 0x0000000a090e7224 */ /* 0x040fe400078e02ff */ /*0860*/ IMAD.HI.U32 R13, P1, R9, R15, R12 ; /* 0x0000000f090d7227 */ /* 0x000fc8000782000c */ /*0870*/ IMAD.HI.U32 R10, R9, R10, RZ ; /* 0x0000000a090a7227 */ /* 0x000fe200078e00ff */ /*0880*/ IADD3 R13, P2, R14, R13, RZ ; /* 0x0000000d0e0d7210 */ /* 0x000fc80007f5e0ff */ /*0890*/ IADD3.X R9, R10, R9, RZ, P0, !PT ; /* 0x000000090a097210 */ /* 0x000fe200007fe4ff */ /*08a0*/ IMAD.HI.U32 R10, R13, R2, RZ ; /* 0x000000020d0a7227 */ /* 0x000fc600078e00ff */ /*08b0*/ IADD3.X R9, RZ, RZ, R9, P2, P1 ; /* 0x000000ffff097210 */ /* 0x000fc600017e2409 */ /*08c0*/ IMAD.WIDE.U32 R10, R13, R8, R10 ; /* 0x000000080d0a7225 */ /* 0x000fc800078e000a */ /*08d0*/ IMAD R13, R9.reuse, R8, RZ ; /* 0x00000008090d7224 */ /* 0x040fe400078e02ff */ /*08e0*/ IMAD.HI.U32 R10, P0, R9, R2, R10 ; /* 0x00000002090a7227 */ /* 0x000fc8000780000a */ /*08f0*/ IMAD.HI.U32 R9, R9, R8, RZ ; /* 0x0000000809097227 */ /* 0x000fe200078e00ff */ /*0900*/ IADD3 R13, P1, R13, R10, RZ ; /* 0x0000000a0d0d7210 */ /* 0x000fc80007f3e0ff */ /*0910*/ IADD3.X R9, R9, RZ, RZ, P0, !PT ; /* 0x000000ff09097210 */ /* 0x000fe200007fe4ff */ /*0920*/ IMAD.WIDE.U32 R10, R13, R6, RZ ; /* 0x000000060d0a7225 */ /* 0x000fc800078e00ff */ /*0930*/ IMAD.X R9, RZ, RZ, R9, P1 ; /* 0x000000ffff097224 */ /* 0x000fe200008e0609 */ /*0940*/ IADD3 R15, P1, -R10, R2, RZ ; /* 0x000000020a0f7210 */ /* 0x000fe20007f3e1ff */ /*0950*/ IMAD R13, R13, R7, R11 ; /* 0x000000070d0d7224 */ /* 0x000fc600078e020b */ /*0960*/ ISETP.GE.U32.AND P0, PT, R15, R6.reuse, PT ; /* 0x000000060f00720c */ /* 0x080fe20003f06070 */ /*0970*/ IMAD R9, R9, R6, R13 ; /* 0x0000000609097224 */ /* 0x000fca00078e020d */ /*0980*/ IADD3.X R2, ~R9, R8, RZ, P1, !PT ; /* 0x0000000809027210 */ /* 0x000fe40000ffe5ff */ /*0990*/ IADD3 R11, P1, R15, -R6, RZ ; /* 0x800000060f0b7210 */ /* 0x000fe40007f3e0ff */ /*09a0*/ ISETP.GE.U32.AND.EX P0, PT, R2, R7, PT, P0 ; /* 0x000000070200720c */ /* 0x000fc60003f06100 */ /*09b0*/ IMAD.X R8, R2, 0x1, ~R7, P1 ; /* 0x0000000102087824 */ /* 0x000fe200008e0e07 */ /*09c0*/ SEL R11, R11, R15, P0 ; /* 0x0000000f0b0b7207 */ /* 0x000fe40000000000 */ /*09d0*/ ISETP.NE.U32.AND P1, PT, R6, RZ, PT ; /* 0x000000ff0600720c */ /* 0x000fe40003f25070 */ /*09e0*/ SEL R8, R8, R2, P0 ; /* 0x0000000208087207 */ /* 0x000fe40000000000 */ /*09f0*/ ISETP.GE.U32.AND P0, PT, R11.reuse, R6.reuse, PT ; /* 0x000000060b00720c */ /* 0x0c0fe40003f06070 */ /*0a00*/ IADD3 R2, P2, R11, -R6, RZ ; /* 0x800000060b027210 */ /* 0x000fe20007f5e0ff */ /*0a10*/ IMAD.MOV.U32 R6, RZ, RZ, R4 ; /* 0x000000ffff067224 */ /* 0x000fe200078e0004 */ /*0a20*/ ISETP.GE.U32.AND.EX P0, PT, R8, R7, PT, P0 ; /* 0x000000070800720c */ /* 0x000fc40003f06100 */ /*0a30*/ IADD3.X R9, ~R7.reuse, R8.reuse, RZ, P2, !PT ; /* 0x0000000807097210 */ /* 0x0c0fe400017fe5ff */ /*0a40*/ ISETP.NE.AND.EX P1, PT, R7, RZ, PT, P1 ; /* 0x000000ff0700720c */ /* 0x000fe40003f25310 */ /*0a50*/ MOV R7, 0x0 ; /* 0x0000000000077802 */ /* 0x000fe40000000f00 */ /*0a60*/ SEL R2, R2, R11, P0 ; /* 0x0000000b02027207 */ /* 0x000fe40000000000 */ /*0a70*/ SEL R9, R9, R8, P0 ; /* 0x0000000809097207 */ /* 0x000fe40000000000 */ /*0a80*/ SEL R2, R2, 0xffffffff, P1 ; /* 0xffffffff02027807 */ /* 0x000fc40000800000 */ /*0a90*/ SEL R9, R9, 0xffffffff, P1 ; /* 0xffffffff09097807 */ /* 0x000fe20000800000 */ /*0aa0*/ RET.REL.NODEC R6 0x0 ; /* 0xfffff55006007950 */ /* 0x000fec0003c3ffff */ /*0ab0*/ BRA 0xab0; /* 0xfffffff000007947 */ /* 0x000fc0000383ffff */ /*0ac0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0ad0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0ae0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0af0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0b00*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0b10*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0b20*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0b30*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0b40*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0b50*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0b60*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0b70*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ ..........
You are a helpful coding assistant specialized in converting CUDA source code to CUDA host assembly.
//The max possible number of roots #define DEV_MAX_ROOTS_NUM 3 //Number of threads per block. #define NT 1024 //The array index in global memory. __device__ int devArrayIndex; //The answer array stored roots in global memory __device__ unsigned long long int devAnswer[DEV_MAX_ROOTS_NUM]; /** * Device kernel to find (a) cube root(s) for a given integer c with a modular n. * <P> * Called with a one-dimensional grid of one-dimensional blocks. * * @param c The mod cube (input). * @param n The module (input). * * @author Junan Zhao * @version 26-Nov-2018 */ extern "C" __global__ void modCubeRoot( int c, int n) { //Determine number of threads and this thread's m (test number). unsigned long long m = blockIdx.x*NT + threadIdx.x; unsigned long long size = gridDim.x*NT; for(; m<n; m+=size) //use loop to cover all range of n if n is a pretty large integer cannot covered by one round { unsigned long long temp = m*m; temp = temp%n; temp = temp*m; temp = temp%n; if(c==(int)temp) //once found a root { int oldIndex = atomicAdd(&devArrayIndex,1); devAnswer[oldIndex] = m; } } }
.file "tmpxft_00091e3e_00000000-6_ModCubeRoot.cudafe1.cpp" .text #APP #NO_APP .type _ZL26__cudaUnregisterBinaryUtilv, @function _ZL26__cudaUnregisterBinaryUtilv: .LFB2029: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaUnregisterFatBinary@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2029: .size _ZL26__cudaUnregisterBinaryUtilv, .-_ZL26__cudaUnregisterBinaryUtilv .globl _Z31__device_stub__Z11modCubeRootiiii .type _Z31__device_stub__Z11modCubeRootiiii, @function _Z31__device_stub__Z11modCubeRootiiii: .LFB2051: .cfi_startproc endbr64 subq $120, %rsp .cfi_def_cfa_offset 128 movl %edi, 12(%rsp) movl %esi, 8(%rsp) movq %fs:40, %rax movq %rax, 104(%rsp) xorl %eax, %eax leaq 12(%rsp), %rax movq %rax, 80(%rsp) leaq 8(%rsp), %rax movq %rax, 88(%rsp) movl $1, 32(%rsp) movl $1, 36(%rsp) movl $1, 40(%rsp) movl $1, 44(%rsp) movl $1, 48(%rsp) movl $1, 52(%rsp) leaq 24(%rsp), %rcx leaq 16(%rsp), %rdx leaq 44(%rsp), %rsi leaq 32(%rsp), %rdi call __cudaPopCallConfiguration@PLT testl %eax, %eax je .L7 .L3: movq 104(%rsp), %rax subq %fs:40, %rax jne .L8 addq $120, %rsp .cfi_remember_state .cfi_def_cfa_offset 8 ret .L7: .cfi_restore_state pushq 24(%rsp) .cfi_def_cfa_offset 136 pushq 24(%rsp) .cfi_def_cfa_offset 144 leaq 96(%rsp), %r9 movq 60(%rsp), %rcx movl 68(%rsp), %r8d movq 48(%rsp), %rsi movl 56(%rsp), %edx leaq modCubeRoot(%rip), %rdi call cudaLaunchKernel@PLT addq $16, %rsp .cfi_def_cfa_offset 128 jmp .L3 .L8: call __stack_chk_fail@PLT .cfi_endproc .LFE2051: .size _Z31__device_stub__Z11modCubeRootiiii, .-_Z31__device_stub__Z11modCubeRootiiii .globl modCubeRoot .type modCubeRoot, @function modCubeRoot: .LFB2052: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 call _Z31__device_stub__Z11modCubeRootiiii addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2052: .size modCubeRoot, .-modCubeRoot .section .rodata.str1.1,"aMS",@progbits,1 .LC0: .string "modCubeRoot" .LC1: .string "devArrayIndex" .LC2: .string "devAnswer" .text .type _ZL24__sti____cudaRegisterAllv, @function _ZL24__sti____cudaRegisterAllv: .LFB2054: .cfi_startproc endbr64 pushq %rbx .cfi_def_cfa_offset 16 .cfi_offset 3, -16 leaq _ZL15__fatDeviceText(%rip), %rdi call __cudaRegisterFatBinary@PLT movq %rax, %rbx movq %rax, _ZL20__cudaFatCubinHandle(%rip) pushq $0 .cfi_def_cfa_offset 24 pushq $0 .cfi_def_cfa_offset 32 pushq $0 .cfi_def_cfa_offset 40 pushq $0 .cfi_def_cfa_offset 48 movl $0, %r9d movl $-1, %r8d leaq .LC0(%rip), %rdx movq %rdx, %rcx leaq modCubeRoot(%rip), %rsi movq %rax, %rdi call __cudaRegisterFunction@PLT addq $32, %rsp .cfi_def_cfa_offset 16 pushq $0 .cfi_def_cfa_offset 24 pushq $0 .cfi_def_cfa_offset 32 movl $4, %r9d movl $0, %r8d leaq .LC1(%rip), %rdx movq %rdx, %rcx leaq _ZL13devArrayIndex(%rip), %rsi movq %rbx, %rdi call __cudaRegisterVar@PLT addq $16, %rsp .cfi_def_cfa_offset 16 pushq $0 .cfi_def_cfa_offset 24 pushq $0 .cfi_def_cfa_offset 32 movl $24, %r9d movl $0, %r8d leaq .LC2(%rip), %rdx movq %rdx, %rcx leaq _ZL9devAnswer(%rip), %rsi movq %rbx, %rdi call __cudaRegisterVar@PLT addq $16, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaRegisterFatBinaryEnd@PLT leaq _ZL26__cudaUnregisterBinaryUtilv(%rip), %rdi call atexit@PLT popq %rbx .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2054: .size _ZL24__sti____cudaRegisterAllv, .-_ZL24__sti____cudaRegisterAllv .section .init_array,"aw" .align 8 .quad _ZL24__sti____cudaRegisterAllv .section .nvFatBinSegment,"aw" .align 8 .type _ZL15__fatDeviceText, @object .size _ZL15__fatDeviceText, 24 _ZL15__fatDeviceText: .long 1180844977 .long 1 .quad fatbinData .quad 0 .local _ZL20__cudaFatCubinHandle .comm _ZL20__cudaFatCubinHandle,8,8 .local _ZL9devAnswer .comm _ZL9devAnswer,24,16 .local _ZL13devArrayIndex .comm _ZL13devArrayIndex,4,4 .ident "GCC: (Ubuntu 13.3.0-6ubuntu2~24.04) 13.3.0" .section .note.GNU-stack,"",@progbits .section .note.gnu.property,"a" .align 8 .long 1f - 0f .long 4f - 1f .long 5 0: .string "GNU" 1: .align 8 .long 0xc0000002 .long 3f - 2f 2: .long 0x3 3: .align 8 4:
You are a helpful coding assistant specialized in converting CUDA source code to HIP/AMD source code.
//The max possible number of roots #define DEV_MAX_ROOTS_NUM 3 //Number of threads per block. #define NT 1024 //The array index in global memory. __device__ int devArrayIndex; //The answer array stored roots in global memory __device__ unsigned long long int devAnswer[DEV_MAX_ROOTS_NUM]; /** * Device kernel to find (a) cube root(s) for a given integer c with a modular n. * <P> * Called with a one-dimensional grid of one-dimensional blocks. * * @param c The mod cube (input). * @param n The module (input). * * @author Junan Zhao * @version 26-Nov-2018 */ extern "C" __global__ void modCubeRoot( int c, int n) { //Determine number of threads and this thread's m (test number). unsigned long long m = blockIdx.x*NT + threadIdx.x; unsigned long long size = gridDim.x*NT; for(; m<n; m+=size) //use loop to cover all range of n if n is a pretty large integer cannot covered by one round { unsigned long long temp = m*m; temp = temp%n; temp = temp*m; temp = temp%n; if(c==(int)temp) //once found a root { int oldIndex = atomicAdd(&devArrayIndex,1); devAnswer[oldIndex] = m; } } }
#include <hip/hip_runtime.h> //The max possible number of roots #define DEV_MAX_ROOTS_NUM 3 //Number of threads per block. #define NT 1024 //The array index in global memory. __device__ int devArrayIndex; //The answer array stored roots in global memory __device__ unsigned long long int devAnswer[DEV_MAX_ROOTS_NUM]; /** * Device kernel to find (a) cube root(s) for a given integer c with a modular n. * <P> * Called with a one-dimensional grid of one-dimensional blocks. * * @param c The mod cube (input). * @param n The module (input). * * @author Junan Zhao * @version 26-Nov-2018 */ extern "C" __global__ void modCubeRoot( int c, int n) { //Determine number of threads and this thread's m (test number). unsigned long long m = blockIdx.x*NT + threadIdx.x; unsigned long long size = gridDim.x*NT; for(; m<n; m+=size) //use loop to cover all range of n if n is a pretty large integer cannot covered by one round { unsigned long long temp = m*m; temp = temp%n; temp = temp*m; temp = temp%n; if(c==(int)temp) //once found a root { int oldIndex = atomicAdd(&devArrayIndex,1); devAnswer[oldIndex] = m; } } }
You are a helpful coding assistant specialized in converting HIP source code to AMD device assembly.
#include <hip/hip_runtime.h> //The max possible number of roots #define DEV_MAX_ROOTS_NUM 3 //Number of threads per block. #define NT 1024 //The array index in global memory. __device__ int devArrayIndex; //The answer array stored roots in global memory __device__ unsigned long long int devAnswer[DEV_MAX_ROOTS_NUM]; /** * Device kernel to find (a) cube root(s) for a given integer c with a modular n. * <P> * Called with a one-dimensional grid of one-dimensional blocks. * * @param c The mod cube (input). * @param n The module (input). * * @author Junan Zhao * @version 26-Nov-2018 */ extern "C" __global__ void modCubeRoot( int c, int n) { //Determine number of threads and this thread's m (test number). unsigned long long m = blockIdx.x*NT + threadIdx.x; unsigned long long size = gridDim.x*NT; for(; m<n; m+=size) //use loop to cover all range of n if n is a pretty large integer cannot covered by one round { unsigned long long temp = m*m; temp = temp%n; temp = temp*m; temp = temp%n; if(c==(int)temp) //once found a root { int oldIndex = atomicAdd(&devArrayIndex,1); devAnswer[oldIndex] = m; } } }
.text .amdgcn_target "amdgcn-amd-amdhsa--gfx1100" .protected modCubeRoot .globl modCubeRoot .p2align 8 .type modCubeRoot,@function modCubeRoot: s_load_b32 s2, s[0:1], 0x4 v_mov_b32_e32 v1, 0 v_lshl_or_b32 v3, s15, 10, v0 s_mov_b32 s4, exec_lo s_delay_alu instid0(VALU_DEP_2) v_mov_b32_e32 v4, v1 s_waitcnt lgkmcnt(0) s_ashr_i32 s3, s2, 31 s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1) v_cmpx_gt_u64_e64 s[2:3], v[3:4] s_cbranch_execz .LBB0_15 v_cvt_f32_u32_e32 v0, s2 v_cvt_f32_u32_e32 v2, s3 s_clause 0x1 s_load_b32 s4, s[0:1], 0x8 s_load_b32 s1, s[0:1], 0x0 s_mov_b32 s5, 0 s_sub_i32 s6, 0, s2 v_fmamk_f32 v0, v2, 0x4f800000, v0 s_mov_b32 s7, s5 s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(VALU_DEP_1) v_rcp_f32_e32 v0, v0 s_waitcnt_depctr 0xfff v_mul_f32_e32 v0, 0x5f7ffffc, v0 s_waitcnt lgkmcnt(0) s_lshl_b32 s4, s4, 10 v_mul_f32_e32 v2, 0x2f800000, v0 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_trunc_f32_e32 v2, v2 v_fmamk_f32 v0, v2, 0xcf800000, v0 v_cvt_u32_f32_e32 v9, v2 s_delay_alu instid0(VALU_DEP_2) v_cvt_u32_f32_e32 v0, v0 s_branch .LBB0_4 .LBB0_2: s_or_b32 exec_lo, exec_lo, s8 s_waitcnt vmcnt(0) v_readfirstlane_b32 s8, v5 s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_1) v_add_nc_u32_e32 v5, s8, v2 s_getpc_b64 s[8:9] s_add_u32 s8, s8, devAnswer@rel32@lo+4 s_addc_u32 s9, s9, devAnswer@rel32@hi+12 v_ashrrev_i32_e32 v6, 31, v5 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_lshlrev_b64 v[5:6], 3, v[5:6] v_add_co_u32 v5, vcc_lo, v5, s8 s_delay_alu instid0(VALU_DEP_2) v_add_co_ci_u32_e32 v6, vcc_lo, s9, v6, vcc_lo global_store_b64 v[5:6], v[3:4], off .LBB0_3: s_or_b32 exec_lo, exec_lo, s0 v_add_co_u32 v3, vcc_lo, v3, s4 v_add_co_ci_u32_e32 v4, vcc_lo, s5, v4, vcc_lo s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1) v_cmp_le_u64_e32 vcc_lo, s[2:3], v[3:4] s_or_b32 s7, vcc_lo, s7 s_and_not1_b32 exec_lo, exec_lo, s7 s_cbranch_execz .LBB0_15 .LBB0_4: v_mul_lo_u32 v2, v3, v4 v_mad_u64_u32 v[5:6], null, v3, v3, 0 s_mov_b32 s0, exec_lo s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_add3_u32 v6, v6, v2, v2 v_or_b32_e32 v2, s3, v6 s_delay_alu instid0(VALU_DEP_1) v_cmpx_ne_u64_e32 0, v[1:2] s_xor_b32 s8, exec_lo, s0 s_cbranch_execz .LBB0_6 s_sub_u32 s0, 0, s2 s_subb_u32 s9, 0, s3 v_mul_hi_u32 v2, s0, v0 v_mul_lo_u32 v7, s0, v9 v_mul_lo_u32 v8, s9, v0 s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2) v_add_nc_u32_e32 v2, v2, v7 v_mul_lo_u32 v7, s0, v0 v_add_nc_u32_e32 v2, v2, v8 s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_mul_hi_u32 v8, v0, v7 v_mul_lo_u32 v10, v0, v2 v_mul_hi_u32 v11, v0, v2 v_mul_hi_u32 v12, v9, v7 v_mul_lo_u32 v7, v9, v7 v_mul_hi_u32 v13, v9, v2 v_mul_lo_u32 v2, v9, v2 v_add_co_u32 v8, vcc_lo, v8, v10 v_add_co_ci_u32_e32 v10, vcc_lo, 0, v11, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_add_co_u32 v7, vcc_lo, v8, v7 v_add_co_ci_u32_e32 v7, vcc_lo, v10, v12, vcc_lo v_add_co_ci_u32_e32 v8, vcc_lo, 0, v13, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_add_co_u32 v2, vcc_lo, v7, v2 v_add_co_ci_u32_e32 v7, vcc_lo, 0, v8, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_add_co_u32 v2, vcc_lo, v0, v2 v_add_co_ci_u32_e32 v7, vcc_lo, v9, v7, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_3) v_mul_hi_u32 v8, s0, v2 v_mul_lo_u32 v11, s9, v2 v_mul_lo_u32 v10, s0, v7 s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2) v_add_nc_u32_e32 v8, v8, v10 v_mul_lo_u32 v10, s0, v2 v_add_nc_u32_e32 v8, v8, v11 s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_mul_hi_u32 v11, v2, v10 v_mul_lo_u32 v12, v2, v8 v_mul_hi_u32 v13, v2, v8 v_mul_hi_u32 v14, v7, v10 v_mul_lo_u32 v10, v7, v10 v_mul_hi_u32 v15, v7, v8 v_mul_lo_u32 v8, v7, v8 v_add_co_u32 v11, vcc_lo, v11, v12 v_add_co_ci_u32_e32 v12, vcc_lo, 0, v13, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_add_co_u32 v10, vcc_lo, v11, v10 v_add_co_ci_u32_e32 v10, vcc_lo, v12, v14, vcc_lo v_add_co_ci_u32_e32 v11, vcc_lo, 0, v15, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_add_co_u32 v8, vcc_lo, v10, v8 v_add_co_ci_u32_e32 v10, vcc_lo, 0, v11, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_add_co_u32 v2, vcc_lo, v2, v8 v_add_co_ci_u32_e32 v14, vcc_lo, v7, v10, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_3) v_mul_hi_u32 v15, v5, v2 v_mad_u64_u32 v[10:11], null, v6, v2, 0 v_mad_u64_u32 v[7:8], null, v5, v14, 0 v_mad_u64_u32 v[12:13], null, v6, v14, 0 s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3) v_add_co_u32 v2, vcc_lo, v15, v7 v_add_co_ci_u32_e32 v7, vcc_lo, 0, v8, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_add_co_u32 v2, vcc_lo, v2, v10 v_add_co_ci_u32_e32 v2, vcc_lo, v7, v11, vcc_lo v_add_co_ci_u32_e32 v7, vcc_lo, 0, v13, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_add_co_u32 v2, vcc_lo, v2, v12 v_add_co_ci_u32_e32 v10, vcc_lo, 0, v7, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_3) v_mul_lo_u32 v11, s3, v2 v_mad_u64_u32 v[7:8], null, s2, v2, 0 v_mul_lo_u32 v2, s2, v10 s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_sub_co_u32 v5, vcc_lo, v5, v7 v_add3_u32 v2, v8, v2, v11 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_sub_nc_u32_e32 v8, v6, v2 v_subrev_co_ci_u32_e64 v7, s0, s3, v8, vcc_lo v_sub_co_ci_u32_e32 v2, vcc_lo, v6, v2, vcc_lo v_sub_co_u32 v6, vcc_lo, v5, s2 s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_3) v_subrev_co_ci_u32_e64 v8, s0, 0, v7, vcc_lo v_cmp_le_u32_e64 s0, s2, v5 v_subrev_co_ci_u32_e32 v7, vcc_lo, s3, v7, vcc_lo v_cmp_le_u32_e32 vcc_lo, s3, v2 v_cndmask_b32_e64 v10, 0, -1, s0 v_cmp_le_u32_e64 s0, s2, v6 v_cndmask_b32_e64 v13, 0, -1, vcc_lo v_cmp_eq_u32_e32 vcc_lo, s3, v8 s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_1) v_cndmask_b32_e64 v11, 0, -1, s0 v_cmp_le_u32_e64 s0, s3, v8 v_cndmask_b32_e64 v12, 0, -1, s0 v_cmp_eq_u32_e64 s0, s3, v2 s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_3) v_cndmask_b32_e32 v11, v12, v11, vcc_lo v_sub_co_u32 v12, vcc_lo, v6, s2 v_subrev_co_ci_u32_e32 v7, vcc_lo, 0, v7, vcc_lo v_cmp_ne_u32_e32 vcc_lo, 0, v11 v_cndmask_b32_e64 v10, v13, v10, s0 s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2) v_dual_cndmask_b32 v7, v8, v7 :: v_dual_cndmask_b32 v6, v6, v12 v_cmp_ne_u32_e32 vcc_lo, 0, v10 s_delay_alu instid0(VALU_DEP_2) v_dual_cndmask_b32 v8, v2, v7 :: v_dual_cndmask_b32 v7, v5, v6 .LBB0_6: s_or_saveexec_b32 s0, s8 v_cvt_f32_u32_e32 v10, s2 s_xor_b32 exec_lo, exec_lo, s0 s_cbranch_execz .LBB0_8 s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_1) v_rcp_iflag_f32_e32 v2, v10 v_mov_b32_e32 v8, v1 s_waitcnt_depctr 0xfff v_mul_f32_e32 v2, 0x4f7ffffe, v2 v_cvt_u32_f32_e32 v2, v2 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_mul_lo_u32 v6, s6, v2 v_mul_hi_u32 v6, v2, v6 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_add_nc_u32_e32 v2, v2, v6 v_mul_hi_u32 v2, v5, v2 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_mul_lo_u32 v2, v2, s2 v_sub_nc_u32_e32 v2, v5, v2 s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2) v_subrev_nc_u32_e32 v5, s2, v2 v_cmp_le_u32_e32 vcc_lo, s2, v2 v_cndmask_b32_e32 v2, v2, v5, vcc_lo s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2) v_subrev_nc_u32_e32 v5, s2, v2 v_cmp_le_u32_e32 vcc_lo, s2, v2 v_cndmask_b32_e32 v7, v2, v5, vcc_lo .LBB0_8: s_or_b32 exec_lo, exec_lo, s0 v_mul_lo_u32 v2, v8, v3 s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1) v_mul_lo_u32 v8, v7, v4 v_mad_u64_u32 v[5:6], null, v7, v3, 0 s_mov_b32 s0, exec_lo v_add3_u32 v6, v6, v8, v2 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_or_b32_e32 v2, s3, v6 v_cmpx_ne_u64_e32 0, v[1:2] s_xor_b32 s8, exec_lo, s0 s_cbranch_execz .LBB0_10 s_sub_u32 s0, 0, s2 s_subb_u32 s9, 0, s3 v_mul_hi_u32 v2, s0, v0 v_mul_lo_u32 v7, s0, v9 v_mul_lo_u32 v8, s9, v0 s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2) v_add_nc_u32_e32 v2, v2, v7 v_mul_lo_u32 v7, s0, v0 v_add_nc_u32_e32 v2, v2, v8 s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_mul_hi_u32 v8, v0, v7 v_mul_lo_u32 v10, v0, v2 v_mul_hi_u32 v11, v0, v2 v_mul_hi_u32 v12, v9, v7 v_mul_lo_u32 v7, v9, v7 v_mul_hi_u32 v13, v9, v2 v_mul_lo_u32 v2, v9, v2 v_add_co_u32 v8, vcc_lo, v8, v10 v_add_co_ci_u32_e32 v10, vcc_lo, 0, v11, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_add_co_u32 v7, vcc_lo, v8, v7 v_add_co_ci_u32_e32 v7, vcc_lo, v10, v12, vcc_lo v_add_co_ci_u32_e32 v8, vcc_lo, 0, v13, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_add_co_u32 v2, vcc_lo, v7, v2 v_add_co_ci_u32_e32 v7, vcc_lo, 0, v8, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_add_co_u32 v2, vcc_lo, v0, v2 v_add_co_ci_u32_e32 v7, vcc_lo, v9, v7, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_3) v_mul_hi_u32 v8, s0, v2 v_mul_lo_u32 v11, s9, v2 v_mul_lo_u32 v10, s0, v7 s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2) v_add_nc_u32_e32 v8, v8, v10 v_mul_lo_u32 v10, s0, v2 v_add_nc_u32_e32 v8, v8, v11 s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_mul_hi_u32 v11, v2, v10 v_mul_lo_u32 v12, v2, v8 v_mul_hi_u32 v13, v2, v8 v_mul_hi_u32 v14, v7, v10 v_mul_lo_u32 v10, v7, v10 v_mul_hi_u32 v15, v7, v8 v_mul_lo_u32 v8, v7, v8 v_add_co_u32 v11, vcc_lo, v11, v12 v_add_co_ci_u32_e32 v12, vcc_lo, 0, v13, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_add_co_u32 v10, vcc_lo, v11, v10 v_add_co_ci_u32_e32 v10, vcc_lo, v12, v14, vcc_lo v_add_co_ci_u32_e32 v11, vcc_lo, 0, v15, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_add_co_u32 v8, vcc_lo, v10, v8 v_add_co_ci_u32_e32 v10, vcc_lo, 0, v11, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_add_co_u32 v2, vcc_lo, v2, v8 v_add_co_ci_u32_e32 v14, vcc_lo, v7, v10, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_3) v_mul_hi_u32 v15, v5, v2 v_mad_u64_u32 v[10:11], null, v6, v2, 0 v_mad_u64_u32 v[7:8], null, v5, v14, 0 v_mad_u64_u32 v[12:13], null, v6, v14, 0 s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3) v_add_co_u32 v2, vcc_lo, v15, v7 v_add_co_ci_u32_e32 v7, vcc_lo, 0, v8, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_add_co_u32 v2, vcc_lo, v2, v10 v_add_co_ci_u32_e32 v2, vcc_lo, v7, v11, vcc_lo v_add_co_ci_u32_e32 v7, vcc_lo, 0, v13, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_add_co_u32 v2, vcc_lo, v2, v12 v_add_co_ci_u32_e32 v10, vcc_lo, 0, v7, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_3) v_mul_lo_u32 v11, s3, v2 v_mad_u64_u32 v[7:8], null, s2, v2, 0 v_mul_lo_u32 v2, s2, v10 s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_sub_co_u32 v5, vcc_lo, v5, v7 v_add3_u32 v2, v8, v2, v11 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_sub_nc_u32_e32 v8, v6, v2 v_subrev_co_ci_u32_e64 v7, s0, s3, v8, vcc_lo s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_1) | instid1(VALU_DEP_3) v_sub_co_u32 v8, s0, v5, s2 v_sub_co_ci_u32_e32 v2, vcc_lo, v6, v2, vcc_lo v_subrev_co_ci_u32_e64 v7, s0, 0, v7, s0 s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_3) v_cmp_le_u32_e32 vcc_lo, s2, v8 v_cndmask_b32_e64 v6, 0, -1, vcc_lo v_cmp_le_u32_e32 vcc_lo, s3, v7 v_cndmask_b32_e64 v10, 0, -1, vcc_lo v_cmp_le_u32_e32 vcc_lo, s2, v5 v_cndmask_b32_e64 v11, 0, -1, vcc_lo v_cmp_le_u32_e32 vcc_lo, s3, v2 v_cndmask_b32_e64 v12, 0, -1, vcc_lo v_cmp_eq_u32_e32 vcc_lo, s3, v7 v_cndmask_b32_e32 v6, v10, v6, vcc_lo v_sub_co_u32 v7, vcc_lo, v8, s2 v_cmp_eq_u32_e32 vcc_lo, s3, v2 v_cndmask_b32_e32 v2, v12, v11, vcc_lo s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4) v_cmp_ne_u32_e32 vcc_lo, 0, v6 v_cndmask_b32_e32 v6, v8, v7, vcc_lo s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2) v_cmp_ne_u32_e32 vcc_lo, 0, v2 v_cndmask_b32_e32 v7, v5, v6, vcc_lo .LBB0_10: s_and_not1_saveexec_b32 s0, s8 s_cbranch_execz .LBB0_12 v_rcp_iflag_f32_e32 v2, v10 s_waitcnt_depctr 0xfff v_mul_f32_e32 v2, 0x4f7ffffe, v2 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_cvt_u32_f32_e32 v2, v2 v_mul_lo_u32 v6, s6, v2 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_mul_hi_u32 v6, v2, v6 v_add_nc_u32_e32 v2, v2, v6 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_mul_hi_u32 v2, v5, v2 v_mul_lo_u32 v2, v2, s2 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_sub_nc_u32_e32 v2, v5, v2 v_subrev_nc_u32_e32 v5, s2, v2 v_cmp_le_u32_e32 vcc_lo, s2, v2 s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1) v_cndmask_b32_e32 v2, v2, v5, vcc_lo v_subrev_nc_u32_e32 v5, s2, v2 v_cmp_le_u32_e32 vcc_lo, s2, v2 s_delay_alu instid0(VALU_DEP_2) v_cndmask_b32_e32 v7, v2, v5, vcc_lo .LBB0_12: s_or_b32 exec_lo, exec_lo, s0 s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1) s_mov_b32 s0, exec_lo v_cmpx_eq_u32_e64 s1, v7 s_cbranch_execz .LBB0_3 s_mov_b32 s9, exec_lo s_mov_b32 s8, exec_lo v_mbcnt_lo_u32_b32 v2, s9, 0 s_delay_alu instid0(VALU_DEP_1) v_cmpx_eq_u32_e32 0, v2 s_cbranch_execz .LBB0_2 s_bcnt1_i32_b32 s9, s9 s_getpc_b64 s[10:11] s_add_u32 s10, s10, devArrayIndex@rel32@lo+4 s_addc_u32 s11, s11, devArrayIndex@rel32@hi+12 v_mov_b32_e32 v5, s9 global_atomic_add_u32 v5, v1, v5, s[10:11] glc s_branch .LBB0_2 .LBB0_15: s_nop 0 s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) s_endpgm .section .rodata,"a",@progbits .p2align 6, 0x0 .amdhsa_kernel modCubeRoot .amdhsa_group_segment_fixed_size 0 .amdhsa_private_segment_fixed_size 0 .amdhsa_kernarg_size 264 .amdhsa_user_sgpr_count 15 .amdhsa_user_sgpr_dispatch_ptr 0 .amdhsa_user_sgpr_queue_ptr 0 .amdhsa_user_sgpr_kernarg_segment_ptr 1 .amdhsa_user_sgpr_dispatch_id 0 .amdhsa_user_sgpr_private_segment_size 0 .amdhsa_wavefront_size32 1 .amdhsa_uses_dynamic_stack 0 .amdhsa_enable_private_segment 0 .amdhsa_system_sgpr_workgroup_id_x 1 .amdhsa_system_sgpr_workgroup_id_y 0 .amdhsa_system_sgpr_workgroup_id_z 0 .amdhsa_system_sgpr_workgroup_info 0 .amdhsa_system_vgpr_workitem_id 0 .amdhsa_next_free_vgpr 16 .amdhsa_next_free_sgpr 16 .amdhsa_float_round_mode_32 0 .amdhsa_float_round_mode_16_64 0 .amdhsa_float_denorm_mode_32 3 .amdhsa_float_denorm_mode_16_64 3 .amdhsa_dx10_clamp 1 .amdhsa_ieee_mode 1 .amdhsa_fp16_overflow 0 .amdhsa_workgroup_processor_mode 1 .amdhsa_memory_ordered 1 .amdhsa_forward_progress 0 .amdhsa_shared_vgpr_count 0 .amdhsa_exception_fp_ieee_invalid_op 0 .amdhsa_exception_fp_denorm_src 0 .amdhsa_exception_fp_ieee_div_zero 0 .amdhsa_exception_fp_ieee_overflow 0 .amdhsa_exception_fp_ieee_underflow 0 .amdhsa_exception_fp_ieee_inexact 0 .amdhsa_exception_int_div_zero 0 .end_amdhsa_kernel .text .Lfunc_end0: .size modCubeRoot, .Lfunc_end0-modCubeRoot .section .AMDGPU.csdata,"",@progbits .text .p2alignl 7, 3214868480 .fill 96, 4, 3214868480 .protected devArrayIndex .type devArrayIndex,@object .section .bss,"aw",@nobits .globl devArrayIndex .p2align 2, 0x0 devArrayIndex: .long 0 .size devArrayIndex, 4 .protected devAnswer .type devAnswer,@object .globl devAnswer .p2align 4, 0x0 devAnswer: .zero 24 .size devAnswer, 24 .type __hip_cuid_,@object .globl __hip_cuid_ __hip_cuid_: .byte 0 .size __hip_cuid_, 1 .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym devArrayIndex .addrsig_sym devAnswer .addrsig_sym __hip_cuid_ .amdgpu_metadata --- amdhsa.kernels: - .args: - .offset: 0 .size: 4 .value_kind: by_value - .offset: 4 .size: 4 .value_kind: by_value - .offset: 8 .size: 4 .value_kind: hidden_block_count_x - .offset: 12 .size: 4 .value_kind: hidden_block_count_y - .offset: 16 .size: 4 .value_kind: hidden_block_count_z - .offset: 20 .size: 2 .value_kind: hidden_group_size_x - .offset: 22 .size: 2 .value_kind: hidden_group_size_y - .offset: 24 .size: 2 .value_kind: hidden_group_size_z - .offset: 26 .size: 2 .value_kind: hidden_remainder_x - .offset: 28 .size: 2 .value_kind: hidden_remainder_y - .offset: 30 .size: 2 .value_kind: hidden_remainder_z - .offset: 48 .size: 8 .value_kind: hidden_global_offset_x - .offset: 56 .size: 8 .value_kind: hidden_global_offset_y - .offset: 64 .size: 8 .value_kind: hidden_global_offset_z - .offset: 72 .size: 2 .value_kind: hidden_grid_dims .group_segment_fixed_size: 0 .kernarg_segment_align: 8 .kernarg_segment_size: 264 .language: OpenCL C .language_version: - 2 - 0 .max_flat_workgroup_size: 1024 .name: modCubeRoot .private_segment_fixed_size: 0 .sgpr_count: 18 .sgpr_spill_count: 0 .symbol: modCubeRoot.kd .uniform_work_group_size: 1 .uses_dynamic_stack: false .vgpr_count: 16 .vgpr_spill_count: 0 .wavefront_size: 32 .workgroup_processor_mode: 1 amdhsa.target: amdgcn-amd-amdhsa--gfx1100 amdhsa.version: - 1 - 2 ... .end_amdgpu_metadata
You are a helpful coding assistant specialized in converting HIP source code to AMD host assembly.
#include <hip/hip_runtime.h> //The max possible number of roots #define DEV_MAX_ROOTS_NUM 3 //Number of threads per block. #define NT 1024 //The array index in global memory. __device__ int devArrayIndex; //The answer array stored roots in global memory __device__ unsigned long long int devAnswer[DEV_MAX_ROOTS_NUM]; /** * Device kernel to find (a) cube root(s) for a given integer c with a modular n. * <P> * Called with a one-dimensional grid of one-dimensional blocks. * * @param c The mod cube (input). * @param n The module (input). * * @author Junan Zhao * @version 26-Nov-2018 */ extern "C" __global__ void modCubeRoot( int c, int n) { //Determine number of threads and this thread's m (test number). unsigned long long m = blockIdx.x*NT + threadIdx.x; unsigned long long size = gridDim.x*NT; for(; m<n; m+=size) //use loop to cover all range of n if n is a pretty large integer cannot covered by one round { unsigned long long temp = m*m; temp = temp%n; temp = temp*m; temp = temp%n; if(c==(int)temp) //once found a root { int oldIndex = atomicAdd(&devArrayIndex,1); devAnswer[oldIndex] = m; } } }
.text .file "ModCubeRoot.hip" .globl __device_stub__modCubeRoot # -- Begin function __device_stub__modCubeRoot .p2align 4, 0x90 .type __device_stub__modCubeRoot,@function __device_stub__modCubeRoot: # @__device_stub__modCubeRoot .cfi_startproc # %bb.0: subq $88, %rsp .cfi_def_cfa_offset 96 movl %edi, 12(%rsp) movl %esi, 8(%rsp) leaq 12(%rsp), %rax movq %rax, 64(%rsp) leaq 8(%rsp), %rax movq %rax, 72(%rsp) leaq 48(%rsp), %rdi leaq 32(%rsp), %rsi leaq 24(%rsp), %rdx leaq 16(%rsp), %rcx callq __hipPopCallConfiguration movq 48(%rsp), %rsi movl 56(%rsp), %edx movq 32(%rsp), %rcx movl 40(%rsp), %r8d leaq 64(%rsp), %r9 movl $modCubeRoot, %edi pushq 16(%rsp) .cfi_adjust_cfa_offset 8 pushq 32(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $104, %rsp .cfi_adjust_cfa_offset -104 retq .Lfunc_end0: .size __device_stub__modCubeRoot, .Lfunc_end0-__device_stub__modCubeRoot .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_ctor .type __hip_module_ctor,@function __hip_module_ctor: # @__hip_module_ctor .cfi_startproc # %bb.0: pushq %rbx .cfi_def_cfa_offset 16 .cfi_offset %rbx, -16 cmpq $0, __hip_gpubin_handle(%rip) jne .LBB1_2 # %bb.1: movl $__hip_fatbin_wrapper, %edi callq __hipRegisterFatBinary movq %rax, __hip_gpubin_handle(%rip) .LBB1_2: movq __hip_gpubin_handle(%rip), %rbx subq $32, %rsp .cfi_adjust_cfa_offset 32 xorps %xmm0, %xmm0 movups %xmm0, 16(%rsp) movups %xmm0, (%rsp) movl $modCubeRoot, %esi movl $.L__unnamed_1, %edx movl $.L__unnamed_1, %ecx movq %rbx, %rdi movl $-1, %r8d xorl %r9d, %r9d callq __hipRegisterFunction addq $32, %rsp .cfi_adjust_cfa_offset -32 movl $devArrayIndex, %esi movl $.L__unnamed_2, %edx movl $.L__unnamed_2, %ecx movl $4, %r9d movq %rbx, %rdi xorl %r8d, %r8d pushq $0 .cfi_adjust_cfa_offset 8 pushq $0 .cfi_adjust_cfa_offset 8 callq __hipRegisterVar addq $16, %rsp .cfi_adjust_cfa_offset -16 movl $devAnswer, %esi movl $.L__unnamed_3, %edx movl $.L__unnamed_3, %ecx movl $24, %r9d movq %rbx, %rdi xorl %r8d, %r8d pushq $0 .cfi_adjust_cfa_offset 8 pushq $0 .cfi_adjust_cfa_offset 8 callq __hipRegisterVar addq $16, %rsp .cfi_adjust_cfa_offset -16 movl $__hip_module_dtor, %edi popq %rbx .cfi_def_cfa_offset 8 jmp atexit # TAILCALL .Lfunc_end1: .size __hip_module_ctor, .Lfunc_end1-__hip_module_ctor .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_dtor .type __hip_module_dtor,@function __hip_module_dtor: # @__hip_module_dtor .cfi_startproc # %bb.0: movq __hip_gpubin_handle(%rip), %rdi testq %rdi, %rdi je .LBB2_2 # %bb.1: pushq %rax .cfi_def_cfa_offset 16 callq __hipUnregisterFatBinary movq $0, __hip_gpubin_handle(%rip) addq $8, %rsp .cfi_def_cfa_offset 8 .LBB2_2: retq .Lfunc_end2: .size __hip_module_dtor, .Lfunc_end2-__hip_module_dtor .cfi_endproc # -- End function .type devArrayIndex,@object # @devArrayIndex .local devArrayIndex .comm devArrayIndex,4,4 .type devAnswer,@object # @devAnswer .local devAnswer .comm devAnswer,24,16 .type modCubeRoot,@object # @modCubeRoot .section .rodata,"a",@progbits .globl modCubeRoot .p2align 3, 0x0 modCubeRoot: .quad __device_stub__modCubeRoot .size modCubeRoot, 8 .type .L__unnamed_1,@object # @0 .section .rodata.str1.1,"aMS",@progbits,1 .L__unnamed_1: .asciz "modCubeRoot" .size .L__unnamed_1, 12 .type .L__unnamed_2,@object # @1 .L__unnamed_2: .asciz "devArrayIndex" .size .L__unnamed_2, 14 .type .L__unnamed_3,@object # @2 .L__unnamed_3: .asciz "devAnswer" .size .L__unnamed_3, 10 .type __hip_fatbin_wrapper,@object # @__hip_fatbin_wrapper .section .hipFatBinSegment,"a",@progbits .p2align 3, 0x0 __hip_fatbin_wrapper: .long 1212764230 # 0x48495046 .long 1 # 0x1 .quad __hip_fatbin .quad 0 .size __hip_fatbin_wrapper, 24 .type __hip_gpubin_handle,@object # @__hip_gpubin_handle .local __hip_gpubin_handle .comm __hip_gpubin_handle,8,8 .section .init_array,"aw",@init_array .p2align 3, 0x0 .quad __hip_module_ctor .type __hip_cuid_,@object # @__hip_cuid_ .bss .globl __hip_cuid_ __hip_cuid_: .byte 0 # 0x0 .size __hip_cuid_, 1 .section ".linker-options","e",@llvm_linker_options .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym __device_stub__modCubeRoot .addrsig_sym __hip_module_ctor .addrsig_sym __hip_module_dtor .addrsig_sym devArrayIndex .addrsig_sym devAnswer .addrsig_sym modCubeRoot .addrsig_sym __hip_fatbin .addrsig_sym __hip_fatbin_wrapper .addrsig_sym __hip_cuid_
You are a helpful coding assistant specialized in converting from CUDA device assembly to AMD device assembly.
code for sm_80 Function : modCubeRoot .headerflags @"EF_CUDA_TEXMODE_UNIFIED EF_CUDA_64BIT_ADDRESS EF_CUDA_SM80 EF_CUDA_VIRTUAL_SM(EF_CUDA_SM80)" /*0000*/ IMAD.MOV.U32 R1, RZ, RZ, c[0x0][0x28] ; /* 0x00000a00ff017624 */ /* 0x000fe400078e00ff */ /*0010*/ S2R R0, SR_CTAID.X ; /* 0x0000000000007919 */ /* 0x000e220000002500 */ /*0020*/ ULDC UR4, c[0x0][0x164] ; /* 0x0000590000047ab9 */ /* 0x000fe40000000800 */ /*0030*/ USHF.R.S32.HI UR4, URZ, 0x1f, UR4 ; /* 0x0000001f3f047899 */ /* 0x000fe20008011404 */ /*0040*/ S2R R3, SR_TID.X ; /* 0x0000000000037919 */ /* 0x000e240000002100 */ /*0050*/ IMAD R0, R0, 0x400, R3 ; /* 0x0000040000007824 */ /* 0x001fca00078e0203 */ /*0060*/ ISETP.GE.U32.AND P0, PT, R0, c[0x0][0x164], PT ; /* 0x0000590000007a0c */ /* 0x000fc80003f06070 */ /*0070*/ ISETP.GE.U32.AND.EX P0, PT, RZ, UR4, PT, P0 ; /* 0x00000004ff007c0c */ /* 0x000fda000bf06100 */ /*0080*/ @P0 EXIT ; /* 0x000000000000094d */ /* 0x000fea0003800000 */ /*0090*/ HFMA2.MMA R3, -RZ, RZ, 0, 0 ; /* 0x00000000ff037435 */ /* 0x000fe200000001ff */ /*00a0*/ IMAD.MOV.U32 R5, RZ, RZ, c[0x0][0xc] ; /* 0x00000300ff057624 */ /* 0x000fe200078e00ff */ /*00b0*/ ULDC.64 UR6, c[0x0][0x118] ; /* 0x0000460000067ab9 */ /* 0x000fd00000000a00 */ /*00c0*/ IMAD R2, R3, R0.reuse, RZ ; /* 0x0000000003027224 */ /* 0x080fe200078e02ff */ /*00d0*/ YIELD ; /* 0x0000000000007946 */ /* 0x000fe20003800000 */ /*00e0*/ IMAD.WIDE.U32 R6, R0.reuse, R0, RZ ; /* 0x0000000000067225 */ /* 0x040fe200078e00ff */ /*00f0*/ BSSY B0, 0x300 ; /* 0x0000020000007945 */ /* 0x000fe60003800000 */ /*0100*/ IMAD R9, R0, R3, R2 ; /* 0x0000000300097224 */ /* 0x000fc800078e0202 */ /*0110*/ IMAD.IADD R8, R7, 0x1, R9 ; /* 0x0000000107087824 */ /* 0x000fca00078e0209 */ /*0120*/ LOP3.LUT R2, R8, UR4, RZ, 0xfc, !PT ; /* 0x0000000408027c12 */ /* 0x000fc8000f8efcff */ /*0130*/ ISETP.NE.U32.AND P0, PT, R2, RZ, PT ; /* 0x000000ff0200720c */ /* 0x000fda0003f05070 */ /*0140*/ @!P0 BRA 0x1c0 ; /* 0x0000007000008947 */ /* 0x000fea0003800000 */ /*0150*/ MOV R2, R6 ; /* 0x0000000600027202 */ /* 0x000fe20000000f00 */ /*0160*/ IMAD.MOV.U32 R6, RZ, RZ, c[0x0][0x164] ; /* 0x00005900ff067624 */ /* 0x000fe200078e00ff */ /*0170*/ MOV R4, 0x1a0 ; /* 0x000001a000047802 */ /* 0x000fe20000000f00 */ /*0180*/ IMAD.U32 R7, RZ, RZ, UR4 ; /* 0x00000004ff077e24 */ /* 0x000fe4000f8e00ff */ /*0190*/ CALL.REL.NOINC 0x6b0 ; /* 0x0000051000007944 */ /* 0x000fea0003c00000 */ /*01a0*/ MOV R8, R2 ; /* 0x0000000200087202 */ /* 0x000fe20000000f00 */ /*01b0*/ BRA 0x2f0 ; /* 0x0000013000007947 */ /* 0x000fea0003800000 */ /*01c0*/ I2F.U32.RP R2, c[0x0][0x164] ; /* 0x0000590000027b06 */ /* 0x000e220000209000 */ /*01d0*/ ISETP.NE.U32.AND P1, PT, RZ, c[0x0][0x164], PT ; /* 0x00005900ff007a0c */ /* 0x000fce0003f25070 */ /*01e0*/ MUFU.RCP R2, R2 ; /* 0x0000000200027308 */ /* 0x001e240000001000 */ /*01f0*/ IADD3 R8, R2, 0xffffffe, RZ ; /* 0x0ffffffe02087810 */ /* 0x001fcc0007ffe0ff */ /*0200*/ F2I.FTZ.U32.TRUNC.NTZ R9, R8 ; /* 0x0000000800097305 */ /* 0x000064000021f000 */ /*0210*/ IMAD.MOV.U32 R8, RZ, RZ, RZ ; /* 0x000000ffff087224 */ /* 0x001fe400078e00ff */ /*0220*/ IMAD.MOV R7, RZ, RZ, -R9 ; /* 0x000000ffff077224 */ /* 0x002fc800078e0a09 */ /*0230*/ IMAD R7, R7, c[0x0][0x164], RZ ; /* 0x0000590007077a24 */ /* 0x000fc800078e02ff */ /*0240*/ IMAD.HI.U32 R9, R9, R7, R8 ; /* 0x0000000709097227 */ /* 0x000fcc00078e0008 */ /*0250*/ IMAD.HI.U32 R9, R9, R6, RZ ; /* 0x0000000609097227 */ /* 0x000fca00078e00ff */ /*0260*/ IADD3 R9, -R9, RZ, RZ ; /* 0x000000ff09097210 */ /* 0x000fca0007ffe1ff */ /*0270*/ IMAD R6, R9, c[0x0][0x164], R6 ; /* 0x0000590009067a24 */ /* 0x000fe400078e0206 */ /*0280*/ IMAD.MOV.U32 R9, RZ, RZ, RZ ; /* 0x000000ffff097224 */ /* 0x000fc600078e00ff */ /*0290*/ ISETP.GE.U32.AND P0, PT, R6, c[0x0][0x164], PT ; /* 0x0000590006007a0c */ /* 0x000fda0003f06070 */ /*02a0*/ @P0 IADD3 R6, R6, -c[0x0][0x164], RZ ; /* 0x8000590006060a10 */ /* 0x000fc80007ffe0ff */ /*02b0*/ ISETP.GE.U32.AND P0, PT, R6, c[0x0][0x164], PT ; /* 0x0000590006007a0c */ /* 0x000fda0003f06070 */ /*02c0*/ @P0 IADD3 R6, R6, -c[0x0][0x164], RZ ; /* 0x8000590006060a10 */ /* 0x000fe40007ffe0ff */ /*02d0*/ @!P1 LOP3.LUT R6, RZ, c[0x0][0x164], RZ, 0x33, !PT ; /* 0x00005900ff069a12 */ /* 0x000fca00078e33ff */ /*02e0*/ IMAD.MOV.U32 R8, RZ, RZ, R6 ; /* 0x000000ffff087224 */ /* 0x000fe400078e0006 */ /*02f0*/ BSYNC B0 ; /* 0x0000000000007941 */ /* 0x000fea0003800000 */ /*0300*/ IMAD R2, R9, R0.reuse, RZ ; /* 0x0000000009027224 */ /* 0x080fe200078e02ff */ /*0310*/ BSSY B0, 0x500 ; /* 0x000001e000007945 */ /* 0x000fe20003800000 */ /*0320*/ IMAD.WIDE.U32 R6, R8, R0, RZ ; /* 0x0000000008067225 */ /* 0x000fc800078e00ff */ /*0330*/ IMAD R9, R3, R8, R2 ; /* 0x0000000803097224 */ /* 0x000fca00078e0202 */ /*0340*/ IADD3 R8, R7, R9, RZ ; /* 0x0000000907087210 */ /* 0x000fc80007ffe0ff */ /*0350*/ LOP3.LUT R2, R8, UR4, RZ, 0xfc, !PT ; /* 0x0000000408027c12 */ /* 0x000fc8000f8efcff */ /*0360*/ ISETP.NE.U32.AND P0, PT, R2, RZ, PT ; /* 0x000000ff0200720c */ /* 0x000fda0003f05070 */ /*0370*/ @!P0 BRA 0x3e0 ; /* 0x0000006000008947 */ /* 0x000fea0003800000 */ /*0380*/ IMAD.MOV.U32 R2, RZ, RZ, R6 ; /* 0x000000ffff027224 */ /* 0x000fe200078e0006 */ /*0390*/ MOV R7, UR4 ; /* 0x0000000400077c02 */ /* 0x000fe20008000f00 */ /*03a0*/ IMAD.MOV.U32 R6, RZ, RZ, c[0x0][0x164] ; /* 0x00005900ff067624 */ /* 0x000fe200078e00ff */ /*03b0*/ MOV R4, 0x3d0 ; /* 0x000003d000047802 */ /* 0x000fe40000000f00 */ /*03c0*/ CALL.REL.NOINC 0x6b0 ; /* 0x000002e000007944 */ /* 0x000fea0003c00000 */ /*03d0*/ BRA 0x4f0 ; /* 0x0000011000007947 */ /* 0x000fea0003800000 */ /*03e0*/ I2F.U32.RP R4, c[0x0][0x164] ; /* 0x0000590000047b06 */ /* 0x000e220000209000 */ /*03f0*/ ISETP.NE.U32.AND P1, PT, RZ, c[0x0][0x164], PT ; /* 0x00005900ff007a0c */ /* 0x000fce0003f25070 */ /*0400*/ MUFU.RCP R4, R4 ; /* 0x0000000400047308 */ /* 0x001e240000001000 */ /*0410*/ IADD3 R8, R4, 0xffffffe, RZ ; /* 0x0ffffffe04087810 */ /* 0x001fcc0007ffe0ff */ /*0420*/ F2I.FTZ.U32.TRUNC.NTZ R9, R8 ; /* 0x0000000800097305 */ /* 0x000064000021f000 */ /*0430*/ IMAD.MOV.U32 R8, RZ, RZ, RZ ; /* 0x000000ffff087224 */ /* 0x001fe400078e00ff */ /*0440*/ IMAD.MOV R7, RZ, RZ, -R9 ; /* 0x000000ffff077224 */ /* 0x002fc800078e0a09 */ /*0450*/ IMAD R7, R7, c[0x0][0x164], RZ ; /* 0x0000590007077a24 */ /* 0x000fc800078e02ff */ /*0460*/ IMAD.HI.U32 R7, R9, R7, R8 ; /* 0x0000000709077227 */ /* 0x000fcc00078e0008 */ /*0470*/ IMAD.HI.U32 R2, R7, R6, RZ ; /* 0x0000000607027227 */ /* 0x000fca00078e00ff */ /*0480*/ IADD3 R7, -R2, RZ, RZ ; /* 0x000000ff02077210 */ /* 0x000fca0007ffe1ff */ /*0490*/ IMAD R2, R7, c[0x0][0x164], R6 ; /* 0x0000590007027a24 */ /* 0x000fca00078e0206 */ /*04a0*/ ISETP.GE.U32.AND P0, PT, R2, c[0x0][0x164], PT ; /* 0x0000590002007a0c */ /* 0x000fda0003f06070 */ /*04b0*/ @P0 IADD3 R2, R2, -c[0x0][0x164], RZ ; /* 0x8000590002020a10 */ /* 0x000fc80007ffe0ff */ /*04c0*/ ISETP.GE.U32.AND P0, PT, R2, c[0x0][0x164], PT ; /* 0x0000590002007a0c */ /* 0x000fda0003f06070 */ /*04d0*/ @P0 IADD3 R2, R2, -c[0x0][0x164], RZ ; /* 0x8000590002020a10 */ /* 0x000fe40007ffe0ff */ /*04e0*/ @!P1 LOP3.LUT R2, RZ, c[0x0][0x164], RZ, 0x33, !PT ; /* 0x00005900ff029a12 */ /* 0x000fe400078e33ff */ /*04f0*/ BSYNC B0 ; /* 0x0000000000007941 */ /* 0x000fea0003800000 */ /*0500*/ ISETP.NE.AND P0, PT, R2, c[0x0][0x160], PT ; /* 0x0000580002007a0c */ /* 0x000fe20003f05270 */ /*0510*/ BSSY B0, 0x650 ; /* 0x0000013000007945 */ /* 0x000fd80003800000 */ /*0520*/ @P0 BRA 0x640 ; /* 0x0000011000000947 */ /* 0x000fea0003800000 */ /*0530*/ S2R R7, SR_LANEID ; /* 0x0000000000077919 */ /* 0x000e220000000000 */ /*0540*/ VOTEU.ANY UR5, UPT, PT ; /* 0x0000000000057886 */ /* 0x000fe200038e0100 */ /*0550*/ IMAD.MOV.U32 R6, RZ, RZ, c[0x4][0x0] ; /* 0x01000000ff067624 */ /* 0x000fe200078e00ff */ /*0560*/ FLO.U32 R2, UR5 ; /* 0x0000000500027d00 */ /* 0x000e3000080e0000 */ /*0570*/ POPC R11, UR5 ; /* 0x00000005000b7d09 */ /* 0x000e620008000000 */ /*0580*/ ISETP.EQ.U32.AND P0, PT, R2, R7, PT ; /* 0x000000070200720c */ /* 0x001fe20003f02070 */ /*0590*/ IMAD.MOV.U32 R7, RZ, RZ, c[0x4][0x4] ; /* 0x01000100ff077624 */ /* 0x000fd800078e00ff */ /*05a0*/ @P0 ATOMG.E.ADD.STRONG.GPU PT, R7, [R6.64], R11 ; /* 0x0000000b060709a8 */ /* 0x002ea200081ee1c6 */ /*05b0*/ HFMA2.MMA R13, -RZ, RZ, 0, 4.76837158203125e-07 ; /* 0x00000008ff0d7435 */ /* 0x000fc600000001ff */ /*05c0*/ S2R R4, SR_LTMASK ; /* 0x0000000000047919 */ /* 0x000e240000003900 */ /*05d0*/ LOP3.LUT R4, R4, UR5, RZ, 0xc0, !PT ; /* 0x0000000504047c12 */ /* 0x001fc8000f8ec0ff */ /*05e0*/ POPC R9, R4 ; /* 0x0000000400097309 */ /* 0x000e220000000000 */ /*05f0*/ SHFL.IDX PT, R8, R7, R2, 0x1f ; /* 0x00001f0207087589 */ /* 0x00422400000e0000 */ /*0600*/ IMAD.MOV.U32 R2, RZ, RZ, R0 ; /* 0x000000ffff027224 */ /* 0x002fe400078e0000 */ /*0610*/ IMAD.IADD R8, R8, 0x1, R9 ; /* 0x0000000108087824 */ /* 0x001fc800078e0209 */ /*0620*/ IMAD.WIDE R8, R8, R13, c[0x4][0x8] ; /* 0x0100020008087625 */ /* 0x000fca00078e020d */ /*0630*/ STG.E.64 [R8.64], R2 ; /* 0x0000000208007986 */ /* 0x0001e4000c101b06 */ /*0640*/ BSYNC B0 ; /* 0x0000000000007941 */ /* 0x000fea0003800000 */ /*0650*/ LEA R0, P0, R5, R0, 0xa ; /* 0x0000000005007211 */ /* 0x000fc800078050ff */ /*0660*/ IADD3.X R3, RZ, R3, RZ, P0, !PT ; /* 0x00000003ff037210 */ /* 0x001fe400007fe4ff */ /*0670*/ ISETP.GE.U32.AND P0, PT, R0, c[0x0][0x164], PT ; /* 0x0000590000007a0c */ /* 0x000fc80003f06070 */ /*0680*/ ISETP.GE.U32.AND.EX P0, PT, R3, UR4, PT, P0 ; /* 0x0000000403007c0c */ /* 0x000fda000bf06100 */ /*0690*/ @!P0 BRA 0xc0 ; /* 0xfffffa2000008947 */ /* 0x000fea000383ffff */ /*06a0*/ EXIT ; /* 0x000000000000794d */ /* 0x000fea0003800000 */ /*06b0*/ I2F.U64.RP R9, R6 ; /* 0x0000000600097312 */ /* 0x000e300000309000 */ /*06c0*/ MUFU.RCP R9, R9 ; /* 0x0000000900097308 */ /* 0x001e240000001000 */ /*06d0*/ IADD3 R10, R9, 0x1ffffffe, RZ ; /* 0x1ffffffe090a7810 */ /* 0x001fcc0007ffe0ff */ /*06e0*/ F2I.U64.TRUNC R10, R10 ; /* 0x0000000a000a7311 */ /* 0x000e24000020d800 */ /*06f0*/ IMAD.WIDE.U32 R12, R10, R6, RZ ; /* 0x000000060a0c7225 */ /* 0x001fc800078e00ff */ /*0700*/ IMAD R13, R10, R7, R13 ; /* 0x000000070a0d7224 */ /* 0x000fe200078e020d */ /*0710*/ IADD3 R15, P0, RZ, -R12, RZ ; /* 0x8000000cff0f7210 */ /* 0x000fc60007f1e0ff */ /*0720*/ IMAD R13, R11, R6, R13 ; /* 0x000000060b0d7224 */ /* 0x000fe400078e020d */ /*0730*/ IMAD.HI.U32 R12, R10, R15, RZ ; /* 0x0000000f0a0c7227 */ /* 0x000fc800078e00ff */ /*0740*/ IMAD.X R17, RZ, RZ, ~R13, P0 ; /* 0x000000ffff117224 */ /* 0x000fe400000e0e0d */ /*0750*/ IMAD.MOV.U32 R13, RZ, RZ, R10 ; /* 0x000000ffff0d7224 */ /* 0x000fe400078e000a */ /*0760*/ IMAD R19, R11, R17.reuse, RZ ; /* 0x000000110b137224 */ /* 0x080fe400078e02ff */ /*0770*/ IMAD.WIDE.U32 R12, P0, R10, R17, R12 ; /* 0x000000110a0c7225 */ /* 0x000fc8000780000c */ /*0780*/ IMAD.HI.U32 R9, R11, R17, RZ ; /* 0x000000110b097227 */ /* 0x000fc800078e00ff */ /*0790*/ IMAD.HI.U32 R12, P1, R11, R15, R12 ; /* 0x0000000f0b0c7227 */ /* 0x000fe2000782000c */ /*07a0*/ IADD3.X R9, R9, R11, RZ, P0, !PT ; /* 0x0000000b09097210 */ /* 0x000fc800007fe4ff */ /*07b0*/ IADD3 R13, P2, R19, R12, RZ ; /* 0x0000000c130d7210 */ /* 0x000fc80007f5e0ff */ /*07c0*/ IADD3.X R9, RZ, RZ, R9, P2, P1 ; /* 0x000000ffff097210 */ /* 0x000fe200017e2409 */ /*07d0*/ IMAD.WIDE.U32 R10, R13, R6, RZ ; /* 0x000000060d0a7225 */ /* 0x000fc800078e00ff */ /*07e0*/ IMAD R11, R13, R7, R11 ; /* 0x000000070d0b7224 */ /* 0x000fe200078e020b */ /*07f0*/ IADD3 R15, P0, RZ, -R10, RZ ; /* 0x8000000aff0f7210 */ /* 0x000fc60007f1e0ff */ /*0800*/ IMAD R11, R9, R6, R11 ; /* 0x00000006090b7224 */ /* 0x000fe400078e020b */ /*0810*/ IMAD.HI.U32 R12, R13, R15, RZ ; /* 0x0000000f0d0c7227 */ /* 0x000fc800078e00ff */ /*0820*/ IMAD.X R10, RZ, RZ, ~R11, P0 ; /* 0x000000ffff0a7224 */ /* 0x000fe400000e0e0b */ /*0830*/ IMAD.MOV.U32 R11, RZ, RZ, RZ ; /* 0x000000ffff0b7224 */ /* 0x000fe400078e00ff */ /*0840*/ IMAD.WIDE.U32 R12, P0, R13, R10, R12 ; /* 0x0000000a0d0c7225 */ /* 0x000fc8000780000c */ /*0850*/ IMAD R14, R9.reuse, R10, RZ ; /* 0x0000000a090e7224 */ /* 0x040fe400078e02ff */ /*0860*/ IMAD.HI.U32 R13, P1, R9, R15, R12 ; /* 0x0000000f090d7227 */ /* 0x000fc8000782000c */ /*0870*/ IMAD.HI.U32 R10, R9, R10, RZ ; /* 0x0000000a090a7227 */ /* 0x000fe200078e00ff */ /*0880*/ IADD3 R13, P2, R14, R13, RZ ; /* 0x0000000d0e0d7210 */ /* 0x000fc80007f5e0ff */ /*0890*/ IADD3.X R9, R10, R9, RZ, P0, !PT ; /* 0x000000090a097210 */ /* 0x000fe200007fe4ff */ /*08a0*/ IMAD.HI.U32 R10, R13, R2, RZ ; /* 0x000000020d0a7227 */ /* 0x000fc600078e00ff */ /*08b0*/ IADD3.X R9, RZ, RZ, R9, P2, P1 ; /* 0x000000ffff097210 */ /* 0x000fc600017e2409 */ /*08c0*/ IMAD.WIDE.U32 R10, R13, R8, R10 ; /* 0x000000080d0a7225 */ /* 0x000fc800078e000a */ /*08d0*/ IMAD R13, R9.reuse, R8, RZ ; /* 0x00000008090d7224 */ /* 0x040fe400078e02ff */ /*08e0*/ IMAD.HI.U32 R10, P0, R9, R2, R10 ; /* 0x00000002090a7227 */ /* 0x000fc8000780000a */ /*08f0*/ IMAD.HI.U32 R9, R9, R8, RZ ; /* 0x0000000809097227 */ /* 0x000fe200078e00ff */ /*0900*/ IADD3 R13, P1, R13, R10, RZ ; /* 0x0000000a0d0d7210 */ /* 0x000fc80007f3e0ff */ /*0910*/ IADD3.X R9, R9, RZ, RZ, P0, !PT ; /* 0x000000ff09097210 */ /* 0x000fe200007fe4ff */ /*0920*/ IMAD.WIDE.U32 R10, R13, R6, RZ ; /* 0x000000060d0a7225 */ /* 0x000fc800078e00ff */ /*0930*/ IMAD.X R9, RZ, RZ, R9, P1 ; /* 0x000000ffff097224 */ /* 0x000fe200008e0609 */ /*0940*/ IADD3 R15, P1, -R10, R2, RZ ; /* 0x000000020a0f7210 */ /* 0x000fe20007f3e1ff */ /*0950*/ IMAD R13, R13, R7, R11 ; /* 0x000000070d0d7224 */ /* 0x000fc600078e020b */ /*0960*/ ISETP.GE.U32.AND P0, PT, R15, R6.reuse, PT ; /* 0x000000060f00720c */ /* 0x080fe20003f06070 */ /*0970*/ IMAD R9, R9, R6, R13 ; /* 0x0000000609097224 */ /* 0x000fca00078e020d */ /*0980*/ IADD3.X R2, ~R9, R8, RZ, P1, !PT ; /* 0x0000000809027210 */ /* 0x000fe40000ffe5ff */ /*0990*/ IADD3 R11, P1, R15, -R6, RZ ; /* 0x800000060f0b7210 */ /* 0x000fe40007f3e0ff */ /*09a0*/ ISETP.GE.U32.AND.EX P0, PT, R2, R7, PT, P0 ; /* 0x000000070200720c */ /* 0x000fc60003f06100 */ /*09b0*/ IMAD.X R8, R2, 0x1, ~R7, P1 ; /* 0x0000000102087824 */ /* 0x000fe200008e0e07 */ /*09c0*/ SEL R11, R11, R15, P0 ; /* 0x0000000f0b0b7207 */ /* 0x000fe40000000000 */ /*09d0*/ ISETP.NE.U32.AND P1, PT, R6, RZ, PT ; /* 0x000000ff0600720c */ /* 0x000fe40003f25070 */ /*09e0*/ SEL R8, R8, R2, P0 ; /* 0x0000000208087207 */ /* 0x000fe40000000000 */ /*09f0*/ ISETP.GE.U32.AND P0, PT, R11.reuse, R6.reuse, PT ; /* 0x000000060b00720c */ /* 0x0c0fe40003f06070 */ /*0a00*/ IADD3 R2, P2, R11, -R6, RZ ; /* 0x800000060b027210 */ /* 0x000fe20007f5e0ff */ /*0a10*/ IMAD.MOV.U32 R6, RZ, RZ, R4 ; /* 0x000000ffff067224 */ /* 0x000fe200078e0004 */ /*0a20*/ ISETP.GE.U32.AND.EX P0, PT, R8, R7, PT, P0 ; /* 0x000000070800720c */ /* 0x000fc40003f06100 */ /*0a30*/ IADD3.X R9, ~R7.reuse, R8.reuse, RZ, P2, !PT ; /* 0x0000000807097210 */ /* 0x0c0fe400017fe5ff */ /*0a40*/ ISETP.NE.AND.EX P1, PT, R7, RZ, PT, P1 ; /* 0x000000ff0700720c */ /* 0x000fe40003f25310 */ /*0a50*/ MOV R7, 0x0 ; /* 0x0000000000077802 */ /* 0x000fe40000000f00 */ /*0a60*/ SEL R2, R2, R11, P0 ; /* 0x0000000b02027207 */ /* 0x000fe40000000000 */ /*0a70*/ SEL R9, R9, R8, P0 ; /* 0x0000000809097207 */ /* 0x000fe40000000000 */ /*0a80*/ SEL R2, R2, 0xffffffff, P1 ; /* 0xffffffff02027807 */ /* 0x000fc40000800000 */ /*0a90*/ SEL R9, R9, 0xffffffff, P1 ; /* 0xffffffff09097807 */ /* 0x000fe20000800000 */ /*0aa0*/ RET.REL.NODEC R6 0x0 ; /* 0xfffff55006007950 */ /* 0x000fec0003c3ffff */ /*0ab0*/ BRA 0xab0; /* 0xfffffff000007947 */ /* 0x000fc0000383ffff */ /*0ac0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0ad0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0ae0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0af0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0b00*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0b10*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0b20*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0b30*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0b40*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0b50*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0b60*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0b70*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ ..........
.text .amdgcn_target "amdgcn-amd-amdhsa--gfx1100" .protected modCubeRoot .globl modCubeRoot .p2align 8 .type modCubeRoot,@function modCubeRoot: s_load_b32 s2, s[0:1], 0x4 v_mov_b32_e32 v1, 0 v_lshl_or_b32 v3, s15, 10, v0 s_mov_b32 s4, exec_lo s_delay_alu instid0(VALU_DEP_2) v_mov_b32_e32 v4, v1 s_waitcnt lgkmcnt(0) s_ashr_i32 s3, s2, 31 s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1) v_cmpx_gt_u64_e64 s[2:3], v[3:4] s_cbranch_execz .LBB0_15 v_cvt_f32_u32_e32 v0, s2 v_cvt_f32_u32_e32 v2, s3 s_clause 0x1 s_load_b32 s4, s[0:1], 0x8 s_load_b32 s1, s[0:1], 0x0 s_mov_b32 s5, 0 s_sub_i32 s6, 0, s2 v_fmamk_f32 v0, v2, 0x4f800000, v0 s_mov_b32 s7, s5 s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(VALU_DEP_1) v_rcp_f32_e32 v0, v0 s_waitcnt_depctr 0xfff v_mul_f32_e32 v0, 0x5f7ffffc, v0 s_waitcnt lgkmcnt(0) s_lshl_b32 s4, s4, 10 v_mul_f32_e32 v2, 0x2f800000, v0 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_trunc_f32_e32 v2, v2 v_fmamk_f32 v0, v2, 0xcf800000, v0 v_cvt_u32_f32_e32 v9, v2 s_delay_alu instid0(VALU_DEP_2) v_cvt_u32_f32_e32 v0, v0 s_branch .LBB0_4 .LBB0_2: s_or_b32 exec_lo, exec_lo, s8 s_waitcnt vmcnt(0) v_readfirstlane_b32 s8, v5 s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_1) v_add_nc_u32_e32 v5, s8, v2 s_getpc_b64 s[8:9] s_add_u32 s8, s8, devAnswer@rel32@lo+4 s_addc_u32 s9, s9, devAnswer@rel32@hi+12 v_ashrrev_i32_e32 v6, 31, v5 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_lshlrev_b64 v[5:6], 3, v[5:6] v_add_co_u32 v5, vcc_lo, v5, s8 s_delay_alu instid0(VALU_DEP_2) v_add_co_ci_u32_e32 v6, vcc_lo, s9, v6, vcc_lo global_store_b64 v[5:6], v[3:4], off .LBB0_3: s_or_b32 exec_lo, exec_lo, s0 v_add_co_u32 v3, vcc_lo, v3, s4 v_add_co_ci_u32_e32 v4, vcc_lo, s5, v4, vcc_lo s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1) v_cmp_le_u64_e32 vcc_lo, s[2:3], v[3:4] s_or_b32 s7, vcc_lo, s7 s_and_not1_b32 exec_lo, exec_lo, s7 s_cbranch_execz .LBB0_15 .LBB0_4: v_mul_lo_u32 v2, v3, v4 v_mad_u64_u32 v[5:6], null, v3, v3, 0 s_mov_b32 s0, exec_lo s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_add3_u32 v6, v6, v2, v2 v_or_b32_e32 v2, s3, v6 s_delay_alu instid0(VALU_DEP_1) v_cmpx_ne_u64_e32 0, v[1:2] s_xor_b32 s8, exec_lo, s0 s_cbranch_execz .LBB0_6 s_sub_u32 s0, 0, s2 s_subb_u32 s9, 0, s3 v_mul_hi_u32 v2, s0, v0 v_mul_lo_u32 v7, s0, v9 v_mul_lo_u32 v8, s9, v0 s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2) v_add_nc_u32_e32 v2, v2, v7 v_mul_lo_u32 v7, s0, v0 v_add_nc_u32_e32 v2, v2, v8 s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_mul_hi_u32 v8, v0, v7 v_mul_lo_u32 v10, v0, v2 v_mul_hi_u32 v11, v0, v2 v_mul_hi_u32 v12, v9, v7 v_mul_lo_u32 v7, v9, v7 v_mul_hi_u32 v13, v9, v2 v_mul_lo_u32 v2, v9, v2 v_add_co_u32 v8, vcc_lo, v8, v10 v_add_co_ci_u32_e32 v10, vcc_lo, 0, v11, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_add_co_u32 v7, vcc_lo, v8, v7 v_add_co_ci_u32_e32 v7, vcc_lo, v10, v12, vcc_lo v_add_co_ci_u32_e32 v8, vcc_lo, 0, v13, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_add_co_u32 v2, vcc_lo, v7, v2 v_add_co_ci_u32_e32 v7, vcc_lo, 0, v8, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_add_co_u32 v2, vcc_lo, v0, v2 v_add_co_ci_u32_e32 v7, vcc_lo, v9, v7, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_3) v_mul_hi_u32 v8, s0, v2 v_mul_lo_u32 v11, s9, v2 v_mul_lo_u32 v10, s0, v7 s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2) v_add_nc_u32_e32 v8, v8, v10 v_mul_lo_u32 v10, s0, v2 v_add_nc_u32_e32 v8, v8, v11 s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_mul_hi_u32 v11, v2, v10 v_mul_lo_u32 v12, v2, v8 v_mul_hi_u32 v13, v2, v8 v_mul_hi_u32 v14, v7, v10 v_mul_lo_u32 v10, v7, v10 v_mul_hi_u32 v15, v7, v8 v_mul_lo_u32 v8, v7, v8 v_add_co_u32 v11, vcc_lo, v11, v12 v_add_co_ci_u32_e32 v12, vcc_lo, 0, v13, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_add_co_u32 v10, vcc_lo, v11, v10 v_add_co_ci_u32_e32 v10, vcc_lo, v12, v14, vcc_lo v_add_co_ci_u32_e32 v11, vcc_lo, 0, v15, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_add_co_u32 v8, vcc_lo, v10, v8 v_add_co_ci_u32_e32 v10, vcc_lo, 0, v11, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_add_co_u32 v2, vcc_lo, v2, v8 v_add_co_ci_u32_e32 v14, vcc_lo, v7, v10, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_3) v_mul_hi_u32 v15, v5, v2 v_mad_u64_u32 v[10:11], null, v6, v2, 0 v_mad_u64_u32 v[7:8], null, v5, v14, 0 v_mad_u64_u32 v[12:13], null, v6, v14, 0 s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3) v_add_co_u32 v2, vcc_lo, v15, v7 v_add_co_ci_u32_e32 v7, vcc_lo, 0, v8, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_add_co_u32 v2, vcc_lo, v2, v10 v_add_co_ci_u32_e32 v2, vcc_lo, v7, v11, vcc_lo v_add_co_ci_u32_e32 v7, vcc_lo, 0, v13, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_add_co_u32 v2, vcc_lo, v2, v12 v_add_co_ci_u32_e32 v10, vcc_lo, 0, v7, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_3) v_mul_lo_u32 v11, s3, v2 v_mad_u64_u32 v[7:8], null, s2, v2, 0 v_mul_lo_u32 v2, s2, v10 s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_sub_co_u32 v5, vcc_lo, v5, v7 v_add3_u32 v2, v8, v2, v11 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_sub_nc_u32_e32 v8, v6, v2 v_subrev_co_ci_u32_e64 v7, s0, s3, v8, vcc_lo v_sub_co_ci_u32_e32 v2, vcc_lo, v6, v2, vcc_lo v_sub_co_u32 v6, vcc_lo, v5, s2 s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_3) v_subrev_co_ci_u32_e64 v8, s0, 0, v7, vcc_lo v_cmp_le_u32_e64 s0, s2, v5 v_subrev_co_ci_u32_e32 v7, vcc_lo, s3, v7, vcc_lo v_cmp_le_u32_e32 vcc_lo, s3, v2 v_cndmask_b32_e64 v10, 0, -1, s0 v_cmp_le_u32_e64 s0, s2, v6 v_cndmask_b32_e64 v13, 0, -1, vcc_lo v_cmp_eq_u32_e32 vcc_lo, s3, v8 s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_1) v_cndmask_b32_e64 v11, 0, -1, s0 v_cmp_le_u32_e64 s0, s3, v8 v_cndmask_b32_e64 v12, 0, -1, s0 v_cmp_eq_u32_e64 s0, s3, v2 s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_3) v_cndmask_b32_e32 v11, v12, v11, vcc_lo v_sub_co_u32 v12, vcc_lo, v6, s2 v_subrev_co_ci_u32_e32 v7, vcc_lo, 0, v7, vcc_lo v_cmp_ne_u32_e32 vcc_lo, 0, v11 v_cndmask_b32_e64 v10, v13, v10, s0 s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2) v_dual_cndmask_b32 v7, v8, v7 :: v_dual_cndmask_b32 v6, v6, v12 v_cmp_ne_u32_e32 vcc_lo, 0, v10 s_delay_alu instid0(VALU_DEP_2) v_dual_cndmask_b32 v8, v2, v7 :: v_dual_cndmask_b32 v7, v5, v6 .LBB0_6: s_or_saveexec_b32 s0, s8 v_cvt_f32_u32_e32 v10, s2 s_xor_b32 exec_lo, exec_lo, s0 s_cbranch_execz .LBB0_8 s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_1) v_rcp_iflag_f32_e32 v2, v10 v_mov_b32_e32 v8, v1 s_waitcnt_depctr 0xfff v_mul_f32_e32 v2, 0x4f7ffffe, v2 v_cvt_u32_f32_e32 v2, v2 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_mul_lo_u32 v6, s6, v2 v_mul_hi_u32 v6, v2, v6 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_add_nc_u32_e32 v2, v2, v6 v_mul_hi_u32 v2, v5, v2 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_mul_lo_u32 v2, v2, s2 v_sub_nc_u32_e32 v2, v5, v2 s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2) v_subrev_nc_u32_e32 v5, s2, v2 v_cmp_le_u32_e32 vcc_lo, s2, v2 v_cndmask_b32_e32 v2, v2, v5, vcc_lo s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2) v_subrev_nc_u32_e32 v5, s2, v2 v_cmp_le_u32_e32 vcc_lo, s2, v2 v_cndmask_b32_e32 v7, v2, v5, vcc_lo .LBB0_8: s_or_b32 exec_lo, exec_lo, s0 v_mul_lo_u32 v2, v8, v3 s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1) v_mul_lo_u32 v8, v7, v4 v_mad_u64_u32 v[5:6], null, v7, v3, 0 s_mov_b32 s0, exec_lo v_add3_u32 v6, v6, v8, v2 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_or_b32_e32 v2, s3, v6 v_cmpx_ne_u64_e32 0, v[1:2] s_xor_b32 s8, exec_lo, s0 s_cbranch_execz .LBB0_10 s_sub_u32 s0, 0, s2 s_subb_u32 s9, 0, s3 v_mul_hi_u32 v2, s0, v0 v_mul_lo_u32 v7, s0, v9 v_mul_lo_u32 v8, s9, v0 s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2) v_add_nc_u32_e32 v2, v2, v7 v_mul_lo_u32 v7, s0, v0 v_add_nc_u32_e32 v2, v2, v8 s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_mul_hi_u32 v8, v0, v7 v_mul_lo_u32 v10, v0, v2 v_mul_hi_u32 v11, v0, v2 v_mul_hi_u32 v12, v9, v7 v_mul_lo_u32 v7, v9, v7 v_mul_hi_u32 v13, v9, v2 v_mul_lo_u32 v2, v9, v2 v_add_co_u32 v8, vcc_lo, v8, v10 v_add_co_ci_u32_e32 v10, vcc_lo, 0, v11, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_add_co_u32 v7, vcc_lo, v8, v7 v_add_co_ci_u32_e32 v7, vcc_lo, v10, v12, vcc_lo v_add_co_ci_u32_e32 v8, vcc_lo, 0, v13, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_add_co_u32 v2, vcc_lo, v7, v2 v_add_co_ci_u32_e32 v7, vcc_lo, 0, v8, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_add_co_u32 v2, vcc_lo, v0, v2 v_add_co_ci_u32_e32 v7, vcc_lo, v9, v7, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_3) v_mul_hi_u32 v8, s0, v2 v_mul_lo_u32 v11, s9, v2 v_mul_lo_u32 v10, s0, v7 s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2) v_add_nc_u32_e32 v8, v8, v10 v_mul_lo_u32 v10, s0, v2 v_add_nc_u32_e32 v8, v8, v11 s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_mul_hi_u32 v11, v2, v10 v_mul_lo_u32 v12, v2, v8 v_mul_hi_u32 v13, v2, v8 v_mul_hi_u32 v14, v7, v10 v_mul_lo_u32 v10, v7, v10 v_mul_hi_u32 v15, v7, v8 v_mul_lo_u32 v8, v7, v8 v_add_co_u32 v11, vcc_lo, v11, v12 v_add_co_ci_u32_e32 v12, vcc_lo, 0, v13, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_add_co_u32 v10, vcc_lo, v11, v10 v_add_co_ci_u32_e32 v10, vcc_lo, v12, v14, vcc_lo v_add_co_ci_u32_e32 v11, vcc_lo, 0, v15, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_add_co_u32 v8, vcc_lo, v10, v8 v_add_co_ci_u32_e32 v10, vcc_lo, 0, v11, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_add_co_u32 v2, vcc_lo, v2, v8 v_add_co_ci_u32_e32 v14, vcc_lo, v7, v10, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_3) v_mul_hi_u32 v15, v5, v2 v_mad_u64_u32 v[10:11], null, v6, v2, 0 v_mad_u64_u32 v[7:8], null, v5, v14, 0 v_mad_u64_u32 v[12:13], null, v6, v14, 0 s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3) v_add_co_u32 v2, vcc_lo, v15, v7 v_add_co_ci_u32_e32 v7, vcc_lo, 0, v8, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_add_co_u32 v2, vcc_lo, v2, v10 v_add_co_ci_u32_e32 v2, vcc_lo, v7, v11, vcc_lo v_add_co_ci_u32_e32 v7, vcc_lo, 0, v13, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_add_co_u32 v2, vcc_lo, v2, v12 v_add_co_ci_u32_e32 v10, vcc_lo, 0, v7, vcc_lo s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_3) v_mul_lo_u32 v11, s3, v2 v_mad_u64_u32 v[7:8], null, s2, v2, 0 v_mul_lo_u32 v2, s2, v10 s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_sub_co_u32 v5, vcc_lo, v5, v7 v_add3_u32 v2, v8, v2, v11 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_sub_nc_u32_e32 v8, v6, v2 v_subrev_co_ci_u32_e64 v7, s0, s3, v8, vcc_lo s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_1) | instid1(VALU_DEP_3) v_sub_co_u32 v8, s0, v5, s2 v_sub_co_ci_u32_e32 v2, vcc_lo, v6, v2, vcc_lo v_subrev_co_ci_u32_e64 v7, s0, 0, v7, s0 s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_3) v_cmp_le_u32_e32 vcc_lo, s2, v8 v_cndmask_b32_e64 v6, 0, -1, vcc_lo v_cmp_le_u32_e32 vcc_lo, s3, v7 v_cndmask_b32_e64 v10, 0, -1, vcc_lo v_cmp_le_u32_e32 vcc_lo, s2, v5 v_cndmask_b32_e64 v11, 0, -1, vcc_lo v_cmp_le_u32_e32 vcc_lo, s3, v2 v_cndmask_b32_e64 v12, 0, -1, vcc_lo v_cmp_eq_u32_e32 vcc_lo, s3, v7 v_cndmask_b32_e32 v6, v10, v6, vcc_lo v_sub_co_u32 v7, vcc_lo, v8, s2 v_cmp_eq_u32_e32 vcc_lo, s3, v2 v_cndmask_b32_e32 v2, v12, v11, vcc_lo s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4) v_cmp_ne_u32_e32 vcc_lo, 0, v6 v_cndmask_b32_e32 v6, v8, v7, vcc_lo s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2) v_cmp_ne_u32_e32 vcc_lo, 0, v2 v_cndmask_b32_e32 v7, v5, v6, vcc_lo .LBB0_10: s_and_not1_saveexec_b32 s0, s8 s_cbranch_execz .LBB0_12 v_rcp_iflag_f32_e32 v2, v10 s_waitcnt_depctr 0xfff v_mul_f32_e32 v2, 0x4f7ffffe, v2 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_cvt_u32_f32_e32 v2, v2 v_mul_lo_u32 v6, s6, v2 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_mul_hi_u32 v6, v2, v6 v_add_nc_u32_e32 v2, v2, v6 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_mul_hi_u32 v2, v5, v2 v_mul_lo_u32 v2, v2, s2 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_sub_nc_u32_e32 v2, v5, v2 v_subrev_nc_u32_e32 v5, s2, v2 v_cmp_le_u32_e32 vcc_lo, s2, v2 s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1) v_cndmask_b32_e32 v2, v2, v5, vcc_lo v_subrev_nc_u32_e32 v5, s2, v2 v_cmp_le_u32_e32 vcc_lo, s2, v2 s_delay_alu instid0(VALU_DEP_2) v_cndmask_b32_e32 v7, v2, v5, vcc_lo .LBB0_12: s_or_b32 exec_lo, exec_lo, s0 s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1) s_mov_b32 s0, exec_lo v_cmpx_eq_u32_e64 s1, v7 s_cbranch_execz .LBB0_3 s_mov_b32 s9, exec_lo s_mov_b32 s8, exec_lo v_mbcnt_lo_u32_b32 v2, s9, 0 s_delay_alu instid0(VALU_DEP_1) v_cmpx_eq_u32_e32 0, v2 s_cbranch_execz .LBB0_2 s_bcnt1_i32_b32 s9, s9 s_getpc_b64 s[10:11] s_add_u32 s10, s10, devArrayIndex@rel32@lo+4 s_addc_u32 s11, s11, devArrayIndex@rel32@hi+12 v_mov_b32_e32 v5, s9 global_atomic_add_u32 v5, v1, v5, s[10:11] glc s_branch .LBB0_2 .LBB0_15: s_nop 0 s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) s_endpgm .section .rodata,"a",@progbits .p2align 6, 0x0 .amdhsa_kernel modCubeRoot .amdhsa_group_segment_fixed_size 0 .amdhsa_private_segment_fixed_size 0 .amdhsa_kernarg_size 264 .amdhsa_user_sgpr_count 15 .amdhsa_user_sgpr_dispatch_ptr 0 .amdhsa_user_sgpr_queue_ptr 0 .amdhsa_user_sgpr_kernarg_segment_ptr 1 .amdhsa_user_sgpr_dispatch_id 0 .amdhsa_user_sgpr_private_segment_size 0 .amdhsa_wavefront_size32 1 .amdhsa_uses_dynamic_stack 0 .amdhsa_enable_private_segment 0 .amdhsa_system_sgpr_workgroup_id_x 1 .amdhsa_system_sgpr_workgroup_id_y 0 .amdhsa_system_sgpr_workgroup_id_z 0 .amdhsa_system_sgpr_workgroup_info 0 .amdhsa_system_vgpr_workitem_id 0 .amdhsa_next_free_vgpr 16 .amdhsa_next_free_sgpr 16 .amdhsa_float_round_mode_32 0 .amdhsa_float_round_mode_16_64 0 .amdhsa_float_denorm_mode_32 3 .amdhsa_float_denorm_mode_16_64 3 .amdhsa_dx10_clamp 1 .amdhsa_ieee_mode 1 .amdhsa_fp16_overflow 0 .amdhsa_workgroup_processor_mode 1 .amdhsa_memory_ordered 1 .amdhsa_forward_progress 0 .amdhsa_shared_vgpr_count 0 .amdhsa_exception_fp_ieee_invalid_op 0 .amdhsa_exception_fp_denorm_src 0 .amdhsa_exception_fp_ieee_div_zero 0 .amdhsa_exception_fp_ieee_overflow 0 .amdhsa_exception_fp_ieee_underflow 0 .amdhsa_exception_fp_ieee_inexact 0 .amdhsa_exception_int_div_zero 0 .end_amdhsa_kernel .text .Lfunc_end0: .size modCubeRoot, .Lfunc_end0-modCubeRoot .section .AMDGPU.csdata,"",@progbits .text .p2alignl 7, 3214868480 .fill 96, 4, 3214868480 .protected devArrayIndex .type devArrayIndex,@object .section .bss,"aw",@nobits .globl devArrayIndex .p2align 2, 0x0 devArrayIndex: .long 0 .size devArrayIndex, 4 .protected devAnswer .type devAnswer,@object .globl devAnswer .p2align 4, 0x0 devAnswer: .zero 24 .size devAnswer, 24 .type __hip_cuid_,@object .globl __hip_cuid_ __hip_cuid_: .byte 0 .size __hip_cuid_, 1 .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym devArrayIndex .addrsig_sym devAnswer .addrsig_sym __hip_cuid_ .amdgpu_metadata --- amdhsa.kernels: - .args: - .offset: 0 .size: 4 .value_kind: by_value - .offset: 4 .size: 4 .value_kind: by_value - .offset: 8 .size: 4 .value_kind: hidden_block_count_x - .offset: 12 .size: 4 .value_kind: hidden_block_count_y - .offset: 16 .size: 4 .value_kind: hidden_block_count_z - .offset: 20 .size: 2 .value_kind: hidden_group_size_x - .offset: 22 .size: 2 .value_kind: hidden_group_size_y - .offset: 24 .size: 2 .value_kind: hidden_group_size_z - .offset: 26 .size: 2 .value_kind: hidden_remainder_x - .offset: 28 .size: 2 .value_kind: hidden_remainder_y - .offset: 30 .size: 2 .value_kind: hidden_remainder_z - .offset: 48 .size: 8 .value_kind: hidden_global_offset_x - .offset: 56 .size: 8 .value_kind: hidden_global_offset_y - .offset: 64 .size: 8 .value_kind: hidden_global_offset_z - .offset: 72 .size: 2 .value_kind: hidden_grid_dims .group_segment_fixed_size: 0 .kernarg_segment_align: 8 .kernarg_segment_size: 264 .language: OpenCL C .language_version: - 2 - 0 .max_flat_workgroup_size: 1024 .name: modCubeRoot .private_segment_fixed_size: 0 .sgpr_count: 18 .sgpr_spill_count: 0 .symbol: modCubeRoot.kd .uniform_work_group_size: 1 .uses_dynamic_stack: false .vgpr_count: 16 .vgpr_spill_count: 0 .wavefront_size: 32 .workgroup_processor_mode: 1 amdhsa.target: amdgcn-amd-amdhsa--gfx1100 amdhsa.version: - 1 - 2 ... .end_amdgpu_metadata
You are a helpful coding assistant specialized in converting from CUDA host assembly to AMD host assembly.
.file "tmpxft_00091e3e_00000000-6_ModCubeRoot.cudafe1.cpp" .text #APP #NO_APP .type _ZL26__cudaUnregisterBinaryUtilv, @function _ZL26__cudaUnregisterBinaryUtilv: .LFB2029: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaUnregisterFatBinary@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2029: .size _ZL26__cudaUnregisterBinaryUtilv, .-_ZL26__cudaUnregisterBinaryUtilv .globl _Z31__device_stub__Z11modCubeRootiiii .type _Z31__device_stub__Z11modCubeRootiiii, @function _Z31__device_stub__Z11modCubeRootiiii: .LFB2051: .cfi_startproc endbr64 subq $120, %rsp .cfi_def_cfa_offset 128 movl %edi, 12(%rsp) movl %esi, 8(%rsp) movq %fs:40, %rax movq %rax, 104(%rsp) xorl %eax, %eax leaq 12(%rsp), %rax movq %rax, 80(%rsp) leaq 8(%rsp), %rax movq %rax, 88(%rsp) movl $1, 32(%rsp) movl $1, 36(%rsp) movl $1, 40(%rsp) movl $1, 44(%rsp) movl $1, 48(%rsp) movl $1, 52(%rsp) leaq 24(%rsp), %rcx leaq 16(%rsp), %rdx leaq 44(%rsp), %rsi leaq 32(%rsp), %rdi call __cudaPopCallConfiguration@PLT testl %eax, %eax je .L7 .L3: movq 104(%rsp), %rax subq %fs:40, %rax jne .L8 addq $120, %rsp .cfi_remember_state .cfi_def_cfa_offset 8 ret .L7: .cfi_restore_state pushq 24(%rsp) .cfi_def_cfa_offset 136 pushq 24(%rsp) .cfi_def_cfa_offset 144 leaq 96(%rsp), %r9 movq 60(%rsp), %rcx movl 68(%rsp), %r8d movq 48(%rsp), %rsi movl 56(%rsp), %edx leaq modCubeRoot(%rip), %rdi call cudaLaunchKernel@PLT addq $16, %rsp .cfi_def_cfa_offset 128 jmp .L3 .L8: call __stack_chk_fail@PLT .cfi_endproc .LFE2051: .size _Z31__device_stub__Z11modCubeRootiiii, .-_Z31__device_stub__Z11modCubeRootiiii .globl modCubeRoot .type modCubeRoot, @function modCubeRoot: .LFB2052: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 call _Z31__device_stub__Z11modCubeRootiiii addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2052: .size modCubeRoot, .-modCubeRoot .section .rodata.str1.1,"aMS",@progbits,1 .LC0: .string "modCubeRoot" .LC1: .string "devArrayIndex" .LC2: .string "devAnswer" .text .type _ZL24__sti____cudaRegisterAllv, @function _ZL24__sti____cudaRegisterAllv: .LFB2054: .cfi_startproc endbr64 pushq %rbx .cfi_def_cfa_offset 16 .cfi_offset 3, -16 leaq _ZL15__fatDeviceText(%rip), %rdi call __cudaRegisterFatBinary@PLT movq %rax, %rbx movq %rax, _ZL20__cudaFatCubinHandle(%rip) pushq $0 .cfi_def_cfa_offset 24 pushq $0 .cfi_def_cfa_offset 32 pushq $0 .cfi_def_cfa_offset 40 pushq $0 .cfi_def_cfa_offset 48 movl $0, %r9d movl $-1, %r8d leaq .LC0(%rip), %rdx movq %rdx, %rcx leaq modCubeRoot(%rip), %rsi movq %rax, %rdi call __cudaRegisterFunction@PLT addq $32, %rsp .cfi_def_cfa_offset 16 pushq $0 .cfi_def_cfa_offset 24 pushq $0 .cfi_def_cfa_offset 32 movl $4, %r9d movl $0, %r8d leaq .LC1(%rip), %rdx movq %rdx, %rcx leaq _ZL13devArrayIndex(%rip), %rsi movq %rbx, %rdi call __cudaRegisterVar@PLT addq $16, %rsp .cfi_def_cfa_offset 16 pushq $0 .cfi_def_cfa_offset 24 pushq $0 .cfi_def_cfa_offset 32 movl $24, %r9d movl $0, %r8d leaq .LC2(%rip), %rdx movq %rdx, %rcx leaq _ZL9devAnswer(%rip), %rsi movq %rbx, %rdi call __cudaRegisterVar@PLT addq $16, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaRegisterFatBinaryEnd@PLT leaq _ZL26__cudaUnregisterBinaryUtilv(%rip), %rdi call atexit@PLT popq %rbx .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2054: .size _ZL24__sti____cudaRegisterAllv, .-_ZL24__sti____cudaRegisterAllv .section .init_array,"aw" .align 8 .quad _ZL24__sti____cudaRegisterAllv .section .nvFatBinSegment,"aw" .align 8 .type _ZL15__fatDeviceText, @object .size _ZL15__fatDeviceText, 24 _ZL15__fatDeviceText: .long 1180844977 .long 1 .quad fatbinData .quad 0 .local _ZL20__cudaFatCubinHandle .comm _ZL20__cudaFatCubinHandle,8,8 .local _ZL9devAnswer .comm _ZL9devAnswer,24,16 .local _ZL13devArrayIndex .comm _ZL13devArrayIndex,4,4 .ident "GCC: (Ubuntu 13.3.0-6ubuntu2~24.04) 13.3.0" .section .note.GNU-stack,"",@progbits .section .note.gnu.property,"a" .align 8 .long 1f - 0f .long 4f - 1f .long 5 0: .string "GNU" 1: .align 8 .long 0xc0000002 .long 3f - 2f 2: .long 0x3 3: .align 8 4:
.text .file "ModCubeRoot.hip" .globl __device_stub__modCubeRoot # -- Begin function __device_stub__modCubeRoot .p2align 4, 0x90 .type __device_stub__modCubeRoot,@function __device_stub__modCubeRoot: # @__device_stub__modCubeRoot .cfi_startproc # %bb.0: subq $88, %rsp .cfi_def_cfa_offset 96 movl %edi, 12(%rsp) movl %esi, 8(%rsp) leaq 12(%rsp), %rax movq %rax, 64(%rsp) leaq 8(%rsp), %rax movq %rax, 72(%rsp) leaq 48(%rsp), %rdi leaq 32(%rsp), %rsi leaq 24(%rsp), %rdx leaq 16(%rsp), %rcx callq __hipPopCallConfiguration movq 48(%rsp), %rsi movl 56(%rsp), %edx movq 32(%rsp), %rcx movl 40(%rsp), %r8d leaq 64(%rsp), %r9 movl $modCubeRoot, %edi pushq 16(%rsp) .cfi_adjust_cfa_offset 8 pushq 32(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $104, %rsp .cfi_adjust_cfa_offset -104 retq .Lfunc_end0: .size __device_stub__modCubeRoot, .Lfunc_end0-__device_stub__modCubeRoot .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_ctor .type __hip_module_ctor,@function __hip_module_ctor: # @__hip_module_ctor .cfi_startproc # %bb.0: pushq %rbx .cfi_def_cfa_offset 16 .cfi_offset %rbx, -16 cmpq $0, __hip_gpubin_handle(%rip) jne .LBB1_2 # %bb.1: movl $__hip_fatbin_wrapper, %edi callq __hipRegisterFatBinary movq %rax, __hip_gpubin_handle(%rip) .LBB1_2: movq __hip_gpubin_handle(%rip), %rbx subq $32, %rsp .cfi_adjust_cfa_offset 32 xorps %xmm0, %xmm0 movups %xmm0, 16(%rsp) movups %xmm0, (%rsp) movl $modCubeRoot, %esi movl $.L__unnamed_1, %edx movl $.L__unnamed_1, %ecx movq %rbx, %rdi movl $-1, %r8d xorl %r9d, %r9d callq __hipRegisterFunction addq $32, %rsp .cfi_adjust_cfa_offset -32 movl $devArrayIndex, %esi movl $.L__unnamed_2, %edx movl $.L__unnamed_2, %ecx movl $4, %r9d movq %rbx, %rdi xorl %r8d, %r8d pushq $0 .cfi_adjust_cfa_offset 8 pushq $0 .cfi_adjust_cfa_offset 8 callq __hipRegisterVar addq $16, %rsp .cfi_adjust_cfa_offset -16 movl $devAnswer, %esi movl $.L__unnamed_3, %edx movl $.L__unnamed_3, %ecx movl $24, %r9d movq %rbx, %rdi xorl %r8d, %r8d pushq $0 .cfi_adjust_cfa_offset 8 pushq $0 .cfi_adjust_cfa_offset 8 callq __hipRegisterVar addq $16, %rsp .cfi_adjust_cfa_offset -16 movl $__hip_module_dtor, %edi popq %rbx .cfi_def_cfa_offset 8 jmp atexit # TAILCALL .Lfunc_end1: .size __hip_module_ctor, .Lfunc_end1-__hip_module_ctor .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_dtor .type __hip_module_dtor,@function __hip_module_dtor: # @__hip_module_dtor .cfi_startproc # %bb.0: movq __hip_gpubin_handle(%rip), %rdi testq %rdi, %rdi je .LBB2_2 # %bb.1: pushq %rax .cfi_def_cfa_offset 16 callq __hipUnregisterFatBinary movq $0, __hip_gpubin_handle(%rip) addq $8, %rsp .cfi_def_cfa_offset 8 .LBB2_2: retq .Lfunc_end2: .size __hip_module_dtor, .Lfunc_end2-__hip_module_dtor .cfi_endproc # -- End function .type devArrayIndex,@object # @devArrayIndex .local devArrayIndex .comm devArrayIndex,4,4 .type devAnswer,@object # @devAnswer .local devAnswer .comm devAnswer,24,16 .type modCubeRoot,@object # @modCubeRoot .section .rodata,"a",@progbits .globl modCubeRoot .p2align 3, 0x0 modCubeRoot: .quad __device_stub__modCubeRoot .size modCubeRoot, 8 .type .L__unnamed_1,@object # @0 .section .rodata.str1.1,"aMS",@progbits,1 .L__unnamed_1: .asciz "modCubeRoot" .size .L__unnamed_1, 12 .type .L__unnamed_2,@object # @1 .L__unnamed_2: .asciz "devArrayIndex" .size .L__unnamed_2, 14 .type .L__unnamed_3,@object # @2 .L__unnamed_3: .asciz "devAnswer" .size .L__unnamed_3, 10 .type __hip_fatbin_wrapper,@object # @__hip_fatbin_wrapper .section .hipFatBinSegment,"a",@progbits .p2align 3, 0x0 __hip_fatbin_wrapper: .long 1212764230 # 0x48495046 .long 1 # 0x1 .quad __hip_fatbin .quad 0 .size __hip_fatbin_wrapper, 24 .type __hip_gpubin_handle,@object # @__hip_gpubin_handle .local __hip_gpubin_handle .comm __hip_gpubin_handle,8,8 .section .init_array,"aw",@init_array .p2align 3, 0x0 .quad __hip_module_ctor .type __hip_cuid_,@object # @__hip_cuid_ .bss .globl __hip_cuid_ __hip_cuid_: .byte 0 # 0x0 .size __hip_cuid_, 1 .section ".linker-options","e",@llvm_linker_options .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym __device_stub__modCubeRoot .addrsig_sym __hip_module_ctor .addrsig_sym __hip_module_dtor .addrsig_sym devArrayIndex .addrsig_sym devAnswer .addrsig_sym modCubeRoot .addrsig_sym __hip_fatbin .addrsig_sym __hip_fatbin_wrapper .addrsig_sym __hip_cuid_
You are a helpful coding assistant specialized in converting CUDA source code to CUDA device assembly.
#include <cstdio> #include <stdio.h> int main(void) { int a[100000]; int b[100000]; int *ary1; for(int i=0;i<100000;i++) { a[i] = i; } cudaMalloc((void**)&ary1 , 100000*sizeof(int)); cudaMemcpy(ary1, a, 100000*sizeof(int),cudaMemcpyHostToDevice); cudaMemcpy(b, ary1 ,100000*sizeof(int),cudaMemcpyDeviceToHost); for(int i =0; i < 100000; i++) { if(i/500 ==0) { printf("%d ",i); } } cudaFree(ary1); return 0; }
code for sm_80
You are a helpful coding assistant specialized in converting CUDA source code to CUDA host assembly.
#include <cstdio> #include <stdio.h> int main(void) { int a[100000]; int b[100000]; int *ary1; for(int i=0;i<100000;i++) { a[i] = i; } cudaMalloc((void**)&ary1 , 100000*sizeof(int)); cudaMemcpy(ary1, a, 100000*sizeof(int),cudaMemcpyHostToDevice); cudaMemcpy(b, ary1 ,100000*sizeof(int),cudaMemcpyDeviceToHost); for(int i =0; i < 100000; i++) { if(i/500 ==0) { printf("%d ",i); } } cudaFree(ary1); return 0; }
.file "tmpxft_00131a50_00000000-6_mem.cudafe1.cpp" .text #APP #NO_APP .type _ZL26__cudaUnregisterBinaryUtilv, @function _ZL26__cudaUnregisterBinaryUtilv: .LFB2060: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaUnregisterFatBinary@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2060: .size _ZL26__cudaUnregisterBinaryUtilv, .-_ZL26__cudaUnregisterBinaryUtilv .section .rodata.str1.1,"aMS",@progbits,1 .LC0: .string "%d " .text .globl main .type main, @function main: .LFB2057: .cfi_startproc endbr64 pushq %rbx .cfi_def_cfa_offset 16 .cfi_offset 3, -16 leaq -798720(%rsp), %r11 .cfi_def_cfa 11, 798736 .LPSRL0: subq $4096, %rsp orq $0, (%rsp) cmpq %r11, %rsp jne .LPSRL0 .cfi_def_cfa_register 7 subq $1312, %rsp .cfi_def_cfa_offset 800048 movq %fs:40, %rax movq %rax, 800024(%rsp) xorl %eax, %eax .L4: movl %eax, 16(%rsp,%rax,4) addq $1, %rax cmpq $100000, %rax jne .L4 leaq 8(%rsp), %rdi movl $400000, %esi call cudaMalloc@PLT leaq 16(%rsp), %rsi movl $1, %ecx movl $400000, %edx movq 8(%rsp), %rdi call cudaMemcpy@PLT leaq 400016(%rsp), %rdi movl $2, %ecx movl $400000, %edx movq 8(%rsp), %rsi call cudaMemcpy@PLT movl $0, %ebx .L8: leal 499(%rbx), %eax cmpl $998, %eax jbe .L12 addl $1, %ebx cmpl $100000, %ebx jne .L8 movq 8(%rsp), %rdi call cudaFree@PLT movq 800024(%rsp), %rax subq %fs:40, %rax jne .L13 movl $0, %eax addq $800032, %rsp .cfi_remember_state .cfi_def_cfa_offset 16 popq %rbx .cfi_def_cfa_offset 8 ret .L12: .cfi_restore_state movl %ebx, %edx leaq .LC0(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT addl $1, %ebx jmp .L8 .L13: call __stack_chk_fail@PLT .cfi_endproc .LFE2057: .size main, .-main .type _ZL24__sti____cudaRegisterAllv, @function _ZL24__sti____cudaRegisterAllv: .LFB2083: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 leaq _ZL15__fatDeviceText(%rip), %rdi call __cudaRegisterFatBinary@PLT movq %rax, %rdi movq %rax, _ZL20__cudaFatCubinHandle(%rip) call __cudaRegisterFatBinaryEnd@PLT leaq _ZL26__cudaUnregisterBinaryUtilv(%rip), %rdi call atexit@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2083: .size _ZL24__sti____cudaRegisterAllv, .-_ZL24__sti____cudaRegisterAllv .section .init_array,"aw" .align 8 .quad _ZL24__sti____cudaRegisterAllv .section .nvFatBinSegment,"aw" .align 8 .type _ZL15__fatDeviceText, @object .size _ZL15__fatDeviceText, 24 _ZL15__fatDeviceText: .long 1180844977 .long 1 .quad fatbinData .quad 0 .local _ZL20__cudaFatCubinHandle .comm _ZL20__cudaFatCubinHandle,8,8 .ident "GCC: (Ubuntu 13.3.0-6ubuntu2~24.04) 13.3.0" .section .note.GNU-stack,"",@progbits .section .note.gnu.property,"a" .align 8 .long 1f - 0f .long 4f - 1f .long 5 0: .string "GNU" 1: .align 8 .long 0xc0000002 .long 3f - 2f 2: .long 0x3 3: .align 8 4:
You are a helpful coding assistant specialized in converting CUDA source code to HIP/AMD source code.
#include <cstdio> #include <stdio.h> int main(void) { int a[100000]; int b[100000]; int *ary1; for(int i=0;i<100000;i++) { a[i] = i; } cudaMalloc((void**)&ary1 , 100000*sizeof(int)); cudaMemcpy(ary1, a, 100000*sizeof(int),cudaMemcpyHostToDevice); cudaMemcpy(b, ary1 ,100000*sizeof(int),cudaMemcpyDeviceToHost); for(int i =0; i < 100000; i++) { if(i/500 ==0) { printf("%d ",i); } } cudaFree(ary1); return 0; }
#include <hip/hip_runtime.h> #include <cstdio> #include <stdio.h> int main(void) { int a[100000]; int b[100000]; int *ary1; for(int i=0;i<100000;i++) { a[i] = i; } hipMalloc((void**)&ary1 , 100000*sizeof(int)); hipMemcpy(ary1, a, 100000*sizeof(int),hipMemcpyHostToDevice); hipMemcpy(b, ary1 ,100000*sizeof(int),hipMemcpyDeviceToHost); for(int i =0; i < 100000; i++) { if(i/500 ==0) { printf("%d ",i); } } hipFree(ary1); return 0; }
You are a helpful coding assistant specialized in converting HIP source code to AMD device assembly.
#include <hip/hip_runtime.h> #include <cstdio> #include <stdio.h> int main(void) { int a[100000]; int b[100000]; int *ary1; for(int i=0;i<100000;i++) { a[i] = i; } hipMalloc((void**)&ary1 , 100000*sizeof(int)); hipMemcpy(ary1, a, 100000*sizeof(int),hipMemcpyHostToDevice); hipMemcpy(b, ary1 ,100000*sizeof(int),hipMemcpyDeviceToHost); for(int i =0; i < 100000; i++) { if(i/500 ==0) { printf("%d ",i); } } hipFree(ary1); return 0; }
.text .p2alignl 7, 3214868480 .fill 96, 4, 3214868480 .type __hip_cuid_,@object .section .bss,"aw",@nobits .globl __hip_cuid_ __hip_cuid_: .byte 0 .size __hip_cuid_, 1 .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym __hip_cuid_ .amdgcn_target "amdgcn-amd-amdhsa--gfx1100" .amdgpu_metadata --- amdhsa.kernels: [] amdhsa.target: amdgcn-amd-amdhsa--gfx1100 amdhsa.version: - 1 - 2 ... .end_amdgpu_metadata
You are a helpful coding assistant specialized in converting HIP source code to AMD host assembly.
#include <hip/hip_runtime.h> #include <cstdio> #include <stdio.h> int main(void) { int a[100000]; int b[100000]; int *ary1; for(int i=0;i<100000;i++) { a[i] = i; } hipMalloc((void**)&ary1 , 100000*sizeof(int)); hipMemcpy(ary1, a, 100000*sizeof(int),hipMemcpyHostToDevice); hipMemcpy(b, ary1 ,100000*sizeof(int),hipMemcpyDeviceToHost); for(int i =0; i < 100000; i++) { if(i/500 ==0) { printf("%d ",i); } } hipFree(ary1); return 0; }
.text .file "mem.hip" .globl main # -- Begin function main .p2align 4, 0x90 .type main,@function main: # @main .cfi_startproc # %bb.0: pushq %rbx .cfi_def_cfa_offset 16 subq $800016, %rsp # imm = 0xC3510 .cfi_def_cfa_offset 800032 .cfi_offset %rbx, -16 xorl %eax, %eax .p2align 4, 0x90 .LBB0_1: # =>This Inner Loop Header: Depth=1 movl %eax, 16(%rsp,%rax,4) incq %rax cmpq $100000, %rax # imm = 0x186A0 jne .LBB0_1 # %bb.2: leaq 8(%rsp), %rdi movl $400000, %esi # imm = 0x61A80 callq hipMalloc movq 8(%rsp), %rdi leaq 16(%rsp), %rsi movl $400000, %edx # imm = 0x61A80 movl $1, %ecx callq hipMemcpy movq 8(%rsp), %rsi leaq 400016(%rsp), %rdi movl $400000, %edx # imm = 0x61A80 movl $2, %ecx callq hipMemcpy xorl %ebx, %ebx jmp .LBB0_3 .p2align 4, 0x90 .LBB0_5: # in Loop: Header=BB0_3 Depth=1 incl %ebx cmpl $100000, %ebx # imm = 0x186A0 je .LBB0_6 .LBB0_3: # =>This Inner Loop Header: Depth=1 cmpl $499, %ebx # imm = 0x1F3 ja .LBB0_5 # %bb.4: # in Loop: Header=BB0_3 Depth=1 movl $.L.str, %edi movl %ebx, %esi xorl %eax, %eax callq printf jmp .LBB0_5 .LBB0_6: movq 8(%rsp), %rdi callq hipFree xorl %eax, %eax addq $800016, %rsp # imm = 0xC3510 .cfi_def_cfa_offset 16 popq %rbx .cfi_def_cfa_offset 8 retq .Lfunc_end0: .size main, .Lfunc_end0-main .cfi_endproc # -- End function .type .L.str,@object # @.str .section .rodata.str1.1,"aMS",@progbits,1 .L.str: .asciz "%d " .size .L.str, 4 .type __hip_cuid_,@object # @__hip_cuid_ .bss .globl __hip_cuid_ __hip_cuid_: .byte 0 # 0x0 .size __hip_cuid_, 1 .section ".linker-options","e",@llvm_linker_options .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym __hip_cuid_
You are a helpful coding assistant specialized in converting from CUDA device assembly to AMD device assembly.
code for sm_80
.text .p2alignl 7, 3214868480 .fill 96, 4, 3214868480 .type __hip_cuid_,@object .section .bss,"aw",@nobits .globl __hip_cuid_ __hip_cuid_: .byte 0 .size __hip_cuid_, 1 .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym __hip_cuid_ .amdgcn_target "amdgcn-amd-amdhsa--gfx1100" .amdgpu_metadata --- amdhsa.kernels: [] amdhsa.target: amdgcn-amd-amdhsa--gfx1100 amdhsa.version: - 1 - 2 ... .end_amdgpu_metadata
You are a helpful coding assistant specialized in converting from CUDA host assembly to AMD host assembly.
.file "tmpxft_00131a50_00000000-6_mem.cudafe1.cpp" .text #APP #NO_APP .type _ZL26__cudaUnregisterBinaryUtilv, @function _ZL26__cudaUnregisterBinaryUtilv: .LFB2060: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaUnregisterFatBinary@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2060: .size _ZL26__cudaUnregisterBinaryUtilv, .-_ZL26__cudaUnregisterBinaryUtilv .section .rodata.str1.1,"aMS",@progbits,1 .LC0: .string "%d " .text .globl main .type main, @function main: .LFB2057: .cfi_startproc endbr64 pushq %rbx .cfi_def_cfa_offset 16 .cfi_offset 3, -16 leaq -798720(%rsp), %r11 .cfi_def_cfa 11, 798736 .LPSRL0: subq $4096, %rsp orq $0, (%rsp) cmpq %r11, %rsp jne .LPSRL0 .cfi_def_cfa_register 7 subq $1312, %rsp .cfi_def_cfa_offset 800048 movq %fs:40, %rax movq %rax, 800024(%rsp) xorl %eax, %eax .L4: movl %eax, 16(%rsp,%rax,4) addq $1, %rax cmpq $100000, %rax jne .L4 leaq 8(%rsp), %rdi movl $400000, %esi call cudaMalloc@PLT leaq 16(%rsp), %rsi movl $1, %ecx movl $400000, %edx movq 8(%rsp), %rdi call cudaMemcpy@PLT leaq 400016(%rsp), %rdi movl $2, %ecx movl $400000, %edx movq 8(%rsp), %rsi call cudaMemcpy@PLT movl $0, %ebx .L8: leal 499(%rbx), %eax cmpl $998, %eax jbe .L12 addl $1, %ebx cmpl $100000, %ebx jne .L8 movq 8(%rsp), %rdi call cudaFree@PLT movq 800024(%rsp), %rax subq %fs:40, %rax jne .L13 movl $0, %eax addq $800032, %rsp .cfi_remember_state .cfi_def_cfa_offset 16 popq %rbx .cfi_def_cfa_offset 8 ret .L12: .cfi_restore_state movl %ebx, %edx leaq .LC0(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT addl $1, %ebx jmp .L8 .L13: call __stack_chk_fail@PLT .cfi_endproc .LFE2057: .size main, .-main .type _ZL24__sti____cudaRegisterAllv, @function _ZL24__sti____cudaRegisterAllv: .LFB2083: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 leaq _ZL15__fatDeviceText(%rip), %rdi call __cudaRegisterFatBinary@PLT movq %rax, %rdi movq %rax, _ZL20__cudaFatCubinHandle(%rip) call __cudaRegisterFatBinaryEnd@PLT leaq _ZL26__cudaUnregisterBinaryUtilv(%rip), %rdi call atexit@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2083: .size _ZL24__sti____cudaRegisterAllv, .-_ZL24__sti____cudaRegisterAllv .section .init_array,"aw" .align 8 .quad _ZL24__sti____cudaRegisterAllv .section .nvFatBinSegment,"aw" .align 8 .type _ZL15__fatDeviceText, @object .size _ZL15__fatDeviceText, 24 _ZL15__fatDeviceText: .long 1180844977 .long 1 .quad fatbinData .quad 0 .local _ZL20__cudaFatCubinHandle .comm _ZL20__cudaFatCubinHandle,8,8 .ident "GCC: (Ubuntu 13.3.0-6ubuntu2~24.04) 13.3.0" .section .note.GNU-stack,"",@progbits .section .note.gnu.property,"a" .align 8 .long 1f - 0f .long 4f - 1f .long 5 0: .string "GNU" 1: .align 8 .long 0xc0000002 .long 3f - 2f 2: .long 0x3 3: .align 8 4:
.text .file "mem.hip" .globl main # -- Begin function main .p2align 4, 0x90 .type main,@function main: # @main .cfi_startproc # %bb.0: pushq %rbx .cfi_def_cfa_offset 16 subq $800016, %rsp # imm = 0xC3510 .cfi_def_cfa_offset 800032 .cfi_offset %rbx, -16 xorl %eax, %eax .p2align 4, 0x90 .LBB0_1: # =>This Inner Loop Header: Depth=1 movl %eax, 16(%rsp,%rax,4) incq %rax cmpq $100000, %rax # imm = 0x186A0 jne .LBB0_1 # %bb.2: leaq 8(%rsp), %rdi movl $400000, %esi # imm = 0x61A80 callq hipMalloc movq 8(%rsp), %rdi leaq 16(%rsp), %rsi movl $400000, %edx # imm = 0x61A80 movl $1, %ecx callq hipMemcpy movq 8(%rsp), %rsi leaq 400016(%rsp), %rdi movl $400000, %edx # imm = 0x61A80 movl $2, %ecx callq hipMemcpy xorl %ebx, %ebx jmp .LBB0_3 .p2align 4, 0x90 .LBB0_5: # in Loop: Header=BB0_3 Depth=1 incl %ebx cmpl $100000, %ebx # imm = 0x186A0 je .LBB0_6 .LBB0_3: # =>This Inner Loop Header: Depth=1 cmpl $499, %ebx # imm = 0x1F3 ja .LBB0_5 # %bb.4: # in Loop: Header=BB0_3 Depth=1 movl $.L.str, %edi movl %ebx, %esi xorl %eax, %eax callq printf jmp .LBB0_5 .LBB0_6: movq 8(%rsp), %rdi callq hipFree xorl %eax, %eax addq $800016, %rsp # imm = 0xC3510 .cfi_def_cfa_offset 16 popq %rbx .cfi_def_cfa_offset 8 retq .Lfunc_end0: .size main, .Lfunc_end0-main .cfi_endproc # -- End function .type .L.str,@object # @.str .section .rodata.str1.1,"aMS",@progbits,1 .L.str: .asciz "%d " .size .L.str, 4 .type __hip_cuid_,@object # @__hip_cuid_ .bss .globl __hip_cuid_ __hip_cuid_: .byte 0 # 0x0 .size __hip_cuid_, 1 .section ".linker-options","e",@llvm_linker_options .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym __hip_cuid_
You are a helpful coding assistant specialized in converting CUDA source code to CUDA host assembly.
#include <stdio.h> #include <stdlib.h> #include <sys/time.h> #include <sys/resource.h> #include <math.h> //134217728 double dwalltime(){ double sec; struct timeval tv; gettimeofday(&tv,NULL); sec = tv.tv_sec + tv.tv_usec/1000000.0; return sec; } __global__ void ecuacion_kernel_outplace_p1(double *d_matA,double *d_matAT,double *d_matB,double *d_matBT, unsigned int n){ int distA = blockIdx.y * blockDim.y + threadIdx.y; //i int distB = blockIdx.x * blockDim.x + threadIdx.x; //j //transpuesta out-place A y B if( (distA<n*n) && (distB<n*n) ){ d_matAT [distB*n + distA] = d_matA[distA*n + distB]; d_matBT [distB*n + distA] = d_matB[distA*n + distB]; } } __global__ void ecuacion_kernel_outplace_p2(double *d_matA,double *d_matB,double *d_matC,double *d_matAT,double *d_matBT, unsigned int n){ int distA = blockIdx.y * blockDim.y + threadIdx.y; //i int distB = blockIdx.x * blockDim.x + threadIdx.x; //j int k; if (distA*n+distB <= (n*n - 1)){ //multiplicacion for(k = 0; k < n ;k++){ d_matC[distA*n+distB] += d_matA[distA*n+k] * d_matBT[distB+k*n]; } //suma d_matC[distA*n+distB] += d_matB[distA*n+distB] + d_matAT[distA*n+distB]; } } __global__ void ecuacion_kernel_inplace_suma (double *d_matA,double *d_matB,double *d_matC, unsigned int n){ int distA = blockIdx.y * blockDim.y + threadIdx.y; //i int distB = blockIdx.x * blockDim.x + threadIdx.x; //j int k; //multiplicacion if (distA*n+distB < (n*n - 1)){ d_matC[distA*n+distB] += d_matB[distA*n+distB] + d_matA[distA+distB*n]; for(k = 0; k < n ;k++){ d_matC[distA*n+distB] += d_matA[distA*n+k] * d_matB[distB*n+k]; } } } __global__ void kernel_sum_Matriz (double *d_matA,double *d_matB,double *d_matC, unsigned int n){ int distA = blockIdx.y * blockDim.y + threadIdx.y; //i int distB = blockIdx.x * blockDim.x + threadIdx.x; //j //suma if (distA*n+distB < (n*n)){ d_matC[distA*n+distB] += d_matA[distA*n+distB] + d_matB[distA+distB*n]; } } __global__ void kernel_transpuesta(double *m, int N){ int tid = blockIdx.x*blockDim.x + threadIdx.x; int i = int((1 + sqrtf(1 + 8*tid)) / 2); int j = tid - (i*(i-1)/2); int aux; if ( (i<N) && (j<N) ){ aux = m[i*N + j] ; m[i*N + j] = m[j*N + i]; m[j*N + i] = aux; } } __global__ void kernel_mult_sum_matriz (double *d_matA,double *d_matB,double *d_matC, unsigned int n){ int distA = blockIdx.y * blockDim.y + threadIdx.y; //i int distB = blockIdx.x * blockDim.x + threadIdx.x; //j int k; //multiplicacion if (distA*n+distB < (n*n)){ for(k = 0; k < n ;k++){ d_matC[distA*n+distB] += d_matA[distA*n+k] * d_matB[distB*n+k]; } } } int main(int argc, char *argv[]){ if (argc != 3){ printf("Falta argumento: N, CUDABLK\n"); return 0; } //declaracion de variables cudaError_t error; unsigned int N = atoi (argv[1]); unsigned long CUDA_BLK = atoi (argv[2]), gridBlock; unsigned long numBytes = sizeof(double)*N*N; double *matA,*matB,*matC,*d_matA,*d_matB,*d_matC,*d_matAT,*d_matBT,timetick; unsigned int i,j,k; //inicializa variables para cpu matA = (double *)malloc(numBytes); matB = (double *)malloc(numBytes); matC = (double *)malloc(numBytes); for (i = 0; i < N*N; i++){ matA[i] = i; matB[i] = i; matC[i] = 0; } //inicializa variables para gpu cudaMalloc((void **) &d_matA, numBytes); cudaMalloc((void **) &d_matAT, numBytes); cudaMalloc((void **) &d_matB, numBytes); cudaMalloc((void **) &d_matBT, numBytes); cudaMalloc((void **) &d_matC, numBytes); gridBlock = (unsigned int)sqrt(N*N/CUDA_BLK/CUDA_BLK); dim3 dimBlock(CUDA_BLK,CUDA_BLK); // Bloque bidimencional de hilos (*cb* hilos) dim3 dimGrid(gridBlock,gridBlock); // Grid bidimencional (*ceil(n/cb)* bloques) //--------------------------------cpu comienza ------------------------------------ //secuencial timetick = dwalltime(); //multiplicacion for(i = 0; i < N; i++){ for(j = 0; j < N; j++){ for(k = 0; k < N ;k++){ matC[i*N+j] += matA[i*N+k] * matB[j*N+k]; //multiplica a matB por fila, eso simula la matB transpuesta } } } //suma for(i = 0; i < N; i++){ for(j = 0; j < N; j++){ matC[i*N+j] += matB[i*N+j] + matA[i+j*N]; } } printf("Tiempo para la ecuacion CPU: %f\n\n",dwalltime() - timetick); /* for(i = 0; i < N; i++){ for(j = 0; j < N; j++){ printf("%f|",matC[i*N+j]); } printf("\n"); } printf("\n"); */ //--------------------------------cpu termina ------------------------------------ for (i = 0; i < N*N; i++){ matA[i] = i; matB[i] = i; matC[i] = 0; } cudaMemcpy(d_matA, matA, numBytes, cudaMemcpyHostToDevice); // CPU -> GPU cudaMemcpy(d_matB, matB, numBytes, cudaMemcpyHostToDevice); // CPU -> GPU cudaMemcpy(d_matC, matC, numBytes, cudaMemcpyHostToDevice); // CPU -> GPU //--------------------------------gpu out-place comienza ------------------------------------ timetick = dwalltime(); ecuacion_kernel_outplace_p1<<<dimGrid, dimBlock>>>(d_matA, d_matAT,d_matB,d_matBT, N); cudaThreadSynchronize(); ecuacion_kernel_outplace_p2<<<dimGrid, dimBlock>>>(d_matA, d_matB,d_matC,d_matAT,d_matBT, N); cudaThreadSynchronize(); printf("Tiempo para la ecuacion out-place GPU: %f\n",dwalltime() - timetick); error = cudaGetLastError(); printf("error: %d\n\n",error); cudaMemcpy(matC, d_matC, numBytes, cudaMemcpyDeviceToHost); // GPU -> CPU /* for(i = 0; i < N; i++){ for(j = 0; j < N; j++){ printf("%f|",matC[i*N+j]); } printf("\n"); } printf("\n"); */ //--------------------------------gpu out-place termina ------------------------------------ cudaFree(d_matA); cudaFree(d_matB); cudaFree(d_matC); cudaFree(d_matAT); cudaFree(d_matBT); for (i = 0; i < N*N; i++){ matA[i] = i; matB[i] = i; matC[i] = 0; } cudaMalloc((void **) &d_matA, numBytes); cudaMalloc((void **) &d_matB, numBytes); cudaMalloc((void **) &d_matC, numBytes); cudaMemcpy(d_matA, matA, numBytes, cudaMemcpyHostToDevice); // CPU -> GPU cudaMemcpy(d_matB, matB, numBytes, cudaMemcpyHostToDevice); // CPU -> GPU cudaMemcpy(d_matC, matC, numBytes, cudaMemcpyHostToDevice); // CPU -> GPU //--------------------------------gpu in-place comienza ------------------------------------ timetick = dwalltime(); kernel_transpuesta<<<dimGrid, dimBlock>>>(d_matA, N); cudaThreadSynchronize(); kernel_sum_Matriz<<<dimGrid, dimBlock>>>(d_matA, d_matB,d_matC, N); cudaThreadSynchronize(); kernel_transpuesta<<<dimGrid, dimBlock>>>(d_matA, N); cudaThreadSynchronize(); kernel_mult_sum_matriz<<<dimGrid, dimBlock>>>(d_matA, d_matB,d_matC, N); cudaThreadSynchronize(); printf("Tiempo para la ecuacion in-place GPU: %f\n",dwalltime() - timetick); error = cudaGetLastError(); printf("error: %d\n\n",error); cudaMemcpy(matC, d_matC, numBytes, cudaMemcpyDeviceToHost); // GPU -> CPU //--------------------------------gpu in-place termina ------------------------------------ cudaFree(d_matA); cudaFree(d_matB); cudaFree(d_matC); /* //imprime la matriz matC for(i = 0; i < N; i++){ for(j = 0; j < N; j++){ printf("%f|",matC[i*N+j]); } printf("\n"); } printf("\n"); */ free(matA); free(matB); free(matC); return 0; }
.file "tmpxft_000d35e5_00000000-6_ejercicio1_sin_tiempo_copia.cudafe1.cpp" .text #APP #NO_APP .type _ZL26__cudaUnregisterBinaryUtilv, @function _ZL26__cudaUnregisterBinaryUtilv: .LFB2061: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaUnregisterFatBinary@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2061: .size _ZL26__cudaUnregisterBinaryUtilv, .-_ZL26__cudaUnregisterBinaryUtilv .globl _Z9dwalltimev .type _Z9dwalltimev, @function _Z9dwalltimev: .LFB2057: .cfi_startproc endbr64 subq $40, %rsp .cfi_def_cfa_offset 48 movq %fs:40, %rax movq %rax, 24(%rsp) xorl %eax, %eax movq %rsp, %rdi movl $0, %esi call gettimeofday@PLT pxor %xmm0, %xmm0 cvtsi2sdq 8(%rsp), %xmm0 divsd .LC0(%rip), %xmm0 pxor %xmm1, %xmm1 cvtsi2sdq (%rsp), %xmm1 addsd %xmm1, %xmm0 movq 24(%rsp), %rax subq %fs:40, %rax jne .L6 addq $40, %rsp .cfi_remember_state .cfi_def_cfa_offset 8 ret .L6: .cfi_restore_state call __stack_chk_fail@PLT .cfi_endproc .LFE2057: .size _Z9dwalltimev, .-_Z9dwalltimev .globl _Z54__device_stub__Z27ecuacion_kernel_outplace_p1PdS_S_S_jPdS_S_S_j .type _Z54__device_stub__Z27ecuacion_kernel_outplace_p1PdS_S_S_jPdS_S_S_j, @function _Z54__device_stub__Z27ecuacion_kernel_outplace_p1PdS_S_S_jPdS_S_S_j: .LFB2083: .cfi_startproc endbr64 subq $168, %rsp .cfi_def_cfa_offset 176 movq %rdi, 40(%rsp) movq %rsi, 32(%rsp) movq %rdx, 24(%rsp) movq %rcx, 16(%rsp) movl %r8d, 12(%rsp) movq %fs:40, %rax movq %rax, 152(%rsp) xorl %eax, %eax leaq 40(%rsp), %rax movq %rax, 112(%rsp) leaq 32(%rsp), %rax movq %rax, 120(%rsp) leaq 24(%rsp), %rax movq %rax, 128(%rsp) leaq 16(%rsp), %rax movq %rax, 136(%rsp) leaq 12(%rsp), %rax movq %rax, 144(%rsp) movl $1, 64(%rsp) movl $1, 68(%rsp) movl $1, 72(%rsp) movl $1, 76(%rsp) movl $1, 80(%rsp) movl $1, 84(%rsp) leaq 56(%rsp), %rcx leaq 48(%rsp), %rdx leaq 76(%rsp), %rsi leaq 64(%rsp), %rdi call __cudaPopCallConfiguration@PLT testl %eax, %eax je .L11 .L7: movq 152(%rsp), %rax subq %fs:40, %rax jne .L12 addq $168, %rsp .cfi_remember_state .cfi_def_cfa_offset 8 ret .L11: .cfi_restore_state pushq 56(%rsp) .cfi_def_cfa_offset 184 pushq 56(%rsp) .cfi_def_cfa_offset 192 leaq 128(%rsp), %r9 movq 92(%rsp), %rcx movl 100(%rsp), %r8d movq 80(%rsp), %rsi movl 88(%rsp), %edx leaq _Z27ecuacion_kernel_outplace_p1PdS_S_S_j(%rip), %rdi call cudaLaunchKernel@PLT addq $16, %rsp .cfi_def_cfa_offset 176 jmp .L7 .L12: call __stack_chk_fail@PLT .cfi_endproc .LFE2083: .size _Z54__device_stub__Z27ecuacion_kernel_outplace_p1PdS_S_S_jPdS_S_S_j, .-_Z54__device_stub__Z27ecuacion_kernel_outplace_p1PdS_S_S_jPdS_S_S_j .globl _Z27ecuacion_kernel_outplace_p1PdS_S_S_j .type _Z27ecuacion_kernel_outplace_p1PdS_S_S_j, @function _Z27ecuacion_kernel_outplace_p1PdS_S_S_j: .LFB2084: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 call _Z54__device_stub__Z27ecuacion_kernel_outplace_p1PdS_S_S_jPdS_S_S_j addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2084: .size _Z27ecuacion_kernel_outplace_p1PdS_S_S_j, .-_Z27ecuacion_kernel_outplace_p1PdS_S_S_j .globl _Z56__device_stub__Z27ecuacion_kernel_outplace_p2PdS_S_S_S_jPdS_S_S_S_j .type _Z56__device_stub__Z27ecuacion_kernel_outplace_p2PdS_S_S_S_jPdS_S_S_S_j, @function _Z56__device_stub__Z27ecuacion_kernel_outplace_p2PdS_S_S_S_jPdS_S_S_S_j: .LFB2085: .cfi_startproc endbr64 subq $184, %rsp .cfi_def_cfa_offset 192 movq %rdi, 40(%rsp) movq %rsi, 32(%rsp) movq %rdx, 24(%rsp) movq %rcx, 16(%rsp) movq %r8, 8(%rsp) movl %r9d, 4(%rsp) movq %fs:40, %rax movq %rax, 168(%rsp) xorl %eax, %eax leaq 40(%rsp), %rax movq %rax, 112(%rsp) leaq 32(%rsp), %rax movq %rax, 120(%rsp) leaq 24(%rsp), %rax movq %rax, 128(%rsp) leaq 16(%rsp), %rax movq %rax, 136(%rsp) leaq 8(%rsp), %rax movq %rax, 144(%rsp) leaq 4(%rsp), %rax movq %rax, 152(%rsp) movl $1, 64(%rsp) movl $1, 68(%rsp) movl $1, 72(%rsp) movl $1, 76(%rsp) movl $1, 80(%rsp) movl $1, 84(%rsp) leaq 56(%rsp), %rcx leaq 48(%rsp), %rdx leaq 76(%rsp), %rsi leaq 64(%rsp), %rdi call __cudaPopCallConfiguration@PLT testl %eax, %eax je .L19 .L15: movq 168(%rsp), %rax subq %fs:40, %rax jne .L20 addq $184, %rsp .cfi_remember_state .cfi_def_cfa_offset 8 ret .L19: .cfi_restore_state pushq 56(%rsp) .cfi_def_cfa_offset 200 pushq 56(%rsp) .cfi_def_cfa_offset 208 leaq 128(%rsp), %r9 movq 92(%rsp), %rcx movl 100(%rsp), %r8d movq 80(%rsp), %rsi movl 88(%rsp), %edx leaq _Z27ecuacion_kernel_outplace_p2PdS_S_S_S_j(%rip), %rdi call cudaLaunchKernel@PLT addq $16, %rsp .cfi_def_cfa_offset 192 jmp .L15 .L20: call __stack_chk_fail@PLT .cfi_endproc .LFE2085: .size _Z56__device_stub__Z27ecuacion_kernel_outplace_p2PdS_S_S_S_jPdS_S_S_S_j, .-_Z56__device_stub__Z27ecuacion_kernel_outplace_p2PdS_S_S_S_jPdS_S_S_S_j .globl _Z27ecuacion_kernel_outplace_p2PdS_S_S_S_j .type _Z27ecuacion_kernel_outplace_p2PdS_S_S_S_j, @function _Z27ecuacion_kernel_outplace_p2PdS_S_S_S_j: .LFB2086: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 call _Z56__device_stub__Z27ecuacion_kernel_outplace_p2PdS_S_S_S_jPdS_S_S_S_j addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2086: .size _Z27ecuacion_kernel_outplace_p2PdS_S_S_S_j, .-_Z27ecuacion_kernel_outplace_p2PdS_S_S_S_j .globl _Z53__device_stub__Z28ecuacion_kernel_inplace_sumaPdS_S_jPdS_S_j .type _Z53__device_stub__Z28ecuacion_kernel_inplace_sumaPdS_S_jPdS_S_j, @function _Z53__device_stub__Z28ecuacion_kernel_inplace_sumaPdS_S_jPdS_S_j: .LFB2087: .cfi_startproc endbr64 subq $152, %rsp .cfi_def_cfa_offset 160 movq %rdi, 24(%rsp) movq %rsi, 16(%rsp) movq %rdx, 8(%rsp) movl %ecx, 4(%rsp) movq %fs:40, %rax movq %rax, 136(%rsp) xorl %eax, %eax leaq 24(%rsp), %rax movq %rax, 96(%rsp) leaq 16(%rsp), %rax movq %rax, 104(%rsp) leaq 8(%rsp), %rax movq %rax, 112(%rsp) leaq 4(%rsp), %rax movq %rax, 120(%rsp) movl $1, 48(%rsp) movl $1, 52(%rsp) movl $1, 56(%rsp) movl $1, 60(%rsp) movl $1, 64(%rsp) movl $1, 68(%rsp) leaq 40(%rsp), %rcx leaq 32(%rsp), %rdx leaq 60(%rsp), %rsi leaq 48(%rsp), %rdi call __cudaPopCallConfiguration@PLT testl %eax, %eax je .L27 .L23: movq 136(%rsp), %rax subq %fs:40, %rax jne .L28 addq $152, %rsp .cfi_remember_state .cfi_def_cfa_offset 8 ret .L27: .cfi_restore_state pushq 40(%rsp) .cfi_def_cfa_offset 168 pushq 40(%rsp) .cfi_def_cfa_offset 176 leaq 112(%rsp), %r9 movq 76(%rsp), %rcx movl 84(%rsp), %r8d movq 64(%rsp), %rsi movl 72(%rsp), %edx leaq _Z28ecuacion_kernel_inplace_sumaPdS_S_j(%rip), %rdi call cudaLaunchKernel@PLT addq $16, %rsp .cfi_def_cfa_offset 160 jmp .L23 .L28: call __stack_chk_fail@PLT .cfi_endproc .LFE2087: .size _Z53__device_stub__Z28ecuacion_kernel_inplace_sumaPdS_S_jPdS_S_j, .-_Z53__device_stub__Z28ecuacion_kernel_inplace_sumaPdS_S_jPdS_S_j .globl _Z28ecuacion_kernel_inplace_sumaPdS_S_j .type _Z28ecuacion_kernel_inplace_sumaPdS_S_j, @function _Z28ecuacion_kernel_inplace_sumaPdS_S_j: .LFB2088: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 call _Z53__device_stub__Z28ecuacion_kernel_inplace_sumaPdS_S_jPdS_S_j addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2088: .size _Z28ecuacion_kernel_inplace_sumaPdS_S_j, .-_Z28ecuacion_kernel_inplace_sumaPdS_S_j .globl _Z42__device_stub__Z17kernel_sum_MatrizPdS_S_jPdS_S_j .type _Z42__device_stub__Z17kernel_sum_MatrizPdS_S_jPdS_S_j, @function _Z42__device_stub__Z17kernel_sum_MatrizPdS_S_jPdS_S_j: .LFB2089: .cfi_startproc endbr64 subq $152, %rsp .cfi_def_cfa_offset 160 movq %rdi, 24(%rsp) movq %rsi, 16(%rsp) movq %rdx, 8(%rsp) movl %ecx, 4(%rsp) movq %fs:40, %rax movq %rax, 136(%rsp) xorl %eax, %eax leaq 24(%rsp), %rax movq %rax, 96(%rsp) leaq 16(%rsp), %rax movq %rax, 104(%rsp) leaq 8(%rsp), %rax movq %rax, 112(%rsp) leaq 4(%rsp), %rax movq %rax, 120(%rsp) movl $1, 48(%rsp) movl $1, 52(%rsp) movl $1, 56(%rsp) movl $1, 60(%rsp) movl $1, 64(%rsp) movl $1, 68(%rsp) leaq 40(%rsp), %rcx leaq 32(%rsp), %rdx leaq 60(%rsp), %rsi leaq 48(%rsp), %rdi call __cudaPopCallConfiguration@PLT testl %eax, %eax je .L35 .L31: movq 136(%rsp), %rax subq %fs:40, %rax jne .L36 addq $152, %rsp .cfi_remember_state .cfi_def_cfa_offset 8 ret .L35: .cfi_restore_state pushq 40(%rsp) .cfi_def_cfa_offset 168 pushq 40(%rsp) .cfi_def_cfa_offset 176 leaq 112(%rsp), %r9 movq 76(%rsp), %rcx movl 84(%rsp), %r8d movq 64(%rsp), %rsi movl 72(%rsp), %edx leaq _Z17kernel_sum_MatrizPdS_S_j(%rip), %rdi call cudaLaunchKernel@PLT addq $16, %rsp .cfi_def_cfa_offset 160 jmp .L31 .L36: call __stack_chk_fail@PLT .cfi_endproc .LFE2089: .size _Z42__device_stub__Z17kernel_sum_MatrizPdS_S_jPdS_S_j, .-_Z42__device_stub__Z17kernel_sum_MatrizPdS_S_jPdS_S_j .globl _Z17kernel_sum_MatrizPdS_S_j .type _Z17kernel_sum_MatrizPdS_S_j, @function _Z17kernel_sum_MatrizPdS_S_j: .LFB2090: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 call _Z42__device_stub__Z17kernel_sum_MatrizPdS_S_jPdS_S_j addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2090: .size _Z17kernel_sum_MatrizPdS_S_j, .-_Z17kernel_sum_MatrizPdS_S_j .globl _Z39__device_stub__Z18kernel_transpuestaPdiPdi .type _Z39__device_stub__Z18kernel_transpuestaPdiPdi, @function _Z39__device_stub__Z18kernel_transpuestaPdiPdi: .LFB2091: .cfi_startproc endbr64 subq $120, %rsp .cfi_def_cfa_offset 128 movq %rdi, 8(%rsp) movl %esi, 4(%rsp) movq %fs:40, %rax movq %rax, 104(%rsp) xorl %eax, %eax leaq 8(%rsp), %rax movq %rax, 80(%rsp) leaq 4(%rsp), %rax movq %rax, 88(%rsp) movl $1, 32(%rsp) movl $1, 36(%rsp) movl $1, 40(%rsp) movl $1, 44(%rsp) movl $1, 48(%rsp) movl $1, 52(%rsp) leaq 24(%rsp), %rcx leaq 16(%rsp), %rdx leaq 44(%rsp), %rsi leaq 32(%rsp), %rdi call __cudaPopCallConfiguration@PLT testl %eax, %eax je .L43 .L39: movq 104(%rsp), %rax subq %fs:40, %rax jne .L44 addq $120, %rsp .cfi_remember_state .cfi_def_cfa_offset 8 ret .L43: .cfi_restore_state pushq 24(%rsp) .cfi_def_cfa_offset 136 pushq 24(%rsp) .cfi_def_cfa_offset 144 leaq 96(%rsp), %r9 movq 60(%rsp), %rcx movl 68(%rsp), %r8d movq 48(%rsp), %rsi movl 56(%rsp), %edx leaq _Z18kernel_transpuestaPdi(%rip), %rdi call cudaLaunchKernel@PLT addq $16, %rsp .cfi_def_cfa_offset 128 jmp .L39 .L44: call __stack_chk_fail@PLT .cfi_endproc .LFE2091: .size _Z39__device_stub__Z18kernel_transpuestaPdiPdi, .-_Z39__device_stub__Z18kernel_transpuestaPdiPdi .globl _Z18kernel_transpuestaPdi .type _Z18kernel_transpuestaPdi, @function _Z18kernel_transpuestaPdi: .LFB2092: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 call _Z39__device_stub__Z18kernel_transpuestaPdiPdi addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2092: .size _Z18kernel_transpuestaPdi, .-_Z18kernel_transpuestaPdi .globl _Z47__device_stub__Z22kernel_mult_sum_matrizPdS_S_jPdS_S_j .type _Z47__device_stub__Z22kernel_mult_sum_matrizPdS_S_jPdS_S_j, @function _Z47__device_stub__Z22kernel_mult_sum_matrizPdS_S_jPdS_S_j: .LFB2093: .cfi_startproc endbr64 subq $152, %rsp .cfi_def_cfa_offset 160 movq %rdi, 24(%rsp) movq %rsi, 16(%rsp) movq %rdx, 8(%rsp) movl %ecx, 4(%rsp) movq %fs:40, %rax movq %rax, 136(%rsp) xorl %eax, %eax leaq 24(%rsp), %rax movq %rax, 96(%rsp) leaq 16(%rsp), %rax movq %rax, 104(%rsp) leaq 8(%rsp), %rax movq %rax, 112(%rsp) leaq 4(%rsp), %rax movq %rax, 120(%rsp) movl $1, 48(%rsp) movl $1, 52(%rsp) movl $1, 56(%rsp) movl $1, 60(%rsp) movl $1, 64(%rsp) movl $1, 68(%rsp) leaq 40(%rsp), %rcx leaq 32(%rsp), %rdx leaq 60(%rsp), %rsi leaq 48(%rsp), %rdi call __cudaPopCallConfiguration@PLT testl %eax, %eax je .L51 .L47: movq 136(%rsp), %rax subq %fs:40, %rax jne .L52 addq $152, %rsp .cfi_remember_state .cfi_def_cfa_offset 8 ret .L51: .cfi_restore_state pushq 40(%rsp) .cfi_def_cfa_offset 168 pushq 40(%rsp) .cfi_def_cfa_offset 176 leaq 112(%rsp), %r9 movq 76(%rsp), %rcx movl 84(%rsp), %r8d movq 64(%rsp), %rsi movl 72(%rsp), %edx leaq _Z22kernel_mult_sum_matrizPdS_S_j(%rip), %rdi call cudaLaunchKernel@PLT addq $16, %rsp .cfi_def_cfa_offset 160 jmp .L47 .L52: call __stack_chk_fail@PLT .cfi_endproc .LFE2093: .size _Z47__device_stub__Z22kernel_mult_sum_matrizPdS_S_jPdS_S_j, .-_Z47__device_stub__Z22kernel_mult_sum_matrizPdS_S_jPdS_S_j .globl _Z22kernel_mult_sum_matrizPdS_S_j .type _Z22kernel_mult_sum_matrizPdS_S_j, @function _Z22kernel_mult_sum_matrizPdS_S_j: .LFB2094: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 call _Z47__device_stub__Z22kernel_mult_sum_matrizPdS_S_jPdS_S_j addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2094: .size _Z22kernel_mult_sum_matrizPdS_S_j, .-_Z22kernel_mult_sum_matrizPdS_S_j .section .rodata.str1.1,"aMS",@progbits,1 .LC1: .string "Falta argumento: N, CUDABLK\n" .section .rodata.str1.8,"aMS",@progbits,1 .align 8 .LC3: .string "Tiempo para la ecuacion CPU: %f\n\n" .align 8 .LC4: .string "Tiempo para la ecuacion out-place GPU: %f\n" .section .rodata.str1.1 .LC5: .string "error: %d\n\n" .section .rodata.str1.8 .align 8 .LC6: .string "Tiempo para la ecuacion in-place GPU: %f\n" .text .globl main .type main, @function main: .LFB2058: .cfi_startproc endbr64 pushq %r15 .cfi_def_cfa_offset 16 .cfi_offset 15, -16 pushq %r14 .cfi_def_cfa_offset 24 .cfi_offset 14, -24 pushq %r13 .cfi_def_cfa_offset 32 .cfi_offset 13, -32 pushq %r12 .cfi_def_cfa_offset 40 .cfi_offset 12, -40 pushq %rbp .cfi_def_cfa_offset 48 .cfi_offset 6, -48 pushq %rbx .cfi_def_cfa_offset 56 .cfi_offset 3, -56 subq $120, %rsp .cfi_def_cfa_offset 176 movq %fs:40, %rax movq %rax, 104(%rsp) xorl %eax, %eax cmpl $3, %edi je .L56 leaq .LC1(%rip), %rsi movl $2, %edi call __printf_chk@PLT .L57: movq 104(%rsp), %rax subq %fs:40, %rax jne .L95 movl $0, %eax addq $120, %rsp .cfi_remember_state .cfi_def_cfa_offset 56 popq %rbx .cfi_def_cfa_offset 48 popq %rbp .cfi_def_cfa_offset 40 popq %r12 .cfi_def_cfa_offset 32 popq %r13 .cfi_def_cfa_offset 24 popq %r14 .cfi_def_cfa_offset 16 popq %r15 .cfi_def_cfa_offset 8 ret .L56: .cfi_restore_state movq %rsi, %rbx movq 8(%rsi), %rdi movl $10, %edx movl $0, %esi call __isoc23_strtol@PLT movq %rax, %r14 movq %rax, (%rsp) movl %eax, 20(%rsp) movl %eax, %r13d movq 16(%rbx), %rdi movl $10, %edx movl $0, %esi call __isoc23_strtol@PLT cltq movq %rax, 8(%rsp) movl %r14d, %eax imulq %rax, %rax leaq 0(,%rax,8), %r15 movq %r15, %rdi call malloc@PLT movq %rax, %rbp movq %r15, %rdi call malloc@PLT movq %rax, %rbx movq %r15, %rdi call malloc@PLT movq %rax, %r12 movl %r14d, %eax imull %r14d, %eax movl %eax, 16(%rsp) testl %eax, %eax je .L58 movl %eax, %ecx movl $0, %eax .L61: movl %eax, %edx pxor %xmm0, %xmm0 cvtsi2sdq %rdx, %xmm0 movsd %xmm0, 0(%rbp,%rax,8) movsd %xmm0, (%rbx,%rax,8) movq $0x000000000, (%r12,%rax,8) addq $1, %rax cmpq %rcx, %rax jne .L61 .L58: leaq 40(%rsp), %rdi movq %r15, %rsi call cudaMalloc@PLT leaq 64(%rsp), %rdi movq %r15, %rsi call cudaMalloc@PLT leaq 48(%rsp), %rdi movq %r15, %rsi call cudaMalloc@PLT leaq 72(%rsp), %rdi movq %r15, %rsi call cudaMalloc@PLT leaq 56(%rsp), %rdi movq %r15, %rsi call cudaMalloc@PLT movl 16(%rsp), %r14d movq %r14, %rax movq 8(%rsp), %rsi movl $0, %edx divq %rsi movl $0, %edx divq %rsi pxor %xmm0, %xmm0 cvtsi2sdq %rax, %xmm0 call sqrt@PLT cvttsd2siq %xmm0, %rax movq 8(%rsp), %rsi movl %esi, 80(%rsp) movl %esi, 84(%rsp) movl $1, 88(%rsp) movl %eax, 92(%rsp) movl %eax, 96(%rsp) movl $1, 100(%rsp) call _Z9dwalltimev movsd %xmm0, 8(%rsp) cmpl $0, (%rsp) je .L65 movl (%rsp), %edi movl $0, %edx movl $0, %r10d movl $0, %r8d movq %r14, 24(%rsp) .L64: movl %edx, %r9d movl %r10d, %esi .L69: movl %esi, %eax leaq (%r12,%rax,8), %r11 movsd (%r11), %xmm1 movl %r10d, %eax .L66: leal (%rax,%r9), %r14d movl %eax, %ecx movsd (%rbx,%r14,8), %xmm0 mulsd 0(%rbp,%rcx,8), %xmm0 addsd %xmm0, %xmm1 addl $1, %eax cmpl %edi, %eax jne .L66 movsd %xmm1, (%r11) addl $1, %esi addl %r13d, %r9d cmpl %edi, %esi jne .L69 leal 1(%r8), %r9d addl %r13d, %edi addl %r13d, %r10d subl %r13d, %edx cmpl %r9d, %r13d je .L87 movl %r9d, %r8d jmp .L64 .L87: movl %r9d, %r10d movl $0, %edi movl $0, %esi .L68: movl %esi, %edx movl %edi, %eax .L70: movl %eax, %r11d leaq (%r12,%r11,8), %rcx movl %edx, %r14d movsd 0(%rbp,%r14,8), %xmm0 addsd (%rbx,%r11,8), %xmm0 addsd (%rcx), %xmm0 movsd %xmm0, (%rcx) addl $1, %eax addl %r9d, %edx cmpl %eax, %r10d jne .L70 leal 1(%rsi), %eax addl %r9d, %edi addl %r9d, %r10d cmpl %esi, %r8d je .L93 movl %eax, %esi jmp .L68 .L93: movq 24(%rsp), %r14 .L65: call _Z9dwalltimev subsd 8(%rsp), %xmm0 leaq .LC3(%rip), %rsi movl $2, %edi movl $1, %eax call __printf_chk@PLT cmpl $0, 16(%rsp) je .L71 movl $0, %eax .L74: movl %eax, %edx pxor %xmm0, %xmm0 cvtsi2sdq %rdx, %xmm0 movsd %xmm0, 0(%rbp,%rax,8) movsd %xmm0, (%rbx,%rax,8) movq $0x000000000, (%r12,%rax,8) addq $1, %rax cmpq %rax, %r14 jne .L74 .L71: movl $1, %ecx movq %r15, %rdx movq %rbp, %rsi movq 40(%rsp), %rdi call cudaMemcpy@PLT movl $1, %ecx movq %r15, %rdx movq %rbx, %rsi movq 48(%rsp), %rdi call cudaMemcpy@PLT movl $1, %ecx movq %r15, %rdx movq %r12, %rsi movq 56(%rsp), %rdi call cudaMemcpy@PLT call _Z9dwalltimev movsd %xmm0, 8(%rsp) movl 88(%rsp), %ecx movl $0, %r9d movl $0, %r8d movq 80(%rsp), %rdx movq 92(%rsp), %rdi movl 100(%rsp), %esi call __cudaPushCallConfiguration@PLT testl %eax, %eax je .L96 .L75: call cudaThreadSynchronize@PLT movl 88(%rsp), %ecx movl $0, %r9d movl $0, %r8d movq 80(%rsp), %rdx movq 92(%rsp), %rdi movl 100(%rsp), %esi call __cudaPushCallConfiguration@PLT testl %eax, %eax je .L97 .L76: call cudaThreadSynchronize@PLT call _Z9dwalltimev subsd 8(%rsp), %xmm0 leaq .LC4(%rip), %rsi movl $2, %edi movl $1, %eax call __printf_chk@PLT call cudaGetLastError@PLT movl %eax, %edx leaq .LC5(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT movl $2, %ecx movq %r15, %rdx movq 56(%rsp), %rsi movq %r12, %rdi call cudaMemcpy@PLT movq 40(%rsp), %rdi call cudaFree@PLT movq 48(%rsp), %rdi call cudaFree@PLT movq 56(%rsp), %rdi call cudaFree@PLT movq 64(%rsp), %rdi call cudaFree@PLT movq 72(%rsp), %rdi call cudaFree@PLT cmpl $0, 16(%rsp) je .L77 movl $0, %eax .L80: movl %eax, %edx pxor %xmm0, %xmm0 cvtsi2sdq %rdx, %xmm0 movsd %xmm0, 0(%rbp,%rax,8) movsd %xmm0, (%rbx,%rax,8) movq $0x000000000, (%r12,%rax,8) addq $1, %rax cmpq %rax, %r14 jne .L80 .L77: leaq 40(%rsp), %rdi movq %r15, %rsi call cudaMalloc@PLT leaq 48(%rsp), %rdi movq %r15, %rsi call cudaMalloc@PLT leaq 56(%rsp), %rdi movq %r15, %rsi call cudaMalloc@PLT movl $1, %ecx movq %r15, %rdx movq %rbp, %rsi movq 40(%rsp), %rdi call cudaMemcpy@PLT movl $1, %ecx movq %r15, %rdx movq %rbx, %rsi movq 48(%rsp), %rdi call cudaMemcpy@PLT movl $1, %ecx movq %r15, %rdx movq %r12, %rsi movq 56(%rsp), %rdi call cudaMemcpy@PLT call _Z9dwalltimev movsd %xmm0, (%rsp) movl 88(%rsp), %ecx movl $0, %r9d movl $0, %r8d movq 80(%rsp), %rdx movq 92(%rsp), %rdi movl 100(%rsp), %esi call __cudaPushCallConfiguration@PLT testl %eax, %eax je .L98 .L81: call cudaThreadSynchronize@PLT movl 88(%rsp), %ecx movl $0, %r9d movl $0, %r8d movq 80(%rsp), %rdx movq 92(%rsp), %rdi movl 100(%rsp), %esi call __cudaPushCallConfiguration@PLT testl %eax, %eax je .L99 .L82: call cudaThreadSynchronize@PLT movl 88(%rsp), %ecx movl $0, %r9d movl $0, %r8d movq 80(%rsp), %rdx movq 92(%rsp), %rdi movl 100(%rsp), %esi call __cudaPushCallConfiguration@PLT testl %eax, %eax je .L100 .L83: call cudaThreadSynchronize@PLT movl 88(%rsp), %ecx movl $0, %r9d movl $0, %r8d movq 80(%rsp), %rdx movq 92(%rsp), %rdi movl 100(%rsp), %esi call __cudaPushCallConfiguration@PLT testl %eax, %eax je .L101 .L84: call cudaThreadSynchronize@PLT call _Z9dwalltimev subsd (%rsp), %xmm0 leaq .LC6(%rip), %rsi movl $2, %edi movl $1, %eax call __printf_chk@PLT call cudaGetLastError@PLT movl %eax, %edx leaq .LC5(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT movl $2, %ecx movq %r15, %rdx movq 56(%rsp), %rsi movq %r12, %rdi call cudaMemcpy@PLT movq 40(%rsp), %rdi call cudaFree@PLT movq 48(%rsp), %rdi call cudaFree@PLT movq 56(%rsp), %rdi call cudaFree@PLT movq %rbp, %rdi call free@PLT movq %rbx, %rdi call free@PLT movq %r12, %rdi call free@PLT jmp .L57 .L96: movl (%rsp), %r8d movq 72(%rsp), %rcx movq 48(%rsp), %rdx movq 64(%rsp), %rsi movq 40(%rsp), %rdi call _Z54__device_stub__Z27ecuacion_kernel_outplace_p1PdS_S_S_jPdS_S_S_j jmp .L75 .L97: movl (%rsp), %r9d movq 72(%rsp), %r8 movq 64(%rsp), %rcx movq 56(%rsp), %rdx movq 48(%rsp), %rsi movq 40(%rsp), %rdi call _Z56__device_stub__Z27ecuacion_kernel_outplace_p2PdS_S_S_S_jPdS_S_S_S_j jmp .L76 .L98: movl 20(%rsp), %esi movq 40(%rsp), %rdi call _Z39__device_stub__Z18kernel_transpuestaPdiPdi jmp .L81 .L99: movl %r13d, %ecx movq 56(%rsp), %rdx movq 48(%rsp), %rsi movq 40(%rsp), %rdi call _Z42__device_stub__Z17kernel_sum_MatrizPdS_S_jPdS_S_j jmp .L82 .L100: movl 20(%rsp), %esi movq 40(%rsp), %rdi call _Z39__device_stub__Z18kernel_transpuestaPdiPdi jmp .L83 .L101: movl %r13d, %ecx movq 56(%rsp), %rdx movq 48(%rsp), %rsi movq 40(%rsp), %rdi call _Z47__device_stub__Z22kernel_mult_sum_matrizPdS_S_jPdS_S_j jmp .L84 .L95: call __stack_chk_fail@PLT .cfi_endproc .LFE2058: .size main, .-main .section .rodata.str1.8 .align 8 .LC7: .string "_Z22kernel_mult_sum_matrizPdS_S_j" .section .rodata.str1.1 .LC8: .string "_Z18kernel_transpuestaPdi" .LC9: .string "_Z17kernel_sum_MatrizPdS_S_j" .section .rodata.str1.8 .align 8 .LC10: .string "_Z28ecuacion_kernel_inplace_sumaPdS_S_j" .align 8 .LC11: .string "_Z27ecuacion_kernel_outplace_p2PdS_S_S_S_j" .align 8 .LC12: .string "_Z27ecuacion_kernel_outplace_p1PdS_S_S_j" .text .type _ZL24__sti____cudaRegisterAllv, @function _ZL24__sti____cudaRegisterAllv: .LFB2096: .cfi_startproc endbr64 pushq %rbx .cfi_def_cfa_offset 16 .cfi_offset 3, -16 leaq _ZL15__fatDeviceText(%rip), %rdi call __cudaRegisterFatBinary@PLT movq %rax, %rbx movq %rax, _ZL20__cudaFatCubinHandle(%rip) pushq $0 .cfi_def_cfa_offset 24 pushq $0 .cfi_def_cfa_offset 32 pushq $0 .cfi_def_cfa_offset 40 pushq $0 .cfi_def_cfa_offset 48 movl $0, %r9d movl $-1, %r8d leaq .LC7(%rip), %rdx movq %rdx, %rcx leaq _Z22kernel_mult_sum_matrizPdS_S_j(%rip), %rsi movq %rax, %rdi call __cudaRegisterFunction@PLT addq $32, %rsp .cfi_def_cfa_offset 16 pushq $0 .cfi_def_cfa_offset 24 pushq $0 .cfi_def_cfa_offset 32 pushq $0 .cfi_def_cfa_offset 40 pushq $0 .cfi_def_cfa_offset 48 movl $0, %r9d movl $-1, %r8d leaq .LC8(%rip), %rdx movq %rdx, %rcx leaq _Z18kernel_transpuestaPdi(%rip), %rsi movq %rbx, %rdi call __cudaRegisterFunction@PLT addq $32, %rsp .cfi_def_cfa_offset 16 pushq $0 .cfi_def_cfa_offset 24 pushq $0 .cfi_def_cfa_offset 32 pushq $0 .cfi_def_cfa_offset 40 pushq $0 .cfi_def_cfa_offset 48 movl $0, %r9d movl $-1, %r8d leaq .LC9(%rip), %rdx movq %rdx, %rcx leaq _Z17kernel_sum_MatrizPdS_S_j(%rip), %rsi movq %rbx, %rdi call __cudaRegisterFunction@PLT addq $32, %rsp .cfi_def_cfa_offset 16 pushq $0 .cfi_def_cfa_offset 24 pushq $0 .cfi_def_cfa_offset 32 pushq $0 .cfi_def_cfa_offset 40 pushq $0 .cfi_def_cfa_offset 48 movl $0, %r9d movl $-1, %r8d leaq .LC10(%rip), %rdx movq %rdx, %rcx leaq _Z28ecuacion_kernel_inplace_sumaPdS_S_j(%rip), %rsi movq %rbx, %rdi call __cudaRegisterFunction@PLT addq $32, %rsp .cfi_def_cfa_offset 16 pushq $0 .cfi_def_cfa_offset 24 pushq $0 .cfi_def_cfa_offset 32 pushq $0 .cfi_def_cfa_offset 40 pushq $0 .cfi_def_cfa_offset 48 movl $0, %r9d movl $-1, %r8d leaq .LC11(%rip), %rdx movq %rdx, %rcx leaq _Z27ecuacion_kernel_outplace_p2PdS_S_S_S_j(%rip), %rsi movq %rbx, %rdi call __cudaRegisterFunction@PLT addq $32, %rsp .cfi_def_cfa_offset 16 pushq $0 .cfi_def_cfa_offset 24 pushq $0 .cfi_def_cfa_offset 32 pushq $0 .cfi_def_cfa_offset 40 pushq $0 .cfi_def_cfa_offset 48 movl $0, %r9d movl $-1, %r8d leaq .LC12(%rip), %rdx movq %rdx, %rcx leaq _Z27ecuacion_kernel_outplace_p1PdS_S_S_j(%rip), %rsi movq %rbx, %rdi call __cudaRegisterFunction@PLT addq $32, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaRegisterFatBinaryEnd@PLT leaq _ZL26__cudaUnregisterBinaryUtilv(%rip), %rdi call atexit@PLT popq %rbx .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2096: .size _ZL24__sti____cudaRegisterAllv, .-_ZL24__sti____cudaRegisterAllv .section .init_array,"aw" .align 8 .quad _ZL24__sti____cudaRegisterAllv .section .nvFatBinSegment,"aw" .align 8 .type _ZL15__fatDeviceText, @object .size _ZL15__fatDeviceText, 24 _ZL15__fatDeviceText: .long 1180844977 .long 1 .quad fatbinData .quad 0 .local _ZL20__cudaFatCubinHandle .comm _ZL20__cudaFatCubinHandle,8,8 .section .rodata.cst8,"aM",@progbits,8 .align 8 .LC0: .long 0 .long 1093567616 .ident "GCC: (Ubuntu 13.3.0-6ubuntu2~24.04) 13.3.0" .section .note.GNU-stack,"",@progbits .section .note.gnu.property,"a" .align 8 .long 1f - 0f .long 4f - 1f .long 5 0: .string "GNU" 1: .align 8 .long 0xc0000002 .long 3f - 2f 2: .long 0x3 3: .align 8 4:
You are a helpful coding assistant specialized in converting CUDA source code to HIP/AMD source code.
#include <stdio.h> #include <stdlib.h> #include <sys/time.h> #include <sys/resource.h> #include <math.h> //134217728 double dwalltime(){ double sec; struct timeval tv; gettimeofday(&tv,NULL); sec = tv.tv_sec + tv.tv_usec/1000000.0; return sec; } __global__ void ecuacion_kernel_outplace_p1(double *d_matA,double *d_matAT,double *d_matB,double *d_matBT, unsigned int n){ int distA = blockIdx.y * blockDim.y + threadIdx.y; //i int distB = blockIdx.x * blockDim.x + threadIdx.x; //j //transpuesta out-place A y B if( (distA<n*n) && (distB<n*n) ){ d_matAT [distB*n + distA] = d_matA[distA*n + distB]; d_matBT [distB*n + distA] = d_matB[distA*n + distB]; } } __global__ void ecuacion_kernel_outplace_p2(double *d_matA,double *d_matB,double *d_matC,double *d_matAT,double *d_matBT, unsigned int n){ int distA = blockIdx.y * blockDim.y + threadIdx.y; //i int distB = blockIdx.x * blockDim.x + threadIdx.x; //j int k; if (distA*n+distB <= (n*n - 1)){ //multiplicacion for(k = 0; k < n ;k++){ d_matC[distA*n+distB] += d_matA[distA*n+k] * d_matBT[distB+k*n]; } //suma d_matC[distA*n+distB] += d_matB[distA*n+distB] + d_matAT[distA*n+distB]; } } __global__ void ecuacion_kernel_inplace_suma (double *d_matA,double *d_matB,double *d_matC, unsigned int n){ int distA = blockIdx.y * blockDim.y + threadIdx.y; //i int distB = blockIdx.x * blockDim.x + threadIdx.x; //j int k; //multiplicacion if (distA*n+distB < (n*n - 1)){ d_matC[distA*n+distB] += d_matB[distA*n+distB] + d_matA[distA+distB*n]; for(k = 0; k < n ;k++){ d_matC[distA*n+distB] += d_matA[distA*n+k] * d_matB[distB*n+k]; } } } __global__ void kernel_sum_Matriz (double *d_matA,double *d_matB,double *d_matC, unsigned int n){ int distA = blockIdx.y * blockDim.y + threadIdx.y; //i int distB = blockIdx.x * blockDim.x + threadIdx.x; //j //suma if (distA*n+distB < (n*n)){ d_matC[distA*n+distB] += d_matA[distA*n+distB] + d_matB[distA+distB*n]; } } __global__ void kernel_transpuesta(double *m, int N){ int tid = blockIdx.x*blockDim.x + threadIdx.x; int i = int((1 + sqrtf(1 + 8*tid)) / 2); int j = tid - (i*(i-1)/2); int aux; if ( (i<N) && (j<N) ){ aux = m[i*N + j] ; m[i*N + j] = m[j*N + i]; m[j*N + i] = aux; } } __global__ void kernel_mult_sum_matriz (double *d_matA,double *d_matB,double *d_matC, unsigned int n){ int distA = blockIdx.y * blockDim.y + threadIdx.y; //i int distB = blockIdx.x * blockDim.x + threadIdx.x; //j int k; //multiplicacion if (distA*n+distB < (n*n)){ for(k = 0; k < n ;k++){ d_matC[distA*n+distB] += d_matA[distA*n+k] * d_matB[distB*n+k]; } } } int main(int argc, char *argv[]){ if (argc != 3){ printf("Falta argumento: N, CUDABLK\n"); return 0; } //declaracion de variables cudaError_t error; unsigned int N = atoi (argv[1]); unsigned long CUDA_BLK = atoi (argv[2]), gridBlock; unsigned long numBytes = sizeof(double)*N*N; double *matA,*matB,*matC,*d_matA,*d_matB,*d_matC,*d_matAT,*d_matBT,timetick; unsigned int i,j,k; //inicializa variables para cpu matA = (double *)malloc(numBytes); matB = (double *)malloc(numBytes); matC = (double *)malloc(numBytes); for (i = 0; i < N*N; i++){ matA[i] = i; matB[i] = i; matC[i] = 0; } //inicializa variables para gpu cudaMalloc((void **) &d_matA, numBytes); cudaMalloc((void **) &d_matAT, numBytes); cudaMalloc((void **) &d_matB, numBytes); cudaMalloc((void **) &d_matBT, numBytes); cudaMalloc((void **) &d_matC, numBytes); gridBlock = (unsigned int)sqrt(N*N/CUDA_BLK/CUDA_BLK); dim3 dimBlock(CUDA_BLK,CUDA_BLK); // Bloque bidimencional de hilos (*cb* hilos) dim3 dimGrid(gridBlock,gridBlock); // Grid bidimencional (*ceil(n/cb)* bloques) //--------------------------------cpu comienza ------------------------------------ //secuencial timetick = dwalltime(); //multiplicacion for(i = 0; i < N; i++){ for(j = 0; j < N; j++){ for(k = 0; k < N ;k++){ matC[i*N+j] += matA[i*N+k] * matB[j*N+k]; //multiplica a matB por fila, eso simula la matB transpuesta } } } //suma for(i = 0; i < N; i++){ for(j = 0; j < N; j++){ matC[i*N+j] += matB[i*N+j] + matA[i+j*N]; } } printf("Tiempo para la ecuacion CPU: %f\n\n",dwalltime() - timetick); /* for(i = 0; i < N; i++){ for(j = 0; j < N; j++){ printf("%f|",matC[i*N+j]); } printf("\n"); } printf("\n"); */ //--------------------------------cpu termina ------------------------------------ for (i = 0; i < N*N; i++){ matA[i] = i; matB[i] = i; matC[i] = 0; } cudaMemcpy(d_matA, matA, numBytes, cudaMemcpyHostToDevice); // CPU -> GPU cudaMemcpy(d_matB, matB, numBytes, cudaMemcpyHostToDevice); // CPU -> GPU cudaMemcpy(d_matC, matC, numBytes, cudaMemcpyHostToDevice); // CPU -> GPU //--------------------------------gpu out-place comienza ------------------------------------ timetick = dwalltime(); ecuacion_kernel_outplace_p1<<<dimGrid, dimBlock>>>(d_matA, d_matAT,d_matB,d_matBT, N); cudaThreadSynchronize(); ecuacion_kernel_outplace_p2<<<dimGrid, dimBlock>>>(d_matA, d_matB,d_matC,d_matAT,d_matBT, N); cudaThreadSynchronize(); printf("Tiempo para la ecuacion out-place GPU: %f\n",dwalltime() - timetick); error = cudaGetLastError(); printf("error: %d\n\n",error); cudaMemcpy(matC, d_matC, numBytes, cudaMemcpyDeviceToHost); // GPU -> CPU /* for(i = 0; i < N; i++){ for(j = 0; j < N; j++){ printf("%f|",matC[i*N+j]); } printf("\n"); } printf("\n"); */ //--------------------------------gpu out-place termina ------------------------------------ cudaFree(d_matA); cudaFree(d_matB); cudaFree(d_matC); cudaFree(d_matAT); cudaFree(d_matBT); for (i = 0; i < N*N; i++){ matA[i] = i; matB[i] = i; matC[i] = 0; } cudaMalloc((void **) &d_matA, numBytes); cudaMalloc((void **) &d_matB, numBytes); cudaMalloc((void **) &d_matC, numBytes); cudaMemcpy(d_matA, matA, numBytes, cudaMemcpyHostToDevice); // CPU -> GPU cudaMemcpy(d_matB, matB, numBytes, cudaMemcpyHostToDevice); // CPU -> GPU cudaMemcpy(d_matC, matC, numBytes, cudaMemcpyHostToDevice); // CPU -> GPU //--------------------------------gpu in-place comienza ------------------------------------ timetick = dwalltime(); kernel_transpuesta<<<dimGrid, dimBlock>>>(d_matA, N); cudaThreadSynchronize(); kernel_sum_Matriz<<<dimGrid, dimBlock>>>(d_matA, d_matB,d_matC, N); cudaThreadSynchronize(); kernel_transpuesta<<<dimGrid, dimBlock>>>(d_matA, N); cudaThreadSynchronize(); kernel_mult_sum_matriz<<<dimGrid, dimBlock>>>(d_matA, d_matB,d_matC, N); cudaThreadSynchronize(); printf("Tiempo para la ecuacion in-place GPU: %f\n",dwalltime() - timetick); error = cudaGetLastError(); printf("error: %d\n\n",error); cudaMemcpy(matC, d_matC, numBytes, cudaMemcpyDeviceToHost); // GPU -> CPU //--------------------------------gpu in-place termina ------------------------------------ cudaFree(d_matA); cudaFree(d_matB); cudaFree(d_matC); /* //imprime la matriz matC for(i = 0; i < N; i++){ for(j = 0; j < N; j++){ printf("%f|",matC[i*N+j]); } printf("\n"); } printf("\n"); */ free(matA); free(matB); free(matC); return 0; }
#include <hip/hip_runtime.h> #include <stdio.h> #include <stdlib.h> #include <sys/time.h> #include <sys/resource.h> #include <math.h> //134217728 double dwalltime(){ double sec; struct timeval tv; gettimeofday(&tv,NULL); sec = tv.tv_sec + tv.tv_usec/1000000.0; return sec; } __global__ void ecuacion_kernel_outplace_p1(double *d_matA,double *d_matAT,double *d_matB,double *d_matBT, unsigned int n){ int distA = blockIdx.y * blockDim.y + threadIdx.y; //i int distB = blockIdx.x * blockDim.x + threadIdx.x; //j //transpuesta out-place A y B if( (distA<n*n) && (distB<n*n) ){ d_matAT [distB*n + distA] = d_matA[distA*n + distB]; d_matBT [distB*n + distA] = d_matB[distA*n + distB]; } } __global__ void ecuacion_kernel_outplace_p2(double *d_matA,double *d_matB,double *d_matC,double *d_matAT,double *d_matBT, unsigned int n){ int distA = blockIdx.y * blockDim.y + threadIdx.y; //i int distB = blockIdx.x * blockDim.x + threadIdx.x; //j int k; if (distA*n+distB <= (n*n - 1)){ //multiplicacion for(k = 0; k < n ;k++){ d_matC[distA*n+distB] += d_matA[distA*n+k] * d_matBT[distB+k*n]; } //suma d_matC[distA*n+distB] += d_matB[distA*n+distB] + d_matAT[distA*n+distB]; } } __global__ void ecuacion_kernel_inplace_suma (double *d_matA,double *d_matB,double *d_matC, unsigned int n){ int distA = blockIdx.y * blockDim.y + threadIdx.y; //i int distB = blockIdx.x * blockDim.x + threadIdx.x; //j int k; //multiplicacion if (distA*n+distB < (n*n - 1)){ d_matC[distA*n+distB] += d_matB[distA*n+distB] + d_matA[distA+distB*n]; for(k = 0; k < n ;k++){ d_matC[distA*n+distB] += d_matA[distA*n+k] * d_matB[distB*n+k]; } } } __global__ void kernel_sum_Matriz (double *d_matA,double *d_matB,double *d_matC, unsigned int n){ int distA = blockIdx.y * blockDim.y + threadIdx.y; //i int distB = blockIdx.x * blockDim.x + threadIdx.x; //j //suma if (distA*n+distB < (n*n)){ d_matC[distA*n+distB] += d_matA[distA*n+distB] + d_matB[distA+distB*n]; } } __global__ void kernel_transpuesta(double *m, int N){ int tid = blockIdx.x*blockDim.x + threadIdx.x; int i = int((1 + sqrtf(1 + 8*tid)) / 2); int j = tid - (i*(i-1)/2); int aux; if ( (i<N) && (j<N) ){ aux = m[i*N + j] ; m[i*N + j] = m[j*N + i]; m[j*N + i] = aux; } } __global__ void kernel_mult_sum_matriz (double *d_matA,double *d_matB,double *d_matC, unsigned int n){ int distA = blockIdx.y * blockDim.y + threadIdx.y; //i int distB = blockIdx.x * blockDim.x + threadIdx.x; //j int k; //multiplicacion if (distA*n+distB < (n*n)){ for(k = 0; k < n ;k++){ d_matC[distA*n+distB] += d_matA[distA*n+k] * d_matB[distB*n+k]; } } } int main(int argc, char *argv[]){ if (argc != 3){ printf("Falta argumento: N, CUDABLK\n"); return 0; } //declaracion de variables hipError_t error; unsigned int N = atoi (argv[1]); unsigned long CUDA_BLK = atoi (argv[2]), gridBlock; unsigned long numBytes = sizeof(double)*N*N; double *matA,*matB,*matC,*d_matA,*d_matB,*d_matC,*d_matAT,*d_matBT,timetick; unsigned int i,j,k; //inicializa variables para cpu matA = (double *)malloc(numBytes); matB = (double *)malloc(numBytes); matC = (double *)malloc(numBytes); for (i = 0; i < N*N; i++){ matA[i] = i; matB[i] = i; matC[i] = 0; } //inicializa variables para gpu hipMalloc((void **) &d_matA, numBytes); hipMalloc((void **) &d_matAT, numBytes); hipMalloc((void **) &d_matB, numBytes); hipMalloc((void **) &d_matBT, numBytes); hipMalloc((void **) &d_matC, numBytes); gridBlock = (unsigned int)sqrt(N*N/CUDA_BLK/CUDA_BLK); dim3 dimBlock(CUDA_BLK,CUDA_BLK); // Bloque bidimencional de hilos (*cb* hilos) dim3 dimGrid(gridBlock,gridBlock); // Grid bidimencional (*ceil(n/cb)* bloques) //--------------------------------cpu comienza ------------------------------------ //secuencial timetick = dwalltime(); //multiplicacion for(i = 0; i < N; i++){ for(j = 0; j < N; j++){ for(k = 0; k < N ;k++){ matC[i*N+j] += matA[i*N+k] * matB[j*N+k]; //multiplica a matB por fila, eso simula la matB transpuesta } } } //suma for(i = 0; i < N; i++){ for(j = 0; j < N; j++){ matC[i*N+j] += matB[i*N+j] + matA[i+j*N]; } } printf("Tiempo para la ecuacion CPU: %f\n\n",dwalltime() - timetick); /* for(i = 0; i < N; i++){ for(j = 0; j < N; j++){ printf("%f|",matC[i*N+j]); } printf("\n"); } printf("\n"); */ //--------------------------------cpu termina ------------------------------------ for (i = 0; i < N*N; i++){ matA[i] = i; matB[i] = i; matC[i] = 0; } hipMemcpy(d_matA, matA, numBytes, hipMemcpyHostToDevice); // CPU -> GPU hipMemcpy(d_matB, matB, numBytes, hipMemcpyHostToDevice); // CPU -> GPU hipMemcpy(d_matC, matC, numBytes, hipMemcpyHostToDevice); // CPU -> GPU //--------------------------------gpu out-place comienza ------------------------------------ timetick = dwalltime(); ecuacion_kernel_outplace_p1<<<dimGrid, dimBlock>>>(d_matA, d_matAT,d_matB,d_matBT, N); hipDeviceSynchronize(); ecuacion_kernel_outplace_p2<<<dimGrid, dimBlock>>>(d_matA, d_matB,d_matC,d_matAT,d_matBT, N); hipDeviceSynchronize(); printf("Tiempo para la ecuacion out-place GPU: %f\n",dwalltime() - timetick); error = hipGetLastError(); printf("error: %d\n\n",error); hipMemcpy(matC, d_matC, numBytes, hipMemcpyDeviceToHost); // GPU -> CPU /* for(i = 0; i < N; i++){ for(j = 0; j < N; j++){ printf("%f|",matC[i*N+j]); } printf("\n"); } printf("\n"); */ //--------------------------------gpu out-place termina ------------------------------------ hipFree(d_matA); hipFree(d_matB); hipFree(d_matC); hipFree(d_matAT); hipFree(d_matBT); for (i = 0; i < N*N; i++){ matA[i] = i; matB[i] = i; matC[i] = 0; } hipMalloc((void **) &d_matA, numBytes); hipMalloc((void **) &d_matB, numBytes); hipMalloc((void **) &d_matC, numBytes); hipMemcpy(d_matA, matA, numBytes, hipMemcpyHostToDevice); // CPU -> GPU hipMemcpy(d_matB, matB, numBytes, hipMemcpyHostToDevice); // CPU -> GPU hipMemcpy(d_matC, matC, numBytes, hipMemcpyHostToDevice); // CPU -> GPU //--------------------------------gpu in-place comienza ------------------------------------ timetick = dwalltime(); kernel_transpuesta<<<dimGrid, dimBlock>>>(d_matA, N); hipDeviceSynchronize(); kernel_sum_Matriz<<<dimGrid, dimBlock>>>(d_matA, d_matB,d_matC, N); hipDeviceSynchronize(); kernel_transpuesta<<<dimGrid, dimBlock>>>(d_matA, N); hipDeviceSynchronize(); kernel_mult_sum_matriz<<<dimGrid, dimBlock>>>(d_matA, d_matB,d_matC, N); hipDeviceSynchronize(); printf("Tiempo para la ecuacion in-place GPU: %f\n",dwalltime() - timetick); error = hipGetLastError(); printf("error: %d\n\n",error); hipMemcpy(matC, d_matC, numBytes, hipMemcpyDeviceToHost); // GPU -> CPU //--------------------------------gpu in-place termina ------------------------------------ hipFree(d_matA); hipFree(d_matB); hipFree(d_matC); /* //imprime la matriz matC for(i = 0; i < N; i++){ for(j = 0; j < N; j++){ printf("%f|",matC[i*N+j]); } printf("\n"); } printf("\n"); */ free(matA); free(matB); free(matC); return 0; }
You are a helpful coding assistant specialized in converting HIP source code to AMD device assembly.
#include <hip/hip_runtime.h> #include <stdio.h> #include <stdlib.h> #include <sys/time.h> #include <sys/resource.h> #include <math.h> //134217728 double dwalltime(){ double sec; struct timeval tv; gettimeofday(&tv,NULL); sec = tv.tv_sec + tv.tv_usec/1000000.0; return sec; } __global__ void ecuacion_kernel_outplace_p1(double *d_matA,double *d_matAT,double *d_matB,double *d_matBT, unsigned int n){ int distA = blockIdx.y * blockDim.y + threadIdx.y; //i int distB = blockIdx.x * blockDim.x + threadIdx.x; //j //transpuesta out-place A y B if( (distA<n*n) && (distB<n*n) ){ d_matAT [distB*n + distA] = d_matA[distA*n + distB]; d_matBT [distB*n + distA] = d_matB[distA*n + distB]; } } __global__ void ecuacion_kernel_outplace_p2(double *d_matA,double *d_matB,double *d_matC,double *d_matAT,double *d_matBT, unsigned int n){ int distA = blockIdx.y * blockDim.y + threadIdx.y; //i int distB = blockIdx.x * blockDim.x + threadIdx.x; //j int k; if (distA*n+distB <= (n*n - 1)){ //multiplicacion for(k = 0; k < n ;k++){ d_matC[distA*n+distB] += d_matA[distA*n+k] * d_matBT[distB+k*n]; } //suma d_matC[distA*n+distB] += d_matB[distA*n+distB] + d_matAT[distA*n+distB]; } } __global__ void ecuacion_kernel_inplace_suma (double *d_matA,double *d_matB,double *d_matC, unsigned int n){ int distA = blockIdx.y * blockDim.y + threadIdx.y; //i int distB = blockIdx.x * blockDim.x + threadIdx.x; //j int k; //multiplicacion if (distA*n+distB < (n*n - 1)){ d_matC[distA*n+distB] += d_matB[distA*n+distB] + d_matA[distA+distB*n]; for(k = 0; k < n ;k++){ d_matC[distA*n+distB] += d_matA[distA*n+k] * d_matB[distB*n+k]; } } } __global__ void kernel_sum_Matriz (double *d_matA,double *d_matB,double *d_matC, unsigned int n){ int distA = blockIdx.y * blockDim.y + threadIdx.y; //i int distB = blockIdx.x * blockDim.x + threadIdx.x; //j //suma if (distA*n+distB < (n*n)){ d_matC[distA*n+distB] += d_matA[distA*n+distB] + d_matB[distA+distB*n]; } } __global__ void kernel_transpuesta(double *m, int N){ int tid = blockIdx.x*blockDim.x + threadIdx.x; int i = int((1 + sqrtf(1 + 8*tid)) / 2); int j = tid - (i*(i-1)/2); int aux; if ( (i<N) && (j<N) ){ aux = m[i*N + j] ; m[i*N + j] = m[j*N + i]; m[j*N + i] = aux; } } __global__ void kernel_mult_sum_matriz (double *d_matA,double *d_matB,double *d_matC, unsigned int n){ int distA = blockIdx.y * blockDim.y + threadIdx.y; //i int distB = blockIdx.x * blockDim.x + threadIdx.x; //j int k; //multiplicacion if (distA*n+distB < (n*n)){ for(k = 0; k < n ;k++){ d_matC[distA*n+distB] += d_matA[distA*n+k] * d_matB[distB*n+k]; } } } int main(int argc, char *argv[]){ if (argc != 3){ printf("Falta argumento: N, CUDABLK\n"); return 0; } //declaracion de variables hipError_t error; unsigned int N = atoi (argv[1]); unsigned long CUDA_BLK = atoi (argv[2]), gridBlock; unsigned long numBytes = sizeof(double)*N*N; double *matA,*matB,*matC,*d_matA,*d_matB,*d_matC,*d_matAT,*d_matBT,timetick; unsigned int i,j,k; //inicializa variables para cpu matA = (double *)malloc(numBytes); matB = (double *)malloc(numBytes); matC = (double *)malloc(numBytes); for (i = 0; i < N*N; i++){ matA[i] = i; matB[i] = i; matC[i] = 0; } //inicializa variables para gpu hipMalloc((void **) &d_matA, numBytes); hipMalloc((void **) &d_matAT, numBytes); hipMalloc((void **) &d_matB, numBytes); hipMalloc((void **) &d_matBT, numBytes); hipMalloc((void **) &d_matC, numBytes); gridBlock = (unsigned int)sqrt(N*N/CUDA_BLK/CUDA_BLK); dim3 dimBlock(CUDA_BLK,CUDA_BLK); // Bloque bidimencional de hilos (*cb* hilos) dim3 dimGrid(gridBlock,gridBlock); // Grid bidimencional (*ceil(n/cb)* bloques) //--------------------------------cpu comienza ------------------------------------ //secuencial timetick = dwalltime(); //multiplicacion for(i = 0; i < N; i++){ for(j = 0; j < N; j++){ for(k = 0; k < N ;k++){ matC[i*N+j] += matA[i*N+k] * matB[j*N+k]; //multiplica a matB por fila, eso simula la matB transpuesta } } } //suma for(i = 0; i < N; i++){ for(j = 0; j < N; j++){ matC[i*N+j] += matB[i*N+j] + matA[i+j*N]; } } printf("Tiempo para la ecuacion CPU: %f\n\n",dwalltime() - timetick); /* for(i = 0; i < N; i++){ for(j = 0; j < N; j++){ printf("%f|",matC[i*N+j]); } printf("\n"); } printf("\n"); */ //--------------------------------cpu termina ------------------------------------ for (i = 0; i < N*N; i++){ matA[i] = i; matB[i] = i; matC[i] = 0; } hipMemcpy(d_matA, matA, numBytes, hipMemcpyHostToDevice); // CPU -> GPU hipMemcpy(d_matB, matB, numBytes, hipMemcpyHostToDevice); // CPU -> GPU hipMemcpy(d_matC, matC, numBytes, hipMemcpyHostToDevice); // CPU -> GPU //--------------------------------gpu out-place comienza ------------------------------------ timetick = dwalltime(); ecuacion_kernel_outplace_p1<<<dimGrid, dimBlock>>>(d_matA, d_matAT,d_matB,d_matBT, N); hipDeviceSynchronize(); ecuacion_kernel_outplace_p2<<<dimGrid, dimBlock>>>(d_matA, d_matB,d_matC,d_matAT,d_matBT, N); hipDeviceSynchronize(); printf("Tiempo para la ecuacion out-place GPU: %f\n",dwalltime() - timetick); error = hipGetLastError(); printf("error: %d\n\n",error); hipMemcpy(matC, d_matC, numBytes, hipMemcpyDeviceToHost); // GPU -> CPU /* for(i = 0; i < N; i++){ for(j = 0; j < N; j++){ printf("%f|",matC[i*N+j]); } printf("\n"); } printf("\n"); */ //--------------------------------gpu out-place termina ------------------------------------ hipFree(d_matA); hipFree(d_matB); hipFree(d_matC); hipFree(d_matAT); hipFree(d_matBT); for (i = 0; i < N*N; i++){ matA[i] = i; matB[i] = i; matC[i] = 0; } hipMalloc((void **) &d_matA, numBytes); hipMalloc((void **) &d_matB, numBytes); hipMalloc((void **) &d_matC, numBytes); hipMemcpy(d_matA, matA, numBytes, hipMemcpyHostToDevice); // CPU -> GPU hipMemcpy(d_matB, matB, numBytes, hipMemcpyHostToDevice); // CPU -> GPU hipMemcpy(d_matC, matC, numBytes, hipMemcpyHostToDevice); // CPU -> GPU //--------------------------------gpu in-place comienza ------------------------------------ timetick = dwalltime(); kernel_transpuesta<<<dimGrid, dimBlock>>>(d_matA, N); hipDeviceSynchronize(); kernel_sum_Matriz<<<dimGrid, dimBlock>>>(d_matA, d_matB,d_matC, N); hipDeviceSynchronize(); kernel_transpuesta<<<dimGrid, dimBlock>>>(d_matA, N); hipDeviceSynchronize(); kernel_mult_sum_matriz<<<dimGrid, dimBlock>>>(d_matA, d_matB,d_matC, N); hipDeviceSynchronize(); printf("Tiempo para la ecuacion in-place GPU: %f\n",dwalltime() - timetick); error = hipGetLastError(); printf("error: %d\n\n",error); hipMemcpy(matC, d_matC, numBytes, hipMemcpyDeviceToHost); // GPU -> CPU //--------------------------------gpu in-place termina ------------------------------------ hipFree(d_matA); hipFree(d_matB); hipFree(d_matC); /* //imprime la matriz matC for(i = 0; i < N; i++){ for(j = 0; j < N; j++){ printf("%f|",matC[i*N+j]); } printf("\n"); } printf("\n"); */ free(matA); free(matB); free(matC); return 0; }
.text .amdgcn_target "amdgcn-amd-amdhsa--gfx1100" .protected _Z27ecuacion_kernel_outplace_p1PdS_S_S_j .globl _Z27ecuacion_kernel_outplace_p1PdS_S_S_j .p2align 8 .type _Z27ecuacion_kernel_outplace_p1PdS_S_S_j,@function _Z27ecuacion_kernel_outplace_p1PdS_S_S_j: s_clause 0x1 s_load_b32 s3, s[0:1], 0x34 s_load_b32 s2, s[0:1], 0x20 v_bfe_u32 v2, v0, 10, 10 v_and_b32_e32 v3, 0x3ff, v0 s_waitcnt lgkmcnt(0) s_lshr_b32 s4, s3, 16 s_and_b32 s3, s3, 0xffff s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1) v_mad_u64_u32 v[0:1], null, s15, s4, v[2:3] v_mad_u64_u32 v[1:2], null, s14, s3, v[3:4] s_mul_i32 s3, s2, s2 v_max_u32_e32 v2, v0, v1 s_delay_alu instid0(VALU_DEP_1) v_cmp_gt_u32_e32 vcc_lo, s3, v2 s_and_saveexec_b32 s3, vcc_lo s_cbranch_execz .LBB0_2 s_load_b256 s[4:11], s[0:1], 0x0 v_mad_u64_u32 v[2:3], null, v0, s2, v[1:2] v_mov_b32_e32 v3, 0 v_mad_u64_u32 v[8:9], null, v1, s2, v[0:1] s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_4) v_mov_b32_e32 v9, v3 v_lshlrev_b64 v[4:5], 3, v[2:3] s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2) v_lshlrev_b64 v[0:1], 3, v[8:9] s_waitcnt lgkmcnt(0) v_add_co_u32 v6, vcc_lo, s4, v4 s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3) v_add_co_ci_u32_e32 v7, vcc_lo, s5, v5, vcc_lo v_add_co_u32 v2, vcc_lo, s6, v0 s_delay_alu instid0(VALU_DEP_4) v_add_co_ci_u32_e32 v3, vcc_lo, s7, v1, vcc_lo global_load_b64 v[6:7], v[6:7], off v_add_co_u32 v4, vcc_lo, s8, v4 v_add_co_ci_u32_e32 v5, vcc_lo, s9, v5, vcc_lo v_add_co_u32 v0, vcc_lo, s10, v0 v_add_co_ci_u32_e32 v1, vcc_lo, s11, v1, vcc_lo s_waitcnt vmcnt(0) global_store_b64 v[2:3], v[6:7], off global_load_b64 v[2:3], v[4:5], off s_waitcnt vmcnt(0) global_store_b64 v[0:1], v[2:3], off .LBB0_2: s_nop 0 s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) s_endpgm .section .rodata,"a",@progbits .p2align 6, 0x0 .amdhsa_kernel _Z27ecuacion_kernel_outplace_p1PdS_S_S_j .amdhsa_group_segment_fixed_size 0 .amdhsa_private_segment_fixed_size 0 .amdhsa_kernarg_size 296 .amdhsa_user_sgpr_count 14 .amdhsa_user_sgpr_dispatch_ptr 0 .amdhsa_user_sgpr_queue_ptr 0 .amdhsa_user_sgpr_kernarg_segment_ptr 1 .amdhsa_user_sgpr_dispatch_id 0 .amdhsa_user_sgpr_private_segment_size 0 .amdhsa_wavefront_size32 1 .amdhsa_uses_dynamic_stack 0 .amdhsa_enable_private_segment 0 .amdhsa_system_sgpr_workgroup_id_x 1 .amdhsa_system_sgpr_workgroup_id_y 1 .amdhsa_system_sgpr_workgroup_id_z 0 .amdhsa_system_sgpr_workgroup_info 0 .amdhsa_system_vgpr_workitem_id 1 .amdhsa_next_free_vgpr 10 .amdhsa_next_free_sgpr 16 .amdhsa_float_round_mode_32 0 .amdhsa_float_round_mode_16_64 0 .amdhsa_float_denorm_mode_32 3 .amdhsa_float_denorm_mode_16_64 3 .amdhsa_dx10_clamp 1 .amdhsa_ieee_mode 1 .amdhsa_fp16_overflow 0 .amdhsa_workgroup_processor_mode 1 .amdhsa_memory_ordered 1 .amdhsa_forward_progress 0 .amdhsa_shared_vgpr_count 0 .amdhsa_exception_fp_ieee_invalid_op 0 .amdhsa_exception_fp_denorm_src 0 .amdhsa_exception_fp_ieee_div_zero 0 .amdhsa_exception_fp_ieee_overflow 0 .amdhsa_exception_fp_ieee_underflow 0 .amdhsa_exception_fp_ieee_inexact 0 .amdhsa_exception_int_div_zero 0 .end_amdhsa_kernel .text .Lfunc_end0: .size _Z27ecuacion_kernel_outplace_p1PdS_S_S_j, .Lfunc_end0-_Z27ecuacion_kernel_outplace_p1PdS_S_S_j .section .AMDGPU.csdata,"",@progbits .text .protected _Z27ecuacion_kernel_outplace_p2PdS_S_S_S_j .globl _Z27ecuacion_kernel_outplace_p2PdS_S_S_S_j .p2align 8 .type _Z27ecuacion_kernel_outplace_p2PdS_S_S_S_j,@function _Z27ecuacion_kernel_outplace_p2PdS_S_S_S_j: s_clause 0x1 s_load_b32 s2, s[0:1], 0x3c s_load_b32 s8, s[0:1], 0x28 v_bfe_u32 v1, v0, 10, 10 v_and_b32_e32 v0, 0x3ff, v0 s_waitcnt lgkmcnt(0) s_lshr_b32 s3, s2, 16 s_and_b32 s2, s2, 0xffff v_mad_u64_u32 v[4:5], null, s15, s3, v[1:2] v_mad_u64_u32 v[2:3], null, s14, s2, v[0:1] s_mul_i32 s2, s8, s8 s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_2) s_add_i32 s2, s2, -1 v_mul_lo_u32 v8, v4, s8 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_add_nc_u32_e32 v0, v8, v2 v_cmp_ge_u32_e32 vcc_lo, s2, v0 s_and_saveexec_b32 s2, vcc_lo s_cbranch_execz .LBB1_5 s_load_b64 s[2:3], s[0:1], 0x10 s_cmp_eq_u32 s8, 0 s_mov_b32 s9, 0 s_cbranch_scc1 .LBB1_4 v_mov_b32_e32 v1, 0 s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1) v_lshlrev_b64 v[4:5], 3, v[0:1] s_waitcnt lgkmcnt(0) v_add_co_u32 v4, vcc_lo, s2, v4 s_delay_alu instid0(VALU_DEP_2) v_add_co_ci_u32_e32 v5, vcc_lo, s3, v5, vcc_lo global_load_b64 v[6:7], v[4:5], off s_clause 0x1 s_load_b64 s[4:5], s[0:1], 0x0 s_load_b64 s[6:7], s[0:1], 0x20 .p2align 6 .LBB1_3: v_dual_mov_b32 v10, 0 :: v_dual_add_nc_u32 v9, s9, v8 s_add_i32 s9, s9, 1 s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1) s_cmp_lg_u32 s8, s9 v_mov_b32_e32 v3, v10 s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_lshlrev_b64 v[9:10], 3, v[9:10] v_lshlrev_b64 v[11:12], 3, v[2:3] v_add_nc_u32_e32 v2, s8, v2 s_waitcnt lgkmcnt(0) s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_4) v_add_co_u32 v9, vcc_lo, s4, v9 v_add_co_ci_u32_e32 v10, vcc_lo, s5, v10, vcc_lo s_delay_alu instid0(VALU_DEP_4) v_add_co_u32 v11, vcc_lo, s6, v11 v_add_co_ci_u32_e32 v12, vcc_lo, s7, v12, vcc_lo global_load_b64 v[9:10], v[9:10], off global_load_b64 v[11:12], v[11:12], off s_waitcnt vmcnt(0) v_fma_f64 v[6:7], v[9:10], v[11:12], v[6:7] global_store_b64 v[4:5], v[6:7], off s_cbranch_scc1 .LBB1_3 .LBB1_4: s_clause 0x1 s_load_b64 s[4:5], s[0:1], 0x8 s_load_b64 s[0:1], s[0:1], 0x18 v_mov_b32_e32 v1, 0 s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1) v_lshlrev_b64 v[0:1], 3, v[0:1] s_waitcnt lgkmcnt(0) v_add_co_u32 v2, vcc_lo, s4, v0 s_delay_alu instid0(VALU_DEP_2) v_add_co_ci_u32_e32 v3, vcc_lo, s5, v1, vcc_lo v_add_co_u32 v4, vcc_lo, s0, v0 v_add_co_ci_u32_e32 v5, vcc_lo, s1, v1, vcc_lo v_add_co_u32 v0, vcc_lo, s2, v0 global_load_b64 v[2:3], v[2:3], off global_load_b64 v[4:5], v[4:5], off v_add_co_ci_u32_e32 v1, vcc_lo, s3, v1, vcc_lo global_load_b64 v[6:7], v[0:1], off s_waitcnt vmcnt(1) v_add_f64 v[2:3], v[2:3], v[4:5] s_waitcnt vmcnt(0) s_delay_alu instid0(VALU_DEP_1) v_add_f64 v[2:3], v[6:7], v[2:3] global_store_b64 v[0:1], v[2:3], off .LBB1_5: s_nop 0 s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) s_endpgm .section .rodata,"a",@progbits .p2align 6, 0x0 .amdhsa_kernel _Z27ecuacion_kernel_outplace_p2PdS_S_S_S_j .amdhsa_group_segment_fixed_size 0 .amdhsa_private_segment_fixed_size 0 .amdhsa_kernarg_size 304 .amdhsa_user_sgpr_count 14 .amdhsa_user_sgpr_dispatch_ptr 0 .amdhsa_user_sgpr_queue_ptr 0 .amdhsa_user_sgpr_kernarg_segment_ptr 1 .amdhsa_user_sgpr_dispatch_id 0 .amdhsa_user_sgpr_private_segment_size 0 .amdhsa_wavefront_size32 1 .amdhsa_uses_dynamic_stack 0 .amdhsa_enable_private_segment 0 .amdhsa_system_sgpr_workgroup_id_x 1 .amdhsa_system_sgpr_workgroup_id_y 1 .amdhsa_system_sgpr_workgroup_id_z 0 .amdhsa_system_sgpr_workgroup_info 0 .amdhsa_system_vgpr_workitem_id 1 .amdhsa_next_free_vgpr 13 .amdhsa_next_free_sgpr 16 .amdhsa_float_round_mode_32 0 .amdhsa_float_round_mode_16_64 0 .amdhsa_float_denorm_mode_32 3 .amdhsa_float_denorm_mode_16_64 3 .amdhsa_dx10_clamp 1 .amdhsa_ieee_mode 1 .amdhsa_fp16_overflow 0 .amdhsa_workgroup_processor_mode 1 .amdhsa_memory_ordered 1 .amdhsa_forward_progress 0 .amdhsa_shared_vgpr_count 0 .amdhsa_exception_fp_ieee_invalid_op 0 .amdhsa_exception_fp_denorm_src 0 .amdhsa_exception_fp_ieee_div_zero 0 .amdhsa_exception_fp_ieee_overflow 0 .amdhsa_exception_fp_ieee_underflow 0 .amdhsa_exception_fp_ieee_inexact 0 .amdhsa_exception_int_div_zero 0 .end_amdhsa_kernel .text .Lfunc_end1: .size _Z27ecuacion_kernel_outplace_p2PdS_S_S_S_j, .Lfunc_end1-_Z27ecuacion_kernel_outplace_p2PdS_S_S_S_j .section .AMDGPU.csdata,"",@progbits .text .protected _Z28ecuacion_kernel_inplace_sumaPdS_S_j .globl _Z28ecuacion_kernel_inplace_sumaPdS_S_j .p2align 8 .type _Z28ecuacion_kernel_inplace_sumaPdS_S_j,@function _Z28ecuacion_kernel_inplace_sumaPdS_S_j: s_clause 0x1 s_load_b32 s3, s[0:1], 0x2c s_load_b32 s2, s[0:1], 0x18 v_bfe_u32 v3, v0, 10, 10 v_and_b32_e32 v0, 0x3ff, v0 s_waitcnt lgkmcnt(0) s_lshr_b32 s4, s3, 16 s_and_b32 s3, s3, 0xffff v_mad_u64_u32 v[1:2], null, s15, s4, v[3:4] s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(SALU_CYCLE_1) v_mad_u64_u32 v[2:3], null, s14, s3, v[0:1] v_mul_lo_u32 v0, v1, s2 s_mul_i32 s3, s2, s2 s_add_i32 s3, s3, -1 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_add_nc_u32_e32 v3, v0, v2 v_cmp_gt_u32_e32 vcc_lo, s3, v3 s_and_saveexec_b32 s3, vcc_lo s_cbranch_execz .LBB2_4 s_load_b128 s[4:7], s[0:1], 0x0 v_mul_lo_u32 v2, v2, s2 v_mov_b32_e32 v4, 0 s_load_b64 s[0:1], s[0:1], 0x10 s_cmp_eq_u32 s2, 0 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_3) v_lshlrev_b64 v[5:6], 3, v[3:4] v_add_nc_u32_e32 v3, v2, v1 s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_3) v_lshlrev_b64 v[3:4], 3, v[3:4] s_waitcnt lgkmcnt(0) v_add_co_u32 v7, vcc_lo, s6, v5 s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_3) v_add_co_ci_u32_e32 v8, vcc_lo, s7, v6, vcc_lo v_add_co_u32 v3, vcc_lo, s4, v3 s_delay_alu instid0(VALU_DEP_4) v_add_co_ci_u32_e32 v4, vcc_lo, s5, v4, vcc_lo global_load_b64 v[7:8], v[7:8], off global_load_b64 v[9:10], v[3:4], off v_add_co_u32 v4, vcc_lo, s0, v5 v_add_co_ci_u32_e32 v5, vcc_lo, s1, v6, vcc_lo global_load_b64 v[11:12], v[4:5], off s_waitcnt vmcnt(1) v_add_f64 v[6:7], v[7:8], v[9:10] s_waitcnt vmcnt(0) s_delay_alu instid0(VALU_DEP_1) v_add_f64 v[6:7], v[11:12], v[6:7] global_store_b64 v[4:5], v[6:7], off s_cbranch_scc1 .LBB2_4 global_load_b64 v[6:7], v[4:5], off .p2align 6 .LBB2_3: v_mov_b32_e32 v1, 0 s_add_i32 s2, s2, -1 s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1) s_cmp_lg_u32 s2, 0 v_mov_b32_e32 v3, v1 s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_2) v_lshlrev_b64 v[10:11], 3, v[2:3] v_add_nc_u32_e32 v2, 1, v2 v_lshlrev_b64 v[8:9], 3, v[0:1] v_add_nc_u32_e32 v0, 1, v0 v_add_co_u32 v8, vcc_lo, s4, v8 s_delay_alu instid0(VALU_DEP_3) v_add_co_ci_u32_e32 v9, vcc_lo, s5, v9, vcc_lo v_add_co_u32 v10, vcc_lo, s6, v10 v_add_co_ci_u32_e32 v11, vcc_lo, s7, v11, vcc_lo global_load_b64 v[8:9], v[8:9], off global_load_b64 v[10:11], v[10:11], off s_waitcnt vmcnt(0) v_fma_f64 v[6:7], v[8:9], v[10:11], v[6:7] global_store_b64 v[4:5], v[6:7], off s_cbranch_scc1 .LBB2_3 .LBB2_4: s_nop 0 s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) s_endpgm .section .rodata,"a",@progbits .p2align 6, 0x0 .amdhsa_kernel _Z28ecuacion_kernel_inplace_sumaPdS_S_j .amdhsa_group_segment_fixed_size 0 .amdhsa_private_segment_fixed_size 0 .amdhsa_kernarg_size 288 .amdhsa_user_sgpr_count 14 .amdhsa_user_sgpr_dispatch_ptr 0 .amdhsa_user_sgpr_queue_ptr 0 .amdhsa_user_sgpr_kernarg_segment_ptr 1 .amdhsa_user_sgpr_dispatch_id 0 .amdhsa_user_sgpr_private_segment_size 0 .amdhsa_wavefront_size32 1 .amdhsa_uses_dynamic_stack 0 .amdhsa_enable_private_segment 0 .amdhsa_system_sgpr_workgroup_id_x 1 .amdhsa_system_sgpr_workgroup_id_y 1 .amdhsa_system_sgpr_workgroup_id_z 0 .amdhsa_system_sgpr_workgroup_info 0 .amdhsa_system_vgpr_workitem_id 1 .amdhsa_next_free_vgpr 13 .amdhsa_next_free_sgpr 16 .amdhsa_float_round_mode_32 0 .amdhsa_float_round_mode_16_64 0 .amdhsa_float_denorm_mode_32 3 .amdhsa_float_denorm_mode_16_64 3 .amdhsa_dx10_clamp 1 .amdhsa_ieee_mode 1 .amdhsa_fp16_overflow 0 .amdhsa_workgroup_processor_mode 1 .amdhsa_memory_ordered 1 .amdhsa_forward_progress 0 .amdhsa_shared_vgpr_count 0 .amdhsa_exception_fp_ieee_invalid_op 0 .amdhsa_exception_fp_denorm_src 0 .amdhsa_exception_fp_ieee_div_zero 0 .amdhsa_exception_fp_ieee_overflow 0 .amdhsa_exception_fp_ieee_underflow 0 .amdhsa_exception_fp_ieee_inexact 0 .amdhsa_exception_int_div_zero 0 .end_amdhsa_kernel .text .Lfunc_end2: .size _Z28ecuacion_kernel_inplace_sumaPdS_S_j, .Lfunc_end2-_Z28ecuacion_kernel_inplace_sumaPdS_S_j .section .AMDGPU.csdata,"",@progbits .text .protected _Z17kernel_sum_MatrizPdS_S_j .globl _Z17kernel_sum_MatrizPdS_S_j .p2align 8 .type _Z17kernel_sum_MatrizPdS_S_j,@function _Z17kernel_sum_MatrizPdS_S_j: s_clause 0x1 s_load_b32 s3, s[0:1], 0x2c s_load_b32 s2, s[0:1], 0x18 v_bfe_u32 v2, v0, 10, 10 v_and_b32_e32 v5, 0x3ff, v0 s_waitcnt lgkmcnt(0) s_lshr_b32 s4, s3, 16 s_and_b32 s3, s3, 0xffff v_mad_u64_u32 v[0:1], null, s15, s4, v[2:3] v_mad_u64_u32 v[3:4], null, s14, s3, v[5:6] s_mul_i32 s3, s2, s2 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_mad_u64_u32 v[1:2], null, v0, s2, v[3:4] v_cmp_gt_u32_e32 vcc_lo, s3, v1 s_and_saveexec_b32 s3, vcc_lo s_cbranch_execz .LBB3_2 s_load_b128 s[4:7], s[0:1], 0x0 v_mov_b32_e32 v2, 0 v_mad_u64_u32 v[4:5], null, v3, s2, v[0:1] s_load_b64 s[0:1], s[0:1], 0x10 s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2) v_mov_b32_e32 v5, v2 v_lshlrev_b64 v[0:1], 3, v[1:2] v_lshlrev_b64 v[2:3], 3, v[4:5] s_waitcnt lgkmcnt(0) s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3) v_add_co_u32 v4, vcc_lo, s4, v0 v_add_co_ci_u32_e32 v5, vcc_lo, s5, v1, vcc_lo s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_4) v_add_co_u32 v2, vcc_lo, s6, v2 v_add_co_ci_u32_e32 v3, vcc_lo, s7, v3, vcc_lo v_add_co_u32 v0, vcc_lo, s0, v0 global_load_b64 v[4:5], v[4:5], off global_load_b64 v[2:3], v[2:3], off v_add_co_ci_u32_e32 v1, vcc_lo, s1, v1, vcc_lo global_load_b64 v[6:7], v[0:1], off s_waitcnt vmcnt(1) v_add_f64 v[2:3], v[4:5], v[2:3] s_waitcnt vmcnt(0) s_delay_alu instid0(VALU_DEP_1) v_add_f64 v[2:3], v[6:7], v[2:3] global_store_b64 v[0:1], v[2:3], off .LBB3_2: s_nop 0 s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) s_endpgm .section .rodata,"a",@progbits .p2align 6, 0x0 .amdhsa_kernel _Z17kernel_sum_MatrizPdS_S_j .amdhsa_group_segment_fixed_size 0 .amdhsa_private_segment_fixed_size 0 .amdhsa_kernarg_size 288 .amdhsa_user_sgpr_count 14 .amdhsa_user_sgpr_dispatch_ptr 0 .amdhsa_user_sgpr_queue_ptr 0 .amdhsa_user_sgpr_kernarg_segment_ptr 1 .amdhsa_user_sgpr_dispatch_id 0 .amdhsa_user_sgpr_private_segment_size 0 .amdhsa_wavefront_size32 1 .amdhsa_uses_dynamic_stack 0 .amdhsa_enable_private_segment 0 .amdhsa_system_sgpr_workgroup_id_x 1 .amdhsa_system_sgpr_workgroup_id_y 1 .amdhsa_system_sgpr_workgroup_id_z 0 .amdhsa_system_sgpr_workgroup_info 0 .amdhsa_system_vgpr_workitem_id 1 .amdhsa_next_free_vgpr 8 .amdhsa_next_free_sgpr 16 .amdhsa_float_round_mode_32 0 .amdhsa_float_round_mode_16_64 0 .amdhsa_float_denorm_mode_32 3 .amdhsa_float_denorm_mode_16_64 3 .amdhsa_dx10_clamp 1 .amdhsa_ieee_mode 1 .amdhsa_fp16_overflow 0 .amdhsa_workgroup_processor_mode 1 .amdhsa_memory_ordered 1 .amdhsa_forward_progress 0 .amdhsa_shared_vgpr_count 0 .amdhsa_exception_fp_ieee_invalid_op 0 .amdhsa_exception_fp_denorm_src 0 .amdhsa_exception_fp_ieee_div_zero 0 .amdhsa_exception_fp_ieee_overflow 0 .amdhsa_exception_fp_ieee_underflow 0 .amdhsa_exception_fp_ieee_inexact 0 .amdhsa_exception_int_div_zero 0 .end_amdhsa_kernel .text .Lfunc_end3: .size _Z17kernel_sum_MatrizPdS_S_j, .Lfunc_end3-_Z17kernel_sum_MatrizPdS_S_j .section .AMDGPU.csdata,"",@progbits .text .protected _Z18kernel_transpuestaPdi .globl _Z18kernel_transpuestaPdi .p2align 8 .type _Z18kernel_transpuestaPdi,@function _Z18kernel_transpuestaPdi: s_load_b32 s2, s[0:1], 0x1c s_mov_b32 s3, exec_lo s_waitcnt lgkmcnt(0) s_and_b32 s2, s2, 0xffff s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_mad_u64_u32 v[1:2], null, s15, s2, v[0:1] v_lshl_or_b32 v0, v1, 3, 1 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_cvt_f32_i32_e32 v0, v0 v_mul_f32_e32 v2, 0x4f800000, v0 v_cmp_gt_f32_e32 vcc_lo, 0xf800000, v0 s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1) v_cndmask_b32_e32 v0, v0, v2, vcc_lo v_sqrt_f32_e32 v2, v0 s_waitcnt_depctr 0xfff v_add_nc_u32_e32 v3, -1, v2 v_add_nc_u32_e32 v4, 1, v2 s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_fma_f32 v5, -v3, v2, v0 v_fma_f32 v6, -v4, v2, v0 s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1) v_cmp_ge_f32_e64 s2, 0, v5 v_cndmask_b32_e64 v2, v2, v3, s2 s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1) v_cmp_lt_f32_e64 s2, 0, v6 v_cndmask_b32_e64 v2, v2, v4, s2 s_load_b32 s2, s[0:1], 0x8 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_mul_f32_e32 v3, 0x37800000, v2 v_cndmask_b32_e32 v2, v2, v3, vcc_lo v_cmp_class_f32_e64 vcc_lo, v0, 0x260 s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1) v_cndmask_b32_e32 v0, v2, v0, vcc_lo v_add_f32_e32 v0, 1.0, v0 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_mul_f32_e32 v0, 0.5, v0 v_cvt_i32_f32_e32 v0, v0 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_add_nc_u32_e32 v2, -1, v0 v_mul_lo_u32 v2, v2, v0 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_lshrrev_b32_e32 v3, 31, v2 v_add_nc_u32_e32 v2, v2, v3 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_ashrrev_i32_e32 v2, 1, v2 v_sub_nc_u32_e32 v1, v1, v2 s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1) v_max_i32_e32 v2, v0, v1 s_waitcnt lgkmcnt(0) v_cmpx_gt_i32_e64 s2, v2 s_cbranch_execz .LBB4_2 s_load_b64 s[0:1], s[0:1], 0x0 v_mad_u64_u32 v[2:3], null, v0, s2, v[1:2] v_mad_u64_u32 v[6:7], null, v1, s2, v[0:1] s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_ashrrev_i32_e32 v3, 31, v2 v_ashrrev_i32_e32 v7, 31, v6 s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_lshlrev_b64 v[2:3], 3, v[2:3] v_lshlrev_b64 v[0:1], 3, v[6:7] s_waitcnt lgkmcnt(0) s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3) v_add_co_u32 v2, vcc_lo, s0, v2 v_add_co_ci_u32_e32 v3, vcc_lo, s1, v3, vcc_lo s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_4) v_add_co_u32 v0, vcc_lo, s0, v0 v_add_co_ci_u32_e32 v1, vcc_lo, s1, v1, vcc_lo s_clause 0x1 global_load_b64 v[4:5], v[2:3], off global_load_b64 v[6:7], v[0:1], off s_waitcnt vmcnt(1) v_cvt_i32_f64_e32 v4, v[4:5] s_delay_alu instid0(VALU_DEP_1) v_cvt_f64_i32_e32 v[4:5], v4 s_waitcnt vmcnt(0) s_clause 0x1 global_store_b64 v[2:3], v[6:7], off global_store_b64 v[0:1], v[4:5], off .LBB4_2: s_nop 0 s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) s_endpgm .section .rodata,"a",@progbits .p2align 6, 0x0 .amdhsa_kernel _Z18kernel_transpuestaPdi .amdhsa_group_segment_fixed_size 0 .amdhsa_private_segment_fixed_size 0 .amdhsa_kernarg_size 272 .amdhsa_user_sgpr_count 15 .amdhsa_user_sgpr_dispatch_ptr 0 .amdhsa_user_sgpr_queue_ptr 0 .amdhsa_user_sgpr_kernarg_segment_ptr 1 .amdhsa_user_sgpr_dispatch_id 0 .amdhsa_user_sgpr_private_segment_size 0 .amdhsa_wavefront_size32 1 .amdhsa_uses_dynamic_stack 0 .amdhsa_enable_private_segment 0 .amdhsa_system_sgpr_workgroup_id_x 1 .amdhsa_system_sgpr_workgroup_id_y 0 .amdhsa_system_sgpr_workgroup_id_z 0 .amdhsa_system_sgpr_workgroup_info 0 .amdhsa_system_vgpr_workitem_id 0 .amdhsa_next_free_vgpr 8 .amdhsa_next_free_sgpr 16 .amdhsa_float_round_mode_32 0 .amdhsa_float_round_mode_16_64 0 .amdhsa_float_denorm_mode_32 3 .amdhsa_float_denorm_mode_16_64 3 .amdhsa_dx10_clamp 1 .amdhsa_ieee_mode 1 .amdhsa_fp16_overflow 0 .amdhsa_workgroup_processor_mode 1 .amdhsa_memory_ordered 1 .amdhsa_forward_progress 0 .amdhsa_shared_vgpr_count 0 .amdhsa_exception_fp_ieee_invalid_op 0 .amdhsa_exception_fp_denorm_src 0 .amdhsa_exception_fp_ieee_div_zero 0 .amdhsa_exception_fp_ieee_overflow 0 .amdhsa_exception_fp_ieee_underflow 0 .amdhsa_exception_fp_ieee_inexact 0 .amdhsa_exception_int_div_zero 0 .end_amdhsa_kernel .text .Lfunc_end4: .size _Z18kernel_transpuestaPdi, .Lfunc_end4-_Z18kernel_transpuestaPdi .section .AMDGPU.csdata,"",@progbits .text .protected _Z22kernel_mult_sum_matrizPdS_S_j .globl _Z22kernel_mult_sum_matrizPdS_S_j .p2align 8 .type _Z22kernel_mult_sum_matrizPdS_S_j,@function _Z22kernel_mult_sum_matrizPdS_S_j: s_clause 0x1 s_load_b32 s2, s[0:1], 0x2c s_load_b32 s4, s[0:1], 0x18 v_bfe_u32 v1, v0, 10, 10 v_and_b32_e32 v0, 0x3ff, v0 s_waitcnt lgkmcnt(0) s_lshr_b32 s3, s2, 16 s_and_b32 s2, s2, 0xffff v_mad_u64_u32 v[3:4], null, s15, s3, v[1:2] v_mad_u64_u32 v[1:2], null, s14, s2, v[0:1] s_mul_i32 s2, s4, s4 s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1) v_mul_lo_u32 v0, v3, s4 v_add_nc_u32_e32 v2, v0, v1 s_delay_alu instid0(VALU_DEP_1) v_cmp_gt_u32_e32 vcc_lo, s2, v2 s_and_saveexec_b32 s2, vcc_lo s_cbranch_execz .LBB5_3 s_load_b64 s[2:3], s[0:1], 0x10 v_mov_b32_e32 v3, 0 v_mul_lo_u32 v6, v1, s4 s_max_u32 s4, s4, 1 s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_1) v_lshlrev_b64 v[2:3], 3, v[2:3] s_waitcnt lgkmcnt(0) v_add_co_u32 v2, vcc_lo, s2, v2 s_delay_alu instid0(VALU_DEP_2) v_add_co_ci_u32_e32 v3, vcc_lo, s3, v3, vcc_lo s_load_b128 s[0:3], s[0:1], 0x0 global_load_b64 v[4:5], v[2:3], off .p2align 6 .LBB5_2: v_mov_b32_e32 v1, 0 s_add_i32 s4, s4, -1 s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1) s_cmp_lg_u32 s4, 0 v_mov_b32_e32 v7, v1 s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(VALU_DEP_2) v_lshlrev_b64 v[10:11], 3, v[6:7] v_add_nc_u32_e32 v6, 1, v6 v_lshlrev_b64 v[8:9], 3, v[0:1] v_add_nc_u32_e32 v0, 1, v0 s_waitcnt lgkmcnt(0) v_add_co_u32 v7, vcc_lo, s0, v8 s_delay_alu instid0(VALU_DEP_3) v_add_co_ci_u32_e32 v8, vcc_lo, s1, v9, vcc_lo v_add_co_u32 v9, vcc_lo, s2, v10 v_add_co_ci_u32_e32 v10, vcc_lo, s3, v11, vcc_lo global_load_b64 v[7:8], v[7:8], off global_load_b64 v[9:10], v[9:10], off s_waitcnt vmcnt(0) v_fma_f64 v[4:5], v[7:8], v[9:10], v[4:5] global_store_b64 v[2:3], v[4:5], off s_cbranch_scc1 .LBB5_2 .LBB5_3: s_nop 0 s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) s_endpgm .section .rodata,"a",@progbits .p2align 6, 0x0 .amdhsa_kernel _Z22kernel_mult_sum_matrizPdS_S_j .amdhsa_group_segment_fixed_size 0 .amdhsa_private_segment_fixed_size 0 .amdhsa_kernarg_size 288 .amdhsa_user_sgpr_count 14 .amdhsa_user_sgpr_dispatch_ptr 0 .amdhsa_user_sgpr_queue_ptr 0 .amdhsa_user_sgpr_kernarg_segment_ptr 1 .amdhsa_user_sgpr_dispatch_id 0 .amdhsa_user_sgpr_private_segment_size 0 .amdhsa_wavefront_size32 1 .amdhsa_uses_dynamic_stack 0 .amdhsa_enable_private_segment 0 .amdhsa_system_sgpr_workgroup_id_x 1 .amdhsa_system_sgpr_workgroup_id_y 1 .amdhsa_system_sgpr_workgroup_id_z 0 .amdhsa_system_sgpr_workgroup_info 0 .amdhsa_system_vgpr_workitem_id 1 .amdhsa_next_free_vgpr 12 .amdhsa_next_free_sgpr 16 .amdhsa_float_round_mode_32 0 .amdhsa_float_round_mode_16_64 0 .amdhsa_float_denorm_mode_32 3 .amdhsa_float_denorm_mode_16_64 3 .amdhsa_dx10_clamp 1 .amdhsa_ieee_mode 1 .amdhsa_fp16_overflow 0 .amdhsa_workgroup_processor_mode 1 .amdhsa_memory_ordered 1 .amdhsa_forward_progress 0 .amdhsa_shared_vgpr_count 0 .amdhsa_exception_fp_ieee_invalid_op 0 .amdhsa_exception_fp_denorm_src 0 .amdhsa_exception_fp_ieee_div_zero 0 .amdhsa_exception_fp_ieee_overflow 0 .amdhsa_exception_fp_ieee_underflow 0 .amdhsa_exception_fp_ieee_inexact 0 .amdhsa_exception_int_div_zero 0 .end_amdhsa_kernel .text .Lfunc_end5: .size _Z22kernel_mult_sum_matrizPdS_S_j, .Lfunc_end5-_Z22kernel_mult_sum_matrizPdS_S_j .section .AMDGPU.csdata,"",@progbits .text .p2alignl 7, 3214868480 .fill 96, 4, 3214868480 .type __hip_cuid_,@object .section .bss,"aw",@nobits .globl __hip_cuid_ __hip_cuid_: .byte 0 .size __hip_cuid_, 1 .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym __hip_cuid_ .amdgpu_metadata --- amdhsa.kernels: - .args: - .address_space: global .offset: 0 .size: 8 .value_kind: global_buffer - .address_space: global .offset: 8 .size: 8 .value_kind: global_buffer - .address_space: global .offset: 16 .size: 8 .value_kind: global_buffer - .address_space: global .offset: 24 .size: 8 .value_kind: global_buffer - .offset: 32 .size: 4 .value_kind: by_value - .offset: 40 .size: 4 .value_kind: hidden_block_count_x - .offset: 44 .size: 4 .value_kind: hidden_block_count_y - .offset: 48 .size: 4 .value_kind: hidden_block_count_z - .offset: 52 .size: 2 .value_kind: hidden_group_size_x - .offset: 54 .size: 2 .value_kind: hidden_group_size_y - .offset: 56 .size: 2 .value_kind: hidden_group_size_z - .offset: 58 .size: 2 .value_kind: hidden_remainder_x - .offset: 60 .size: 2 .value_kind: hidden_remainder_y - .offset: 62 .size: 2 .value_kind: hidden_remainder_z - .offset: 80 .size: 8 .value_kind: hidden_global_offset_x - .offset: 88 .size: 8 .value_kind: hidden_global_offset_y - .offset: 96 .size: 8 .value_kind: hidden_global_offset_z - .offset: 104 .size: 2 .value_kind: hidden_grid_dims .group_segment_fixed_size: 0 .kernarg_segment_align: 8 .kernarg_segment_size: 296 .language: OpenCL C .language_version: - 2 - 0 .max_flat_workgroup_size: 1024 .name: _Z27ecuacion_kernel_outplace_p1PdS_S_S_j .private_segment_fixed_size: 0 .sgpr_count: 18 .sgpr_spill_count: 0 .symbol: _Z27ecuacion_kernel_outplace_p1PdS_S_S_j.kd .uniform_work_group_size: 1 .uses_dynamic_stack: false .vgpr_count: 10 .vgpr_spill_count: 0 .wavefront_size: 32 .workgroup_processor_mode: 1 - .args: - .address_space: global .offset: 0 .size: 8 .value_kind: global_buffer - .address_space: global .offset: 8 .size: 8 .value_kind: global_buffer - .address_space: global .offset: 16 .size: 8 .value_kind: global_buffer - .address_space: global .offset: 24 .size: 8 .value_kind: global_buffer - .address_space: global .offset: 32 .size: 8 .value_kind: global_buffer - .offset: 40 .size: 4 .value_kind: by_value - .offset: 48 .size: 4 .value_kind: hidden_block_count_x - .offset: 52 .size: 4 .value_kind: hidden_block_count_y - .offset: 56 .size: 4 .value_kind: hidden_block_count_z - .offset: 60 .size: 2 .value_kind: hidden_group_size_x - .offset: 62 .size: 2 .value_kind: hidden_group_size_y - .offset: 64 .size: 2 .value_kind: hidden_group_size_z - .offset: 66 .size: 2 .value_kind: hidden_remainder_x - .offset: 68 .size: 2 .value_kind: hidden_remainder_y - .offset: 70 .size: 2 .value_kind: hidden_remainder_z - .offset: 88 .size: 8 .value_kind: hidden_global_offset_x - .offset: 96 .size: 8 .value_kind: hidden_global_offset_y - .offset: 104 .size: 8 .value_kind: hidden_global_offset_z - .offset: 112 .size: 2 .value_kind: hidden_grid_dims .group_segment_fixed_size: 0 .kernarg_segment_align: 8 .kernarg_segment_size: 304 .language: OpenCL C .language_version: - 2 - 0 .max_flat_workgroup_size: 1024 .name: _Z27ecuacion_kernel_outplace_p2PdS_S_S_S_j .private_segment_fixed_size: 0 .sgpr_count: 18 .sgpr_spill_count: 0 .symbol: _Z27ecuacion_kernel_outplace_p2PdS_S_S_S_j.kd .uniform_work_group_size: 1 .uses_dynamic_stack: false .vgpr_count: 13 .vgpr_spill_count: 0 .wavefront_size: 32 .workgroup_processor_mode: 1 - .args: - .address_space: global .offset: 0 .size: 8 .value_kind: global_buffer - .address_space: global .offset: 8 .size: 8 .value_kind: global_buffer - .address_space: global .offset: 16 .size: 8 .value_kind: global_buffer - .offset: 24 .size: 4 .value_kind: by_value - .offset: 32 .size: 4 .value_kind: hidden_block_count_x - .offset: 36 .size: 4 .value_kind: hidden_block_count_y - .offset: 40 .size: 4 .value_kind: hidden_block_count_z - .offset: 44 .size: 2 .value_kind: hidden_group_size_x - .offset: 46 .size: 2 .value_kind: hidden_group_size_y - .offset: 48 .size: 2 .value_kind: hidden_group_size_z - .offset: 50 .size: 2 .value_kind: hidden_remainder_x - .offset: 52 .size: 2 .value_kind: hidden_remainder_y - .offset: 54 .size: 2 .value_kind: hidden_remainder_z - .offset: 72 .size: 8 .value_kind: hidden_global_offset_x - .offset: 80 .size: 8 .value_kind: hidden_global_offset_y - .offset: 88 .size: 8 .value_kind: hidden_global_offset_z - .offset: 96 .size: 2 .value_kind: hidden_grid_dims .group_segment_fixed_size: 0 .kernarg_segment_align: 8 .kernarg_segment_size: 288 .language: OpenCL C .language_version: - 2 - 0 .max_flat_workgroup_size: 1024 .name: _Z28ecuacion_kernel_inplace_sumaPdS_S_j .private_segment_fixed_size: 0 .sgpr_count: 18 .sgpr_spill_count: 0 .symbol: _Z28ecuacion_kernel_inplace_sumaPdS_S_j.kd .uniform_work_group_size: 1 .uses_dynamic_stack: false .vgpr_count: 13 .vgpr_spill_count: 0 .wavefront_size: 32 .workgroup_processor_mode: 1 - .args: - .address_space: global .offset: 0 .size: 8 .value_kind: global_buffer - .address_space: global .offset: 8 .size: 8 .value_kind: global_buffer - .address_space: global .offset: 16 .size: 8 .value_kind: global_buffer - .offset: 24 .size: 4 .value_kind: by_value - .offset: 32 .size: 4 .value_kind: hidden_block_count_x - .offset: 36 .size: 4 .value_kind: hidden_block_count_y - .offset: 40 .size: 4 .value_kind: hidden_block_count_z - .offset: 44 .size: 2 .value_kind: hidden_group_size_x - .offset: 46 .size: 2 .value_kind: hidden_group_size_y - .offset: 48 .size: 2 .value_kind: hidden_group_size_z - .offset: 50 .size: 2 .value_kind: hidden_remainder_x - .offset: 52 .size: 2 .value_kind: hidden_remainder_y - .offset: 54 .size: 2 .value_kind: hidden_remainder_z - .offset: 72 .size: 8 .value_kind: hidden_global_offset_x - .offset: 80 .size: 8 .value_kind: hidden_global_offset_y - .offset: 88 .size: 8 .value_kind: hidden_global_offset_z - .offset: 96 .size: 2 .value_kind: hidden_grid_dims .group_segment_fixed_size: 0 .kernarg_segment_align: 8 .kernarg_segment_size: 288 .language: OpenCL C .language_version: - 2 - 0 .max_flat_workgroup_size: 1024 .name: _Z17kernel_sum_MatrizPdS_S_j .private_segment_fixed_size: 0 .sgpr_count: 18 .sgpr_spill_count: 0 .symbol: _Z17kernel_sum_MatrizPdS_S_j.kd .uniform_work_group_size: 1 .uses_dynamic_stack: false .vgpr_count: 8 .vgpr_spill_count: 0 .wavefront_size: 32 .workgroup_processor_mode: 1 - .args: - .address_space: global .offset: 0 .size: 8 .value_kind: global_buffer - .offset: 8 .size: 4 .value_kind: by_value - .offset: 16 .size: 4 .value_kind: hidden_block_count_x - .offset: 20 .size: 4 .value_kind: hidden_block_count_y - .offset: 24 .size: 4 .value_kind: hidden_block_count_z - .offset: 28 .size: 2 .value_kind: hidden_group_size_x - .offset: 30 .size: 2 .value_kind: hidden_group_size_y - .offset: 32 .size: 2 .value_kind: hidden_group_size_z - .offset: 34 .size: 2 .value_kind: hidden_remainder_x - .offset: 36 .size: 2 .value_kind: hidden_remainder_y - .offset: 38 .size: 2 .value_kind: hidden_remainder_z - .offset: 56 .size: 8 .value_kind: hidden_global_offset_x - .offset: 64 .size: 8 .value_kind: hidden_global_offset_y - .offset: 72 .size: 8 .value_kind: hidden_global_offset_z - .offset: 80 .size: 2 .value_kind: hidden_grid_dims .group_segment_fixed_size: 0 .kernarg_segment_align: 8 .kernarg_segment_size: 272 .language: OpenCL C .language_version: - 2 - 0 .max_flat_workgroup_size: 1024 .name: _Z18kernel_transpuestaPdi .private_segment_fixed_size: 0 .sgpr_count: 18 .sgpr_spill_count: 0 .symbol: _Z18kernel_transpuestaPdi.kd .uniform_work_group_size: 1 .uses_dynamic_stack: false .vgpr_count: 8 .vgpr_spill_count: 0 .wavefront_size: 32 .workgroup_processor_mode: 1 - .args: - .address_space: global .offset: 0 .size: 8 .value_kind: global_buffer - .address_space: global .offset: 8 .size: 8 .value_kind: global_buffer - .address_space: global .offset: 16 .size: 8 .value_kind: global_buffer - .offset: 24 .size: 4 .value_kind: by_value - .offset: 32 .size: 4 .value_kind: hidden_block_count_x - .offset: 36 .size: 4 .value_kind: hidden_block_count_y - .offset: 40 .size: 4 .value_kind: hidden_block_count_z - .offset: 44 .size: 2 .value_kind: hidden_group_size_x - .offset: 46 .size: 2 .value_kind: hidden_group_size_y - .offset: 48 .size: 2 .value_kind: hidden_group_size_z - .offset: 50 .size: 2 .value_kind: hidden_remainder_x - .offset: 52 .size: 2 .value_kind: hidden_remainder_y - .offset: 54 .size: 2 .value_kind: hidden_remainder_z - .offset: 72 .size: 8 .value_kind: hidden_global_offset_x - .offset: 80 .size: 8 .value_kind: hidden_global_offset_y - .offset: 88 .size: 8 .value_kind: hidden_global_offset_z - .offset: 96 .size: 2 .value_kind: hidden_grid_dims .group_segment_fixed_size: 0 .kernarg_segment_align: 8 .kernarg_segment_size: 288 .language: OpenCL C .language_version: - 2 - 0 .max_flat_workgroup_size: 1024 .name: _Z22kernel_mult_sum_matrizPdS_S_j .private_segment_fixed_size: 0 .sgpr_count: 18 .sgpr_spill_count: 0 .symbol: _Z22kernel_mult_sum_matrizPdS_S_j.kd .uniform_work_group_size: 1 .uses_dynamic_stack: false .vgpr_count: 12 .vgpr_spill_count: 0 .wavefront_size: 32 .workgroup_processor_mode: 1 amdhsa.target: amdgcn-amd-amdhsa--gfx1100 amdhsa.version: - 1 - 2 ... .end_amdgpu_metadata
You are a helpful coding assistant specialized in converting HIP source code to AMD host assembly.
#include <hip/hip_runtime.h> #include <stdio.h> #include <stdlib.h> #include <sys/time.h> #include <sys/resource.h> #include <math.h> //134217728 double dwalltime(){ double sec; struct timeval tv; gettimeofday(&tv,NULL); sec = tv.tv_sec + tv.tv_usec/1000000.0; return sec; } __global__ void ecuacion_kernel_outplace_p1(double *d_matA,double *d_matAT,double *d_matB,double *d_matBT, unsigned int n){ int distA = blockIdx.y * blockDim.y + threadIdx.y; //i int distB = blockIdx.x * blockDim.x + threadIdx.x; //j //transpuesta out-place A y B if( (distA<n*n) && (distB<n*n) ){ d_matAT [distB*n + distA] = d_matA[distA*n + distB]; d_matBT [distB*n + distA] = d_matB[distA*n + distB]; } } __global__ void ecuacion_kernel_outplace_p2(double *d_matA,double *d_matB,double *d_matC,double *d_matAT,double *d_matBT, unsigned int n){ int distA = blockIdx.y * blockDim.y + threadIdx.y; //i int distB = blockIdx.x * blockDim.x + threadIdx.x; //j int k; if (distA*n+distB <= (n*n - 1)){ //multiplicacion for(k = 0; k < n ;k++){ d_matC[distA*n+distB] += d_matA[distA*n+k] * d_matBT[distB+k*n]; } //suma d_matC[distA*n+distB] += d_matB[distA*n+distB] + d_matAT[distA*n+distB]; } } __global__ void ecuacion_kernel_inplace_suma (double *d_matA,double *d_matB,double *d_matC, unsigned int n){ int distA = blockIdx.y * blockDim.y + threadIdx.y; //i int distB = blockIdx.x * blockDim.x + threadIdx.x; //j int k; //multiplicacion if (distA*n+distB < (n*n - 1)){ d_matC[distA*n+distB] += d_matB[distA*n+distB] + d_matA[distA+distB*n]; for(k = 0; k < n ;k++){ d_matC[distA*n+distB] += d_matA[distA*n+k] * d_matB[distB*n+k]; } } } __global__ void kernel_sum_Matriz (double *d_matA,double *d_matB,double *d_matC, unsigned int n){ int distA = blockIdx.y * blockDim.y + threadIdx.y; //i int distB = blockIdx.x * blockDim.x + threadIdx.x; //j //suma if (distA*n+distB < (n*n)){ d_matC[distA*n+distB] += d_matA[distA*n+distB] + d_matB[distA+distB*n]; } } __global__ void kernel_transpuesta(double *m, int N){ int tid = blockIdx.x*blockDim.x + threadIdx.x; int i = int((1 + sqrtf(1 + 8*tid)) / 2); int j = tid - (i*(i-1)/2); int aux; if ( (i<N) && (j<N) ){ aux = m[i*N + j] ; m[i*N + j] = m[j*N + i]; m[j*N + i] = aux; } } __global__ void kernel_mult_sum_matriz (double *d_matA,double *d_matB,double *d_matC, unsigned int n){ int distA = blockIdx.y * blockDim.y + threadIdx.y; //i int distB = blockIdx.x * blockDim.x + threadIdx.x; //j int k; //multiplicacion if (distA*n+distB < (n*n)){ for(k = 0; k < n ;k++){ d_matC[distA*n+distB] += d_matA[distA*n+k] * d_matB[distB*n+k]; } } } int main(int argc, char *argv[]){ if (argc != 3){ printf("Falta argumento: N, CUDABLK\n"); return 0; } //declaracion de variables hipError_t error; unsigned int N = atoi (argv[1]); unsigned long CUDA_BLK = atoi (argv[2]), gridBlock; unsigned long numBytes = sizeof(double)*N*N; double *matA,*matB,*matC,*d_matA,*d_matB,*d_matC,*d_matAT,*d_matBT,timetick; unsigned int i,j,k; //inicializa variables para cpu matA = (double *)malloc(numBytes); matB = (double *)malloc(numBytes); matC = (double *)malloc(numBytes); for (i = 0; i < N*N; i++){ matA[i] = i; matB[i] = i; matC[i] = 0; } //inicializa variables para gpu hipMalloc((void **) &d_matA, numBytes); hipMalloc((void **) &d_matAT, numBytes); hipMalloc((void **) &d_matB, numBytes); hipMalloc((void **) &d_matBT, numBytes); hipMalloc((void **) &d_matC, numBytes); gridBlock = (unsigned int)sqrt(N*N/CUDA_BLK/CUDA_BLK); dim3 dimBlock(CUDA_BLK,CUDA_BLK); // Bloque bidimencional de hilos (*cb* hilos) dim3 dimGrid(gridBlock,gridBlock); // Grid bidimencional (*ceil(n/cb)* bloques) //--------------------------------cpu comienza ------------------------------------ //secuencial timetick = dwalltime(); //multiplicacion for(i = 0; i < N; i++){ for(j = 0; j < N; j++){ for(k = 0; k < N ;k++){ matC[i*N+j] += matA[i*N+k] * matB[j*N+k]; //multiplica a matB por fila, eso simula la matB transpuesta } } } //suma for(i = 0; i < N; i++){ for(j = 0; j < N; j++){ matC[i*N+j] += matB[i*N+j] + matA[i+j*N]; } } printf("Tiempo para la ecuacion CPU: %f\n\n",dwalltime() - timetick); /* for(i = 0; i < N; i++){ for(j = 0; j < N; j++){ printf("%f|",matC[i*N+j]); } printf("\n"); } printf("\n"); */ //--------------------------------cpu termina ------------------------------------ for (i = 0; i < N*N; i++){ matA[i] = i; matB[i] = i; matC[i] = 0; } hipMemcpy(d_matA, matA, numBytes, hipMemcpyHostToDevice); // CPU -> GPU hipMemcpy(d_matB, matB, numBytes, hipMemcpyHostToDevice); // CPU -> GPU hipMemcpy(d_matC, matC, numBytes, hipMemcpyHostToDevice); // CPU -> GPU //--------------------------------gpu out-place comienza ------------------------------------ timetick = dwalltime(); ecuacion_kernel_outplace_p1<<<dimGrid, dimBlock>>>(d_matA, d_matAT,d_matB,d_matBT, N); hipDeviceSynchronize(); ecuacion_kernel_outplace_p2<<<dimGrid, dimBlock>>>(d_matA, d_matB,d_matC,d_matAT,d_matBT, N); hipDeviceSynchronize(); printf("Tiempo para la ecuacion out-place GPU: %f\n",dwalltime() - timetick); error = hipGetLastError(); printf("error: %d\n\n",error); hipMemcpy(matC, d_matC, numBytes, hipMemcpyDeviceToHost); // GPU -> CPU /* for(i = 0; i < N; i++){ for(j = 0; j < N; j++){ printf("%f|",matC[i*N+j]); } printf("\n"); } printf("\n"); */ //--------------------------------gpu out-place termina ------------------------------------ hipFree(d_matA); hipFree(d_matB); hipFree(d_matC); hipFree(d_matAT); hipFree(d_matBT); for (i = 0; i < N*N; i++){ matA[i] = i; matB[i] = i; matC[i] = 0; } hipMalloc((void **) &d_matA, numBytes); hipMalloc((void **) &d_matB, numBytes); hipMalloc((void **) &d_matC, numBytes); hipMemcpy(d_matA, matA, numBytes, hipMemcpyHostToDevice); // CPU -> GPU hipMemcpy(d_matB, matB, numBytes, hipMemcpyHostToDevice); // CPU -> GPU hipMemcpy(d_matC, matC, numBytes, hipMemcpyHostToDevice); // CPU -> GPU //--------------------------------gpu in-place comienza ------------------------------------ timetick = dwalltime(); kernel_transpuesta<<<dimGrid, dimBlock>>>(d_matA, N); hipDeviceSynchronize(); kernel_sum_Matriz<<<dimGrid, dimBlock>>>(d_matA, d_matB,d_matC, N); hipDeviceSynchronize(); kernel_transpuesta<<<dimGrid, dimBlock>>>(d_matA, N); hipDeviceSynchronize(); kernel_mult_sum_matriz<<<dimGrid, dimBlock>>>(d_matA, d_matB,d_matC, N); hipDeviceSynchronize(); printf("Tiempo para la ecuacion in-place GPU: %f\n",dwalltime() - timetick); error = hipGetLastError(); printf("error: %d\n\n",error); hipMemcpy(matC, d_matC, numBytes, hipMemcpyDeviceToHost); // GPU -> CPU //--------------------------------gpu in-place termina ------------------------------------ hipFree(d_matA); hipFree(d_matB); hipFree(d_matC); /* //imprime la matriz matC for(i = 0; i < N; i++){ for(j = 0; j < N; j++){ printf("%f|",matC[i*N+j]); } printf("\n"); } printf("\n"); */ free(matA); free(matB); free(matC); return 0; }
.text .file "ejercicio1_sin_tiempo_copia.hip" .section .rodata.cst8,"aM",@progbits,8 .p2align 3, 0x0 # -- Begin function _Z9dwalltimev .LCPI0_0: .quad 0x412e848000000000 # double 1.0E+6 .text .globl _Z9dwalltimev .p2align 4, 0x90 .type _Z9dwalltimev,@function _Z9dwalltimev: # @_Z9dwalltimev .cfi_startproc # %bb.0: subq $24, %rsp .cfi_def_cfa_offset 32 leaq 8(%rsp), %rdi xorl %esi, %esi callq gettimeofday cvtsi2sdq 8(%rsp), %xmm1 cvtsi2sdq 16(%rsp), %xmm0 divsd .LCPI0_0(%rip), %xmm0 addsd %xmm1, %xmm0 addq $24, %rsp .cfi_def_cfa_offset 8 retq .Lfunc_end0: .size _Z9dwalltimev, .Lfunc_end0-_Z9dwalltimev .cfi_endproc # -- End function .globl _Z42__device_stub__ecuacion_kernel_outplace_p1PdS_S_S_j # -- Begin function _Z42__device_stub__ecuacion_kernel_outplace_p1PdS_S_S_j .p2align 4, 0x90 .type _Z42__device_stub__ecuacion_kernel_outplace_p1PdS_S_S_j,@function _Z42__device_stub__ecuacion_kernel_outplace_p1PdS_S_S_j: # @_Z42__device_stub__ecuacion_kernel_outplace_p1PdS_S_S_j .cfi_startproc # %bb.0: subq $136, %rsp .cfi_def_cfa_offset 144 movq %rdi, 88(%rsp) movq %rsi, 80(%rsp) movq %rdx, 72(%rsp) movq %rcx, 64(%rsp) movl %r8d, 12(%rsp) leaq 88(%rsp), %rax movq %rax, 96(%rsp) leaq 80(%rsp), %rax movq %rax, 104(%rsp) leaq 72(%rsp), %rax movq %rax, 112(%rsp) leaq 64(%rsp), %rax movq %rax, 120(%rsp) leaq 12(%rsp), %rax movq %rax, 128(%rsp) leaq 48(%rsp), %rdi leaq 32(%rsp), %rsi leaq 24(%rsp), %rdx leaq 16(%rsp), %rcx callq __hipPopCallConfiguration movq 48(%rsp), %rsi movl 56(%rsp), %edx movq 32(%rsp), %rcx movl 40(%rsp), %r8d leaq 96(%rsp), %r9 movl $_Z27ecuacion_kernel_outplace_p1PdS_S_S_j, %edi pushq 16(%rsp) .cfi_adjust_cfa_offset 8 pushq 32(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $152, %rsp .cfi_adjust_cfa_offset -152 retq .Lfunc_end1: .size _Z42__device_stub__ecuacion_kernel_outplace_p1PdS_S_S_j, .Lfunc_end1-_Z42__device_stub__ecuacion_kernel_outplace_p1PdS_S_S_j .cfi_endproc # -- End function .globl _Z42__device_stub__ecuacion_kernel_outplace_p2PdS_S_S_S_j # -- Begin function _Z42__device_stub__ecuacion_kernel_outplace_p2PdS_S_S_S_j .p2align 4, 0x90 .type _Z42__device_stub__ecuacion_kernel_outplace_p2PdS_S_S_S_j,@function _Z42__device_stub__ecuacion_kernel_outplace_p2PdS_S_S_S_j: # @_Z42__device_stub__ecuacion_kernel_outplace_p2PdS_S_S_S_j .cfi_startproc # %bb.0: subq $152, %rsp .cfi_def_cfa_offset 160 movq %rdi, 88(%rsp) movq %rsi, 80(%rsp) movq %rdx, 72(%rsp) movq %rcx, 64(%rsp) movq %r8, 56(%rsp) movl %r9d, 4(%rsp) leaq 88(%rsp), %rax movq %rax, 96(%rsp) leaq 80(%rsp), %rax movq %rax, 104(%rsp) leaq 72(%rsp), %rax movq %rax, 112(%rsp) leaq 64(%rsp), %rax movq %rax, 120(%rsp) leaq 56(%rsp), %rax movq %rax, 128(%rsp) leaq 4(%rsp), %rax movq %rax, 136(%rsp) leaq 40(%rsp), %rdi leaq 24(%rsp), %rsi leaq 16(%rsp), %rdx leaq 8(%rsp), %rcx callq __hipPopCallConfiguration movq 40(%rsp), %rsi movl 48(%rsp), %edx movq 24(%rsp), %rcx movl 32(%rsp), %r8d leaq 96(%rsp), %r9 movl $_Z27ecuacion_kernel_outplace_p2PdS_S_S_S_j, %edi pushq 8(%rsp) .cfi_adjust_cfa_offset 8 pushq 24(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $168, %rsp .cfi_adjust_cfa_offset -168 retq .Lfunc_end2: .size _Z42__device_stub__ecuacion_kernel_outplace_p2PdS_S_S_S_j, .Lfunc_end2-_Z42__device_stub__ecuacion_kernel_outplace_p2PdS_S_S_S_j .cfi_endproc # -- End function .globl _Z43__device_stub__ecuacion_kernel_inplace_sumaPdS_S_j # -- Begin function _Z43__device_stub__ecuacion_kernel_inplace_sumaPdS_S_j .p2align 4, 0x90 .type _Z43__device_stub__ecuacion_kernel_inplace_sumaPdS_S_j,@function _Z43__device_stub__ecuacion_kernel_inplace_sumaPdS_S_j: # @_Z43__device_stub__ecuacion_kernel_inplace_sumaPdS_S_j .cfi_startproc # %bb.0: subq $120, %rsp .cfi_def_cfa_offset 128 movq %rdi, 72(%rsp) movq %rsi, 64(%rsp) movq %rdx, 56(%rsp) movl %ecx, 4(%rsp) leaq 72(%rsp), %rax movq %rax, 80(%rsp) leaq 64(%rsp), %rax movq %rax, 88(%rsp) leaq 56(%rsp), %rax movq %rax, 96(%rsp) leaq 4(%rsp), %rax movq %rax, 104(%rsp) leaq 40(%rsp), %rdi leaq 24(%rsp), %rsi leaq 16(%rsp), %rdx leaq 8(%rsp), %rcx callq __hipPopCallConfiguration movq 40(%rsp), %rsi movl 48(%rsp), %edx movq 24(%rsp), %rcx movl 32(%rsp), %r8d leaq 80(%rsp), %r9 movl $_Z28ecuacion_kernel_inplace_sumaPdS_S_j, %edi pushq 8(%rsp) .cfi_adjust_cfa_offset 8 pushq 24(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $136, %rsp .cfi_adjust_cfa_offset -136 retq .Lfunc_end3: .size _Z43__device_stub__ecuacion_kernel_inplace_sumaPdS_S_j, .Lfunc_end3-_Z43__device_stub__ecuacion_kernel_inplace_sumaPdS_S_j .cfi_endproc # -- End function .globl _Z32__device_stub__kernel_sum_MatrizPdS_S_j # -- Begin function _Z32__device_stub__kernel_sum_MatrizPdS_S_j .p2align 4, 0x90 .type _Z32__device_stub__kernel_sum_MatrizPdS_S_j,@function _Z32__device_stub__kernel_sum_MatrizPdS_S_j: # @_Z32__device_stub__kernel_sum_MatrizPdS_S_j .cfi_startproc # %bb.0: subq $120, %rsp .cfi_def_cfa_offset 128 movq %rdi, 72(%rsp) movq %rsi, 64(%rsp) movq %rdx, 56(%rsp) movl %ecx, 4(%rsp) leaq 72(%rsp), %rax movq %rax, 80(%rsp) leaq 64(%rsp), %rax movq %rax, 88(%rsp) leaq 56(%rsp), %rax movq %rax, 96(%rsp) leaq 4(%rsp), %rax movq %rax, 104(%rsp) leaq 40(%rsp), %rdi leaq 24(%rsp), %rsi leaq 16(%rsp), %rdx leaq 8(%rsp), %rcx callq __hipPopCallConfiguration movq 40(%rsp), %rsi movl 48(%rsp), %edx movq 24(%rsp), %rcx movl 32(%rsp), %r8d leaq 80(%rsp), %r9 movl $_Z17kernel_sum_MatrizPdS_S_j, %edi pushq 8(%rsp) .cfi_adjust_cfa_offset 8 pushq 24(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $136, %rsp .cfi_adjust_cfa_offset -136 retq .Lfunc_end4: .size _Z32__device_stub__kernel_sum_MatrizPdS_S_j, .Lfunc_end4-_Z32__device_stub__kernel_sum_MatrizPdS_S_j .cfi_endproc # -- End function .globl _Z33__device_stub__kernel_transpuestaPdi # -- Begin function _Z33__device_stub__kernel_transpuestaPdi .p2align 4, 0x90 .type _Z33__device_stub__kernel_transpuestaPdi,@function _Z33__device_stub__kernel_transpuestaPdi: # @_Z33__device_stub__kernel_transpuestaPdi .cfi_startproc # %bb.0: subq $88, %rsp .cfi_def_cfa_offset 96 movq %rdi, 56(%rsp) movl %esi, 4(%rsp) leaq 56(%rsp), %rax movq %rax, 64(%rsp) leaq 4(%rsp), %rax movq %rax, 72(%rsp) leaq 40(%rsp), %rdi leaq 24(%rsp), %rsi leaq 16(%rsp), %rdx leaq 8(%rsp), %rcx callq __hipPopCallConfiguration movq 40(%rsp), %rsi movl 48(%rsp), %edx movq 24(%rsp), %rcx movl 32(%rsp), %r8d leaq 64(%rsp), %r9 movl $_Z18kernel_transpuestaPdi, %edi pushq 8(%rsp) .cfi_adjust_cfa_offset 8 pushq 24(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $104, %rsp .cfi_adjust_cfa_offset -104 retq .Lfunc_end5: .size _Z33__device_stub__kernel_transpuestaPdi, .Lfunc_end5-_Z33__device_stub__kernel_transpuestaPdi .cfi_endproc # -- End function .globl _Z37__device_stub__kernel_mult_sum_matrizPdS_S_j # -- Begin function _Z37__device_stub__kernel_mult_sum_matrizPdS_S_j .p2align 4, 0x90 .type _Z37__device_stub__kernel_mult_sum_matrizPdS_S_j,@function _Z37__device_stub__kernel_mult_sum_matrizPdS_S_j: # @_Z37__device_stub__kernel_mult_sum_matrizPdS_S_j .cfi_startproc # %bb.0: subq $120, %rsp .cfi_def_cfa_offset 128 movq %rdi, 72(%rsp) movq %rsi, 64(%rsp) movq %rdx, 56(%rsp) movl %ecx, 4(%rsp) leaq 72(%rsp), %rax movq %rax, 80(%rsp) leaq 64(%rsp), %rax movq %rax, 88(%rsp) leaq 56(%rsp), %rax movq %rax, 96(%rsp) leaq 4(%rsp), %rax movq %rax, 104(%rsp) leaq 40(%rsp), %rdi leaq 24(%rsp), %rsi leaq 16(%rsp), %rdx leaq 8(%rsp), %rcx callq __hipPopCallConfiguration movq 40(%rsp), %rsi movl 48(%rsp), %edx movq 24(%rsp), %rcx movl 32(%rsp), %r8d leaq 80(%rsp), %r9 movl $_Z22kernel_mult_sum_matrizPdS_S_j, %edi pushq 8(%rsp) .cfi_adjust_cfa_offset 8 pushq 24(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $136, %rsp .cfi_adjust_cfa_offset -136 retq .Lfunc_end6: .size _Z37__device_stub__kernel_mult_sum_matrizPdS_S_j, .Lfunc_end6-_Z37__device_stub__kernel_mult_sum_matrizPdS_S_j .cfi_endproc # -- End function .section .rodata.cst8,"aM",@progbits,8 .p2align 3, 0x0 # -- Begin function main .LCPI7_0: .quad 0x412e848000000000 # double 1.0E+6 .text .globl main .p2align 4, 0x90 .type main,@function main: # @main .cfi_startproc # %bb.0: pushq %rbp .cfi_def_cfa_offset 16 pushq %r15 .cfi_def_cfa_offset 24 pushq %r14 .cfi_def_cfa_offset 32 pushq %r13 .cfi_def_cfa_offset 40 pushq %r12 .cfi_def_cfa_offset 48 pushq %rbx .cfi_def_cfa_offset 56 subq $264, %rsp # imm = 0x108 .cfi_def_cfa_offset 320 .cfi_offset %rbx, -56 .cfi_offset %r12, -48 .cfi_offset %r13, -40 .cfi_offset %r14, -32 .cfi_offset %r15, -24 .cfi_offset %rbp, -16 cmpl $3, %edi jne .LBB7_1 # %bb.2: movq 8(%rsi), %rdi movq %rsi, %rbx xorl %esi, %esi movl $10, %edx callq __isoc23_strtol movq %rax, %r13 movq 16(%rbx), %rdi xorl %esi, %esi movl $10, %edx callq __isoc23_strtol movq %rax, 96(%rsp) # 8-byte Spill cltq movq %rax, 184(%rsp) # 8-byte Spill movl %r13d, %ebp imulq %rbp, %rbp shlq $3, %rbp movq %rbp, %rdi callq malloc movq %rax, %rbx movq %rbp, %rdi callq malloc movq %rax, %r14 movq %rbp, %rdi callq malloc movq %rax, %r15 movl %r13d, %ecx imull %r13d, %ecx movl %ecx, %eax movl %ecx, 196(%rsp) # 4-byte Spill testl %ecx, %ecx movq %rax, 8(%rsp) # 8-byte Spill je .LBB7_5 # %bb.3: # %.lr.ph.preheader leaq (,%rax,8), %rdx xorl %r12d, %r12d movq %r15, %rdi xorl %esi, %esi callq memset@PLT movq 8(%rsp), %rcx # 8-byte Reload .p2align 4, 0x90 .LBB7_4: # %.lr.ph # =>This Inner Loop Header: Depth=1 movl %r12d, %eax xorps %xmm0, %xmm0 cvtsi2sd %rax, %xmm0 movsd %xmm0, (%rbx,%r12,8) movsd %xmm0, (%r14,%r12,8) incq %r12 cmpq %r12, %rcx jne .LBB7_4 .LBB7_5: # %._crit_edge leaq 16(%rsp), %rdi movq %rbp, %rsi callq hipMalloc leaq 232(%rsp), %rdi movq %rbp, %rsi callq hipMalloc leaq 104(%rsp), %rdi movq %rbp, %rsi callq hipMalloc leaq 224(%rsp), %rdi movq %rbp, %rsi callq hipMalloc leaq 88(%rsp), %rdi movq %rbp, %rsi callq hipMalloc movq 8(%rsp), %rax # 8-byte Reload xorl %edx, %edx movq 184(%rsp), %rcx # 8-byte Reload divq %rcx xorl %edx, %edx divq %rcx xorps %xmm0, %xmm0 cvtsi2sd %rax, %xmm0 xorpd %xmm1, %xmm1 ucomisd %xmm1, %xmm0 jb .LBB7_7 # %bb.6: sqrtsd %xmm0, %xmm0 jmp .LBB7_8 .LBB7_1: movl $.Lstr, %edi callq puts@PLT jmp .LBB7_38 .LBB7_7: # %call.sqrt callq sqrt .LBB7_8: # %._crit_edge.split cvttsd2si %xmm0, %rdx movq 96(%rsp), %rcx # 8-byte Reload movl %ecx, %eax movq %rax, 184(%rsp) # 8-byte Spill shlq $32, %rcx movq %rcx, 96(%rsp) # 8-byte Spill movl %edx, %eax movq %rax, 216(%rsp) # 8-byte Spill shlq $32, %rdx movq %rdx, 200(%rsp) # 8-byte Spill xorl %r12d, %r12d leaq 128(%rsp), %rdi xorl %esi, %esi callq gettimeofday xorps %xmm0, %xmm0 cvtsi2sdq 128(%rsp), %xmm0 xorps %xmm1, %xmm1 cvtsi2sdq 136(%rsp), %xmm1 divsd .LCPI7_0(%rip), %xmm1 addsd %xmm0, %xmm1 movsd %xmm1, 256(%rsp) # 8-byte Spill testl %r13d, %r13d movq %rbp, 208(%rsp) # 8-byte Spill je .LBB7_19 # %bb.9: # %.preheader241.preheader movl %r13d, %eax xorl %ecx, %ecx .p2align 4, 0x90 .LBB7_10: # %.preheader241 # =>This Loop Header: Depth=1 # Child Loop BB7_11 Depth 2 # Child Loop BB7_12 Depth 3 movl %ecx, %edx imull %r13d, %edx xorl %esi, %esi xorl %edi, %edi .p2align 4, 0x90 .LBB7_11: # %.preheader240 # Parent Loop BB7_10 Depth=1 # => This Loop Header: Depth=2 # Child Loop BB7_12 Depth 3 leal (%rdx,%rdi), %r8d movsd (%r15,%r8,8), %xmm0 # xmm0 = mem[0],zero movq %rax, %r9 movl %r12d, %r10d movq %rsi, %r11 .p2align 4, 0x90 .LBB7_12: # Parent Loop BB7_10 Depth=1 # Parent Loop BB7_11 Depth=2 # => This Inner Loop Header: Depth=3 movl %r10d, %ebp movsd (%rbx,%rbp,8), %xmm1 # xmm1 = mem[0],zero movl %r11d, %ebp mulsd (%r14,%rbp,8), %xmm1 addsd %xmm1, %xmm0 incq %r11 incl %r10d decq %r9 jne .LBB7_12 # %bb.13: # in Loop: Header=BB7_11 Depth=2 movsd %xmm0, (%r15,%r8,8) incq %rdi addq %r13, %rsi cmpq %rax, %rdi jne .LBB7_11 # %bb.14: # in Loop: Header=BB7_10 Depth=1 incl %ecx addl %r13d, %r12d cmpl %r13d, %ecx jne .LBB7_10 # %bb.15: # %.preheader.preheader xorl %ecx, %ecx xorl %edx, %edx movq 208(%rsp), %rbp # 8-byte Reload .p2align 4, 0x90 .LBB7_16: # %.preheader # =>This Loop Header: Depth=1 # Child Loop BB7_17 Depth 2 movq %rax, %rsi movl %ecx, %edi movl %edx, %r8d .p2align 4, 0x90 .LBB7_17: # Parent Loop BB7_16 Depth=1 # => This Inner Loop Header: Depth=2 movl %edi, %r9d movsd (%r14,%r9,8), %xmm0 # xmm0 = mem[0],zero movl %r8d, %r10d addsd (%rbx,%r10,8), %xmm0 addsd (%r15,%r9,8), %xmm0 movsd %xmm0, (%r15,%r9,8) addl %r13d, %r8d incl %edi decq %rsi jne .LBB7_17 # %bb.18: # in Loop: Header=BB7_16 Depth=1 incl %edx addl %r13d, %ecx cmpl %r13d, %edx jne .LBB7_16 .LBB7_19: # %._crit_edge249 movq 96(%rsp), %rax # 8-byte Reload addq %rax, 184(%rsp) # 8-byte Folded Spill movq 200(%rsp), %rax # 8-byte Reload addq %rax, 216(%rsp) # 8-byte Folded Spill leaq 128(%rsp), %rdi xorl %esi, %esi callq gettimeofday xorps %xmm1, %xmm1 cvtsi2sdq 128(%rsp), %xmm1 xorps %xmm0, %xmm0 cvtsi2sdq 136(%rsp), %xmm0 divsd .LCPI7_0(%rip), %xmm0 addsd %xmm1, %xmm0 subsd 256(%rsp), %xmm0 # 8-byte Folded Reload movl $.L.str.1, %edi movb $1, %al callq printf cmpl $0, 196(%rsp) # 4-byte Folded Reload movq 8(%rsp), %rax # 8-byte Reload je .LBB7_22 # %bb.20: # %.lr.ph252.preheader leaq (,%rax,8), %rdx xorl %r12d, %r12d movq %r15, %rdi xorl %esi, %esi callq memset@PLT movq 8(%rsp), %rcx # 8-byte Reload .p2align 4, 0x90 .LBB7_21: # %.lr.ph252 # =>This Inner Loop Header: Depth=1 movl %r12d, %eax xorps %xmm0, %xmm0 cvtsi2sd %rax, %xmm0 movsd %xmm0, (%rbx,%r12,8) movsd %xmm0, (%r14,%r12,8) incq %r12 cmpq %r12, %rcx jne .LBB7_21 .LBB7_22: # %._crit_edge253 movq 16(%rsp), %rdi movq %rbx, %rsi movq %rbp, %rdx movl $1, %ecx callq hipMemcpy movq 104(%rsp), %rdi movq %r14, %rsi movq %rbp, %rdx movl $1, %ecx callq hipMemcpy movq 88(%rsp), %rdi movq %r15, %rsi movq %rbp, %rdx movl $1, %ecx callq hipMemcpy leaq 128(%rsp), %rdi xorl %esi, %esi callq gettimeofday xorps %xmm0, %xmm0 cvtsi2sdq 128(%rsp), %xmm0 movsd %xmm0, 200(%rsp) # 8-byte Spill xorps %xmm0, %xmm0 cvtsi2sdq 136(%rsp), %xmm0 divsd .LCPI7_0(%rip), %xmm0 movsd %xmm0, 96(%rsp) # 8-byte Spill movq 216(%rsp), %rbp # 8-byte Reload movq %rbp, %rdi movl $1, %esi movq 184(%rsp), %r12 # 8-byte Reload movq %r12, %rdx movl $1, %ecx xorl %r8d, %r8d xorl %r9d, %r9d callq __hipPushCallConfiguration testl %eax, %eax jne .LBB7_24 # %bb.23: movq 16(%rsp), %rax movq 232(%rsp), %rcx movq 104(%rsp), %rdx movq 224(%rsp), %rsi movq %rax, 80(%rsp) movq %rcx, 72(%rsp) movq %rdx, 64(%rsp) movq %rsi, 24(%rsp) movl %r13d, 240(%rsp) leaq 80(%rsp), %rax movq %rax, 128(%rsp) leaq 72(%rsp), %rax movq %rax, 136(%rsp) leaq 64(%rsp), %rax movq %rax, 144(%rsp) leaq 24(%rsp), %rax movq %rax, 152(%rsp) leaq 240(%rsp), %rax movq %rax, 160(%rsp) leaq 48(%rsp), %rdi leaq 32(%rsp), %rsi leaq 120(%rsp), %rdx leaq 112(%rsp), %rcx callq __hipPopCallConfiguration movq 48(%rsp), %rsi movl 56(%rsp), %edx movq 32(%rsp), %rcx movl 40(%rsp), %r8d leaq 128(%rsp), %r9 movl $_Z27ecuacion_kernel_outplace_p1PdS_S_S_j, %edi pushq 112(%rsp) .cfi_adjust_cfa_offset 8 pushq 128(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $16, %rsp .cfi_adjust_cfa_offset -16 .LBB7_24: movsd 96(%rsp), %xmm0 # 8-byte Reload # xmm0 = mem[0],zero addsd 200(%rsp), %xmm0 # 8-byte Folded Reload movsd %xmm0, 96(%rsp) # 8-byte Spill callq hipDeviceSynchronize movq %rbp, %rdi movl $1, %esi movq %r12, %rdx movl $1, %ecx xorl %r8d, %r8d xorl %r9d, %r9d callq __hipPushCallConfiguration testl %eax, %eax jne .LBB7_26 # %bb.25: movq 16(%rsp), %rax movq 104(%rsp), %rcx movq 88(%rsp), %rdx movq 232(%rsp), %rsi movq 224(%rsp), %rdi movq %rax, 80(%rsp) movq %rcx, 72(%rsp) movq %rdx, 64(%rsp) movq %rsi, 24(%rsp) movq %rdi, 120(%rsp) movl %r13d, 252(%rsp) leaq 80(%rsp), %rax movq %rax, 128(%rsp) leaq 72(%rsp), %rax movq %rax, 136(%rsp) leaq 64(%rsp), %rax movq %rax, 144(%rsp) leaq 24(%rsp), %rax movq %rax, 152(%rsp) leaq 120(%rsp), %rax movq %rax, 160(%rsp) leaq 252(%rsp), %rax movq %rax, 168(%rsp) leaq 48(%rsp), %rdi leaq 32(%rsp), %rsi leaq 112(%rsp), %rdx leaq 240(%rsp), %rcx callq __hipPopCallConfiguration movq 48(%rsp), %rsi movl 56(%rsp), %edx movq 32(%rsp), %rcx movl 40(%rsp), %r8d leaq 128(%rsp), %r9 movl $_Z27ecuacion_kernel_outplace_p2PdS_S_S_S_j, %edi pushq 240(%rsp) .cfi_adjust_cfa_offset 8 pushq 120(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $16, %rsp .cfi_adjust_cfa_offset -16 .LBB7_26: callq hipDeviceSynchronize leaq 128(%rsp), %rdi xorl %esi, %esi callq gettimeofday xorps %xmm1, %xmm1 cvtsi2sdq 128(%rsp), %xmm1 xorps %xmm0, %xmm0 cvtsi2sdq 136(%rsp), %xmm0 divsd .LCPI7_0(%rip), %xmm0 addsd %xmm1, %xmm0 subsd 96(%rsp), %xmm0 # 8-byte Folded Reload movl $.L.str.2, %edi movb $1, %al callq printf callq hipGetLastError movl $.L.str.3, %edi movl %eax, %esi xorl %eax, %eax callq printf movq 88(%rsp), %rsi movq %r15, %rdi movq 208(%rsp), %rbp # 8-byte Reload movq %rbp, %rdx movl $2, %ecx callq hipMemcpy movq 16(%rsp), %rdi callq hipFree movq 104(%rsp), %rdi callq hipFree movq 88(%rsp), %rdi callq hipFree movq 232(%rsp), %rdi callq hipFree movq 224(%rsp), %rdi callq hipFree cmpl $0, 196(%rsp) # 4-byte Folded Reload movq 8(%rsp), %rax # 8-byte Reload je .LBB7_29 # %bb.27: # %.lr.ph256.preheader leaq (,%rax,8), %rdx xorl %r12d, %r12d movq %r15, %rdi xorl %esi, %esi callq memset@PLT movq 8(%rsp), %rcx # 8-byte Reload .p2align 4, 0x90 .LBB7_28: # %.lr.ph256 # =>This Inner Loop Header: Depth=1 movl %r12d, %eax xorps %xmm0, %xmm0 cvtsi2sd %rax, %xmm0 movsd %xmm0, (%rbx,%r12,8) movsd %xmm0, (%r14,%r12,8) incq %r12 cmpq %r12, %rcx jne .LBB7_28 .LBB7_29: # %._crit_edge257 leaq 16(%rsp), %rdi movq %rbp, %rsi callq hipMalloc leaq 104(%rsp), %rdi movq %rbp, %rsi callq hipMalloc leaq 88(%rsp), %rdi movq %rbp, %rsi callq hipMalloc movq 16(%rsp), %rdi movq %rbx, %rsi movq %rbp, %rdx movl $1, %ecx callq hipMemcpy movq 104(%rsp), %rdi movq %r14, %rsi movq %rbp, %rdx movl $1, %ecx callq hipMemcpy movq 88(%rsp), %rdi movq %r15, %rsi movq %rbp, %rdx movl $1, %ecx callq hipMemcpy leaq 128(%rsp), %rdi xorl %esi, %esi callq gettimeofday xorps %xmm0, %xmm0 cvtsi2sdq 128(%rsp), %xmm0 movsd %xmm0, 96(%rsp) # 8-byte Spill xorps %xmm0, %xmm0 cvtsi2sdq 136(%rsp), %xmm0 movsd %xmm0, 8(%rsp) # 8-byte Spill movq 216(%rsp), %rbp # 8-byte Reload movq %rbp, %rdi movl $1, %esi movq 184(%rsp), %r12 # 8-byte Reload movq %r12, %rdx movl $1, %ecx xorl %r8d, %r8d xorl %r9d, %r9d callq __hipPushCallConfiguration testl %eax, %eax jne .LBB7_31 # %bb.30: movq 16(%rsp), %rax movq %rax, 80(%rsp) movl %r13d, 24(%rsp) leaq 80(%rsp), %rax movq %rax, 128(%rsp) leaq 24(%rsp), %rax movq %rax, 136(%rsp) leaq 48(%rsp), %rdi leaq 32(%rsp), %rsi leaq 72(%rsp), %rdx leaq 64(%rsp), %rcx callq __hipPopCallConfiguration movq 48(%rsp), %rsi movl 56(%rsp), %edx movq 32(%rsp), %rcx movl 40(%rsp), %r8d leaq 128(%rsp), %r9 movl $_Z18kernel_transpuestaPdi, %edi pushq 64(%rsp) .cfi_adjust_cfa_offset 8 pushq 80(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $16, %rsp .cfi_adjust_cfa_offset -16 .LBB7_31: callq hipDeviceSynchronize movq %rbp, %rdi movl $1, %esi movq %r12, %rdx movl $1, %ecx xorl %r8d, %r8d xorl %r9d, %r9d callq __hipPushCallConfiguration testl %eax, %eax jne .LBB7_33 # %bb.32: movq 16(%rsp), %rax movq 104(%rsp), %rcx movq 88(%rsp), %rdx movq %rax, 80(%rsp) movq %rcx, 72(%rsp) movq %rdx, 64(%rsp) movl %r13d, 112(%rsp) leaq 80(%rsp), %rax movq %rax, 128(%rsp) leaq 72(%rsp), %rax movq %rax, 136(%rsp) leaq 64(%rsp), %rax movq %rax, 144(%rsp) leaq 112(%rsp), %rax movq %rax, 152(%rsp) leaq 48(%rsp), %rdi leaq 32(%rsp), %rsi leaq 24(%rsp), %rdx leaq 120(%rsp), %rcx callq __hipPopCallConfiguration movq 48(%rsp), %rsi movl 56(%rsp), %edx movq 32(%rsp), %rcx movl 40(%rsp), %r8d leaq 128(%rsp), %r9 movl $_Z17kernel_sum_MatrizPdS_S_j, %edi pushq 120(%rsp) .cfi_adjust_cfa_offset 8 pushq 32(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $16, %rsp .cfi_adjust_cfa_offset -16 .LBB7_33: movsd 8(%rsp), %xmm0 # 8-byte Reload # xmm0 = mem[0],zero divsd .LCPI7_0(%rip), %xmm0 movsd %xmm0, 8(%rsp) # 8-byte Spill callq hipDeviceSynchronize movq %rbp, %rdi movl $1, %esi movq %r12, %rdx movl $1, %ecx xorl %r8d, %r8d xorl %r9d, %r9d callq __hipPushCallConfiguration testl %eax, %eax jne .LBB7_35 # %bb.34: movq 16(%rsp), %rax movq %rax, 80(%rsp) movl %r13d, 24(%rsp) leaq 80(%rsp), %rax movq %rax, 128(%rsp) leaq 24(%rsp), %rax movq %rax, 136(%rsp) leaq 48(%rsp), %rdi leaq 32(%rsp), %rsi leaq 72(%rsp), %rdx leaq 64(%rsp), %rcx callq __hipPopCallConfiguration movq 48(%rsp), %rsi movl 56(%rsp), %edx movq 32(%rsp), %rcx movl 40(%rsp), %r8d leaq 128(%rsp), %r9 movl $_Z18kernel_transpuestaPdi, %edi pushq 64(%rsp) .cfi_adjust_cfa_offset 8 pushq 80(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $16, %rsp .cfi_adjust_cfa_offset -16 .LBB7_35: movsd 8(%rsp), %xmm0 # 8-byte Reload # xmm0 = mem[0],zero addsd 96(%rsp), %xmm0 # 8-byte Folded Reload movsd %xmm0, 8(%rsp) # 8-byte Spill callq hipDeviceSynchronize movq %rbp, %rdi movl $1, %esi movq %r12, %rdx movl $1, %ecx xorl %r8d, %r8d xorl %r9d, %r9d callq __hipPushCallConfiguration testl %eax, %eax jne .LBB7_37 # %bb.36: movq 16(%rsp), %rax movq 104(%rsp), %rcx movq 88(%rsp), %rdx movq %rax, 80(%rsp) movq %rcx, 72(%rsp) movq %rdx, 64(%rsp) movl %r13d, 112(%rsp) leaq 80(%rsp), %rax movq %rax, 128(%rsp) leaq 72(%rsp), %rax movq %rax, 136(%rsp) leaq 64(%rsp), %rax movq %rax, 144(%rsp) leaq 112(%rsp), %rax movq %rax, 152(%rsp) leaq 48(%rsp), %rdi leaq 32(%rsp), %rsi leaq 24(%rsp), %rdx leaq 120(%rsp), %rcx callq __hipPopCallConfiguration movq 48(%rsp), %rsi movl 56(%rsp), %edx movq 32(%rsp), %rcx movl 40(%rsp), %r8d leaq 128(%rsp), %r9 movl $_Z22kernel_mult_sum_matrizPdS_S_j, %edi pushq 120(%rsp) .cfi_adjust_cfa_offset 8 pushq 32(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $16, %rsp .cfi_adjust_cfa_offset -16 .LBB7_37: callq hipDeviceSynchronize leaq 128(%rsp), %rdi xorl %esi, %esi callq gettimeofday cvtsi2sdq 128(%rsp), %xmm1 xorps %xmm0, %xmm0 cvtsi2sdq 136(%rsp), %xmm0 divsd .LCPI7_0(%rip), %xmm0 addsd %xmm1, %xmm0 subsd 8(%rsp), %xmm0 # 8-byte Folded Reload movl $.L.str.4, %edi movb $1, %al callq printf callq hipGetLastError movl $.L.str.3, %edi movl %eax, %esi xorl %eax, %eax callq printf movq 88(%rsp), %rsi movq %r15, %rdi movq 208(%rsp), %rdx # 8-byte Reload movl $2, %ecx callq hipMemcpy movq 16(%rsp), %rdi callq hipFree movq 104(%rsp), %rdi callq hipFree movq 88(%rsp), %rdi callq hipFree movq %rbx, %rdi callq free movq %r14, %rdi callq free movq %r15, %rdi callq free .LBB7_38: xorl %eax, %eax addq $264, %rsp # imm = 0x108 .cfi_def_cfa_offset 56 popq %rbx .cfi_def_cfa_offset 48 popq %r12 .cfi_def_cfa_offset 40 popq %r13 .cfi_def_cfa_offset 32 popq %r14 .cfi_def_cfa_offset 24 popq %r15 .cfi_def_cfa_offset 16 popq %rbp .cfi_def_cfa_offset 8 retq .Lfunc_end7: .size main, .Lfunc_end7-main .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_ctor .type __hip_module_ctor,@function __hip_module_ctor: # @__hip_module_ctor .cfi_startproc # %bb.0: pushq %rbx .cfi_def_cfa_offset 16 subq $32, %rsp .cfi_def_cfa_offset 48 .cfi_offset %rbx, -16 cmpq $0, __hip_gpubin_handle(%rip) jne .LBB8_2 # %bb.1: movl $__hip_fatbin_wrapper, %edi callq __hipRegisterFatBinary movq %rax, __hip_gpubin_handle(%rip) .LBB8_2: movq __hip_gpubin_handle(%rip), %rbx xorps %xmm0, %xmm0 movups %xmm0, 16(%rsp) movups %xmm0, (%rsp) movl $_Z27ecuacion_kernel_outplace_p1PdS_S_S_j, %esi movl $.L__unnamed_1, %edx movl $.L__unnamed_1, %ecx movq %rbx, %rdi movl $-1, %r8d xorl %r9d, %r9d callq __hipRegisterFunction xorps %xmm0, %xmm0 movups %xmm0, 16(%rsp) movups %xmm0, (%rsp) movl $_Z27ecuacion_kernel_outplace_p2PdS_S_S_S_j, %esi movl $.L__unnamed_2, %edx movl $.L__unnamed_2, %ecx movq %rbx, %rdi movl $-1, %r8d xorl %r9d, %r9d callq __hipRegisterFunction xorps %xmm0, %xmm0 movups %xmm0, 16(%rsp) movups %xmm0, (%rsp) movl $_Z28ecuacion_kernel_inplace_sumaPdS_S_j, %esi movl $.L__unnamed_3, %edx movl $.L__unnamed_3, %ecx movq %rbx, %rdi movl $-1, %r8d xorl %r9d, %r9d callq __hipRegisterFunction xorps %xmm0, %xmm0 movups %xmm0, 16(%rsp) movups %xmm0, (%rsp) movl $_Z17kernel_sum_MatrizPdS_S_j, %esi movl $.L__unnamed_4, %edx movl $.L__unnamed_4, %ecx movq %rbx, %rdi movl $-1, %r8d xorl %r9d, %r9d callq __hipRegisterFunction xorps %xmm0, %xmm0 movups %xmm0, 16(%rsp) movups %xmm0, (%rsp) movl $_Z18kernel_transpuestaPdi, %esi movl $.L__unnamed_5, %edx movl $.L__unnamed_5, %ecx movq %rbx, %rdi movl $-1, %r8d xorl %r9d, %r9d callq __hipRegisterFunction xorps %xmm0, %xmm0 movups %xmm0, 16(%rsp) movups %xmm0, (%rsp) movl $_Z22kernel_mult_sum_matrizPdS_S_j, %esi movl $.L__unnamed_6, %edx movl $.L__unnamed_6, %ecx movq %rbx, %rdi movl $-1, %r8d xorl %r9d, %r9d callq __hipRegisterFunction movl $__hip_module_dtor, %edi addq $32, %rsp .cfi_def_cfa_offset 16 popq %rbx .cfi_def_cfa_offset 8 jmp atexit # TAILCALL .Lfunc_end8: .size __hip_module_ctor, .Lfunc_end8-__hip_module_ctor .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_dtor .type __hip_module_dtor,@function __hip_module_dtor: # @__hip_module_dtor .cfi_startproc # %bb.0: movq __hip_gpubin_handle(%rip), %rdi testq %rdi, %rdi je .LBB9_2 # %bb.1: pushq %rax .cfi_def_cfa_offset 16 callq __hipUnregisterFatBinary movq $0, __hip_gpubin_handle(%rip) addq $8, %rsp .cfi_def_cfa_offset 8 .LBB9_2: retq .Lfunc_end9: .size __hip_module_dtor, .Lfunc_end9-__hip_module_dtor .cfi_endproc # -- End function .type _Z27ecuacion_kernel_outplace_p1PdS_S_S_j,@object # @_Z27ecuacion_kernel_outplace_p1PdS_S_S_j .section .rodata,"a",@progbits .globl _Z27ecuacion_kernel_outplace_p1PdS_S_S_j .p2align 3, 0x0 _Z27ecuacion_kernel_outplace_p1PdS_S_S_j: .quad _Z42__device_stub__ecuacion_kernel_outplace_p1PdS_S_S_j .size _Z27ecuacion_kernel_outplace_p1PdS_S_S_j, 8 .type _Z27ecuacion_kernel_outplace_p2PdS_S_S_S_j,@object # @_Z27ecuacion_kernel_outplace_p2PdS_S_S_S_j .globl _Z27ecuacion_kernel_outplace_p2PdS_S_S_S_j .p2align 3, 0x0 _Z27ecuacion_kernel_outplace_p2PdS_S_S_S_j: .quad _Z42__device_stub__ecuacion_kernel_outplace_p2PdS_S_S_S_j .size _Z27ecuacion_kernel_outplace_p2PdS_S_S_S_j, 8 .type _Z28ecuacion_kernel_inplace_sumaPdS_S_j,@object # @_Z28ecuacion_kernel_inplace_sumaPdS_S_j .globl _Z28ecuacion_kernel_inplace_sumaPdS_S_j .p2align 3, 0x0 _Z28ecuacion_kernel_inplace_sumaPdS_S_j: .quad _Z43__device_stub__ecuacion_kernel_inplace_sumaPdS_S_j .size _Z28ecuacion_kernel_inplace_sumaPdS_S_j, 8 .type _Z17kernel_sum_MatrizPdS_S_j,@object # @_Z17kernel_sum_MatrizPdS_S_j .globl _Z17kernel_sum_MatrizPdS_S_j .p2align 3, 0x0 _Z17kernel_sum_MatrizPdS_S_j: .quad _Z32__device_stub__kernel_sum_MatrizPdS_S_j .size _Z17kernel_sum_MatrizPdS_S_j, 8 .type _Z18kernel_transpuestaPdi,@object # @_Z18kernel_transpuestaPdi .globl _Z18kernel_transpuestaPdi .p2align 3, 0x0 _Z18kernel_transpuestaPdi: .quad _Z33__device_stub__kernel_transpuestaPdi .size _Z18kernel_transpuestaPdi, 8 .type _Z22kernel_mult_sum_matrizPdS_S_j,@object # @_Z22kernel_mult_sum_matrizPdS_S_j .globl _Z22kernel_mult_sum_matrizPdS_S_j .p2align 3, 0x0 _Z22kernel_mult_sum_matrizPdS_S_j: .quad _Z37__device_stub__kernel_mult_sum_matrizPdS_S_j .size _Z22kernel_mult_sum_matrizPdS_S_j, 8 .type .L.str.1,@object # @.str.1 .section .rodata.str1.1,"aMS",@progbits,1 .L.str.1: .asciz "Tiempo para la ecuacion CPU: %f\n\n" .size .L.str.1, 34 .type .L.str.2,@object # @.str.2 .L.str.2: .asciz "Tiempo para la ecuacion out-place GPU: %f\n" .size .L.str.2, 43 .type .L.str.3,@object # @.str.3 .L.str.3: .asciz "error: %d\n\n" .size .L.str.3, 12 .type .L.str.4,@object # @.str.4 .L.str.4: .asciz "Tiempo para la ecuacion in-place GPU: %f\n" .size .L.str.4, 42 .type .L__unnamed_1,@object # @0 .L__unnamed_1: .asciz "_Z27ecuacion_kernel_outplace_p1PdS_S_S_j" .size .L__unnamed_1, 41 .type .L__unnamed_2,@object # @1 .L__unnamed_2: .asciz "_Z27ecuacion_kernel_outplace_p2PdS_S_S_S_j" .size .L__unnamed_2, 43 .type .L__unnamed_3,@object # @2 .L__unnamed_3: .asciz "_Z28ecuacion_kernel_inplace_sumaPdS_S_j" .size .L__unnamed_3, 40 .type .L__unnamed_4,@object # @3 .L__unnamed_4: .asciz "_Z17kernel_sum_MatrizPdS_S_j" .size .L__unnamed_4, 29 .type .L__unnamed_5,@object # @4 .L__unnamed_5: .asciz "_Z18kernel_transpuestaPdi" .size .L__unnamed_5, 26 .type .L__unnamed_6,@object # @5 .L__unnamed_6: .asciz "_Z22kernel_mult_sum_matrizPdS_S_j" .size .L__unnamed_6, 34 .type __hip_fatbin_wrapper,@object # @__hip_fatbin_wrapper .section .hipFatBinSegment,"a",@progbits .p2align 3, 0x0 __hip_fatbin_wrapper: .long 1212764230 # 0x48495046 .long 1 # 0x1 .quad __hip_fatbin .quad 0 .size __hip_fatbin_wrapper, 24 .type __hip_gpubin_handle,@object # @__hip_gpubin_handle .local __hip_gpubin_handle .comm __hip_gpubin_handle,8,8 .section .init_array,"aw",@init_array .p2align 3, 0x0 .quad __hip_module_ctor .type __hip_cuid_,@object # @__hip_cuid_ .bss .globl __hip_cuid_ __hip_cuid_: .byte 0 # 0x0 .size __hip_cuid_, 1 .type .Lstr,@object # @str .section .rodata.str1.1,"aMS",@progbits,1 .Lstr: .asciz "Falta argumento: N, CUDABLK" .size .Lstr, 28 .section ".linker-options","e",@llvm_linker_options .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym _Z42__device_stub__ecuacion_kernel_outplace_p1PdS_S_S_j .addrsig_sym _Z42__device_stub__ecuacion_kernel_outplace_p2PdS_S_S_S_j .addrsig_sym _Z43__device_stub__ecuacion_kernel_inplace_sumaPdS_S_j .addrsig_sym _Z32__device_stub__kernel_sum_MatrizPdS_S_j .addrsig_sym _Z33__device_stub__kernel_transpuestaPdi .addrsig_sym _Z37__device_stub__kernel_mult_sum_matrizPdS_S_j .addrsig_sym __hip_module_ctor .addrsig_sym __hip_module_dtor .addrsig_sym _Z27ecuacion_kernel_outplace_p1PdS_S_S_j .addrsig_sym _Z27ecuacion_kernel_outplace_p2PdS_S_S_S_j .addrsig_sym _Z28ecuacion_kernel_inplace_sumaPdS_S_j .addrsig_sym _Z17kernel_sum_MatrizPdS_S_j .addrsig_sym _Z18kernel_transpuestaPdi .addrsig_sym _Z22kernel_mult_sum_matrizPdS_S_j .addrsig_sym __hip_fatbin .addrsig_sym __hip_fatbin_wrapper .addrsig_sym __hip_cuid_
You are a helpful coding assistant specialized in converting CUDA source code to CUDA device assembly.
#include <stdio.h> #define DIM 4 #define NUM_ELEMS DIM*DIM __global__ void transpose(int *a, int *b) { int row = blockIdx.x * DIM/2 + threadIdx.x; int col = blockIdx.y * DIM/2 + threadIdx.y; int newIndex = row * DIM + col; int oldIndex = col * DIM + row; b[newIndex] = a[oldIndex]; } int main() { //device memory int *device1, *device2; //host memory int host[NUM_ELEMS]; int output[NUM_ELEMS]; size_t numBytes = NUM_ELEMS * sizeof(int); int i = 0; //loop counter //Load host1 and host2 with values. for (i = 0; i < NUM_ELEMS; i++) { host[i] = i+1; } //Allocate memory for device vars. cudaMalloc((void **)&device1, numBytes); cudaMalloc((void **)&device2, numBytes); //Transfer values from host to device. cudaMemcpy(device1, &host, numBytes, cudaMemcpyHostToDevice); //Launch transpose kernel on GPU with given parameters. dim3 grid(DIM/2, DIM/2); //# of thread blocks dim3 block(DIM/2, DIM/2); //# of threads per thread block transpose<<<grid,block>>>(device1, device2); //Get result from device to host. cudaMemcpy(&output, device2, numBytes, cudaMemcpyDeviceToHost); //Print out values. printf("["); for (i = 0; i < NUM_ELEMS; i++) { printf("%d ", output[i]); } printf("]\n"); //Free all variables. cudaFree(device1); cudaFree(device2); return 0; }
code for sm_80 Function : _Z9transposePiS_ .headerflags @"EF_CUDA_TEXMODE_UNIFIED EF_CUDA_64BIT_ADDRESS EF_CUDA_SM80 EF_CUDA_VIRTUAL_SM(EF_CUDA_SM80)" /*0000*/ IMAD.MOV.U32 R1, RZ, RZ, c[0x0][0x28] ; /* 0x00000a00ff017624 */ /* 0x000fe400078e00ff */ /*0010*/ S2UR UR4, SR_CTAID.X ; /* 0x00000000000479c3 */ /* 0x000e220000002500 */ /*0020*/ S2R R0, SR_TID.X ; /* 0x0000000000007919 */ /* 0x000e620000002100 */ /*0030*/ IMAD.MOV.U32 R4, RZ, RZ, 0x4 ; /* 0x00000004ff047424 */ /* 0x000fc600078e00ff */ /*0040*/ S2R R5, SR_TID.Y ; /* 0x0000000000057919 */ /* 0x000ea60000002200 */ /*0050*/ S2UR UR5, SR_CTAID.Y ; /* 0x00000000000579c3 */ /* 0x000ee20000002600 */ /*0060*/ USHF.L.U32 UR4, UR4, 0x1, URZ ; /* 0x0000000104047899 */ /* 0x001fc8000800063f */ /*0070*/ ULOP3.LUT UR4, UR4, 0x7ffffffe, URZ, 0xc0, !UPT ; /* 0x7ffffffe04047892 */ /* 0x000fe4000f8ec03f */ /*0080*/ USHF.L.U32 UR5, UR5, 0x1, URZ ; /* 0x0000000105057899 */ /* 0x008fc8000800063f */ /*0090*/ IADD3 R0, R0, UR4, RZ ; /* 0x0000000400007c10 */ /* 0x002fe2000fffe0ff */ /*00a0*/ ULOP3.LUT UR5, UR5, 0x7ffffffe, URZ, 0xc0, !UPT ; /* 0x7ffffffe05057892 */ /* 0x000fcc000f8ec03f */ /*00b0*/ IADD3 R5, R5, UR5, RZ ; /* 0x0000000505057c10 */ /* 0x004fe2000fffe0ff */ /*00c0*/ ULDC.64 UR4, c[0x0][0x118] ; /* 0x0000460000047ab9 */ /* 0x000fc60000000a00 */ /*00d0*/ LEA R2, R5, R0, 0x2 ; /* 0x0000000005027211 */ /* 0x000fca00078e10ff */ /*00e0*/ IMAD.WIDE R2, R2, R4, c[0x0][0x160] ; /* 0x0000580002027625 */ /* 0x000fcc00078e0204 */ /*00f0*/ LDG.E R3, [R2.64] ; /* 0x0000000402037981 */ /* 0x000ea2000c1e1900 */ /*0100*/ LEA R5, R0, R5, 0x2 ; /* 0x0000000500057211 */ /* 0x000fca00078e10ff */ /*0110*/ IMAD.WIDE R4, R5, R4, c[0x0][0x168] ; /* 0x00005a0005047625 */ /* 0x000fca00078e0204 */ /*0120*/ STG.E [R4.64], R3 ; /* 0x0000000304007986 */ /* 0x004fe2000c101904 */ /*0130*/ EXIT ; /* 0x000000000000794d */ /* 0x000fea0003800000 */ /*0140*/ BRA 0x140; /* 0xfffffff000007947 */ /* 0x000fc0000383ffff */ /*0150*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0160*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0170*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0180*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0190*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*01a0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*01b0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*01c0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*01d0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*01e0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*01f0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ ..........
You are a helpful coding assistant specialized in converting CUDA source code to CUDA host assembly.
#include <stdio.h> #define DIM 4 #define NUM_ELEMS DIM*DIM __global__ void transpose(int *a, int *b) { int row = blockIdx.x * DIM/2 + threadIdx.x; int col = blockIdx.y * DIM/2 + threadIdx.y; int newIndex = row * DIM + col; int oldIndex = col * DIM + row; b[newIndex] = a[oldIndex]; } int main() { //device memory int *device1, *device2; //host memory int host[NUM_ELEMS]; int output[NUM_ELEMS]; size_t numBytes = NUM_ELEMS * sizeof(int); int i = 0; //loop counter //Load host1 and host2 with values. for (i = 0; i < NUM_ELEMS; i++) { host[i] = i+1; } //Allocate memory for device vars. cudaMalloc((void **)&device1, numBytes); cudaMalloc((void **)&device2, numBytes); //Transfer values from host to device. cudaMemcpy(device1, &host, numBytes, cudaMemcpyHostToDevice); //Launch transpose kernel on GPU with given parameters. dim3 grid(DIM/2, DIM/2); //# of thread blocks dim3 block(DIM/2, DIM/2); //# of threads per thread block transpose<<<grid,block>>>(device1, device2); //Get result from device to host. cudaMemcpy(&output, device2, numBytes, cudaMemcpyDeviceToHost); //Print out values. printf("["); for (i = 0; i < NUM_ELEMS; i++) { printf("%d ", output[i]); } printf("]\n"); //Free all variables. cudaFree(device1); cudaFree(device2); return 0; }
.file "tmpxft_0010dbbf_00000000-6_transpose.cudafe1.cpp" .text #APP #NO_APP .type _ZL26__cudaUnregisterBinaryUtilv, @function _ZL26__cudaUnregisterBinaryUtilv: .LFB2060: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaUnregisterFatBinary@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2060: .size _ZL26__cudaUnregisterBinaryUtilv, .-_ZL26__cudaUnregisterBinaryUtilv .globl _Z30__device_stub__Z9transposePiS_PiS_ .type _Z30__device_stub__Z9transposePiS_PiS_, @function _Z30__device_stub__Z9transposePiS_PiS_: .LFB2082: .cfi_startproc endbr64 subq $120, %rsp .cfi_def_cfa_offset 128 movq %rdi, 8(%rsp) movq %rsi, (%rsp) movq %fs:40, %rax movq %rax, 104(%rsp) xorl %eax, %eax leaq 8(%rsp), %rax movq %rax, 80(%rsp) movq %rsp, %rax movq %rax, 88(%rsp) movl $1, 32(%rsp) movl $1, 36(%rsp) movl $1, 40(%rsp) movl $1, 44(%rsp) movl $1, 48(%rsp) movl $1, 52(%rsp) leaq 24(%rsp), %rcx leaq 16(%rsp), %rdx leaq 44(%rsp), %rsi leaq 32(%rsp), %rdi call __cudaPopCallConfiguration@PLT testl %eax, %eax je .L7 .L3: movq 104(%rsp), %rax subq %fs:40, %rax jne .L8 addq $120, %rsp .cfi_remember_state .cfi_def_cfa_offset 8 ret .L7: .cfi_restore_state pushq 24(%rsp) .cfi_def_cfa_offset 136 pushq 24(%rsp) .cfi_def_cfa_offset 144 leaq 96(%rsp), %r9 movq 60(%rsp), %rcx movl 68(%rsp), %r8d movq 48(%rsp), %rsi movl 56(%rsp), %edx leaq _Z9transposePiS_(%rip), %rdi call cudaLaunchKernel@PLT addq $16, %rsp .cfi_def_cfa_offset 128 jmp .L3 .L8: call __stack_chk_fail@PLT .cfi_endproc .LFE2082: .size _Z30__device_stub__Z9transposePiS_PiS_, .-_Z30__device_stub__Z9transposePiS_PiS_ .globl _Z9transposePiS_ .type _Z9transposePiS_, @function _Z9transposePiS_: .LFB2083: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 call _Z30__device_stub__Z9transposePiS_PiS_ addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2083: .size _Z9transposePiS_, .-_Z9transposePiS_ .section .rodata.str1.1,"aMS",@progbits,1 .LC0: .string "[" .LC1: .string "%d " .LC2: .string "]\n" .text .globl main .type main, @function main: .LFB2057: .cfi_startproc endbr64 pushq %r12 .cfi_def_cfa_offset 16 .cfi_offset 12, -16 pushq %rbp .cfi_def_cfa_offset 24 .cfi_offset 6, -24 pushq %rbx .cfi_def_cfa_offset 32 .cfi_offset 3, -32 subq $192, %rsp .cfi_def_cfa_offset 224 movq %fs:40, %rax movq %rax, 184(%rsp) xorl %eax, %eax movl $1, %eax .L12: movl %eax, 44(%rsp,%rax,4) addq $1, %rax cmpq $17, %rax jne .L12 leaq 8(%rsp), %rdi movl $64, %esi call cudaMalloc@PLT leaq 16(%rsp), %rdi movl $64, %esi call cudaMalloc@PLT leaq 48(%rsp), %rsi movl $1, %ecx movl $64, %edx movq 8(%rsp), %rdi call cudaMemcpy@PLT movl $2, 24(%rsp) movl $2, 28(%rsp) movl $1, 32(%rsp) movl $2, 36(%rsp) movl $2, 40(%rsp) movl $1, 44(%rsp) movl $0, %r9d movl $0, %r8d movq 36(%rsp), %rdx movl $1, %ecx movq 24(%rsp), %rdi movl $1, %esi call __cudaPushCallConfiguration@PLT testl %eax, %eax je .L19 .L13: leaq 112(%rsp), %rbx movl $2, %ecx movl $64, %edx movq 16(%rsp), %rsi movq %rbx, %rdi call cudaMemcpy@PLT leaq .LC0(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT leaq 176(%rsp), %r12 leaq .LC1(%rip), %rbp .L14: movl (%rbx), %edx movq %rbp, %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT addq $4, %rbx cmpq %r12, %rbx jne .L14 leaq .LC2(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT movq 8(%rsp), %rdi call cudaFree@PLT movq 16(%rsp), %rdi call cudaFree@PLT movq 184(%rsp), %rax subq %fs:40, %rax jne .L20 movl $0, %eax addq $192, %rsp .cfi_remember_state .cfi_def_cfa_offset 32 popq %rbx .cfi_def_cfa_offset 24 popq %rbp .cfi_def_cfa_offset 16 popq %r12 .cfi_def_cfa_offset 8 ret .L19: .cfi_restore_state movq 16(%rsp), %rsi movq 8(%rsp), %rdi call _Z30__device_stub__Z9transposePiS_PiS_ jmp .L13 .L20: call __stack_chk_fail@PLT .cfi_endproc .LFE2057: .size main, .-main .section .rodata.str1.1 .LC3: .string "_Z9transposePiS_" .text .type _ZL24__sti____cudaRegisterAllv, @function _ZL24__sti____cudaRegisterAllv: .LFB2085: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 leaq _ZL15__fatDeviceText(%rip), %rdi call __cudaRegisterFatBinary@PLT movq %rax, %rdi movq %rax, _ZL20__cudaFatCubinHandle(%rip) pushq $0 .cfi_def_cfa_offset 24 pushq $0 .cfi_def_cfa_offset 32 pushq $0 .cfi_def_cfa_offset 40 pushq $0 .cfi_def_cfa_offset 48 movl $0, %r9d movl $-1, %r8d leaq .LC3(%rip), %rdx movq %rdx, %rcx leaq _Z9transposePiS_(%rip), %rsi call __cudaRegisterFunction@PLT addq $32, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaRegisterFatBinaryEnd@PLT leaq _ZL26__cudaUnregisterBinaryUtilv(%rip), %rdi call atexit@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2085: .size _ZL24__sti____cudaRegisterAllv, .-_ZL24__sti____cudaRegisterAllv .section .init_array,"aw" .align 8 .quad _ZL24__sti____cudaRegisterAllv .section .nvFatBinSegment,"aw" .align 8 .type _ZL15__fatDeviceText, @object .size _ZL15__fatDeviceText, 24 _ZL15__fatDeviceText: .long 1180844977 .long 1 .quad fatbinData .quad 0 .local _ZL20__cudaFatCubinHandle .comm _ZL20__cudaFatCubinHandle,8,8 .ident "GCC: (Ubuntu 13.3.0-6ubuntu2~24.04) 13.3.0" .section .note.GNU-stack,"",@progbits .section .note.gnu.property,"a" .align 8 .long 1f - 0f .long 4f - 1f .long 5 0: .string "GNU" 1: .align 8 .long 0xc0000002 .long 3f - 2f 2: .long 0x3 3: .align 8 4:
You are a helpful coding assistant specialized in converting CUDA source code to HIP/AMD source code.
#include <stdio.h> #define DIM 4 #define NUM_ELEMS DIM*DIM __global__ void transpose(int *a, int *b) { int row = blockIdx.x * DIM/2 + threadIdx.x; int col = blockIdx.y * DIM/2 + threadIdx.y; int newIndex = row * DIM + col; int oldIndex = col * DIM + row; b[newIndex] = a[oldIndex]; } int main() { //device memory int *device1, *device2; //host memory int host[NUM_ELEMS]; int output[NUM_ELEMS]; size_t numBytes = NUM_ELEMS * sizeof(int); int i = 0; //loop counter //Load host1 and host2 with values. for (i = 0; i < NUM_ELEMS; i++) { host[i] = i+1; } //Allocate memory for device vars. cudaMalloc((void **)&device1, numBytes); cudaMalloc((void **)&device2, numBytes); //Transfer values from host to device. cudaMemcpy(device1, &host, numBytes, cudaMemcpyHostToDevice); //Launch transpose kernel on GPU with given parameters. dim3 grid(DIM/2, DIM/2); //# of thread blocks dim3 block(DIM/2, DIM/2); //# of threads per thread block transpose<<<grid,block>>>(device1, device2); //Get result from device to host. cudaMemcpy(&output, device2, numBytes, cudaMemcpyDeviceToHost); //Print out values. printf("["); for (i = 0; i < NUM_ELEMS; i++) { printf("%d ", output[i]); } printf("]\n"); //Free all variables. cudaFree(device1); cudaFree(device2); return 0; }
#include <hip/hip_runtime.h> #include <stdio.h> #define DIM 4 #define NUM_ELEMS DIM*DIM __global__ void transpose(int *a, int *b) { int row = blockIdx.x * DIM/2 + threadIdx.x; int col = blockIdx.y * DIM/2 + threadIdx.y; int newIndex = row * DIM + col; int oldIndex = col * DIM + row; b[newIndex] = a[oldIndex]; } int main() { //device memory int *device1, *device2; //host memory int host[NUM_ELEMS]; int output[NUM_ELEMS]; size_t numBytes = NUM_ELEMS * sizeof(int); int i = 0; //loop counter //Load host1 and host2 with values. for (i = 0; i < NUM_ELEMS; i++) { host[i] = i+1; } //Allocate memory for device vars. hipMalloc((void **)&device1, numBytes); hipMalloc((void **)&device2, numBytes); //Transfer values from host to device. hipMemcpy(device1, &host, numBytes, hipMemcpyHostToDevice); //Launch transpose kernel on GPU with given parameters. dim3 grid(DIM/2, DIM/2); //# of thread blocks dim3 block(DIM/2, DIM/2); //# of threads per thread block transpose<<<grid,block>>>(device1, device2); //Get result from device to host. hipMemcpy(&output, device2, numBytes, hipMemcpyDeviceToHost); //Print out values. printf("["); for (i = 0; i < NUM_ELEMS; i++) { printf("%d ", output[i]); } printf("]\n"); //Free all variables. hipFree(device1); hipFree(device2); return 0; }
You are a helpful coding assistant specialized in converting HIP source code to AMD device assembly.
#include <hip/hip_runtime.h> #include <stdio.h> #define DIM 4 #define NUM_ELEMS DIM*DIM __global__ void transpose(int *a, int *b) { int row = blockIdx.x * DIM/2 + threadIdx.x; int col = blockIdx.y * DIM/2 + threadIdx.y; int newIndex = row * DIM + col; int oldIndex = col * DIM + row; b[newIndex] = a[oldIndex]; } int main() { //device memory int *device1, *device2; //host memory int host[NUM_ELEMS]; int output[NUM_ELEMS]; size_t numBytes = NUM_ELEMS * sizeof(int); int i = 0; //loop counter //Load host1 and host2 with values. for (i = 0; i < NUM_ELEMS; i++) { host[i] = i+1; } //Allocate memory for device vars. hipMalloc((void **)&device1, numBytes); hipMalloc((void **)&device2, numBytes); //Transfer values from host to device. hipMemcpy(device1, &host, numBytes, hipMemcpyHostToDevice); //Launch transpose kernel on GPU with given parameters. dim3 grid(DIM/2, DIM/2); //# of thread blocks dim3 block(DIM/2, DIM/2); //# of threads per thread block transpose<<<grid,block>>>(device1, device2); //Get result from device to host. hipMemcpy(&output, device2, numBytes, hipMemcpyDeviceToHost); //Print out values. printf("["); for (i = 0; i < NUM_ELEMS; i++) { printf("%d ", output[i]); } printf("]\n"); //Free all variables. hipFree(device1); hipFree(device2); return 0; }
.text .amdgcn_target "amdgcn-amd-amdhsa--gfx1100" .protected _Z9transposePiS_ .globl _Z9transposePiS_ .p2align 8 .type _Z9transposePiS_,@function _Z9transposePiS_: v_and_b32_e32 v1, 0x3ff, v0 v_bfe_u32 v0, v0, 10, 10 s_lshl_b32 s2, s14, 1 s_lshl_b32 s3, s15, 1 s_and_b32 s2, s2, 0x7ffffffe s_and_b32 s3, s3, 0x7ffffffe v_add_nc_u32_e32 v2, s2, v1 v_add_nc_u32_e32 v3, s3, v0 s_load_b128 s[0:3], s[0:1], 0x0 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_lshl_add_u32 v0, v3, 2, v2 v_ashrrev_i32_e32 v1, 31, v0 s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1) v_lshlrev_b64 v[0:1], 2, v[0:1] s_waitcnt lgkmcnt(0) v_add_co_u32 v0, vcc_lo, s0, v0 s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1) v_add_co_ci_u32_e32 v1, vcc_lo, s1, v1, vcc_lo global_load_b32 v4, v[0:1], off v_lshl_add_u32 v0, v2, 2, v3 v_ashrrev_i32_e32 v1, 31, v0 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_lshlrev_b64 v[0:1], 2, v[0:1] v_add_co_u32 v0, vcc_lo, s2, v0 s_delay_alu instid0(VALU_DEP_2) v_add_co_ci_u32_e32 v1, vcc_lo, s3, v1, vcc_lo s_waitcnt vmcnt(0) global_store_b32 v[0:1], v4, off s_nop 0 s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) s_endpgm .section .rodata,"a",@progbits .p2align 6, 0x0 .amdhsa_kernel _Z9transposePiS_ .amdhsa_group_segment_fixed_size 0 .amdhsa_private_segment_fixed_size 0 .amdhsa_kernarg_size 16 .amdhsa_user_sgpr_count 14 .amdhsa_user_sgpr_dispatch_ptr 0 .amdhsa_user_sgpr_queue_ptr 0 .amdhsa_user_sgpr_kernarg_segment_ptr 1 .amdhsa_user_sgpr_dispatch_id 0 .amdhsa_user_sgpr_private_segment_size 0 .amdhsa_wavefront_size32 1 .amdhsa_uses_dynamic_stack 0 .amdhsa_enable_private_segment 0 .amdhsa_system_sgpr_workgroup_id_x 1 .amdhsa_system_sgpr_workgroup_id_y 1 .amdhsa_system_sgpr_workgroup_id_z 0 .amdhsa_system_sgpr_workgroup_info 0 .amdhsa_system_vgpr_workitem_id 1 .amdhsa_next_free_vgpr 5 .amdhsa_next_free_sgpr 16 .amdhsa_float_round_mode_32 0 .amdhsa_float_round_mode_16_64 0 .amdhsa_float_denorm_mode_32 3 .amdhsa_float_denorm_mode_16_64 3 .amdhsa_dx10_clamp 1 .amdhsa_ieee_mode 1 .amdhsa_fp16_overflow 0 .amdhsa_workgroup_processor_mode 1 .amdhsa_memory_ordered 1 .amdhsa_forward_progress 0 .amdhsa_shared_vgpr_count 0 .amdhsa_exception_fp_ieee_invalid_op 0 .amdhsa_exception_fp_denorm_src 0 .amdhsa_exception_fp_ieee_div_zero 0 .amdhsa_exception_fp_ieee_overflow 0 .amdhsa_exception_fp_ieee_underflow 0 .amdhsa_exception_fp_ieee_inexact 0 .amdhsa_exception_int_div_zero 0 .end_amdhsa_kernel .text .Lfunc_end0: .size _Z9transposePiS_, .Lfunc_end0-_Z9transposePiS_ .section .AMDGPU.csdata,"",@progbits .text .p2alignl 7, 3214868480 .fill 96, 4, 3214868480 .type __hip_cuid_,@object .section .bss,"aw",@nobits .globl __hip_cuid_ __hip_cuid_: .byte 0 .size __hip_cuid_, 1 .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym __hip_cuid_ .amdgpu_metadata --- amdhsa.kernels: - .args: - .address_space: global .offset: 0 .size: 8 .value_kind: global_buffer - .address_space: global .offset: 8 .size: 8 .value_kind: global_buffer .group_segment_fixed_size: 0 .kernarg_segment_align: 8 .kernarg_segment_size: 16 .language: OpenCL C .language_version: - 2 - 0 .max_flat_workgroup_size: 1024 .name: _Z9transposePiS_ .private_segment_fixed_size: 0 .sgpr_count: 18 .sgpr_spill_count: 0 .symbol: _Z9transposePiS_.kd .uniform_work_group_size: 1 .uses_dynamic_stack: false .vgpr_count: 5 .vgpr_spill_count: 0 .wavefront_size: 32 .workgroup_processor_mode: 1 amdhsa.target: amdgcn-amd-amdhsa--gfx1100 amdhsa.version: - 1 - 2 ... .end_amdgpu_metadata
You are a helpful coding assistant specialized in converting HIP source code to AMD host assembly.
#include <hip/hip_runtime.h> #include <stdio.h> #define DIM 4 #define NUM_ELEMS DIM*DIM __global__ void transpose(int *a, int *b) { int row = blockIdx.x * DIM/2 + threadIdx.x; int col = blockIdx.y * DIM/2 + threadIdx.y; int newIndex = row * DIM + col; int oldIndex = col * DIM + row; b[newIndex] = a[oldIndex]; } int main() { //device memory int *device1, *device2; //host memory int host[NUM_ELEMS]; int output[NUM_ELEMS]; size_t numBytes = NUM_ELEMS * sizeof(int); int i = 0; //loop counter //Load host1 and host2 with values. for (i = 0; i < NUM_ELEMS; i++) { host[i] = i+1; } //Allocate memory for device vars. hipMalloc((void **)&device1, numBytes); hipMalloc((void **)&device2, numBytes); //Transfer values from host to device. hipMemcpy(device1, &host, numBytes, hipMemcpyHostToDevice); //Launch transpose kernel on GPU with given parameters. dim3 grid(DIM/2, DIM/2); //# of thread blocks dim3 block(DIM/2, DIM/2); //# of threads per thread block transpose<<<grid,block>>>(device1, device2); //Get result from device to host. hipMemcpy(&output, device2, numBytes, hipMemcpyDeviceToHost); //Print out values. printf("["); for (i = 0; i < NUM_ELEMS; i++) { printf("%d ", output[i]); } printf("]\n"); //Free all variables. hipFree(device1); hipFree(device2); return 0; }
.text .file "transpose.hip" .globl _Z24__device_stub__transposePiS_ # -- Begin function _Z24__device_stub__transposePiS_ .p2align 4, 0x90 .type _Z24__device_stub__transposePiS_,@function _Z24__device_stub__transposePiS_: # @_Z24__device_stub__transposePiS_ .cfi_startproc # %bb.0: subq $88, %rsp .cfi_def_cfa_offset 96 movq %rdi, 56(%rsp) movq %rsi, 48(%rsp) leaq 56(%rsp), %rax movq %rax, 64(%rsp) leaq 48(%rsp), %rax movq %rax, 72(%rsp) leaq 32(%rsp), %rdi leaq 16(%rsp), %rsi leaq 8(%rsp), %rdx movq %rsp, %rcx callq __hipPopCallConfiguration movq 32(%rsp), %rsi movl 40(%rsp), %edx movq 16(%rsp), %rcx movl 24(%rsp), %r8d leaq 64(%rsp), %r9 movl $_Z9transposePiS_, %edi pushq (%rsp) .cfi_adjust_cfa_offset 8 pushq 16(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $104, %rsp .cfi_adjust_cfa_offset -104 retq .Lfunc_end0: .size _Z24__device_stub__transposePiS_, .Lfunc_end0-_Z24__device_stub__transposePiS_ .cfi_endproc # -- End function .globl main # -- Begin function main .p2align 4, 0x90 .type main,@function main: # @main .cfi_startproc # %bb.0: pushq %rbx .cfi_def_cfa_offset 16 subq $208, %rsp .cfi_def_cfa_offset 224 .cfi_offset %rbx, -16 xorl %eax, %eax .p2align 4, 0x90 .LBB1_1: # =>This Inner Loop Header: Depth=1 leaq 1(%rax), %rcx movl %ecx, 144(%rsp,%rax,4) movq %rcx, %rax cmpq $16, %rcx jne .LBB1_1 # %bb.2: leaq 8(%rsp), %rdi movl $64, %esi callq hipMalloc movq %rsp, %rdi movl $64, %esi callq hipMalloc movq 8(%rsp), %rdi leaq 144(%rsp), %rsi movl $64, %edx movl $1, %ecx callq hipMemcpy movabsq $8589934594, %rdi # imm = 0x200000002 movl $1, %esi movq %rdi, %rdx movl $1, %ecx xorl %r8d, %r8d xorl %r9d, %r9d callq __hipPushCallConfiguration testl %eax, %eax jne .LBB1_4 # %bb.3: movq 8(%rsp), %rax movq (%rsp), %rcx movq %rax, 72(%rsp) movq %rcx, 64(%rsp) leaq 72(%rsp), %rax movq %rax, 80(%rsp) leaq 64(%rsp), %rax movq %rax, 88(%rsp) leaq 48(%rsp), %rdi leaq 32(%rsp), %rsi leaq 24(%rsp), %rdx leaq 16(%rsp), %rcx callq __hipPopCallConfiguration movq 48(%rsp), %rsi movl 56(%rsp), %edx movq 32(%rsp), %rcx movl 40(%rsp), %r8d leaq 80(%rsp), %r9 movl $_Z9transposePiS_, %edi pushq 16(%rsp) .cfi_adjust_cfa_offset 8 pushq 32(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $16, %rsp .cfi_adjust_cfa_offset -16 .LBB1_4: movq (%rsp), %rsi leaq 80(%rsp), %rdi movl $64, %edx movl $2, %ecx callq hipMemcpy movl $91, %edi callq putchar@PLT xorl %ebx, %ebx .p2align 4, 0x90 .LBB1_5: # =>This Inner Loop Header: Depth=1 movl 80(%rsp,%rbx,4), %esi movl $.L.str.1, %edi xorl %eax, %eax callq printf incq %rbx cmpq $16, %rbx jne .LBB1_5 # %bb.6: movl $.Lstr, %edi callq puts@PLT movq 8(%rsp), %rdi callq hipFree movq (%rsp), %rdi callq hipFree xorl %eax, %eax addq $208, %rsp .cfi_def_cfa_offset 16 popq %rbx .cfi_def_cfa_offset 8 retq .Lfunc_end1: .size main, .Lfunc_end1-main .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_ctor .type __hip_module_ctor,@function __hip_module_ctor: # @__hip_module_ctor .cfi_startproc # %bb.0: subq $40, %rsp .cfi_def_cfa_offset 48 cmpq $0, __hip_gpubin_handle(%rip) jne .LBB2_2 # %bb.1: movl $__hip_fatbin_wrapper, %edi callq __hipRegisterFatBinary movq %rax, __hip_gpubin_handle(%rip) .LBB2_2: movq __hip_gpubin_handle(%rip), %rdi xorps %xmm0, %xmm0 movups %xmm0, 16(%rsp) movups %xmm0, (%rsp) movl $_Z9transposePiS_, %esi movl $.L__unnamed_1, %edx movl $.L__unnamed_1, %ecx movl $-1, %r8d xorl %r9d, %r9d callq __hipRegisterFunction movl $__hip_module_dtor, %edi addq $40, %rsp .cfi_def_cfa_offset 8 jmp atexit # TAILCALL .Lfunc_end2: .size __hip_module_ctor, .Lfunc_end2-__hip_module_ctor .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_dtor .type __hip_module_dtor,@function __hip_module_dtor: # @__hip_module_dtor .cfi_startproc # %bb.0: movq __hip_gpubin_handle(%rip), %rdi testq %rdi, %rdi je .LBB3_2 # %bb.1: pushq %rax .cfi_def_cfa_offset 16 callq __hipUnregisterFatBinary movq $0, __hip_gpubin_handle(%rip) addq $8, %rsp .cfi_def_cfa_offset 8 .LBB3_2: retq .Lfunc_end3: .size __hip_module_dtor, .Lfunc_end3-__hip_module_dtor .cfi_endproc # -- End function .type _Z9transposePiS_,@object # @_Z9transposePiS_ .section .rodata,"a",@progbits .globl _Z9transposePiS_ .p2align 3, 0x0 _Z9transposePiS_: .quad _Z24__device_stub__transposePiS_ .size _Z9transposePiS_, 8 .type .L.str.1,@object # @.str.1 .section .rodata.str1.1,"aMS",@progbits,1 .L.str.1: .asciz "%d " .size .L.str.1, 4 .type .L__unnamed_1,@object # @0 .L__unnamed_1: .asciz "_Z9transposePiS_" .size .L__unnamed_1, 17 .type __hip_fatbin_wrapper,@object # @__hip_fatbin_wrapper .section .hipFatBinSegment,"a",@progbits .p2align 3, 0x0 __hip_fatbin_wrapper: .long 1212764230 # 0x48495046 .long 1 # 0x1 .quad __hip_fatbin .quad 0 .size __hip_fatbin_wrapper, 24 .type __hip_gpubin_handle,@object # @__hip_gpubin_handle .local __hip_gpubin_handle .comm __hip_gpubin_handle,8,8 .section .init_array,"aw",@init_array .p2align 3, 0x0 .quad __hip_module_ctor .type __hip_cuid_,@object # @__hip_cuid_ .bss .globl __hip_cuid_ __hip_cuid_: .byte 0 # 0x0 .size __hip_cuid_, 1 .type .Lstr,@object # @str .section .rodata.str1.1,"aMS",@progbits,1 .Lstr: .asciz "]" .size .Lstr, 2 .section ".linker-options","e",@llvm_linker_options .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym _Z24__device_stub__transposePiS_ .addrsig_sym __hip_module_ctor .addrsig_sym __hip_module_dtor .addrsig_sym _Z9transposePiS_ .addrsig_sym __hip_fatbin .addrsig_sym __hip_fatbin_wrapper .addrsig_sym __hip_cuid_
You are a helpful coding assistant specialized in converting from CUDA device assembly to AMD device assembly.
code for sm_80 Function : _Z9transposePiS_ .headerflags @"EF_CUDA_TEXMODE_UNIFIED EF_CUDA_64BIT_ADDRESS EF_CUDA_SM80 EF_CUDA_VIRTUAL_SM(EF_CUDA_SM80)" /*0000*/ IMAD.MOV.U32 R1, RZ, RZ, c[0x0][0x28] ; /* 0x00000a00ff017624 */ /* 0x000fe400078e00ff */ /*0010*/ S2UR UR4, SR_CTAID.X ; /* 0x00000000000479c3 */ /* 0x000e220000002500 */ /*0020*/ S2R R0, SR_TID.X ; /* 0x0000000000007919 */ /* 0x000e620000002100 */ /*0030*/ IMAD.MOV.U32 R4, RZ, RZ, 0x4 ; /* 0x00000004ff047424 */ /* 0x000fc600078e00ff */ /*0040*/ S2R R5, SR_TID.Y ; /* 0x0000000000057919 */ /* 0x000ea60000002200 */ /*0050*/ S2UR UR5, SR_CTAID.Y ; /* 0x00000000000579c3 */ /* 0x000ee20000002600 */ /*0060*/ USHF.L.U32 UR4, UR4, 0x1, URZ ; /* 0x0000000104047899 */ /* 0x001fc8000800063f */ /*0070*/ ULOP3.LUT UR4, UR4, 0x7ffffffe, URZ, 0xc0, !UPT ; /* 0x7ffffffe04047892 */ /* 0x000fe4000f8ec03f */ /*0080*/ USHF.L.U32 UR5, UR5, 0x1, URZ ; /* 0x0000000105057899 */ /* 0x008fc8000800063f */ /*0090*/ IADD3 R0, R0, UR4, RZ ; /* 0x0000000400007c10 */ /* 0x002fe2000fffe0ff */ /*00a0*/ ULOP3.LUT UR5, UR5, 0x7ffffffe, URZ, 0xc0, !UPT ; /* 0x7ffffffe05057892 */ /* 0x000fcc000f8ec03f */ /*00b0*/ IADD3 R5, R5, UR5, RZ ; /* 0x0000000505057c10 */ /* 0x004fe2000fffe0ff */ /*00c0*/ ULDC.64 UR4, c[0x0][0x118] ; /* 0x0000460000047ab9 */ /* 0x000fc60000000a00 */ /*00d0*/ LEA R2, R5, R0, 0x2 ; /* 0x0000000005027211 */ /* 0x000fca00078e10ff */ /*00e0*/ IMAD.WIDE R2, R2, R4, c[0x0][0x160] ; /* 0x0000580002027625 */ /* 0x000fcc00078e0204 */ /*00f0*/ LDG.E R3, [R2.64] ; /* 0x0000000402037981 */ /* 0x000ea2000c1e1900 */ /*0100*/ LEA R5, R0, R5, 0x2 ; /* 0x0000000500057211 */ /* 0x000fca00078e10ff */ /*0110*/ IMAD.WIDE R4, R5, R4, c[0x0][0x168] ; /* 0x00005a0005047625 */ /* 0x000fca00078e0204 */ /*0120*/ STG.E [R4.64], R3 ; /* 0x0000000304007986 */ /* 0x004fe2000c101904 */ /*0130*/ EXIT ; /* 0x000000000000794d */ /* 0x000fea0003800000 */ /*0140*/ BRA 0x140; /* 0xfffffff000007947 */ /* 0x000fc0000383ffff */ /*0150*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0160*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0170*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0180*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0190*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*01a0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*01b0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*01c0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*01d0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*01e0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*01f0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ ..........
.text .amdgcn_target "amdgcn-amd-amdhsa--gfx1100" .protected _Z9transposePiS_ .globl _Z9transposePiS_ .p2align 8 .type _Z9transposePiS_,@function _Z9transposePiS_: v_and_b32_e32 v1, 0x3ff, v0 v_bfe_u32 v0, v0, 10, 10 s_lshl_b32 s2, s14, 1 s_lshl_b32 s3, s15, 1 s_and_b32 s2, s2, 0x7ffffffe s_and_b32 s3, s3, 0x7ffffffe v_add_nc_u32_e32 v2, s2, v1 v_add_nc_u32_e32 v3, s3, v0 s_load_b128 s[0:3], s[0:1], 0x0 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_lshl_add_u32 v0, v3, 2, v2 v_ashrrev_i32_e32 v1, 31, v0 s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1) v_lshlrev_b64 v[0:1], 2, v[0:1] s_waitcnt lgkmcnt(0) v_add_co_u32 v0, vcc_lo, s0, v0 s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1) v_add_co_ci_u32_e32 v1, vcc_lo, s1, v1, vcc_lo global_load_b32 v4, v[0:1], off v_lshl_add_u32 v0, v2, 2, v3 v_ashrrev_i32_e32 v1, 31, v0 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_lshlrev_b64 v[0:1], 2, v[0:1] v_add_co_u32 v0, vcc_lo, s2, v0 s_delay_alu instid0(VALU_DEP_2) v_add_co_ci_u32_e32 v1, vcc_lo, s3, v1, vcc_lo s_waitcnt vmcnt(0) global_store_b32 v[0:1], v4, off s_nop 0 s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) s_endpgm .section .rodata,"a",@progbits .p2align 6, 0x0 .amdhsa_kernel _Z9transposePiS_ .amdhsa_group_segment_fixed_size 0 .amdhsa_private_segment_fixed_size 0 .amdhsa_kernarg_size 16 .amdhsa_user_sgpr_count 14 .amdhsa_user_sgpr_dispatch_ptr 0 .amdhsa_user_sgpr_queue_ptr 0 .amdhsa_user_sgpr_kernarg_segment_ptr 1 .amdhsa_user_sgpr_dispatch_id 0 .amdhsa_user_sgpr_private_segment_size 0 .amdhsa_wavefront_size32 1 .amdhsa_uses_dynamic_stack 0 .amdhsa_enable_private_segment 0 .amdhsa_system_sgpr_workgroup_id_x 1 .amdhsa_system_sgpr_workgroup_id_y 1 .amdhsa_system_sgpr_workgroup_id_z 0 .amdhsa_system_sgpr_workgroup_info 0 .amdhsa_system_vgpr_workitem_id 1 .amdhsa_next_free_vgpr 5 .amdhsa_next_free_sgpr 16 .amdhsa_float_round_mode_32 0 .amdhsa_float_round_mode_16_64 0 .amdhsa_float_denorm_mode_32 3 .amdhsa_float_denorm_mode_16_64 3 .amdhsa_dx10_clamp 1 .amdhsa_ieee_mode 1 .amdhsa_fp16_overflow 0 .amdhsa_workgroup_processor_mode 1 .amdhsa_memory_ordered 1 .amdhsa_forward_progress 0 .amdhsa_shared_vgpr_count 0 .amdhsa_exception_fp_ieee_invalid_op 0 .amdhsa_exception_fp_denorm_src 0 .amdhsa_exception_fp_ieee_div_zero 0 .amdhsa_exception_fp_ieee_overflow 0 .amdhsa_exception_fp_ieee_underflow 0 .amdhsa_exception_fp_ieee_inexact 0 .amdhsa_exception_int_div_zero 0 .end_amdhsa_kernel .text .Lfunc_end0: .size _Z9transposePiS_, .Lfunc_end0-_Z9transposePiS_ .section .AMDGPU.csdata,"",@progbits .text .p2alignl 7, 3214868480 .fill 96, 4, 3214868480 .type __hip_cuid_,@object .section .bss,"aw",@nobits .globl __hip_cuid_ __hip_cuid_: .byte 0 .size __hip_cuid_, 1 .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym __hip_cuid_ .amdgpu_metadata --- amdhsa.kernels: - .args: - .address_space: global .offset: 0 .size: 8 .value_kind: global_buffer - .address_space: global .offset: 8 .size: 8 .value_kind: global_buffer .group_segment_fixed_size: 0 .kernarg_segment_align: 8 .kernarg_segment_size: 16 .language: OpenCL C .language_version: - 2 - 0 .max_flat_workgroup_size: 1024 .name: _Z9transposePiS_ .private_segment_fixed_size: 0 .sgpr_count: 18 .sgpr_spill_count: 0 .symbol: _Z9transposePiS_.kd .uniform_work_group_size: 1 .uses_dynamic_stack: false .vgpr_count: 5 .vgpr_spill_count: 0 .wavefront_size: 32 .workgroup_processor_mode: 1 amdhsa.target: amdgcn-amd-amdhsa--gfx1100 amdhsa.version: - 1 - 2 ... .end_amdgpu_metadata
You are a helpful coding assistant specialized in converting from CUDA host assembly to AMD host assembly.
.file "tmpxft_0010dbbf_00000000-6_transpose.cudafe1.cpp" .text #APP #NO_APP .type _ZL26__cudaUnregisterBinaryUtilv, @function _ZL26__cudaUnregisterBinaryUtilv: .LFB2060: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaUnregisterFatBinary@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2060: .size _ZL26__cudaUnregisterBinaryUtilv, .-_ZL26__cudaUnregisterBinaryUtilv .globl _Z30__device_stub__Z9transposePiS_PiS_ .type _Z30__device_stub__Z9transposePiS_PiS_, @function _Z30__device_stub__Z9transposePiS_PiS_: .LFB2082: .cfi_startproc endbr64 subq $120, %rsp .cfi_def_cfa_offset 128 movq %rdi, 8(%rsp) movq %rsi, (%rsp) movq %fs:40, %rax movq %rax, 104(%rsp) xorl %eax, %eax leaq 8(%rsp), %rax movq %rax, 80(%rsp) movq %rsp, %rax movq %rax, 88(%rsp) movl $1, 32(%rsp) movl $1, 36(%rsp) movl $1, 40(%rsp) movl $1, 44(%rsp) movl $1, 48(%rsp) movl $1, 52(%rsp) leaq 24(%rsp), %rcx leaq 16(%rsp), %rdx leaq 44(%rsp), %rsi leaq 32(%rsp), %rdi call __cudaPopCallConfiguration@PLT testl %eax, %eax je .L7 .L3: movq 104(%rsp), %rax subq %fs:40, %rax jne .L8 addq $120, %rsp .cfi_remember_state .cfi_def_cfa_offset 8 ret .L7: .cfi_restore_state pushq 24(%rsp) .cfi_def_cfa_offset 136 pushq 24(%rsp) .cfi_def_cfa_offset 144 leaq 96(%rsp), %r9 movq 60(%rsp), %rcx movl 68(%rsp), %r8d movq 48(%rsp), %rsi movl 56(%rsp), %edx leaq _Z9transposePiS_(%rip), %rdi call cudaLaunchKernel@PLT addq $16, %rsp .cfi_def_cfa_offset 128 jmp .L3 .L8: call __stack_chk_fail@PLT .cfi_endproc .LFE2082: .size _Z30__device_stub__Z9transposePiS_PiS_, .-_Z30__device_stub__Z9transposePiS_PiS_ .globl _Z9transposePiS_ .type _Z9transposePiS_, @function _Z9transposePiS_: .LFB2083: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 call _Z30__device_stub__Z9transposePiS_PiS_ addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2083: .size _Z9transposePiS_, .-_Z9transposePiS_ .section .rodata.str1.1,"aMS",@progbits,1 .LC0: .string "[" .LC1: .string "%d " .LC2: .string "]\n" .text .globl main .type main, @function main: .LFB2057: .cfi_startproc endbr64 pushq %r12 .cfi_def_cfa_offset 16 .cfi_offset 12, -16 pushq %rbp .cfi_def_cfa_offset 24 .cfi_offset 6, -24 pushq %rbx .cfi_def_cfa_offset 32 .cfi_offset 3, -32 subq $192, %rsp .cfi_def_cfa_offset 224 movq %fs:40, %rax movq %rax, 184(%rsp) xorl %eax, %eax movl $1, %eax .L12: movl %eax, 44(%rsp,%rax,4) addq $1, %rax cmpq $17, %rax jne .L12 leaq 8(%rsp), %rdi movl $64, %esi call cudaMalloc@PLT leaq 16(%rsp), %rdi movl $64, %esi call cudaMalloc@PLT leaq 48(%rsp), %rsi movl $1, %ecx movl $64, %edx movq 8(%rsp), %rdi call cudaMemcpy@PLT movl $2, 24(%rsp) movl $2, 28(%rsp) movl $1, 32(%rsp) movl $2, 36(%rsp) movl $2, 40(%rsp) movl $1, 44(%rsp) movl $0, %r9d movl $0, %r8d movq 36(%rsp), %rdx movl $1, %ecx movq 24(%rsp), %rdi movl $1, %esi call __cudaPushCallConfiguration@PLT testl %eax, %eax je .L19 .L13: leaq 112(%rsp), %rbx movl $2, %ecx movl $64, %edx movq 16(%rsp), %rsi movq %rbx, %rdi call cudaMemcpy@PLT leaq .LC0(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT leaq 176(%rsp), %r12 leaq .LC1(%rip), %rbp .L14: movl (%rbx), %edx movq %rbp, %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT addq $4, %rbx cmpq %r12, %rbx jne .L14 leaq .LC2(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT movq 8(%rsp), %rdi call cudaFree@PLT movq 16(%rsp), %rdi call cudaFree@PLT movq 184(%rsp), %rax subq %fs:40, %rax jne .L20 movl $0, %eax addq $192, %rsp .cfi_remember_state .cfi_def_cfa_offset 32 popq %rbx .cfi_def_cfa_offset 24 popq %rbp .cfi_def_cfa_offset 16 popq %r12 .cfi_def_cfa_offset 8 ret .L19: .cfi_restore_state movq 16(%rsp), %rsi movq 8(%rsp), %rdi call _Z30__device_stub__Z9transposePiS_PiS_ jmp .L13 .L20: call __stack_chk_fail@PLT .cfi_endproc .LFE2057: .size main, .-main .section .rodata.str1.1 .LC3: .string "_Z9transposePiS_" .text .type _ZL24__sti____cudaRegisterAllv, @function _ZL24__sti____cudaRegisterAllv: .LFB2085: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 leaq _ZL15__fatDeviceText(%rip), %rdi call __cudaRegisterFatBinary@PLT movq %rax, %rdi movq %rax, _ZL20__cudaFatCubinHandle(%rip) pushq $0 .cfi_def_cfa_offset 24 pushq $0 .cfi_def_cfa_offset 32 pushq $0 .cfi_def_cfa_offset 40 pushq $0 .cfi_def_cfa_offset 48 movl $0, %r9d movl $-1, %r8d leaq .LC3(%rip), %rdx movq %rdx, %rcx leaq _Z9transposePiS_(%rip), %rsi call __cudaRegisterFunction@PLT addq $32, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaRegisterFatBinaryEnd@PLT leaq _ZL26__cudaUnregisterBinaryUtilv(%rip), %rdi call atexit@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2085: .size _ZL24__sti____cudaRegisterAllv, .-_ZL24__sti____cudaRegisterAllv .section .init_array,"aw" .align 8 .quad _ZL24__sti____cudaRegisterAllv .section .nvFatBinSegment,"aw" .align 8 .type _ZL15__fatDeviceText, @object .size _ZL15__fatDeviceText, 24 _ZL15__fatDeviceText: .long 1180844977 .long 1 .quad fatbinData .quad 0 .local _ZL20__cudaFatCubinHandle .comm _ZL20__cudaFatCubinHandle,8,8 .ident "GCC: (Ubuntu 13.3.0-6ubuntu2~24.04) 13.3.0" .section .note.GNU-stack,"",@progbits .section .note.gnu.property,"a" .align 8 .long 1f - 0f .long 4f - 1f .long 5 0: .string "GNU" 1: .align 8 .long 0xc0000002 .long 3f - 2f 2: .long 0x3 3: .align 8 4:
.text .file "transpose.hip" .globl _Z24__device_stub__transposePiS_ # -- Begin function _Z24__device_stub__transposePiS_ .p2align 4, 0x90 .type _Z24__device_stub__transposePiS_,@function _Z24__device_stub__transposePiS_: # @_Z24__device_stub__transposePiS_ .cfi_startproc # %bb.0: subq $88, %rsp .cfi_def_cfa_offset 96 movq %rdi, 56(%rsp) movq %rsi, 48(%rsp) leaq 56(%rsp), %rax movq %rax, 64(%rsp) leaq 48(%rsp), %rax movq %rax, 72(%rsp) leaq 32(%rsp), %rdi leaq 16(%rsp), %rsi leaq 8(%rsp), %rdx movq %rsp, %rcx callq __hipPopCallConfiguration movq 32(%rsp), %rsi movl 40(%rsp), %edx movq 16(%rsp), %rcx movl 24(%rsp), %r8d leaq 64(%rsp), %r9 movl $_Z9transposePiS_, %edi pushq (%rsp) .cfi_adjust_cfa_offset 8 pushq 16(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $104, %rsp .cfi_adjust_cfa_offset -104 retq .Lfunc_end0: .size _Z24__device_stub__transposePiS_, .Lfunc_end0-_Z24__device_stub__transposePiS_ .cfi_endproc # -- End function .globl main # -- Begin function main .p2align 4, 0x90 .type main,@function main: # @main .cfi_startproc # %bb.0: pushq %rbx .cfi_def_cfa_offset 16 subq $208, %rsp .cfi_def_cfa_offset 224 .cfi_offset %rbx, -16 xorl %eax, %eax .p2align 4, 0x90 .LBB1_1: # =>This Inner Loop Header: Depth=1 leaq 1(%rax), %rcx movl %ecx, 144(%rsp,%rax,4) movq %rcx, %rax cmpq $16, %rcx jne .LBB1_1 # %bb.2: leaq 8(%rsp), %rdi movl $64, %esi callq hipMalloc movq %rsp, %rdi movl $64, %esi callq hipMalloc movq 8(%rsp), %rdi leaq 144(%rsp), %rsi movl $64, %edx movl $1, %ecx callq hipMemcpy movabsq $8589934594, %rdi # imm = 0x200000002 movl $1, %esi movq %rdi, %rdx movl $1, %ecx xorl %r8d, %r8d xorl %r9d, %r9d callq __hipPushCallConfiguration testl %eax, %eax jne .LBB1_4 # %bb.3: movq 8(%rsp), %rax movq (%rsp), %rcx movq %rax, 72(%rsp) movq %rcx, 64(%rsp) leaq 72(%rsp), %rax movq %rax, 80(%rsp) leaq 64(%rsp), %rax movq %rax, 88(%rsp) leaq 48(%rsp), %rdi leaq 32(%rsp), %rsi leaq 24(%rsp), %rdx leaq 16(%rsp), %rcx callq __hipPopCallConfiguration movq 48(%rsp), %rsi movl 56(%rsp), %edx movq 32(%rsp), %rcx movl 40(%rsp), %r8d leaq 80(%rsp), %r9 movl $_Z9transposePiS_, %edi pushq 16(%rsp) .cfi_adjust_cfa_offset 8 pushq 32(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $16, %rsp .cfi_adjust_cfa_offset -16 .LBB1_4: movq (%rsp), %rsi leaq 80(%rsp), %rdi movl $64, %edx movl $2, %ecx callq hipMemcpy movl $91, %edi callq putchar@PLT xorl %ebx, %ebx .p2align 4, 0x90 .LBB1_5: # =>This Inner Loop Header: Depth=1 movl 80(%rsp,%rbx,4), %esi movl $.L.str.1, %edi xorl %eax, %eax callq printf incq %rbx cmpq $16, %rbx jne .LBB1_5 # %bb.6: movl $.Lstr, %edi callq puts@PLT movq 8(%rsp), %rdi callq hipFree movq (%rsp), %rdi callq hipFree xorl %eax, %eax addq $208, %rsp .cfi_def_cfa_offset 16 popq %rbx .cfi_def_cfa_offset 8 retq .Lfunc_end1: .size main, .Lfunc_end1-main .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_ctor .type __hip_module_ctor,@function __hip_module_ctor: # @__hip_module_ctor .cfi_startproc # %bb.0: subq $40, %rsp .cfi_def_cfa_offset 48 cmpq $0, __hip_gpubin_handle(%rip) jne .LBB2_2 # %bb.1: movl $__hip_fatbin_wrapper, %edi callq __hipRegisterFatBinary movq %rax, __hip_gpubin_handle(%rip) .LBB2_2: movq __hip_gpubin_handle(%rip), %rdi xorps %xmm0, %xmm0 movups %xmm0, 16(%rsp) movups %xmm0, (%rsp) movl $_Z9transposePiS_, %esi movl $.L__unnamed_1, %edx movl $.L__unnamed_1, %ecx movl $-1, %r8d xorl %r9d, %r9d callq __hipRegisterFunction movl $__hip_module_dtor, %edi addq $40, %rsp .cfi_def_cfa_offset 8 jmp atexit # TAILCALL .Lfunc_end2: .size __hip_module_ctor, .Lfunc_end2-__hip_module_ctor .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_dtor .type __hip_module_dtor,@function __hip_module_dtor: # @__hip_module_dtor .cfi_startproc # %bb.0: movq __hip_gpubin_handle(%rip), %rdi testq %rdi, %rdi je .LBB3_2 # %bb.1: pushq %rax .cfi_def_cfa_offset 16 callq __hipUnregisterFatBinary movq $0, __hip_gpubin_handle(%rip) addq $8, %rsp .cfi_def_cfa_offset 8 .LBB3_2: retq .Lfunc_end3: .size __hip_module_dtor, .Lfunc_end3-__hip_module_dtor .cfi_endproc # -- End function .type _Z9transposePiS_,@object # @_Z9transposePiS_ .section .rodata,"a",@progbits .globl _Z9transposePiS_ .p2align 3, 0x0 _Z9transposePiS_: .quad _Z24__device_stub__transposePiS_ .size _Z9transposePiS_, 8 .type .L.str.1,@object # @.str.1 .section .rodata.str1.1,"aMS",@progbits,1 .L.str.1: .asciz "%d " .size .L.str.1, 4 .type .L__unnamed_1,@object # @0 .L__unnamed_1: .asciz "_Z9transposePiS_" .size .L__unnamed_1, 17 .type __hip_fatbin_wrapper,@object # @__hip_fatbin_wrapper .section .hipFatBinSegment,"a",@progbits .p2align 3, 0x0 __hip_fatbin_wrapper: .long 1212764230 # 0x48495046 .long 1 # 0x1 .quad __hip_fatbin .quad 0 .size __hip_fatbin_wrapper, 24 .type __hip_gpubin_handle,@object # @__hip_gpubin_handle .local __hip_gpubin_handle .comm __hip_gpubin_handle,8,8 .section .init_array,"aw",@init_array .p2align 3, 0x0 .quad __hip_module_ctor .type __hip_cuid_,@object # @__hip_cuid_ .bss .globl __hip_cuid_ __hip_cuid_: .byte 0 # 0x0 .size __hip_cuid_, 1 .type .Lstr,@object # @str .section .rodata.str1.1,"aMS",@progbits,1 .Lstr: .asciz "]" .size .Lstr, 2 .section ".linker-options","e",@llvm_linker_options .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym _Z24__device_stub__transposePiS_ .addrsig_sym __hip_module_ctor .addrsig_sym __hip_module_dtor .addrsig_sym _Z9transposePiS_ .addrsig_sym __hip_fatbin .addrsig_sym __hip_fatbin_wrapper .addrsig_sym __hip_cuid_
You are a helpful coding assistant specialized in converting CUDA source code to CUDA device assembly.
#include<iostream> using namespace std; int n = 100; __host__ __device__ bool read(int n) { return n != 0; } __host__ __device__ bool read0(int n) { return n == 0; } __global__ void test(int n) { if(read0(n)) { printf("true\n"); } else if(read(n)){ printf("false\n"); } } int main() { dim3 block(8,2); test<<<1,block>>>(n); cudaDeviceSynchronize(); }
code for sm_80 Function : _Z4testi .headerflags @"EF_CUDA_TEXMODE_UNIFIED EF_CUDA_64BIT_ADDRESS EF_CUDA_SM80 EF_CUDA_VIRTUAL_SM(EF_CUDA_SM80)" /*0000*/ IMAD.MOV.U32 R1, RZ, RZ, c[0x0][0x28] ; /* 0x00000a00ff017624 */ /* 0x000fe400078e00ff */ /*0010*/ ISETP.NE.AND P0, PT, RZ, c[0x0][0x160], PT ; /* 0x00005800ff007a0c */ /* 0x000fda0003f05270 */ /*0020*/ @!P0 BRA 0x110 ; /* 0x000000e000008947 */ /* 0x000fea0003800000 */ /*0030*/ MOV R0, 0x0 ; /* 0x0000000000007802 */ /* 0x000fe20000000f00 */ /*0040*/ IMAD.MOV.U32 R4, RZ, RZ, c[0x4][0x10] ; /* 0x01000400ff047624 */ /* 0x000fe200078e00ff */ /*0050*/ CS2R R6, SRZ ; /* 0x0000000000067805 */ /* 0x000fe2000001ff00 */ /*0060*/ IMAD.MOV.U32 R5, RZ, RZ, c[0x4][0x14] ; /* 0x01000500ff057624 */ /* 0x000fe200078e00ff */ /*0070*/ LDC.64 R2, c[0x4][R0] ; /* 0x0100000000027b82 */ /* 0x00006c0000000a00 */ /*0080*/ LEPC R8 ; /* 0x000000000008734e */ /* 0x000fe40000000000 */ /*0090*/ MOV R11, 0x100 ; /* 0x00000100000b7802 */ /* 0x000fe40000000f00 */ /*00a0*/ MOV R20, 0x80 ; /* 0x0000008000147802 */ /* 0x000fe40000000f00 */ /*00b0*/ MOV R21, 0x0 ; /* 0x0000000000157802 */ /* 0x000fe40000000f00 */ /*00c0*/ MOV R0, 0x0 ; /* 0x0000000000007802 */ /* 0x001fc40000000f00 */ /*00d0*/ IADD3 R20, P0, P1, -R20, R11, R8 ; /* 0x0000000b14147210 */ /* 0x000fc8000791e108 */ /*00e0*/ IADD3.X R21, ~R0, R21, R9, P0, P1 ; /* 0x0000001500157210 */ /* 0x000fc800007e2509 */ /*00f0*/ CALL.ABS.NOINC R2 ; /* 0x0000000002007343 */ /* 0x002fea0003c00000 */ /*0100*/ EXIT ; /* 0x000000000000794d */ /* 0x000fea0003800000 */ /*0110*/ MOV R0, 0x0 ; /* 0x0000000000007802 */ /* 0x000fe20000000f00 */ /*0120*/ IMAD.MOV.U32 R4, RZ, RZ, c[0x4][0x8] ; /* 0x01000200ff047624 */ /* 0x000fe200078e00ff */ /*0130*/ CS2R R6, SRZ ; /* 0x0000000000067805 */ /* 0x000fe2000001ff00 */ /*0140*/ IMAD.MOV.U32 R5, RZ, RZ, c[0x4][0xc] ; /* 0x01000300ff057624 */ /* 0x000fe200078e00ff */ /*0150*/ LDC.64 R2, c[0x4][R0] ; /* 0x0100000000027b82 */ /* 0x00006c0000000a00 */ /*0160*/ LEPC R8 ; /* 0x000000000008734e */ /* 0x000fe40000000000 */ /*0170*/ MOV R11, 0x1e0 ; /* 0x000001e0000b7802 */ /* 0x000fe40000000f00 */ /*0180*/ MOV R20, 0x160 ; /* 0x0000016000147802 */ /* 0x000fe40000000f00 */ /*0190*/ MOV R21, 0x0 ; /* 0x0000000000157802 */ /* 0x000fe40000000f00 */ /*01a0*/ MOV R0, 0x0 ; /* 0x0000000000007802 */ /* 0x001fc40000000f00 */ /*01b0*/ IADD3 R20, P0, P1, -R20, R11, R8 ; /* 0x0000000b14147210 */ /* 0x000fc8000791e108 */ /*01c0*/ IADD3.X R21, ~R0, R21, R9, P0, P1 ; /* 0x0000001500157210 */ /* 0x000fc800007e2509 */ /*01d0*/ CALL.ABS.NOINC R2 ; /* 0x0000000002007343 */ /* 0x002fea0003c00000 */ /*01e0*/ EXIT ; /* 0x000000000000794d */ /* 0x000fea0003800000 */ /*01f0*/ BRA 0x1f0; /* 0xfffffff000007947 */ /* 0x000fc0000383ffff */ /*0200*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0210*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0220*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0230*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0240*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0250*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0260*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0270*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ ..........
You are a helpful coding assistant specialized in converting CUDA source code to CUDA host assembly.
#include<iostream> using namespace std; int n = 100; __host__ __device__ bool read(int n) { return n != 0; } __host__ __device__ bool read0(int n) { return n == 0; } __global__ void test(int n) { if(read0(n)) { printf("true\n"); } else if(read(n)){ printf("false\n"); } } int main() { dim3 block(8,2); test<<<1,block>>>(n); cudaDeviceSynchronize(); }
.file "tmpxft_00163fbb_00000000-6_0.cudafe1.cpp" .text #APP .globl _ZSt21ios_base_library_initv #NO_APP .type _ZL26__cudaUnregisterBinaryUtilv, @function _ZL26__cudaUnregisterBinaryUtilv: .LFB3674: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaUnregisterFatBinary@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE3674: .size _ZL26__cudaUnregisterBinaryUtilv, .-_ZL26__cudaUnregisterBinaryUtilv .globl _Z4readi .type _Z4readi, @function _Z4readi: .LFB3669: .cfi_startproc endbr64 testl %edi, %edi setne %al ret .cfi_endproc .LFE3669: .size _Z4readi, .-_Z4readi .globl _Z5read0i .type _Z5read0i, @function _Z5read0i: .LFB3670: .cfi_startproc endbr64 testl %edi, %edi sete %al ret .cfi_endproc .LFE3670: .size _Z5read0i, .-_Z5read0i .globl _Z22__device_stub__Z4testii .type _Z22__device_stub__Z4testii, @function _Z22__device_stub__Z4testii: .LFB3696: .cfi_startproc endbr64 subq $104, %rsp .cfi_def_cfa_offset 112 movl %edi, 12(%rsp) movq %fs:40, %rax movq %rax, 88(%rsp) xorl %eax, %eax leaq 12(%rsp), %rax movq %rax, 80(%rsp) movl $1, 32(%rsp) movl $1, 36(%rsp) movl $1, 40(%rsp) movl $1, 44(%rsp) movl $1, 48(%rsp) movl $1, 52(%rsp) leaq 24(%rsp), %rcx leaq 16(%rsp), %rdx leaq 44(%rsp), %rsi leaq 32(%rsp), %rdi call __cudaPopCallConfiguration@PLT testl %eax, %eax je .L9 .L5: movq 88(%rsp), %rax subq %fs:40, %rax jne .L10 addq $104, %rsp .cfi_remember_state .cfi_def_cfa_offset 8 ret .L9: .cfi_restore_state pushq 24(%rsp) .cfi_def_cfa_offset 120 pushq 24(%rsp) .cfi_def_cfa_offset 128 leaq 96(%rsp), %r9 movq 60(%rsp), %rcx movl 68(%rsp), %r8d movq 48(%rsp), %rsi movl 56(%rsp), %edx leaq _Z4testi(%rip), %rdi call cudaLaunchKernel@PLT addq $16, %rsp .cfi_def_cfa_offset 112 jmp .L5 .L10: call __stack_chk_fail@PLT .cfi_endproc .LFE3696: .size _Z22__device_stub__Z4testii, .-_Z22__device_stub__Z4testii .globl _Z4testi .type _Z4testi, @function _Z4testi: .LFB3697: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 call _Z22__device_stub__Z4testii addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE3697: .size _Z4testi, .-_Z4testi .globl main .type main, @function main: .LFB3671: .cfi_startproc endbr64 subq $40, %rsp .cfi_def_cfa_offset 48 movl $8, 8(%rsp) movl $2, 12(%rsp) movl $1, 20(%rsp) movl $1, 24(%rsp) movl $0, %r9d movl $0, %r8d movq 8(%rsp), %rdx movl $1, %ecx movq 20(%rsp), %rdi movl $1, %esi call __cudaPushCallConfiguration@PLT testl %eax, %eax je .L16 .L14: call cudaDeviceSynchronize@PLT movl $0, %eax addq $40, %rsp .cfi_remember_state .cfi_def_cfa_offset 8 ret .L16: .cfi_restore_state movl n(%rip), %edi call _Z22__device_stub__Z4testii jmp .L14 .cfi_endproc .LFE3671: .size main, .-main .section .rodata.str1.1,"aMS",@progbits,1 .LC0: .string "_Z4testi" .text .type _ZL24__sti____cudaRegisterAllv, @function _ZL24__sti____cudaRegisterAllv: .LFB3699: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 leaq _ZL15__fatDeviceText(%rip), %rdi call __cudaRegisterFatBinary@PLT movq %rax, %rdi movq %rax, _ZL20__cudaFatCubinHandle(%rip) pushq $0 .cfi_def_cfa_offset 24 pushq $0 .cfi_def_cfa_offset 32 pushq $0 .cfi_def_cfa_offset 40 pushq $0 .cfi_def_cfa_offset 48 movl $0, %r9d movl $-1, %r8d leaq .LC0(%rip), %rdx movq %rdx, %rcx leaq _Z4testi(%rip), %rsi call __cudaRegisterFunction@PLT addq $32, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaRegisterFatBinaryEnd@PLT leaq _ZL26__cudaUnregisterBinaryUtilv(%rip), %rdi call atexit@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE3699: .size _ZL24__sti____cudaRegisterAllv, .-_ZL24__sti____cudaRegisterAllv .section .init_array,"aw" .align 8 .quad _ZL24__sti____cudaRegisterAllv .section .nvFatBinSegment,"aw" .align 8 .type _ZL15__fatDeviceText, @object .size _ZL15__fatDeviceText, 24 _ZL15__fatDeviceText: .long 1180844977 .long 1 .quad fatbinData .quad 0 .local _ZL20__cudaFatCubinHandle .comm _ZL20__cudaFatCubinHandle,8,8 .globl n .data .align 4 .type n, @object .size n, 4 n: .long 100 .ident "GCC: (Ubuntu 13.3.0-6ubuntu2~24.04) 13.3.0" .section .note.GNU-stack,"",@progbits .section .note.gnu.property,"a" .align 8 .long 1f - 0f .long 4f - 1f .long 5 0: .string "GNU" 1: .align 8 .long 0xc0000002 .long 3f - 2f 2: .long 0x3 3: .align 8 4:
You are a helpful coding assistant specialized in converting CUDA source code to HIP/AMD source code.
#include<iostream> using namespace std; int n = 100; __host__ __device__ bool read(int n) { return n != 0; } __host__ __device__ bool read0(int n) { return n == 0; } __global__ void test(int n) { if(read0(n)) { printf("true\n"); } else if(read(n)){ printf("false\n"); } } int main() { dim3 block(8,2); test<<<1,block>>>(n); cudaDeviceSynchronize(); }
#include <hip/hip_runtime.h> #include<iostream> using namespace std; int n = 100; __host__ __device__ bool read(int n) { return n != 0; } __host__ __device__ bool read0(int n) { return n == 0; } __global__ void test(int n) { if(read0(n)) { printf("true\n"); } else if(read(n)){ printf("false\n"); } } int main() { dim3 block(8,2); test<<<1,block>>>(n); hipDeviceSynchronize(); }
You are a helpful coding assistant specialized in converting HIP source code to AMD host assembly.
#include <hip/hip_runtime.h> #include<iostream> using namespace std; int n = 100; __host__ __device__ bool read(int n) { return n != 0; } __host__ __device__ bool read0(int n) { return n == 0; } __global__ void test(int n) { if(read0(n)) { printf("true\n"); } else if(read(n)){ printf("false\n"); } } int main() { dim3 block(8,2); test<<<1,block>>>(n); hipDeviceSynchronize(); }
.text .file "0.hip" # Start of file scope inline assembly .globl _ZSt21ios_base_library_initv # End of file scope inline assembly .globl _Z4readi # -- Begin function _Z4readi .p2align 4, 0x90 .type _Z4readi,@function _Z4readi: # @_Z4readi .cfi_startproc # %bb.0: testl %edi, %edi setne %al retq .Lfunc_end0: .size _Z4readi, .Lfunc_end0-_Z4readi .cfi_endproc # -- End function .globl _Z5read0i # -- Begin function _Z5read0i .p2align 4, 0x90 .type _Z5read0i,@function _Z5read0i: # @_Z5read0i .cfi_startproc # %bb.0: testl %edi, %edi sete %al retq .Lfunc_end1: .size _Z5read0i, .Lfunc_end1-_Z5read0i .cfi_endproc # -- End function .globl _Z19__device_stub__testi # -- Begin function _Z19__device_stub__testi .p2align 4, 0x90 .type _Z19__device_stub__testi,@function _Z19__device_stub__testi: # @_Z19__device_stub__testi .cfi_startproc # %bb.0: subq $72, %rsp .cfi_def_cfa_offset 80 movl %edi, 12(%rsp) leaq 12(%rsp), %rax movq %rax, 16(%rsp) leaq 56(%rsp), %rdi leaq 40(%rsp), %rsi leaq 32(%rsp), %rdx leaq 24(%rsp), %rcx callq __hipPopCallConfiguration movq 56(%rsp), %rsi movl 64(%rsp), %edx movq 40(%rsp), %rcx movl 48(%rsp), %r8d leaq 16(%rsp), %r9 movl $_Z4testi, %edi pushq 24(%rsp) .cfi_adjust_cfa_offset 8 pushq 40(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $88, %rsp .cfi_adjust_cfa_offset -88 retq .Lfunc_end2: .size _Z19__device_stub__testi, .Lfunc_end2-_Z19__device_stub__testi .cfi_endproc # -- End function .globl main # -- Begin function main .p2align 4, 0x90 .type main,@function main: # @main .cfi_startproc # %bb.0: subq $72, %rsp .cfi_def_cfa_offset 80 movabsq $4294967297, %rdi # imm = 0x100000001 movabsq $8589934600, %rdx # imm = 0x200000008 movl $1, %esi movl $1, %ecx xorl %r8d, %r8d xorl %r9d, %r9d callq __hipPushCallConfiguration testl %eax, %eax jne .LBB3_2 # %bb.1: movl n(%rip), %eax movl %eax, 12(%rsp) leaq 12(%rsp), %rax movq %rax, 16(%rsp) leaq 56(%rsp), %rdi leaq 40(%rsp), %rsi leaq 32(%rsp), %rdx leaq 24(%rsp), %rcx callq __hipPopCallConfiguration movq 56(%rsp), %rsi movl 64(%rsp), %edx movq 40(%rsp), %rcx movl 48(%rsp), %r8d leaq 16(%rsp), %r9 movl $_Z4testi, %edi pushq 24(%rsp) .cfi_adjust_cfa_offset 8 pushq 40(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $16, %rsp .cfi_adjust_cfa_offset -16 .LBB3_2: callq hipDeviceSynchronize xorl %eax, %eax addq $72, %rsp .cfi_def_cfa_offset 8 retq .Lfunc_end3: .size main, .Lfunc_end3-main .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_ctor .type __hip_module_ctor,@function __hip_module_ctor: # @__hip_module_ctor .cfi_startproc # %bb.0: subq $40, %rsp .cfi_def_cfa_offset 48 cmpq $0, __hip_gpubin_handle(%rip) jne .LBB4_2 # %bb.1: movl $__hip_fatbin_wrapper, %edi callq __hipRegisterFatBinary movq %rax, __hip_gpubin_handle(%rip) .LBB4_2: movq __hip_gpubin_handle(%rip), %rdi xorps %xmm0, %xmm0 movups %xmm0, 16(%rsp) movups %xmm0, (%rsp) movl $_Z4testi, %esi movl $.L__unnamed_1, %edx movl $.L__unnamed_1, %ecx movl $-1, %r8d xorl %r9d, %r9d callq __hipRegisterFunction movl $__hip_module_dtor, %edi addq $40, %rsp .cfi_def_cfa_offset 8 jmp atexit # TAILCALL .Lfunc_end4: .size __hip_module_ctor, .Lfunc_end4-__hip_module_ctor .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_dtor .type __hip_module_dtor,@function __hip_module_dtor: # @__hip_module_dtor .cfi_startproc # %bb.0: movq __hip_gpubin_handle(%rip), %rdi testq %rdi, %rdi je .LBB5_2 # %bb.1: pushq %rax .cfi_def_cfa_offset 16 callq __hipUnregisterFatBinary movq $0, __hip_gpubin_handle(%rip) addq $8, %rsp .cfi_def_cfa_offset 8 .LBB5_2: retq .Lfunc_end5: .size __hip_module_dtor, .Lfunc_end5-__hip_module_dtor .cfi_endproc # -- End function .type n,@object # @n .data .globl n .p2align 2, 0x0 n: .long 100 # 0x64 .size n, 4 .type _Z4testi,@object # @_Z4testi .section .rodata,"a",@progbits .globl _Z4testi .p2align 3, 0x0 _Z4testi: .quad _Z19__device_stub__testi .size _Z4testi, 8 .type .L__unnamed_1,@object # @0 .section .rodata.str1.1,"aMS",@progbits,1 .L__unnamed_1: .asciz "_Z4testi" .size .L__unnamed_1, 9 .type __hip_fatbin_wrapper,@object # @__hip_fatbin_wrapper .section .hipFatBinSegment,"a",@progbits .p2align 3, 0x0 __hip_fatbin_wrapper: .long 1212764230 # 0x48495046 .long 1 # 0x1 .quad __hip_fatbin .quad 0 .size __hip_fatbin_wrapper, 24 .type __hip_gpubin_handle,@object # @__hip_gpubin_handle .local __hip_gpubin_handle .comm __hip_gpubin_handle,8,8 .section .init_array,"aw",@init_array .p2align 3, 0x0 .quad __hip_module_ctor .type __hip_cuid_,@object # @__hip_cuid_ .bss .globl __hip_cuid_ __hip_cuid_: .byte 0 # 0x0 .size __hip_cuid_, 1 .section ".linker-options","e",@llvm_linker_options .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym _Z19__device_stub__testi .addrsig_sym __hip_module_ctor .addrsig_sym __hip_module_dtor .addrsig_sym _Z4testi .addrsig_sym __hip_fatbin .addrsig_sym __hip_fatbin_wrapper .addrsig_sym __hip_cuid_
You are a helpful coding assistant specialized in converting from CUDA host assembly to AMD host assembly.
.file "tmpxft_00163fbb_00000000-6_0.cudafe1.cpp" .text #APP .globl _ZSt21ios_base_library_initv #NO_APP .type _ZL26__cudaUnregisterBinaryUtilv, @function _ZL26__cudaUnregisterBinaryUtilv: .LFB3674: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaUnregisterFatBinary@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE3674: .size _ZL26__cudaUnregisterBinaryUtilv, .-_ZL26__cudaUnregisterBinaryUtilv .globl _Z4readi .type _Z4readi, @function _Z4readi: .LFB3669: .cfi_startproc endbr64 testl %edi, %edi setne %al ret .cfi_endproc .LFE3669: .size _Z4readi, .-_Z4readi .globl _Z5read0i .type _Z5read0i, @function _Z5read0i: .LFB3670: .cfi_startproc endbr64 testl %edi, %edi sete %al ret .cfi_endproc .LFE3670: .size _Z5read0i, .-_Z5read0i .globl _Z22__device_stub__Z4testii .type _Z22__device_stub__Z4testii, @function _Z22__device_stub__Z4testii: .LFB3696: .cfi_startproc endbr64 subq $104, %rsp .cfi_def_cfa_offset 112 movl %edi, 12(%rsp) movq %fs:40, %rax movq %rax, 88(%rsp) xorl %eax, %eax leaq 12(%rsp), %rax movq %rax, 80(%rsp) movl $1, 32(%rsp) movl $1, 36(%rsp) movl $1, 40(%rsp) movl $1, 44(%rsp) movl $1, 48(%rsp) movl $1, 52(%rsp) leaq 24(%rsp), %rcx leaq 16(%rsp), %rdx leaq 44(%rsp), %rsi leaq 32(%rsp), %rdi call __cudaPopCallConfiguration@PLT testl %eax, %eax je .L9 .L5: movq 88(%rsp), %rax subq %fs:40, %rax jne .L10 addq $104, %rsp .cfi_remember_state .cfi_def_cfa_offset 8 ret .L9: .cfi_restore_state pushq 24(%rsp) .cfi_def_cfa_offset 120 pushq 24(%rsp) .cfi_def_cfa_offset 128 leaq 96(%rsp), %r9 movq 60(%rsp), %rcx movl 68(%rsp), %r8d movq 48(%rsp), %rsi movl 56(%rsp), %edx leaq _Z4testi(%rip), %rdi call cudaLaunchKernel@PLT addq $16, %rsp .cfi_def_cfa_offset 112 jmp .L5 .L10: call __stack_chk_fail@PLT .cfi_endproc .LFE3696: .size _Z22__device_stub__Z4testii, .-_Z22__device_stub__Z4testii .globl _Z4testi .type _Z4testi, @function _Z4testi: .LFB3697: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 call _Z22__device_stub__Z4testii addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE3697: .size _Z4testi, .-_Z4testi .globl main .type main, @function main: .LFB3671: .cfi_startproc endbr64 subq $40, %rsp .cfi_def_cfa_offset 48 movl $8, 8(%rsp) movl $2, 12(%rsp) movl $1, 20(%rsp) movl $1, 24(%rsp) movl $0, %r9d movl $0, %r8d movq 8(%rsp), %rdx movl $1, %ecx movq 20(%rsp), %rdi movl $1, %esi call __cudaPushCallConfiguration@PLT testl %eax, %eax je .L16 .L14: call cudaDeviceSynchronize@PLT movl $0, %eax addq $40, %rsp .cfi_remember_state .cfi_def_cfa_offset 8 ret .L16: .cfi_restore_state movl n(%rip), %edi call _Z22__device_stub__Z4testii jmp .L14 .cfi_endproc .LFE3671: .size main, .-main .section .rodata.str1.1,"aMS",@progbits,1 .LC0: .string "_Z4testi" .text .type _ZL24__sti____cudaRegisterAllv, @function _ZL24__sti____cudaRegisterAllv: .LFB3699: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 leaq _ZL15__fatDeviceText(%rip), %rdi call __cudaRegisterFatBinary@PLT movq %rax, %rdi movq %rax, _ZL20__cudaFatCubinHandle(%rip) pushq $0 .cfi_def_cfa_offset 24 pushq $0 .cfi_def_cfa_offset 32 pushq $0 .cfi_def_cfa_offset 40 pushq $0 .cfi_def_cfa_offset 48 movl $0, %r9d movl $-1, %r8d leaq .LC0(%rip), %rdx movq %rdx, %rcx leaq _Z4testi(%rip), %rsi call __cudaRegisterFunction@PLT addq $32, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaRegisterFatBinaryEnd@PLT leaq _ZL26__cudaUnregisterBinaryUtilv(%rip), %rdi call atexit@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE3699: .size _ZL24__sti____cudaRegisterAllv, .-_ZL24__sti____cudaRegisterAllv .section .init_array,"aw" .align 8 .quad _ZL24__sti____cudaRegisterAllv .section .nvFatBinSegment,"aw" .align 8 .type _ZL15__fatDeviceText, @object .size _ZL15__fatDeviceText, 24 _ZL15__fatDeviceText: .long 1180844977 .long 1 .quad fatbinData .quad 0 .local _ZL20__cudaFatCubinHandle .comm _ZL20__cudaFatCubinHandle,8,8 .globl n .data .align 4 .type n, @object .size n, 4 n: .long 100 .ident "GCC: (Ubuntu 13.3.0-6ubuntu2~24.04) 13.3.0" .section .note.GNU-stack,"",@progbits .section .note.gnu.property,"a" .align 8 .long 1f - 0f .long 4f - 1f .long 5 0: .string "GNU" 1: .align 8 .long 0xc0000002 .long 3f - 2f 2: .long 0x3 3: .align 8 4:
.text .file "0.hip" # Start of file scope inline assembly .globl _ZSt21ios_base_library_initv # End of file scope inline assembly .globl _Z4readi # -- Begin function _Z4readi .p2align 4, 0x90 .type _Z4readi,@function _Z4readi: # @_Z4readi .cfi_startproc # %bb.0: testl %edi, %edi setne %al retq .Lfunc_end0: .size _Z4readi, .Lfunc_end0-_Z4readi .cfi_endproc # -- End function .globl _Z5read0i # -- Begin function _Z5read0i .p2align 4, 0x90 .type _Z5read0i,@function _Z5read0i: # @_Z5read0i .cfi_startproc # %bb.0: testl %edi, %edi sete %al retq .Lfunc_end1: .size _Z5read0i, .Lfunc_end1-_Z5read0i .cfi_endproc # -- End function .globl _Z19__device_stub__testi # -- Begin function _Z19__device_stub__testi .p2align 4, 0x90 .type _Z19__device_stub__testi,@function _Z19__device_stub__testi: # @_Z19__device_stub__testi .cfi_startproc # %bb.0: subq $72, %rsp .cfi_def_cfa_offset 80 movl %edi, 12(%rsp) leaq 12(%rsp), %rax movq %rax, 16(%rsp) leaq 56(%rsp), %rdi leaq 40(%rsp), %rsi leaq 32(%rsp), %rdx leaq 24(%rsp), %rcx callq __hipPopCallConfiguration movq 56(%rsp), %rsi movl 64(%rsp), %edx movq 40(%rsp), %rcx movl 48(%rsp), %r8d leaq 16(%rsp), %r9 movl $_Z4testi, %edi pushq 24(%rsp) .cfi_adjust_cfa_offset 8 pushq 40(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $88, %rsp .cfi_adjust_cfa_offset -88 retq .Lfunc_end2: .size _Z19__device_stub__testi, .Lfunc_end2-_Z19__device_stub__testi .cfi_endproc # -- End function .globl main # -- Begin function main .p2align 4, 0x90 .type main,@function main: # @main .cfi_startproc # %bb.0: subq $72, %rsp .cfi_def_cfa_offset 80 movabsq $4294967297, %rdi # imm = 0x100000001 movabsq $8589934600, %rdx # imm = 0x200000008 movl $1, %esi movl $1, %ecx xorl %r8d, %r8d xorl %r9d, %r9d callq __hipPushCallConfiguration testl %eax, %eax jne .LBB3_2 # %bb.1: movl n(%rip), %eax movl %eax, 12(%rsp) leaq 12(%rsp), %rax movq %rax, 16(%rsp) leaq 56(%rsp), %rdi leaq 40(%rsp), %rsi leaq 32(%rsp), %rdx leaq 24(%rsp), %rcx callq __hipPopCallConfiguration movq 56(%rsp), %rsi movl 64(%rsp), %edx movq 40(%rsp), %rcx movl 48(%rsp), %r8d leaq 16(%rsp), %r9 movl $_Z4testi, %edi pushq 24(%rsp) .cfi_adjust_cfa_offset 8 pushq 40(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $16, %rsp .cfi_adjust_cfa_offset -16 .LBB3_2: callq hipDeviceSynchronize xorl %eax, %eax addq $72, %rsp .cfi_def_cfa_offset 8 retq .Lfunc_end3: .size main, .Lfunc_end3-main .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_ctor .type __hip_module_ctor,@function __hip_module_ctor: # @__hip_module_ctor .cfi_startproc # %bb.0: subq $40, %rsp .cfi_def_cfa_offset 48 cmpq $0, __hip_gpubin_handle(%rip) jne .LBB4_2 # %bb.1: movl $__hip_fatbin_wrapper, %edi callq __hipRegisterFatBinary movq %rax, __hip_gpubin_handle(%rip) .LBB4_2: movq __hip_gpubin_handle(%rip), %rdi xorps %xmm0, %xmm0 movups %xmm0, 16(%rsp) movups %xmm0, (%rsp) movl $_Z4testi, %esi movl $.L__unnamed_1, %edx movl $.L__unnamed_1, %ecx movl $-1, %r8d xorl %r9d, %r9d callq __hipRegisterFunction movl $__hip_module_dtor, %edi addq $40, %rsp .cfi_def_cfa_offset 8 jmp atexit # TAILCALL .Lfunc_end4: .size __hip_module_ctor, .Lfunc_end4-__hip_module_ctor .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_dtor .type __hip_module_dtor,@function __hip_module_dtor: # @__hip_module_dtor .cfi_startproc # %bb.0: movq __hip_gpubin_handle(%rip), %rdi testq %rdi, %rdi je .LBB5_2 # %bb.1: pushq %rax .cfi_def_cfa_offset 16 callq __hipUnregisterFatBinary movq $0, __hip_gpubin_handle(%rip) addq $8, %rsp .cfi_def_cfa_offset 8 .LBB5_2: retq .Lfunc_end5: .size __hip_module_dtor, .Lfunc_end5-__hip_module_dtor .cfi_endproc # -- End function .type n,@object # @n .data .globl n .p2align 2, 0x0 n: .long 100 # 0x64 .size n, 4 .type _Z4testi,@object # @_Z4testi .section .rodata,"a",@progbits .globl _Z4testi .p2align 3, 0x0 _Z4testi: .quad _Z19__device_stub__testi .size _Z4testi, 8 .type .L__unnamed_1,@object # @0 .section .rodata.str1.1,"aMS",@progbits,1 .L__unnamed_1: .asciz "_Z4testi" .size .L__unnamed_1, 9 .type __hip_fatbin_wrapper,@object # @__hip_fatbin_wrapper .section .hipFatBinSegment,"a",@progbits .p2align 3, 0x0 __hip_fatbin_wrapper: .long 1212764230 # 0x48495046 .long 1 # 0x1 .quad __hip_fatbin .quad 0 .size __hip_fatbin_wrapper, 24 .type __hip_gpubin_handle,@object # @__hip_gpubin_handle .local __hip_gpubin_handle .comm __hip_gpubin_handle,8,8 .section .init_array,"aw",@init_array .p2align 3, 0x0 .quad __hip_module_ctor .type __hip_cuid_,@object # @__hip_cuid_ .bss .globl __hip_cuid_ __hip_cuid_: .byte 0 # 0x0 .size __hip_cuid_, 1 .section ".linker-options","e",@llvm_linker_options .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym _Z19__device_stub__testi .addrsig_sym __hip_module_ctor .addrsig_sym __hip_module_dtor .addrsig_sym _Z4testi .addrsig_sym __hip_fatbin .addrsig_sym __hip_fatbin_wrapper .addrsig_sym __hip_cuid_
You are a helpful coding assistant specialized in converting CUDA source code to HIP/AMD source code.
#include <thrust/sort.h> #include <thrust/host_vector.h> #include <thrust/device_vector.h> #include <iostream> #include <vector> #include <algorithm> #include <numeric> #include <random> #include <chrono> using namespace std; template<typename T> std::vector<std::size_t> tag_sort(const std::vector<T>& v) { std::vector<std::size_t> result(v.size()); std::iota(std::begin(result), std::end(result), 0); std::sort(std::begin(result), std::end(result), [&v](const double & lhs, const double & rhs) { return v[lhs] < v[rhs]; } ); return result; } int main(){ int vec_size = 10000000; vector<double> values; vector<int> indices; double lower_bound = 0; double upper_bound = 10000; std::default_random_engine re; std::uniform_real_distribution<double> unif(lower_bound,upper_bound); for(int i = 0; i < vec_size; i++){ double a_random_double = unif(re); values.push_back(a_random_double); indices.push_back(i); } // generating values //for(double i : values) // cout << i << " "; //cout << endl; vector<double> values_cpu(values); // sort using c++ stl vector auto start_cpu = std::chrono::high_resolution_clock::now(); auto idices_cpu = tag_sort(values_cpu); auto finish_cpu = std::chrono::high_resolution_clock::now(); std::chrono::duration<double> elapsed_cpu = finish_cpu - start_cpu; cout <<" CPU single thread executing time: " << elapsed_cpu.count() << endl; // output //for (auto && elem:idxs) // std::cout << elem << " : " << values[elem] << std::endl; thrust::device_vector<double> values_gpu(values); thrust::device_vector<int> indices_gpu(indices); // sort using cuda gpu auto start_gpu = std::chrono::high_resolution_clock::now(); thrust::sort_by_key(values_gpu.begin(), values_gpu.end(), indices_gpu.begin()); auto finish_gpu = std::chrono::high_resolution_clock::now(); std::chrono::duration<double> elapsed_gpu = finish_gpu - start_gpu; cout <<" GPU CUDA executing time: " << elapsed_gpu.count() << endl; //for(int i = 0; i < vec_size; i++) // cout << indices[i] << " : " << values[i] << endl; bool are_equal = true; for(int i = 0; i < vec_size; i++) if (idices_cpu[i] != indices_gpu[i]){ are_equal = false; break; } cout << are_equal << endl; return 0; }
#include <hip/hip_runtime.h> #include <thrust/sort.h> #include <thrust/host_vector.h> #include <thrust/device_vector.h> #include <iostream> #include <vector> #include <algorithm> #include <numeric> #include <random> #include <chrono> using namespace std; template<typename T> std::vector<std::size_t> tag_sort(const std::vector<T>& v) { std::vector<std::size_t> result(v.size()); std::iota(std::begin(result), std::end(result), 0); std::sort(std::begin(result), std::end(result), [&v](const double & lhs, const double & rhs) { return v[lhs] < v[rhs]; } ); return result; } int main(){ int vec_size = 10000000; vector<double> values; vector<int> indices; double lower_bound = 0; double upper_bound = 10000; std::default_random_engine re; std::uniform_real_distribution<double> unif(lower_bound,upper_bound); for(int i = 0; i < vec_size; i++){ double a_random_double = unif(re); values.push_back(a_random_double); indices.push_back(i); } // generating values //for(double i : values) // cout << i << " "; //cout << endl; vector<double> values_cpu(values); // sort using c++ stl vector auto start_cpu = std::chrono::high_resolution_clock::now(); auto idices_cpu = tag_sort(values_cpu); auto finish_cpu = std::chrono::high_resolution_clock::now(); std::chrono::duration<double> elapsed_cpu = finish_cpu - start_cpu; cout <<" CPU single thread executing time: " << elapsed_cpu.count() << endl; // output //for (auto && elem:idxs) // std::cout << elem << " : " << values[elem] << std::endl; thrust::device_vector<double> values_gpu(values); thrust::device_vector<int> indices_gpu(indices); // sort using cuda gpu auto start_gpu = std::chrono::high_resolution_clock::now(); thrust::sort_by_key(values_gpu.begin(), values_gpu.end(), indices_gpu.begin()); auto finish_gpu = std::chrono::high_resolution_clock::now(); std::chrono::duration<double> elapsed_gpu = finish_gpu - start_gpu; cout <<" GPU CUDA executing time: " << elapsed_gpu.count() << endl; //for(int i = 0; i < vec_size; i++) // cout << indices[i] << " : " << values[i] << endl; bool are_equal = true; for(int i = 0; i < vec_size; i++) if (idices_cpu[i] != indices_gpu[i]){ are_equal = false; break; } cout << are_equal << endl; return 0; }
You are a helpful coding assistant specialized in converting CUDA source code to CUDA device assembly.
#include "cuda_runtime.h" #include "device_launch_parameters.h" #include <stdio.h> __global__ void mykernel() { int i = blockIdx.x; int j = threadIdx.x; printf("Hello world from Kernel\tBlock id: %d\tThread id: %d\n", i, j); } int main() { printf("Hello world from CPU\n"); mykernel<<<4,5>>>(); cudaDeviceReset(); while(1); return 0; }
code for sm_80 Function : _Z8mykernelv .headerflags @"EF_CUDA_TEXMODE_UNIFIED EF_CUDA_64BIT_ADDRESS EF_CUDA_SM80 EF_CUDA_VIRTUAL_SM(EF_CUDA_SM80)" /*0000*/ IMAD.MOV.U32 R1, RZ, RZ, c[0x0][0x28] ; /* 0x00000a00ff017624 */ /* 0x000fc800078e00ff */ /*0010*/ S2R R9, SR_TID.X ; /* 0x0000000000097919 */ /* 0x000e220000002100 */ /*0020*/ IADD3 R1, R1, -0x8, RZ ; /* 0xfffffff801017810 */ /* 0x000fe20007ffe0ff */ /*0030*/ IMAD.MOV.U32 R4, RZ, RZ, c[0x4][0x8] ; /* 0x01000200ff047624 */ /* 0x000fe200078e00ff */ /*0040*/ MOV R0, 0x0 ; /* 0x0000000000007802 */ /* 0x000fe20000000f00 */ /*0050*/ S2R R8, SR_CTAID.X ; /* 0x0000000000087919 */ /* 0x000e220000002500 */ /*0060*/ IADD3 R6, P0, R1, c[0x0][0x20], RZ ; /* 0x0000080001067a10 */ /* 0x000fe20007f1e0ff */ /*0070*/ IMAD.MOV.U32 R5, RZ, RZ, c[0x4][0xc] ; /* 0x01000300ff057624 */ /* 0x000fe200078e00ff */ /*0080*/ LDC.64 R2, c[0x4][R0] ; /* 0x0100000000027b82 */ /* 0x0002a60000000a00 */ /*0090*/ IMAD.X R7, RZ, RZ, c[0x0][0x24], P0 ; /* 0x00000900ff077624 */ /* 0x000fe200000e06ff */ /*00a0*/ STL.64 [R1], R8 ; /* 0x0000000801007387 */ /* 0x0013e80000100a00 */ /*00b0*/ LEPC R8 ; /* 0x000000000008734e */ /* 0x002fc60000000000 */ /*00c0*/ MOV R11, 0x130 ; /* 0x00000130000b7802 */ /* 0x000fe40000000f00 */ /*00d0*/ MOV R20, 0xb0 ; /* 0x000000b000147802 */ /* 0x000fc40000000f00 */ /*00e0*/ MOV R21, 0x0 ; /* 0x0000000000157802 */ /* 0x000fe40000000f00 */ /*00f0*/ MOV R0, 0x0 ; /* 0x0000000000007802 */ /* 0x000fe40000000f00 */ /*0100*/ IADD3 R20, P0, P1, -R20, R11, R8 ; /* 0x0000000b14147210 */ /* 0x000fc8000791e108 */ /*0110*/ IADD3.X R21, ~R0, R21, R9, P0, P1 ; /* 0x0000001500157210 */ /* 0x000fc800007e2509 */ /*0120*/ CALL.ABS.NOINC R2 ; /* 0x0000000002007343 */ /* 0x004fea0003c00000 */ /*0130*/ EXIT ; /* 0x000000000000794d */ /* 0x000fea0003800000 */ /*0140*/ BRA 0x140; /* 0xfffffff000007947 */ /* 0x000fc0000383ffff */ /*0150*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0160*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0170*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0180*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0190*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*01a0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*01b0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*01c0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*01d0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*01e0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*01f0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ ..........
You are a helpful coding assistant specialized in converting CUDA source code to CUDA host assembly.
#include "cuda_runtime.h" #include "device_launch_parameters.h" #include <stdio.h> __global__ void mykernel() { int i = blockIdx.x; int j = threadIdx.x; printf("Hello world from Kernel\tBlock id: %d\tThread id: %d\n", i, j); } int main() { printf("Hello world from CPU\n"); mykernel<<<4,5>>>(); cudaDeviceReset(); while(1); return 0; }
.file "tmpxft_0004f92f_00000000-6_kernel.cudafe1.cpp" .text #APP #NO_APP .type _ZL26__cudaUnregisterBinaryUtilv, @function _ZL26__cudaUnregisterBinaryUtilv: .LFB2060: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaUnregisterFatBinary@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2060: .size _ZL26__cudaUnregisterBinaryUtilv, .-_ZL26__cudaUnregisterBinaryUtilv .globl _Z26__device_stub__Z8mykernelvv .type _Z26__device_stub__Z8mykernelvv, @function _Z26__device_stub__Z8mykernelvv: .LFB2082: .cfi_startproc endbr64 subq $88, %rsp .cfi_def_cfa_offset 96 movq %fs:40, %rax movq %rax, 72(%rsp) xorl %eax, %eax movl $1, 16(%rsp) movl $1, 20(%rsp) movl $1, 24(%rsp) movl $1, 28(%rsp) movl $1, 32(%rsp) movl $1, 36(%rsp) leaq 8(%rsp), %rcx movq %rsp, %rdx leaq 28(%rsp), %rsi leaq 16(%rsp), %rdi call __cudaPopCallConfiguration@PLT testl %eax, %eax je .L7 .L3: movq 72(%rsp), %rax subq %fs:40, %rax jne .L8 addq $88, %rsp .cfi_remember_state .cfi_def_cfa_offset 8 ret .L7: .cfi_restore_state pushq 8(%rsp) .cfi_def_cfa_offset 104 pushq 8(%rsp) .cfi_def_cfa_offset 112 leaq 80(%rsp), %r9 movq 44(%rsp), %rcx movl 52(%rsp), %r8d movq 32(%rsp), %rsi movl 40(%rsp), %edx leaq _Z8mykernelv(%rip), %rdi call cudaLaunchKernel@PLT addq $16, %rsp .cfi_def_cfa_offset 96 jmp .L3 .L8: call __stack_chk_fail@PLT .cfi_endproc .LFE2082: .size _Z26__device_stub__Z8mykernelvv, .-_Z26__device_stub__Z8mykernelvv .globl _Z8mykernelv .type _Z8mykernelv, @function _Z8mykernelv: .LFB2083: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 call _Z26__device_stub__Z8mykernelvv addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2083: .size _Z8mykernelv, .-_Z8mykernelv .section .rodata.str1.1,"aMS",@progbits,1 .LC0: .string "Hello world from CPU\n" .text .globl main .type main, @function main: .LFB2057: .cfi_startproc endbr64 pushq %rax .cfi_def_cfa_offset 16 popq %rax .cfi_def_cfa_offset 8 subq $40, %rsp .cfi_def_cfa_offset 48 leaq .LC0(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT movl $5, 20(%rsp) movl $1, 24(%rsp) movl $4, 8(%rsp) movl $1, 12(%rsp) movl $0, %r9d movl $0, %r8d movq 20(%rsp), %rdx movl $1, %ecx movq 8(%rsp), %rdi movl $1, %esi call __cudaPushCallConfiguration@PLT testl %eax, %eax je .L15 .L12: call cudaDeviceReset@PLT .L13: jmp .L13 .L15: call _Z26__device_stub__Z8mykernelvv jmp .L12 .cfi_endproc .LFE2057: .size main, .-main .section .rodata.str1.1 .LC1: .string "_Z8mykernelv" .text .type _ZL24__sti____cudaRegisterAllv, @function _ZL24__sti____cudaRegisterAllv: .LFB2085: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 leaq _ZL15__fatDeviceText(%rip), %rdi call __cudaRegisterFatBinary@PLT movq %rax, %rdi movq %rax, _ZL20__cudaFatCubinHandle(%rip) pushq $0 .cfi_def_cfa_offset 24 pushq $0 .cfi_def_cfa_offset 32 pushq $0 .cfi_def_cfa_offset 40 pushq $0 .cfi_def_cfa_offset 48 movl $0, %r9d movl $-1, %r8d leaq .LC1(%rip), %rdx movq %rdx, %rcx leaq _Z8mykernelv(%rip), %rsi call __cudaRegisterFunction@PLT addq $32, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaRegisterFatBinaryEnd@PLT leaq _ZL26__cudaUnregisterBinaryUtilv(%rip), %rdi call atexit@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2085: .size _ZL24__sti____cudaRegisterAllv, .-_ZL24__sti____cudaRegisterAllv .section .init_array,"aw" .align 8 .quad _ZL24__sti____cudaRegisterAllv .section .nvFatBinSegment,"aw" .align 8 .type _ZL15__fatDeviceText, @object .size _ZL15__fatDeviceText, 24 _ZL15__fatDeviceText: .long 1180844977 .long 1 .quad fatbinData .quad 0 .local _ZL20__cudaFatCubinHandle .comm _ZL20__cudaFatCubinHandle,8,8 .ident "GCC: (Ubuntu 13.3.0-6ubuntu2~24.04) 13.3.0" .section .note.GNU-stack,"",@progbits .section .note.gnu.property,"a" .align 8 .long 1f - 0f .long 4f - 1f .long 5 0: .string "GNU" 1: .align 8 .long 0xc0000002 .long 3f - 2f 2: .long 0x3 3: .align 8 4:
You are a helpful coding assistant specialized in converting CUDA source code to HIP/AMD source code.
#include "cuda_runtime.h" #include "device_launch_parameters.h" #include <stdio.h> __global__ void mykernel() { int i = blockIdx.x; int j = threadIdx.x; printf("Hello world from Kernel\tBlock id: %d\tThread id: %d\n", i, j); } int main() { printf("Hello world from CPU\n"); mykernel<<<4,5>>>(); cudaDeviceReset(); while(1); return 0; }
#include "hip/hip_runtime.h" #include <stdio.h> __global__ void mykernel() { int i = blockIdx.x; int j = threadIdx.x; printf("Hello world from Kernel\tBlock id: %d\tThread id: %d\n", i, j); } int main() { printf("Hello world from CPU\n"); mykernel<<<4,5>>>(); hipDeviceReset(); while(1); return 0; }
You are a helpful coding assistant specialized in converting HIP source code to AMD host assembly.
#include "hip/hip_runtime.h" #include <stdio.h> __global__ void mykernel() { int i = blockIdx.x; int j = threadIdx.x; printf("Hello world from Kernel\tBlock id: %d\tThread id: %d\n", i, j); } int main() { printf("Hello world from CPU\n"); mykernel<<<4,5>>>(); hipDeviceReset(); while(1); return 0; }
.text .file "kernel.hip" .globl _Z23__device_stub__mykernelv # -- Begin function _Z23__device_stub__mykernelv .p2align 4, 0x90 .type _Z23__device_stub__mykernelv,@function _Z23__device_stub__mykernelv: # @_Z23__device_stub__mykernelv .cfi_startproc # %bb.0: subq $56, %rsp .cfi_def_cfa_offset 64 leaq 32(%rsp), %rdi leaq 16(%rsp), %rsi leaq 8(%rsp), %rdx movq %rsp, %rcx callq __hipPopCallConfiguration movq 32(%rsp), %rsi movl 40(%rsp), %edx movq 16(%rsp), %rcx movl 24(%rsp), %r8d leaq 48(%rsp), %r9 movl $_Z8mykernelv, %edi pushq (%rsp) .cfi_adjust_cfa_offset 8 pushq 16(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $72, %rsp .cfi_adjust_cfa_offset -72 retq .Lfunc_end0: .size _Z23__device_stub__mykernelv, .Lfunc_end0-_Z23__device_stub__mykernelv .cfi_endproc # -- End function .globl main # -- Begin function main .p2align 4, 0x90 .type main,@function main: # @main .cfi_startproc # %bb.0: subq $56, %rsp .cfi_def_cfa_offset 64 movl $.Lstr, %edi callq puts@PLT movabsq $4294967300, %rdi # imm = 0x100000004 leaq 1(%rdi), %rdx movl $1, %esi movl $1, %ecx xorl %r8d, %r8d xorl %r9d, %r9d callq __hipPushCallConfiguration testl %eax, %eax je .LBB1_1 # %bb.2: callq hipDeviceReset .LBB1_1: leaq 32(%rsp), %rdi leaq 16(%rsp), %rsi leaq 8(%rsp), %rdx movq %rsp, %rcx callq __hipPopCallConfiguration movq 32(%rsp), %rsi movl 40(%rsp), %edx movq 16(%rsp), %rcx movl 24(%rsp), %r8d leaq 48(%rsp), %r9 movl $_Z8mykernelv, %edi pushq (%rsp) .cfi_adjust_cfa_offset 8 pushq 16(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $16, %rsp .cfi_adjust_cfa_offset -16 callq hipDeviceReset .Lfunc_end1: .size main, .Lfunc_end1-main .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_ctor .type __hip_module_ctor,@function __hip_module_ctor: # @__hip_module_ctor .cfi_startproc # %bb.0: subq $40, %rsp .cfi_def_cfa_offset 48 cmpq $0, __hip_gpubin_handle(%rip) jne .LBB2_2 # %bb.1: movl $__hip_fatbin_wrapper, %edi callq __hipRegisterFatBinary movq %rax, __hip_gpubin_handle(%rip) .LBB2_2: movq __hip_gpubin_handle(%rip), %rdi xorps %xmm0, %xmm0 movups %xmm0, 16(%rsp) movups %xmm0, (%rsp) movl $_Z8mykernelv, %esi movl $.L__unnamed_1, %edx movl $.L__unnamed_1, %ecx movl $-1, %r8d xorl %r9d, %r9d callq __hipRegisterFunction movl $__hip_module_dtor, %edi addq $40, %rsp .cfi_def_cfa_offset 8 jmp atexit # TAILCALL .Lfunc_end2: .size __hip_module_ctor, .Lfunc_end2-__hip_module_ctor .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_dtor .type __hip_module_dtor,@function __hip_module_dtor: # @__hip_module_dtor .cfi_startproc # %bb.0: movq __hip_gpubin_handle(%rip), %rdi testq %rdi, %rdi je .LBB3_2 # %bb.1: pushq %rax .cfi_def_cfa_offset 16 callq __hipUnregisterFatBinary movq $0, __hip_gpubin_handle(%rip) addq $8, %rsp .cfi_def_cfa_offset 8 .LBB3_2: retq .Lfunc_end3: .size __hip_module_dtor, .Lfunc_end3-__hip_module_dtor .cfi_endproc # -- End function .type _Z8mykernelv,@object # @_Z8mykernelv .section .rodata,"a",@progbits .globl _Z8mykernelv .p2align 3, 0x0 _Z8mykernelv: .quad _Z23__device_stub__mykernelv .size _Z8mykernelv, 8 .type .L__unnamed_1,@object # @0 .section .rodata.str1.1,"aMS",@progbits,1 .L__unnamed_1: .asciz "_Z8mykernelv" .size .L__unnamed_1, 13 .type __hip_fatbin_wrapper,@object # @__hip_fatbin_wrapper .section .hipFatBinSegment,"a",@progbits .p2align 3, 0x0 __hip_fatbin_wrapper: .long 1212764230 # 0x48495046 .long 1 # 0x1 .quad __hip_fatbin .quad 0 .size __hip_fatbin_wrapper, 24 .type __hip_gpubin_handle,@object # @__hip_gpubin_handle .local __hip_gpubin_handle .comm __hip_gpubin_handle,8,8 .section .init_array,"aw",@init_array .p2align 3, 0x0 .quad __hip_module_ctor .type __hip_cuid_,@object # @__hip_cuid_ .bss .globl __hip_cuid_ __hip_cuid_: .byte 0 # 0x0 .size __hip_cuid_, 1 .type .Lstr,@object # @str .section .rodata.str1.1,"aMS",@progbits,1 .Lstr: .asciz "Hello world from CPU" .size .Lstr, 21 .section ".linker-options","e",@llvm_linker_options .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym _Z23__device_stub__mykernelv .addrsig_sym __hip_module_ctor .addrsig_sym __hip_module_dtor .addrsig_sym _Z8mykernelv .addrsig_sym __hip_fatbin .addrsig_sym __hip_fatbin_wrapper .addrsig_sym __hip_cuid_
You are a helpful coding assistant specialized in converting from CUDA host assembly to AMD host assembly.
.file "tmpxft_0004f92f_00000000-6_kernel.cudafe1.cpp" .text #APP #NO_APP .type _ZL26__cudaUnregisterBinaryUtilv, @function _ZL26__cudaUnregisterBinaryUtilv: .LFB2060: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaUnregisterFatBinary@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2060: .size _ZL26__cudaUnregisterBinaryUtilv, .-_ZL26__cudaUnregisterBinaryUtilv .globl _Z26__device_stub__Z8mykernelvv .type _Z26__device_stub__Z8mykernelvv, @function _Z26__device_stub__Z8mykernelvv: .LFB2082: .cfi_startproc endbr64 subq $88, %rsp .cfi_def_cfa_offset 96 movq %fs:40, %rax movq %rax, 72(%rsp) xorl %eax, %eax movl $1, 16(%rsp) movl $1, 20(%rsp) movl $1, 24(%rsp) movl $1, 28(%rsp) movl $1, 32(%rsp) movl $1, 36(%rsp) leaq 8(%rsp), %rcx movq %rsp, %rdx leaq 28(%rsp), %rsi leaq 16(%rsp), %rdi call __cudaPopCallConfiguration@PLT testl %eax, %eax je .L7 .L3: movq 72(%rsp), %rax subq %fs:40, %rax jne .L8 addq $88, %rsp .cfi_remember_state .cfi_def_cfa_offset 8 ret .L7: .cfi_restore_state pushq 8(%rsp) .cfi_def_cfa_offset 104 pushq 8(%rsp) .cfi_def_cfa_offset 112 leaq 80(%rsp), %r9 movq 44(%rsp), %rcx movl 52(%rsp), %r8d movq 32(%rsp), %rsi movl 40(%rsp), %edx leaq _Z8mykernelv(%rip), %rdi call cudaLaunchKernel@PLT addq $16, %rsp .cfi_def_cfa_offset 96 jmp .L3 .L8: call __stack_chk_fail@PLT .cfi_endproc .LFE2082: .size _Z26__device_stub__Z8mykernelvv, .-_Z26__device_stub__Z8mykernelvv .globl _Z8mykernelv .type _Z8mykernelv, @function _Z8mykernelv: .LFB2083: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 call _Z26__device_stub__Z8mykernelvv addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2083: .size _Z8mykernelv, .-_Z8mykernelv .section .rodata.str1.1,"aMS",@progbits,1 .LC0: .string "Hello world from CPU\n" .text .globl main .type main, @function main: .LFB2057: .cfi_startproc endbr64 pushq %rax .cfi_def_cfa_offset 16 popq %rax .cfi_def_cfa_offset 8 subq $40, %rsp .cfi_def_cfa_offset 48 leaq .LC0(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT movl $5, 20(%rsp) movl $1, 24(%rsp) movl $4, 8(%rsp) movl $1, 12(%rsp) movl $0, %r9d movl $0, %r8d movq 20(%rsp), %rdx movl $1, %ecx movq 8(%rsp), %rdi movl $1, %esi call __cudaPushCallConfiguration@PLT testl %eax, %eax je .L15 .L12: call cudaDeviceReset@PLT .L13: jmp .L13 .L15: call _Z26__device_stub__Z8mykernelvv jmp .L12 .cfi_endproc .LFE2057: .size main, .-main .section .rodata.str1.1 .LC1: .string "_Z8mykernelv" .text .type _ZL24__sti____cudaRegisterAllv, @function _ZL24__sti____cudaRegisterAllv: .LFB2085: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 leaq _ZL15__fatDeviceText(%rip), %rdi call __cudaRegisterFatBinary@PLT movq %rax, %rdi movq %rax, _ZL20__cudaFatCubinHandle(%rip) pushq $0 .cfi_def_cfa_offset 24 pushq $0 .cfi_def_cfa_offset 32 pushq $0 .cfi_def_cfa_offset 40 pushq $0 .cfi_def_cfa_offset 48 movl $0, %r9d movl $-1, %r8d leaq .LC1(%rip), %rdx movq %rdx, %rcx leaq _Z8mykernelv(%rip), %rsi call __cudaRegisterFunction@PLT addq $32, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaRegisterFatBinaryEnd@PLT leaq _ZL26__cudaUnregisterBinaryUtilv(%rip), %rdi call atexit@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2085: .size _ZL24__sti____cudaRegisterAllv, .-_ZL24__sti____cudaRegisterAllv .section .init_array,"aw" .align 8 .quad _ZL24__sti____cudaRegisterAllv .section .nvFatBinSegment,"aw" .align 8 .type _ZL15__fatDeviceText, @object .size _ZL15__fatDeviceText, 24 _ZL15__fatDeviceText: .long 1180844977 .long 1 .quad fatbinData .quad 0 .local _ZL20__cudaFatCubinHandle .comm _ZL20__cudaFatCubinHandle,8,8 .ident "GCC: (Ubuntu 13.3.0-6ubuntu2~24.04) 13.3.0" .section .note.GNU-stack,"",@progbits .section .note.gnu.property,"a" .align 8 .long 1f - 0f .long 4f - 1f .long 5 0: .string "GNU" 1: .align 8 .long 0xc0000002 .long 3f - 2f 2: .long 0x3 3: .align 8 4:
.text .file "kernel.hip" .globl _Z23__device_stub__mykernelv # -- Begin function _Z23__device_stub__mykernelv .p2align 4, 0x90 .type _Z23__device_stub__mykernelv,@function _Z23__device_stub__mykernelv: # @_Z23__device_stub__mykernelv .cfi_startproc # %bb.0: subq $56, %rsp .cfi_def_cfa_offset 64 leaq 32(%rsp), %rdi leaq 16(%rsp), %rsi leaq 8(%rsp), %rdx movq %rsp, %rcx callq __hipPopCallConfiguration movq 32(%rsp), %rsi movl 40(%rsp), %edx movq 16(%rsp), %rcx movl 24(%rsp), %r8d leaq 48(%rsp), %r9 movl $_Z8mykernelv, %edi pushq (%rsp) .cfi_adjust_cfa_offset 8 pushq 16(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $72, %rsp .cfi_adjust_cfa_offset -72 retq .Lfunc_end0: .size _Z23__device_stub__mykernelv, .Lfunc_end0-_Z23__device_stub__mykernelv .cfi_endproc # -- End function .globl main # -- Begin function main .p2align 4, 0x90 .type main,@function main: # @main .cfi_startproc # %bb.0: subq $56, %rsp .cfi_def_cfa_offset 64 movl $.Lstr, %edi callq puts@PLT movabsq $4294967300, %rdi # imm = 0x100000004 leaq 1(%rdi), %rdx movl $1, %esi movl $1, %ecx xorl %r8d, %r8d xorl %r9d, %r9d callq __hipPushCallConfiguration testl %eax, %eax je .LBB1_1 # %bb.2: callq hipDeviceReset .LBB1_1: leaq 32(%rsp), %rdi leaq 16(%rsp), %rsi leaq 8(%rsp), %rdx movq %rsp, %rcx callq __hipPopCallConfiguration movq 32(%rsp), %rsi movl 40(%rsp), %edx movq 16(%rsp), %rcx movl 24(%rsp), %r8d leaq 48(%rsp), %r9 movl $_Z8mykernelv, %edi pushq (%rsp) .cfi_adjust_cfa_offset 8 pushq 16(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $16, %rsp .cfi_adjust_cfa_offset -16 callq hipDeviceReset .Lfunc_end1: .size main, .Lfunc_end1-main .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_ctor .type __hip_module_ctor,@function __hip_module_ctor: # @__hip_module_ctor .cfi_startproc # %bb.0: subq $40, %rsp .cfi_def_cfa_offset 48 cmpq $0, __hip_gpubin_handle(%rip) jne .LBB2_2 # %bb.1: movl $__hip_fatbin_wrapper, %edi callq __hipRegisterFatBinary movq %rax, __hip_gpubin_handle(%rip) .LBB2_2: movq __hip_gpubin_handle(%rip), %rdi xorps %xmm0, %xmm0 movups %xmm0, 16(%rsp) movups %xmm0, (%rsp) movl $_Z8mykernelv, %esi movl $.L__unnamed_1, %edx movl $.L__unnamed_1, %ecx movl $-1, %r8d xorl %r9d, %r9d callq __hipRegisterFunction movl $__hip_module_dtor, %edi addq $40, %rsp .cfi_def_cfa_offset 8 jmp atexit # TAILCALL .Lfunc_end2: .size __hip_module_ctor, .Lfunc_end2-__hip_module_ctor .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_dtor .type __hip_module_dtor,@function __hip_module_dtor: # @__hip_module_dtor .cfi_startproc # %bb.0: movq __hip_gpubin_handle(%rip), %rdi testq %rdi, %rdi je .LBB3_2 # %bb.1: pushq %rax .cfi_def_cfa_offset 16 callq __hipUnregisterFatBinary movq $0, __hip_gpubin_handle(%rip) addq $8, %rsp .cfi_def_cfa_offset 8 .LBB3_2: retq .Lfunc_end3: .size __hip_module_dtor, .Lfunc_end3-__hip_module_dtor .cfi_endproc # -- End function .type _Z8mykernelv,@object # @_Z8mykernelv .section .rodata,"a",@progbits .globl _Z8mykernelv .p2align 3, 0x0 _Z8mykernelv: .quad _Z23__device_stub__mykernelv .size _Z8mykernelv, 8 .type .L__unnamed_1,@object # @0 .section .rodata.str1.1,"aMS",@progbits,1 .L__unnamed_1: .asciz "_Z8mykernelv" .size .L__unnamed_1, 13 .type __hip_fatbin_wrapper,@object # @__hip_fatbin_wrapper .section .hipFatBinSegment,"a",@progbits .p2align 3, 0x0 __hip_fatbin_wrapper: .long 1212764230 # 0x48495046 .long 1 # 0x1 .quad __hip_fatbin .quad 0 .size __hip_fatbin_wrapper, 24 .type __hip_gpubin_handle,@object # @__hip_gpubin_handle .local __hip_gpubin_handle .comm __hip_gpubin_handle,8,8 .section .init_array,"aw",@init_array .p2align 3, 0x0 .quad __hip_module_ctor .type __hip_cuid_,@object # @__hip_cuid_ .bss .globl __hip_cuid_ __hip_cuid_: .byte 0 # 0x0 .size __hip_cuid_, 1 .type .Lstr,@object # @str .section .rodata.str1.1,"aMS",@progbits,1 .Lstr: .asciz "Hello world from CPU" .size .Lstr, 21 .section ".linker-options","e",@llvm_linker_options .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym _Z23__device_stub__mykernelv .addrsig_sym __hip_module_ctor .addrsig_sym __hip_module_dtor .addrsig_sym _Z8mykernelv .addrsig_sym __hip_fatbin .addrsig_sym __hip_fatbin_wrapper .addrsig_sym __hip_cuid_
You are a helpful coding assistant specialized in converting CUDA source code to CUDA device assembly.
#include "cuda_runtime.h" #include "device_launch_parameters.h" #include <iostream> #include <numeric> #include <math.h> using namespace std; #define BLOCK_SIZE 4; __global__ void sum(int* input, int n) // global call to cuda function (host to device) { const int tid = threadIdx.x; // get thread ID int step_size = 1; int number_of_threads = blockDim.x; // initiate step size and number of threads while (number_of_threads > 0) { if (tid < number_of_threads) { const int fst = tid * step_size * 2; const int snd = fst + step_size; // calculate indices of first and second element to be added if(snd < n) { input[fst] += input[snd]; // add elements } } step_size <<= 1; // multiply step size by 2 if(number_of_threads == 1) break; number_of_threads = (int)ceil((float)number_of_threads/2.0); // divide number of threads by 2 __syncthreads(); } } int main() { int count=0; int result; int *d; cout<<"\nEnter the number of elements : "; cin>>count; const int size = count * sizeof(int); int *h; h = new int[count]; cout<<"\nEnter the elements : \n"; for(int i=0;i<count;i++) cin>>h[i]; cudaMalloc(&d, size); // allocate device variable memory cudaMemcpy(d, h, size, cudaMemcpyHostToDevice); // copy array from host to device //cout<<ceil((float)count/2.0); sum <<<1, ceil((float)count/2.0) >>>(d,count); // function call func_name<<<no_of_blocks,no_of_threads>>>(args) cudaMemcpy(&result, d, sizeof(int), cudaMemcpyDeviceToHost); // copy result back from device to host cout << "Sum is " << result << endl; getchar(); cudaFree(d); // free device memory delete[] h; return 0; } /* PS D:\MyFiles\Projects\LP1-LabAsg\2-HPC> nvcc ParRedSum.cu -o ParRedSum ParRedSum.cu Creating library ParRedSum.lib and object ParRedSum.exp PS D:\MyFiles\Projects\LP1-LabAsg\2-HPC> nvprof ./ParRedSum Enter the number of elements : 4 Enter the elements : 2 49 12 54 ==4900== NVPROF is profiling process 4900, command: ./ParRedSum Sum is 117 ==4900== Profiling application: ./ParRedSum ==4900== Profiling result: Type Time(%) Time Calls Avg Min Max Name GPU activities: 60.16% 2.4640us 1 2.4640us 2.4640us 2.4640us sum(int*, int) 25.00% 1.0240us 1 1.0240us 1.0240us 1.0240us [CUDA memcpy HtoD] 14.84% 608ns 1 608ns 608ns 608ns [CUDA memcpy DtoH] API calls: 82.75% 203.24ms 1 203.24ms 203.24ms 203.24ms cudaMalloc 16.83% 41.338ms 1 41.338ms 41.338ms 41.338ms cuDevicePrimaryCtxRelease 0.16% 392.40us 97 4.0450us 100ns 220.50us cuDeviceGetAttribute 0.10% 243.20us 1 243.20us 243.20us 243.20us cudaFree 0.07% 170.80us 2 85.400us 62.900us 107.90us cudaMemcpy 0.06% 151.20us 1 151.20us 151.20us 151.20us cuModuleUnload 0.01% 29.200us 1 29.200us 29.200us 29.200us cudaLaunchKernel 0.01% 18.800us 1 18.800us 18.800us 18.800us cuDeviceTotalMem 0.00% 9.8000us 1 9.8000us 9.8000us 9.8000us cuDeviceGetPCIBusId 0.00% 1.3000us 3 433ns 200ns 800ns cuDeviceGetCount 0.00% 900ns 1 900ns 900ns 900ns cuDeviceGetName 0.00% 700ns 2 350ns 100ns 600ns cuDeviceGet 0.00% 400ns 1 400ns 400ns 400ns cuDeviceGetLuid 0.00% 200ns 1 200ns 200ns 200ns cuDeviceGetUuid */
code for sm_80 Function : _Z3sumPii .headerflags @"EF_CUDA_TEXMODE_UNIFIED EF_CUDA_64BIT_ADDRESS EF_CUDA_SM80 EF_CUDA_VIRTUAL_SM(EF_CUDA_SM80)" /*0000*/ IMAD.MOV.U32 R1, RZ, RZ, c[0x0][0x28] ; /* 0x00000a00ff017624 */ /* 0x000fe400078e00ff */ /*0010*/ IMAD.MOV.U32 R0, RZ, RZ, c[0x0][0x0] ; /* 0x00000000ff007624 */ /* 0x000fca00078e00ff */ /*0020*/ ISETP.GE.AND P0, PT, R0, 0x1, PT ; /* 0x000000010000780c */ /* 0x000fda0003f06270 */ /*0030*/ @!P0 EXIT ; /* 0x000000000000894d */ /* 0x000fea0003800000 */ /*0040*/ S2R R9, SR_TID.X ; /* 0x0000000000097919 */ /* 0x000e220000002100 */ /*0050*/ MOV R0, c[0x0][0x0] ; /* 0x0000000000007a02 */ /* 0x000fe20000000f00 */ /*0060*/ UMOV UR4, 0x1 ; /* 0x0000000100047882 */ /* 0x000fe40000000000 */ /*0070*/ ULDC.64 UR6, c[0x0][0x118] ; /* 0x0000460000067ab9 */ /* 0x000fe20000000a00 */ /*0080*/ IMAD.SHL.U32 R8, R9, 0x2, RZ ; /* 0x0000000209087824 */ /* 0x001fe400078e00ff */ /*0090*/ ISETP.GE.AND P1, PT, R9, R0, PT ; /* 0x000000000900720c */ /* 0x000fe20003f26270 */ /*00a0*/ BSSY B0, 0x190 ; /* 0x000000e000007945 */ /* 0x000fe20003800000 */ /*00b0*/ ISETP.NE.AND P0, PT, R0, 0x1, PT ; /* 0x000000010000780c */ /* 0x000fd60003f05270 */ /*00c0*/ @P1 BRA 0x180 ; /* 0x000000b000001947 */ /* 0x000fea0003800000 */ /*00d0*/ IMAD R5, R8, UR4, RZ ; /* 0x0000000408057c24 */ /* 0x000fca000f8e02ff */ /*00e0*/ IADD3 R3, R5, UR4, RZ ; /* 0x0000000405037c10 */ /* 0x000fc8000fffe0ff */ /*00f0*/ ISETP.GE.AND P1, PT, R3, c[0x0][0x168], PT ; /* 0x00005a0003007a0c */ /* 0x000fda0003f26270 */ /*0100*/ @P1 BRA 0x180 ; /* 0x0000007000001947 */ /* 0x000fea0003800000 */ /*0110*/ HFMA2.MMA R4, -RZ, RZ, 0, 2.384185791015625e-07 ; /* 0x00000004ff047435 */ /* 0x000fd400000001ff */ /*0120*/ IMAD.WIDE R2, R3, R4, c[0x0][0x160] ; /* 0x0000580003027625 */ /* 0x000fc800078e0204 */ /*0130*/ IMAD.WIDE R4, R5, R4, c[0x0][0x160] ; /* 0x0000580005047625 */ /* 0x000fe400078e0204 */ /*0140*/ LDG.E R3, [R2.64] ; /* 0x0000000602037981 */ /* 0x000ea8000c1e1900 */ /*0150*/ LDG.E R6, [R4.64] ; /* 0x0000000604067981 */ /* 0x000ea4000c1e1900 */ /*0160*/ IMAD.IADD R7, R6, 0x1, R3 ; /* 0x0000000106077824 */ /* 0x004fca00078e0203 */ /*0170*/ STG.E [R4.64], R7 ; /* 0x0000000704007986 */ /* 0x0001e4000c101906 */ /*0180*/ BSYNC B0 ; /* 0x0000000000007941 */ /* 0x000fea0003800000 */ /*0190*/ @!P0 EXIT ; /* 0x000000000000894d */ /* 0x000fea0003800000 */ /*01a0*/ I2F R4, R0 ; /* 0x0000000000047306 */ /* 0x001e220000201400 */ /*01b0*/ BAR.SYNC.DEFER_BLOCKING 0x0 ; /* 0x0000000000007b1d */ /* 0x000fe20000010000 */ /*01c0*/ USHF.L.U32 UR4, UR4, 0x1, URZ ; /* 0x0000000104047899 */ /* 0x000fcc000800063f */ /*01d0*/ F2F.F64.F32 R2, R4 ; /* 0x0000000400027310 */ /* 0x001e240000201800 */ /*01e0*/ DMUL R2, R2, 0.5 ; /* 0x3fe0000002027828 */ /* 0x001e0c0000000000 */ /*01f0*/ F2I.F64.CEIL R0, R2 ; /* 0x0000000200007311 */ /* 0x001e240000309100 */ /*0200*/ ISETP.GT.AND P0, PT, R0, RZ, PT ; /* 0x000000ff0000720c */ /* 0x001fda0003f04270 */ /*0210*/ @P0 BRA 0x90 ; /* 0xfffffe7000000947 */ /* 0x000fea000383ffff */ /*0220*/ EXIT ; /* 0x000000000000794d */ /* 0x000fea0003800000 */ /*0230*/ BRA 0x230; /* 0xfffffff000007947 */ /* 0x000fc0000383ffff */ /*0240*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0250*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0260*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0270*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0280*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0290*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*02a0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*02b0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*02c0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*02d0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*02e0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*02f0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ ..........
You are a helpful coding assistant specialized in converting CUDA source code to CUDA host assembly.
#include "cuda_runtime.h" #include "device_launch_parameters.h" #include <iostream> #include <numeric> #include <math.h> using namespace std; #define BLOCK_SIZE 4; __global__ void sum(int* input, int n) // global call to cuda function (host to device) { const int tid = threadIdx.x; // get thread ID int step_size = 1; int number_of_threads = blockDim.x; // initiate step size and number of threads while (number_of_threads > 0) { if (tid < number_of_threads) { const int fst = tid * step_size * 2; const int snd = fst + step_size; // calculate indices of first and second element to be added if(snd < n) { input[fst] += input[snd]; // add elements } } step_size <<= 1; // multiply step size by 2 if(number_of_threads == 1) break; number_of_threads = (int)ceil((float)number_of_threads/2.0); // divide number of threads by 2 __syncthreads(); } } int main() { int count=0; int result; int *d; cout<<"\nEnter the number of elements : "; cin>>count; const int size = count * sizeof(int); int *h; h = new int[count]; cout<<"\nEnter the elements : \n"; for(int i=0;i<count;i++) cin>>h[i]; cudaMalloc(&d, size); // allocate device variable memory cudaMemcpy(d, h, size, cudaMemcpyHostToDevice); // copy array from host to device //cout<<ceil((float)count/2.0); sum <<<1, ceil((float)count/2.0) >>>(d,count); // function call func_name<<<no_of_blocks,no_of_threads>>>(args) cudaMemcpy(&result, d, sizeof(int), cudaMemcpyDeviceToHost); // copy result back from device to host cout << "Sum is " << result << endl; getchar(); cudaFree(d); // free device memory delete[] h; return 0; } /* PS D:\MyFiles\Projects\LP1-LabAsg\2-HPC> nvcc ParRedSum.cu -o ParRedSum ParRedSum.cu Creating library ParRedSum.lib and object ParRedSum.exp PS D:\MyFiles\Projects\LP1-LabAsg\2-HPC> nvprof ./ParRedSum Enter the number of elements : 4 Enter the elements : 2 49 12 54 ==4900== NVPROF is profiling process 4900, command: ./ParRedSum Sum is 117 ==4900== Profiling application: ./ParRedSum ==4900== Profiling result: Type Time(%) Time Calls Avg Min Max Name GPU activities: 60.16% 2.4640us 1 2.4640us 2.4640us 2.4640us sum(int*, int) 25.00% 1.0240us 1 1.0240us 1.0240us 1.0240us [CUDA memcpy HtoD] 14.84% 608ns 1 608ns 608ns 608ns [CUDA memcpy DtoH] API calls: 82.75% 203.24ms 1 203.24ms 203.24ms 203.24ms cudaMalloc 16.83% 41.338ms 1 41.338ms 41.338ms 41.338ms cuDevicePrimaryCtxRelease 0.16% 392.40us 97 4.0450us 100ns 220.50us cuDeviceGetAttribute 0.10% 243.20us 1 243.20us 243.20us 243.20us cudaFree 0.07% 170.80us 2 85.400us 62.900us 107.90us cudaMemcpy 0.06% 151.20us 1 151.20us 151.20us 151.20us cuModuleUnload 0.01% 29.200us 1 29.200us 29.200us 29.200us cudaLaunchKernel 0.01% 18.800us 1 18.800us 18.800us 18.800us cuDeviceTotalMem 0.00% 9.8000us 1 9.8000us 9.8000us 9.8000us cuDeviceGetPCIBusId 0.00% 1.3000us 3 433ns 200ns 800ns cuDeviceGetCount 0.00% 900ns 1 900ns 900ns 900ns cuDeviceGetName 0.00% 700ns 2 350ns 100ns 600ns cuDeviceGet 0.00% 400ns 1 400ns 400ns 400ns cuDeviceGetLuid 0.00% 200ns 1 200ns 200ns 200ns cuDeviceGetUuid */
.file "tmpxft_0002ac37_00000000-6_ParRedSum.cudafe1.cpp" .text #APP .globl _ZSt21ios_base_library_initv #NO_APP .type _ZL26__cudaUnregisterBinaryUtilv, @function _ZL26__cudaUnregisterBinaryUtilv: .LFB3711: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaUnregisterFatBinary@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE3711: .size _ZL26__cudaUnregisterBinaryUtilv, .-_ZL26__cudaUnregisterBinaryUtilv .globl _Z23__device_stub__Z3sumPiiPii .type _Z23__device_stub__Z3sumPiiPii, @function _Z23__device_stub__Z3sumPiiPii: .LFB3733: .cfi_startproc endbr64 subq $120, %rsp .cfi_def_cfa_offset 128 movq %rdi, 8(%rsp) movl %esi, 4(%rsp) movq %fs:40, %rax movq %rax, 104(%rsp) xorl %eax, %eax leaq 8(%rsp), %rax movq %rax, 80(%rsp) leaq 4(%rsp), %rax movq %rax, 88(%rsp) movl $1, 32(%rsp) movl $1, 36(%rsp) movl $1, 40(%rsp) movl $1, 44(%rsp) movl $1, 48(%rsp) movl $1, 52(%rsp) leaq 24(%rsp), %rcx leaq 16(%rsp), %rdx leaq 44(%rsp), %rsi leaq 32(%rsp), %rdi call __cudaPopCallConfiguration@PLT testl %eax, %eax je .L7 .L3: movq 104(%rsp), %rax subq %fs:40, %rax jne .L8 addq $120, %rsp .cfi_remember_state .cfi_def_cfa_offset 8 ret .L7: .cfi_restore_state pushq 24(%rsp) .cfi_def_cfa_offset 136 pushq 24(%rsp) .cfi_def_cfa_offset 144 leaq 96(%rsp), %r9 movq 60(%rsp), %rcx movl 68(%rsp), %r8d movq 48(%rsp), %rsi movl 56(%rsp), %edx leaq _Z3sumPii(%rip), %rdi call cudaLaunchKernel@PLT addq $16, %rsp .cfi_def_cfa_offset 128 jmp .L3 .L8: call __stack_chk_fail@PLT .cfi_endproc .LFE3733: .size _Z23__device_stub__Z3sumPiiPii, .-_Z23__device_stub__Z3sumPiiPii .globl _Z3sumPii .type _Z3sumPii, @function _Z3sumPii: .LFB3734: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 call _Z23__device_stub__Z3sumPiiPii addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE3734: .size _Z3sumPii, .-_Z3sumPii .section .rodata.str1.8,"aMS",@progbits,1 .align 8 .LC0: .string "\nEnter the number of elements : " .section .rodata.str1.1,"aMS",@progbits,1 .LC1: .string "\nEnter the elements : \n" .LC6: .string "Sum is " .text .globl main .type main, @function main: .LFB3708: .cfi_startproc endbr64 pushq %r14 .cfi_def_cfa_offset 16 .cfi_offset 14, -16 pushq %r13 .cfi_def_cfa_offset 24 .cfi_offset 13, -24 pushq %r12 .cfi_def_cfa_offset 32 .cfi_offset 12, -32 pushq %rbp .cfi_def_cfa_offset 40 .cfi_offset 6, -40 pushq %rbx .cfi_def_cfa_offset 48 .cfi_offset 3, -48 subq $48, %rsp .cfi_def_cfa_offset 96 movq %fs:40, %rax movq %rax, 40(%rsp) xorl %eax, %eax movl $0, 4(%rsp) leaq .LC0(%rip), %rsi leaq _ZSt4cout(%rip), %rdi call _ZStlsISt11char_traitsIcEERSt13basic_ostreamIcT_ES5_PKc@PLT leaq 4(%rsp), %rsi leaq _ZSt3cin(%rip), %rdi call _ZNSirsERi@PLT movl 4(%rsp), %ebx movslq %ebx, %rdi movabsq $2305843009213693950, %rax cmpq %rdi, %rax jb .L12 leal 0(,%rbx,4), %r14d salq $2, %rdi call _Znam@PLT movq %rax, %r13 leaq .LC1(%rip), %rsi leaq _ZSt4cout(%rip), %rdi call _ZStlsISt11char_traitsIcEERSt13basic_ostreamIcT_ES5_PKc@PLT movq %r13, %rbp movl $0, %ebx leaq _ZSt3cin(%rip), %r12 cmpl $0, 4(%rsp) jle .L14 .L16: movq %rbp, %rsi movq %r12, %rdi call _ZNSirsERi@PLT addl $1, %ebx addq $4, %rbp cmpl %ebx, 4(%rsp) jg .L16 .L14: movslq %r14d, %r14 leaq 8(%rsp), %rdi movq %r14, %rsi call cudaMalloc@PLT movl $1, %ecx movq %r14, %rdx movq %r13, %rsi movq 8(%rsp), %rdi call cudaMemcpy@PLT pxor %xmm0, %xmm0 cvtsi2ssl 4(%rsp), %xmm0 cvtss2sd %xmm0, %xmm0 mulsd .LC2(%rip), %xmm0 movapd %xmm0, %xmm3 movsd .LC7(%rip), %xmm2 movapd %xmm0, %xmm1 andpd %xmm2, %xmm1 movsd .LC3(%rip), %xmm4 ucomisd %xmm1, %xmm4 jbe .L17 cvttsd2siq %xmm0, %rax pxor %xmm1, %xmm1 cvtsi2sdq %rax, %xmm1 cmpnlesd %xmm1, %xmm3 movsd .LC5(%rip), %xmm4 andpd %xmm4, %xmm3 addsd %xmm1, %xmm3 andnpd %xmm0, %xmm2 orpd %xmm2, %xmm3 .L17: cvttsd2siq %xmm3, %rax movl %eax, 28(%rsp) movl $1, 32(%rsp) movl $1, 36(%rsp) movl $1, 16(%rsp) movl $1, 20(%rsp) movl $0, %r9d movl $0, %r8d movq 28(%rsp), %rdx movl $1, %ecx movq 16(%rsp), %rdi movl $1, %esi call __cudaPushCallConfiguration@PLT testl %eax, %eax je .L23 .L18: leaq 28(%rsp), %rdi movl $2, %ecx movl $4, %edx movq 8(%rsp), %rsi call cudaMemcpy@PLT leaq .LC6(%rip), %rsi leaq _ZSt4cout(%rip), %rdi call _ZStlsISt11char_traitsIcEERSt13basic_ostreamIcT_ES5_PKc@PLT movq %rax, %rdi movl 28(%rsp), %esi call _ZNSolsEi@PLT movq %rax, %rdi call _ZSt4endlIcSt11char_traitsIcEERSt13basic_ostreamIT_T0_ES6_@PLT movq stdin(%rip), %rdi call getc@PLT movq 8(%rsp), %rdi call cudaFree@PLT movq %r13, %rdi call _ZdaPv@PLT movq 40(%rsp), %rax subq %fs:40, %rax jne .L24 movl $0, %eax addq $48, %rsp .cfi_remember_state .cfi_def_cfa_offset 48 popq %rbx .cfi_def_cfa_offset 40 popq %rbp .cfi_def_cfa_offset 32 popq %r12 .cfi_def_cfa_offset 24 popq %r13 .cfi_def_cfa_offset 16 popq %r14 .cfi_def_cfa_offset 8 ret .L12: .cfi_restore_state movq 40(%rsp), %rax subq %fs:40, %rax je .L15 call __stack_chk_fail@PLT .L15: call __cxa_throw_bad_array_new_length@PLT .L23: movl 4(%rsp), %esi movq 8(%rsp), %rdi call _Z23__device_stub__Z3sumPiiPii jmp .L18 .L24: call __stack_chk_fail@PLT .cfi_endproc .LFE3708: .size main, .-main .section .rodata.str1.1 .LC8: .string "_Z3sumPii" .text .type _ZL24__sti____cudaRegisterAllv, @function _ZL24__sti____cudaRegisterAllv: .LFB3736: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 leaq _ZL15__fatDeviceText(%rip), %rdi call __cudaRegisterFatBinary@PLT movq %rax, %rdi movq %rax, _ZL20__cudaFatCubinHandle(%rip) pushq $0 .cfi_def_cfa_offset 24 pushq $0 .cfi_def_cfa_offset 32 pushq $0 .cfi_def_cfa_offset 40 pushq $0 .cfi_def_cfa_offset 48 movl $0, %r9d movl $-1, %r8d leaq .LC8(%rip), %rdx movq %rdx, %rcx leaq _Z3sumPii(%rip), %rsi call __cudaRegisterFunction@PLT addq $32, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaRegisterFatBinaryEnd@PLT leaq _ZL26__cudaUnregisterBinaryUtilv(%rip), %rdi call atexit@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE3736: .size _ZL24__sti____cudaRegisterAllv, .-_ZL24__sti____cudaRegisterAllv .section .init_array,"aw" .align 8 .quad _ZL24__sti____cudaRegisterAllv .section .nvFatBinSegment,"aw" .align 8 .type _ZL15__fatDeviceText, @object .size _ZL15__fatDeviceText, 24 _ZL15__fatDeviceText: .long 1180844977 .long 1 .quad fatbinData .quad 0 .local _ZL20__cudaFatCubinHandle .comm _ZL20__cudaFatCubinHandle,8,8 .section .rodata.cst8,"aM",@progbits,8 .align 8 .LC2: .long 0 .long 1071644672 .align 8 .LC3: .long 0 .long 1127219200 .align 8 .LC5: .long 0 .long 1072693248 .align 8 .LC7: .long -1 .long 2147483647 .ident "GCC: (Ubuntu 13.3.0-6ubuntu2~24.04) 13.3.0" .section .note.GNU-stack,"",@progbits .section .note.gnu.property,"a" .align 8 .long 1f - 0f .long 4f - 1f .long 5 0: .string "GNU" 1: .align 8 .long 0xc0000002 .long 3f - 2f 2: .long 0x3 3: .align 8 4:
You are a helpful coding assistant specialized in converting CUDA source code to HIP/AMD source code.
#include "cuda_runtime.h" #include "device_launch_parameters.h" #include <iostream> #include <numeric> #include <math.h> using namespace std; #define BLOCK_SIZE 4; __global__ void sum(int* input, int n) // global call to cuda function (host to device) { const int tid = threadIdx.x; // get thread ID int step_size = 1; int number_of_threads = blockDim.x; // initiate step size and number of threads while (number_of_threads > 0) { if (tid < number_of_threads) { const int fst = tid * step_size * 2; const int snd = fst + step_size; // calculate indices of first and second element to be added if(snd < n) { input[fst] += input[snd]; // add elements } } step_size <<= 1; // multiply step size by 2 if(number_of_threads == 1) break; number_of_threads = (int)ceil((float)number_of_threads/2.0); // divide number of threads by 2 __syncthreads(); } } int main() { int count=0; int result; int *d; cout<<"\nEnter the number of elements : "; cin>>count; const int size = count * sizeof(int); int *h; h = new int[count]; cout<<"\nEnter the elements : \n"; for(int i=0;i<count;i++) cin>>h[i]; cudaMalloc(&d, size); // allocate device variable memory cudaMemcpy(d, h, size, cudaMemcpyHostToDevice); // copy array from host to device //cout<<ceil((float)count/2.0); sum <<<1, ceil((float)count/2.0) >>>(d,count); // function call func_name<<<no_of_blocks,no_of_threads>>>(args) cudaMemcpy(&result, d, sizeof(int), cudaMemcpyDeviceToHost); // copy result back from device to host cout << "Sum is " << result << endl; getchar(); cudaFree(d); // free device memory delete[] h; return 0; } /* PS D:\MyFiles\Projects\LP1-LabAsg\2-HPC> nvcc ParRedSum.cu -o ParRedSum ParRedSum.cu Creating library ParRedSum.lib and object ParRedSum.exp PS D:\MyFiles\Projects\LP1-LabAsg\2-HPC> nvprof ./ParRedSum Enter the number of elements : 4 Enter the elements : 2 49 12 54 ==4900== NVPROF is profiling process 4900, command: ./ParRedSum Sum is 117 ==4900== Profiling application: ./ParRedSum ==4900== Profiling result: Type Time(%) Time Calls Avg Min Max Name GPU activities: 60.16% 2.4640us 1 2.4640us 2.4640us 2.4640us sum(int*, int) 25.00% 1.0240us 1 1.0240us 1.0240us 1.0240us [CUDA memcpy HtoD] 14.84% 608ns 1 608ns 608ns 608ns [CUDA memcpy DtoH] API calls: 82.75% 203.24ms 1 203.24ms 203.24ms 203.24ms cudaMalloc 16.83% 41.338ms 1 41.338ms 41.338ms 41.338ms cuDevicePrimaryCtxRelease 0.16% 392.40us 97 4.0450us 100ns 220.50us cuDeviceGetAttribute 0.10% 243.20us 1 243.20us 243.20us 243.20us cudaFree 0.07% 170.80us 2 85.400us 62.900us 107.90us cudaMemcpy 0.06% 151.20us 1 151.20us 151.20us 151.20us cuModuleUnload 0.01% 29.200us 1 29.200us 29.200us 29.200us cudaLaunchKernel 0.01% 18.800us 1 18.800us 18.800us 18.800us cuDeviceTotalMem 0.00% 9.8000us 1 9.8000us 9.8000us 9.8000us cuDeviceGetPCIBusId 0.00% 1.3000us 3 433ns 200ns 800ns cuDeviceGetCount 0.00% 900ns 1 900ns 900ns 900ns cuDeviceGetName 0.00% 700ns 2 350ns 100ns 600ns cuDeviceGet 0.00% 400ns 1 400ns 400ns 400ns cuDeviceGetLuid 0.00% 200ns 1 200ns 200ns 200ns cuDeviceGetUuid */
#include "hip/hip_runtime.h" #include <iostream> #include <numeric> #include <math.h> using namespace std; #define BLOCK_SIZE 4; __global__ void sum(int* input, int n) // global call to cuda function (host to device) { const int tid = threadIdx.x; // get thread ID int step_size = 1; int number_of_threads = blockDim.x; // initiate step size and number of threads while (number_of_threads > 0) { if (tid < number_of_threads) { const int fst = tid * step_size * 2; const int snd = fst + step_size; // calculate indices of first and second element to be added if(snd < n) { input[fst] += input[snd]; // add elements } } step_size <<= 1; // multiply step size by 2 if(number_of_threads == 1) break; number_of_threads = (int)ceil((float)number_of_threads/2.0); // divide number of threads by 2 __syncthreads(); } } int main() { int count=0; int result; int *d; cout<<"\nEnter the number of elements : "; cin>>count; const int size = count * sizeof(int); int *h; h = new int[count]; cout<<"\nEnter the elements : \n"; for(int i=0;i<count;i++) cin>>h[i]; hipMalloc(&d, size); // allocate device variable memory hipMemcpy(d, h, size, hipMemcpyHostToDevice); // copy array from host to device //cout<<ceil((float)count/2.0); sum <<<1, ceil((float)count/2.0) >>>(d,count); // function call func_name<<<no_of_blocks,no_of_threads>>>(args) hipMemcpy(&result, d, sizeof(int), hipMemcpyDeviceToHost); // copy result back from device to host cout << "Sum is " << result << endl; getchar(); hipFree(d); // free device memory delete[] h; return 0; } /* PS D:\MyFiles\Projects\LP1-LabAsg\2-HPC> nvcc ParRedSum.cu -o ParRedSum ParRedSum.cu Creating library ParRedSum.lib and object ParRedSum.exp PS D:\MyFiles\Projects\LP1-LabAsg\2-HPC> nvprof ./ParRedSum Enter the number of elements : 4 Enter the elements : 2 49 12 54 ==4900== NVPROF is profiling process 4900, command: ./ParRedSum Sum is 117 ==4900== Profiling application: ./ParRedSum ==4900== Profiling result: Type Time(%) Time Calls Avg Min Max Name GPU activities: 60.16% 2.4640us 1 2.4640us 2.4640us 2.4640us sum(int*, int) 25.00% 1.0240us 1 1.0240us 1.0240us 1.0240us [CUDA memcpy HtoD] 14.84% 608ns 1 608ns 608ns 608ns [CUDA memcpy DtoH] API calls: 82.75% 203.24ms 1 203.24ms 203.24ms 203.24ms cudaMalloc 16.83% 41.338ms 1 41.338ms 41.338ms 41.338ms cuDevicePrimaryCtxRelease 0.16% 392.40us 97 4.0450us 100ns 220.50us cuDeviceGetAttribute 0.10% 243.20us 1 243.20us 243.20us 243.20us cudaFree 0.07% 170.80us 2 85.400us 62.900us 107.90us cudaMemcpy 0.06% 151.20us 1 151.20us 151.20us 151.20us cuModuleUnload 0.01% 29.200us 1 29.200us 29.200us 29.200us cudaLaunchKernel 0.01% 18.800us 1 18.800us 18.800us 18.800us cuDeviceTotalMem 0.00% 9.8000us 1 9.8000us 9.8000us 9.8000us cuDeviceGetPCIBusId 0.00% 1.3000us 3 433ns 200ns 800ns cuDeviceGetCount 0.00% 900ns 1 900ns 900ns 900ns cuDeviceGetName 0.00% 700ns 2 350ns 100ns 600ns cuDeviceGet 0.00% 400ns 1 400ns 400ns 400ns cuDeviceGetLuid 0.00% 200ns 1 200ns 200ns 200ns cuDeviceGetUuid */
You are a helpful coding assistant specialized in converting HIP source code to AMD device assembly.
#include "hip/hip_runtime.h" #include <iostream> #include <numeric> #include <math.h> using namespace std; #define BLOCK_SIZE 4; __global__ void sum(int* input, int n) // global call to cuda function (host to device) { const int tid = threadIdx.x; // get thread ID int step_size = 1; int number_of_threads = blockDim.x; // initiate step size and number of threads while (number_of_threads > 0) { if (tid < number_of_threads) { const int fst = tid * step_size * 2; const int snd = fst + step_size; // calculate indices of first and second element to be added if(snd < n) { input[fst] += input[snd]; // add elements } } step_size <<= 1; // multiply step size by 2 if(number_of_threads == 1) break; number_of_threads = (int)ceil((float)number_of_threads/2.0); // divide number of threads by 2 __syncthreads(); } } int main() { int count=0; int result; int *d; cout<<"\nEnter the number of elements : "; cin>>count; const int size = count * sizeof(int); int *h; h = new int[count]; cout<<"\nEnter the elements : \n"; for(int i=0;i<count;i++) cin>>h[i]; hipMalloc(&d, size); // allocate device variable memory hipMemcpy(d, h, size, hipMemcpyHostToDevice); // copy array from host to device //cout<<ceil((float)count/2.0); sum <<<1, ceil((float)count/2.0) >>>(d,count); // function call func_name<<<no_of_blocks,no_of_threads>>>(args) hipMemcpy(&result, d, sizeof(int), hipMemcpyDeviceToHost); // copy result back from device to host cout << "Sum is " << result << endl; getchar(); hipFree(d); // free device memory delete[] h; return 0; } /* PS D:\MyFiles\Projects\LP1-LabAsg\2-HPC> nvcc ParRedSum.cu -o ParRedSum ParRedSum.cu Creating library ParRedSum.lib and object ParRedSum.exp PS D:\MyFiles\Projects\LP1-LabAsg\2-HPC> nvprof ./ParRedSum Enter the number of elements : 4 Enter the elements : 2 49 12 54 ==4900== NVPROF is profiling process 4900, command: ./ParRedSum Sum is 117 ==4900== Profiling application: ./ParRedSum ==4900== Profiling result: Type Time(%) Time Calls Avg Min Max Name GPU activities: 60.16% 2.4640us 1 2.4640us 2.4640us 2.4640us sum(int*, int) 25.00% 1.0240us 1 1.0240us 1.0240us 1.0240us [CUDA memcpy HtoD] 14.84% 608ns 1 608ns 608ns 608ns [CUDA memcpy DtoH] API calls: 82.75% 203.24ms 1 203.24ms 203.24ms 203.24ms cudaMalloc 16.83% 41.338ms 1 41.338ms 41.338ms 41.338ms cuDevicePrimaryCtxRelease 0.16% 392.40us 97 4.0450us 100ns 220.50us cuDeviceGetAttribute 0.10% 243.20us 1 243.20us 243.20us 243.20us cudaFree 0.07% 170.80us 2 85.400us 62.900us 107.90us cudaMemcpy 0.06% 151.20us 1 151.20us 151.20us 151.20us cuModuleUnload 0.01% 29.200us 1 29.200us 29.200us 29.200us cudaLaunchKernel 0.01% 18.800us 1 18.800us 18.800us 18.800us cuDeviceTotalMem 0.00% 9.8000us 1 9.8000us 9.8000us 9.8000us cuDeviceGetPCIBusId 0.00% 1.3000us 3 433ns 200ns 800ns cuDeviceGetCount 0.00% 900ns 1 900ns 900ns 900ns cuDeviceGetName 0.00% 700ns 2 350ns 100ns 600ns cuDeviceGet 0.00% 400ns 1 400ns 400ns 400ns cuDeviceGetLuid 0.00% 200ns 1 200ns 200ns 200ns cuDeviceGetUuid */
.text .amdgcn_target "amdgcn-amd-amdhsa--gfx1100" .protected _Z3sumPii .globl _Z3sumPii .p2align 8 .type _Z3sumPii,@function _Z3sumPii: s_load_b32 s4, s[0:1], 0x1c s_waitcnt lgkmcnt(0) v_cmp_eq_u16_e64 s2, s4, 0 s_delay_alu instid0(VALU_DEP_1) s_and_b32 vcc_lo, exec_lo, s2 s_cbranch_vccnz .LBB0_7 s_clause 0x1 s_load_b64 s[2:3], s[0:1], 0x0 s_load_b32 s1, s[0:1], 0x8 s_and_b32 s0, 0xffff, s4 s_delay_alu instid0(SALU_CYCLE_1) v_dual_mov_b32 v6, s0 :: v_dual_lshlrev_b32 v5, 1, v0 s_mov_b32 s4, 1 s_branch .LBB0_3 .p2align 6 .LBB0_2: v_cvt_f32_i32_e32 v1, v6 s_lshl_b32 s4, s4, 1 s_waitcnt_vscnt null, 0x0 s_barrier buffer_gl0_inv v_cvt_f64_f32_e32 v[1:2], v1 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_mul_f64 v[1:2], v[1:2], 0.5 v_ceil_f64_e32 v[1:2], v[1:2] s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_cvt_i32_f64_e32 v6, v[1:2] v_cmp_gt_i32_e64 s0, 1, v6 s_delay_alu instid0(VALU_DEP_1) s_and_not1_b32 vcc_lo, exec_lo, s0 s_cbranch_vccz .LBB0_7 .LBB0_3: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2) v_mul_lo_u32 v1, v5, s4 v_cmp_lt_u32_e32 vcc_lo, v0, v6 s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_1) v_add_nc_u32_e32 v3, s4, v1 s_waitcnt lgkmcnt(0) v_cmp_gt_i32_e64 s0, s1, v3 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(SALU_CYCLE_1) s_and_b32 s5, vcc_lo, s0 s_and_saveexec_b32 s0, s5 s_cbranch_execz .LBB0_5 v_ashrrev_i32_e32 v4, 31, v3 v_ashrrev_i32_e32 v2, 31, v1 s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_lshlrev_b64 v[3:4], 2, v[3:4] v_lshlrev_b64 v[1:2], 2, v[1:2] s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3) v_add_co_u32 v3, vcc_lo, s2, v3 v_add_co_ci_u32_e32 v4, vcc_lo, s3, v4, vcc_lo s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_4) v_add_co_u32 v1, vcc_lo, s2, v1 v_add_co_ci_u32_e32 v2, vcc_lo, s3, v2, vcc_lo s_clause 0x1 global_load_b32 v3, v[3:4], off global_load_b32 v4, v[1:2], off s_waitcnt vmcnt(0) v_add_nc_u32_e32 v3, v4, v3 global_store_b32 v[1:2], v3, off .LBB0_5: s_or_b32 exec_lo, exec_lo, s0 v_cmp_eq_u32_e32 vcc_lo, 1, v6 s_cbranch_vccz .LBB0_2 .LBB0_7: s_nop 0 s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) s_endpgm .section .rodata,"a",@progbits .p2align 6, 0x0 .amdhsa_kernel _Z3sumPii .amdhsa_group_segment_fixed_size 0 .amdhsa_private_segment_fixed_size 0 .amdhsa_kernarg_size 272 .amdhsa_user_sgpr_count 15 .amdhsa_user_sgpr_dispatch_ptr 0 .amdhsa_user_sgpr_queue_ptr 0 .amdhsa_user_sgpr_kernarg_segment_ptr 1 .amdhsa_user_sgpr_dispatch_id 0 .amdhsa_user_sgpr_private_segment_size 0 .amdhsa_wavefront_size32 1 .amdhsa_uses_dynamic_stack 0 .amdhsa_enable_private_segment 0 .amdhsa_system_sgpr_workgroup_id_x 1 .amdhsa_system_sgpr_workgroup_id_y 0 .amdhsa_system_sgpr_workgroup_id_z 0 .amdhsa_system_sgpr_workgroup_info 0 .amdhsa_system_vgpr_workitem_id 0 .amdhsa_next_free_vgpr 7 .amdhsa_next_free_sgpr 6 .amdhsa_float_round_mode_32 0 .amdhsa_float_round_mode_16_64 0 .amdhsa_float_denorm_mode_32 3 .amdhsa_float_denorm_mode_16_64 3 .amdhsa_dx10_clamp 1 .amdhsa_ieee_mode 1 .amdhsa_fp16_overflow 0 .amdhsa_workgroup_processor_mode 1 .amdhsa_memory_ordered 1 .amdhsa_forward_progress 0 .amdhsa_shared_vgpr_count 0 .amdhsa_exception_fp_ieee_invalid_op 0 .amdhsa_exception_fp_denorm_src 0 .amdhsa_exception_fp_ieee_div_zero 0 .amdhsa_exception_fp_ieee_overflow 0 .amdhsa_exception_fp_ieee_underflow 0 .amdhsa_exception_fp_ieee_inexact 0 .amdhsa_exception_int_div_zero 0 .end_amdhsa_kernel .text .Lfunc_end0: .size _Z3sumPii, .Lfunc_end0-_Z3sumPii .section .AMDGPU.csdata,"",@progbits .text .p2alignl 7, 3214868480 .fill 96, 4, 3214868480 .type __hip_cuid_,@object .section .bss,"aw",@nobits .globl __hip_cuid_ __hip_cuid_: .byte 0 .size __hip_cuid_, 1 .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym __hip_cuid_ .amdgpu_metadata --- amdhsa.kernels: - .args: - .address_space: global .offset: 0 .size: 8 .value_kind: global_buffer - .offset: 8 .size: 4 .value_kind: by_value - .offset: 16 .size: 4 .value_kind: hidden_block_count_x - .offset: 20 .size: 4 .value_kind: hidden_block_count_y - .offset: 24 .size: 4 .value_kind: hidden_block_count_z - .offset: 28 .size: 2 .value_kind: hidden_group_size_x - .offset: 30 .size: 2 .value_kind: hidden_group_size_y - .offset: 32 .size: 2 .value_kind: hidden_group_size_z - .offset: 34 .size: 2 .value_kind: hidden_remainder_x - .offset: 36 .size: 2 .value_kind: hidden_remainder_y - .offset: 38 .size: 2 .value_kind: hidden_remainder_z - .offset: 56 .size: 8 .value_kind: hidden_global_offset_x - .offset: 64 .size: 8 .value_kind: hidden_global_offset_y - .offset: 72 .size: 8 .value_kind: hidden_global_offset_z - .offset: 80 .size: 2 .value_kind: hidden_grid_dims .group_segment_fixed_size: 0 .kernarg_segment_align: 8 .kernarg_segment_size: 272 .language: OpenCL C .language_version: - 2 - 0 .max_flat_workgroup_size: 1024 .name: _Z3sumPii .private_segment_fixed_size: 0 .sgpr_count: 8 .sgpr_spill_count: 0 .symbol: _Z3sumPii.kd .uniform_work_group_size: 1 .uses_dynamic_stack: false .vgpr_count: 7 .vgpr_spill_count: 0 .wavefront_size: 32 .workgroup_processor_mode: 1 amdhsa.target: amdgcn-amd-amdhsa--gfx1100 amdhsa.version: - 1 - 2 ... .end_amdgpu_metadata
You are a helpful coding assistant specialized in converting HIP source code to AMD host assembly.
#include "hip/hip_runtime.h" #include <iostream> #include <numeric> #include <math.h> using namespace std; #define BLOCK_SIZE 4; __global__ void sum(int* input, int n) // global call to cuda function (host to device) { const int tid = threadIdx.x; // get thread ID int step_size = 1; int number_of_threads = blockDim.x; // initiate step size and number of threads while (number_of_threads > 0) { if (tid < number_of_threads) { const int fst = tid * step_size * 2; const int snd = fst + step_size; // calculate indices of first and second element to be added if(snd < n) { input[fst] += input[snd]; // add elements } } step_size <<= 1; // multiply step size by 2 if(number_of_threads == 1) break; number_of_threads = (int)ceil((float)number_of_threads/2.0); // divide number of threads by 2 __syncthreads(); } } int main() { int count=0; int result; int *d; cout<<"\nEnter the number of elements : "; cin>>count; const int size = count * sizeof(int); int *h; h = new int[count]; cout<<"\nEnter the elements : \n"; for(int i=0;i<count;i++) cin>>h[i]; hipMalloc(&d, size); // allocate device variable memory hipMemcpy(d, h, size, hipMemcpyHostToDevice); // copy array from host to device //cout<<ceil((float)count/2.0); sum <<<1, ceil((float)count/2.0) >>>(d,count); // function call func_name<<<no_of_blocks,no_of_threads>>>(args) hipMemcpy(&result, d, sizeof(int), hipMemcpyDeviceToHost); // copy result back from device to host cout << "Sum is " << result << endl; getchar(); hipFree(d); // free device memory delete[] h; return 0; } /* PS D:\MyFiles\Projects\LP1-LabAsg\2-HPC> nvcc ParRedSum.cu -o ParRedSum ParRedSum.cu Creating library ParRedSum.lib and object ParRedSum.exp PS D:\MyFiles\Projects\LP1-LabAsg\2-HPC> nvprof ./ParRedSum Enter the number of elements : 4 Enter the elements : 2 49 12 54 ==4900== NVPROF is profiling process 4900, command: ./ParRedSum Sum is 117 ==4900== Profiling application: ./ParRedSum ==4900== Profiling result: Type Time(%) Time Calls Avg Min Max Name GPU activities: 60.16% 2.4640us 1 2.4640us 2.4640us 2.4640us sum(int*, int) 25.00% 1.0240us 1 1.0240us 1.0240us 1.0240us [CUDA memcpy HtoD] 14.84% 608ns 1 608ns 608ns 608ns [CUDA memcpy DtoH] API calls: 82.75% 203.24ms 1 203.24ms 203.24ms 203.24ms cudaMalloc 16.83% 41.338ms 1 41.338ms 41.338ms 41.338ms cuDevicePrimaryCtxRelease 0.16% 392.40us 97 4.0450us 100ns 220.50us cuDeviceGetAttribute 0.10% 243.20us 1 243.20us 243.20us 243.20us cudaFree 0.07% 170.80us 2 85.400us 62.900us 107.90us cudaMemcpy 0.06% 151.20us 1 151.20us 151.20us 151.20us cuModuleUnload 0.01% 29.200us 1 29.200us 29.200us 29.200us cudaLaunchKernel 0.01% 18.800us 1 18.800us 18.800us 18.800us cuDeviceTotalMem 0.00% 9.8000us 1 9.8000us 9.8000us 9.8000us cuDeviceGetPCIBusId 0.00% 1.3000us 3 433ns 200ns 800ns cuDeviceGetCount 0.00% 900ns 1 900ns 900ns 900ns cuDeviceGetName 0.00% 700ns 2 350ns 100ns 600ns cuDeviceGet 0.00% 400ns 1 400ns 400ns 400ns cuDeviceGetLuid 0.00% 200ns 1 200ns 200ns 200ns cuDeviceGetUuid */
.text .file "ParRedSum.hip" # Start of file scope inline assembly .globl _ZSt21ios_base_library_initv # End of file scope inline assembly .globl _Z18__device_stub__sumPii # -- Begin function _Z18__device_stub__sumPii .p2align 4, 0x90 .type _Z18__device_stub__sumPii,@function _Z18__device_stub__sumPii: # @_Z18__device_stub__sumPii .cfi_startproc # %bb.0: subq $88, %rsp .cfi_def_cfa_offset 96 movq %rdi, 56(%rsp) movl %esi, 4(%rsp) leaq 56(%rsp), %rax movq %rax, 64(%rsp) leaq 4(%rsp), %rax movq %rax, 72(%rsp) leaq 40(%rsp), %rdi leaq 24(%rsp), %rsi leaq 16(%rsp), %rdx leaq 8(%rsp), %rcx callq __hipPopCallConfiguration movq 40(%rsp), %rsi movl 48(%rsp), %edx movq 24(%rsp), %rcx movl 32(%rsp), %r8d leaq 64(%rsp), %r9 movl $_Z3sumPii, %edi pushq 8(%rsp) .cfi_adjust_cfa_offset 8 pushq 24(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $104, %rsp .cfi_adjust_cfa_offset -104 retq .Lfunc_end0: .size _Z18__device_stub__sumPii, .Lfunc_end0-_Z18__device_stub__sumPii .cfi_endproc # -- End function .section .rodata.cst8,"aM",@progbits,8 .p2align 3, 0x0 # -- Begin function main .LCPI1_0: .quad 0x3fe0000000000000 # double 0.5 .text .globl main .p2align 4, 0x90 .type main,@function main: # @main .cfi_startproc # %bb.0: pushq %rbp .cfi_def_cfa_offset 16 pushq %r15 .cfi_def_cfa_offset 24 pushq %r14 .cfi_def_cfa_offset 32 pushq %rbx .cfi_def_cfa_offset 40 subq $104, %rsp .cfi_def_cfa_offset 144 .cfi_offset %rbx, -40 .cfi_offset %r14, -32 .cfi_offset %r15, -24 .cfi_offset %rbp, -16 movl $0, 12(%rsp) movl $_ZSt4cout, %edi movl $.L.str, %esi movl $32, %edx callq _ZSt16__ostream_insertIcSt11char_traitsIcEERSt13basic_ostreamIT_T0_ES6_PKS3_l leaq 12(%rsp), %rsi movl $_ZSt3cin, %edi callq _ZNSirsERi movslq 12(%rsp), %rax leal (,%rax,4), %ebp leaq (,%rax,4), %rcx testq %rax, %rax movq $-1, %rdi cmovnsq %rcx, %rdi callq _Znam movq %rax, %rbx movl $_ZSt4cout, %edi movl $.L.str.1, %esi movl $23, %edx callq _ZSt16__ostream_insertIcSt11char_traitsIcEERSt13basic_ostreamIT_T0_ES6_PKS3_l cmpl $0, 12(%rsp) jle .LBB1_3 # %bb.1: # %.lr.ph.preheader xorl %r15d, %r15d movq %rbx, %r14 .p2align 4, 0x90 .LBB1_2: # %.lr.ph # =>This Inner Loop Header: Depth=1 movl $_ZSt3cin, %edi movq %r14, %rsi callq _ZNSirsERi incq %r15 movslq 12(%rsp), %rax addq $4, %r14 cmpq %rax, %r15 jl .LBB1_2 .LBB1_3: # %._crit_edge movslq %ebp, %r14 leaq 16(%rsp), %rdi movq %r14, %rsi callq hipMalloc movq 16(%rsp), %rdi movq %rbx, %rsi movq %r14, %rdx movl $1, %ecx callq hipMemcpy cvtsi2ssl 12(%rsp), %xmm0 cvtss2sd %xmm0, %xmm0 mulsd .LCPI1_0(%rip), %xmm0 callq ceil@PLT cvttsd2si %xmm0, %rax movl %eax, %edx movabsq $4294967296, %rdi # imm = 0x100000000 orq %rdi, %rdx orq $1, %rdi movl $1, %esi movl $1, %ecx xorl %r8d, %r8d xorl %r9d, %r9d callq __hipPushCallConfiguration testl %eax, %eax jne .LBB1_5 # %bb.4: movq 16(%rsp), %rax movl 12(%rsp), %ecx movq %rax, 96(%rsp) movl %ecx, 28(%rsp) leaq 96(%rsp), %rax movq %rax, 32(%rsp) leaq 28(%rsp), %rax movq %rax, 40(%rsp) leaq 80(%rsp), %rdi leaq 64(%rsp), %rsi leaq 56(%rsp), %rdx leaq 48(%rsp), %rcx callq __hipPopCallConfiguration movq 80(%rsp), %rsi movl 88(%rsp), %edx movq 64(%rsp), %rcx movl 72(%rsp), %r8d leaq 32(%rsp), %r9 movl $_Z3sumPii, %edi pushq 48(%rsp) .cfi_adjust_cfa_offset 8 pushq 64(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $16, %rsp .cfi_adjust_cfa_offset -16 .LBB1_5: movq 16(%rsp), %rsi leaq 32(%rsp), %rdi movl $4, %edx movl $2, %ecx callq hipMemcpy movl $_ZSt4cout, %edi movl $.L.str.2, %esi movl $7, %edx callq _ZSt16__ostream_insertIcSt11char_traitsIcEERSt13basic_ostreamIT_T0_ES6_PKS3_l movl 32(%rsp), %esi movl $_ZSt4cout, %edi callq _ZNSolsEi movq (%rax), %rcx movq -24(%rcx), %rcx movq 240(%rax,%rcx), %r14 testq %r14, %r14 je .LBB1_10 # %bb.6: # %_ZSt13__check_facetISt5ctypeIcEERKT_PS3_.exit.i.i cmpb $0, 56(%r14) je .LBB1_8 # %bb.7: movzbl 67(%r14), %ecx jmp .LBB1_9 .LBB1_8: movq %r14, %rdi movq %rax, %r15 callq _ZNKSt5ctypeIcE13_M_widen_initEv movq (%r14), %rax movq %r14, %rdi movl $10, %esi callq *48(%rax) movl %eax, %ecx movq %r15, %rax .LBB1_9: # %_ZSt4endlIcSt11char_traitsIcEERSt13basic_ostreamIT_T0_ES6_.exit movsbl %cl, %esi movq %rax, %rdi callq _ZNSo3putEc movq %rax, %rdi callq _ZNSo5flushEv movq stdin(%rip), %rdi callq getc movq 16(%rsp), %rdi callq hipFree movq %rbx, %rdi callq _ZdaPv xorl %eax, %eax addq $104, %rsp .cfi_def_cfa_offset 40 popq %rbx .cfi_def_cfa_offset 32 popq %r14 .cfi_def_cfa_offset 24 popq %r15 .cfi_def_cfa_offset 16 popq %rbp .cfi_def_cfa_offset 8 retq .LBB1_10: .cfi_def_cfa_offset 144 callq _ZSt16__throw_bad_castv .Lfunc_end1: .size main, .Lfunc_end1-main .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_ctor .type __hip_module_ctor,@function __hip_module_ctor: # @__hip_module_ctor .cfi_startproc # %bb.0: subq $40, %rsp .cfi_def_cfa_offset 48 cmpq $0, __hip_gpubin_handle(%rip) jne .LBB2_2 # %bb.1: movl $__hip_fatbin_wrapper, %edi callq __hipRegisterFatBinary movq %rax, __hip_gpubin_handle(%rip) .LBB2_2: movq __hip_gpubin_handle(%rip), %rdi xorps %xmm0, %xmm0 movups %xmm0, 16(%rsp) movups %xmm0, (%rsp) movl $_Z3sumPii, %esi movl $.L__unnamed_1, %edx movl $.L__unnamed_1, %ecx movl $-1, %r8d xorl %r9d, %r9d callq __hipRegisterFunction movl $__hip_module_dtor, %edi addq $40, %rsp .cfi_def_cfa_offset 8 jmp atexit # TAILCALL .Lfunc_end2: .size __hip_module_ctor, .Lfunc_end2-__hip_module_ctor .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_dtor .type __hip_module_dtor,@function __hip_module_dtor: # @__hip_module_dtor .cfi_startproc # %bb.0: movq __hip_gpubin_handle(%rip), %rdi testq %rdi, %rdi je .LBB3_2 # %bb.1: pushq %rax .cfi_def_cfa_offset 16 callq __hipUnregisterFatBinary movq $0, __hip_gpubin_handle(%rip) addq $8, %rsp .cfi_def_cfa_offset 8 .LBB3_2: retq .Lfunc_end3: .size __hip_module_dtor, .Lfunc_end3-__hip_module_dtor .cfi_endproc # -- End function .type _Z3sumPii,@object # @_Z3sumPii .section .rodata,"a",@progbits .globl _Z3sumPii .p2align 3, 0x0 _Z3sumPii: .quad _Z18__device_stub__sumPii .size _Z3sumPii, 8 .type .L.str,@object # @.str .section .rodata.str1.1,"aMS",@progbits,1 .L.str: .asciz "\nEnter the number of elements : " .size .L.str, 33 .type .L.str.1,@object # @.str.1 .L.str.1: .asciz "\nEnter the elements : \n" .size .L.str.1, 24 .type .L.str.2,@object # @.str.2 .L.str.2: .asciz "Sum is " .size .L.str.2, 8 .type .L__unnamed_1,@object # @0 .L__unnamed_1: .asciz "_Z3sumPii" .size .L__unnamed_1, 10 .type __hip_fatbin_wrapper,@object # @__hip_fatbin_wrapper .section .hipFatBinSegment,"a",@progbits .p2align 3, 0x0 __hip_fatbin_wrapper: .long 1212764230 # 0x48495046 .long 1 # 0x1 .quad __hip_fatbin .quad 0 .size __hip_fatbin_wrapper, 24 .type __hip_gpubin_handle,@object # @__hip_gpubin_handle .local __hip_gpubin_handle .comm __hip_gpubin_handle,8,8 .section .init_array,"aw",@init_array .p2align 3, 0x0 .quad __hip_module_ctor .type __hip_cuid_,@object # @__hip_cuid_ .bss .globl __hip_cuid_ __hip_cuid_: .byte 0 # 0x0 .size __hip_cuid_, 1 .section ".linker-options","e",@llvm_linker_options .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym _Z18__device_stub__sumPii .addrsig_sym __hip_module_ctor .addrsig_sym __hip_module_dtor .addrsig_sym _Z3sumPii .addrsig_sym _ZSt4cout .addrsig_sym _ZSt3cin .addrsig_sym __hip_fatbin .addrsig_sym __hip_fatbin_wrapper .addrsig_sym __hip_cuid_
You are a helpful coding assistant specialized in converting from CUDA device assembly to AMD device assembly.
code for sm_80 Function : _Z3sumPii .headerflags @"EF_CUDA_TEXMODE_UNIFIED EF_CUDA_64BIT_ADDRESS EF_CUDA_SM80 EF_CUDA_VIRTUAL_SM(EF_CUDA_SM80)" /*0000*/ IMAD.MOV.U32 R1, RZ, RZ, c[0x0][0x28] ; /* 0x00000a00ff017624 */ /* 0x000fe400078e00ff */ /*0010*/ IMAD.MOV.U32 R0, RZ, RZ, c[0x0][0x0] ; /* 0x00000000ff007624 */ /* 0x000fca00078e00ff */ /*0020*/ ISETP.GE.AND P0, PT, R0, 0x1, PT ; /* 0x000000010000780c */ /* 0x000fda0003f06270 */ /*0030*/ @!P0 EXIT ; /* 0x000000000000894d */ /* 0x000fea0003800000 */ /*0040*/ S2R R9, SR_TID.X ; /* 0x0000000000097919 */ /* 0x000e220000002100 */ /*0050*/ MOV R0, c[0x0][0x0] ; /* 0x0000000000007a02 */ /* 0x000fe20000000f00 */ /*0060*/ UMOV UR4, 0x1 ; /* 0x0000000100047882 */ /* 0x000fe40000000000 */ /*0070*/ ULDC.64 UR6, c[0x0][0x118] ; /* 0x0000460000067ab9 */ /* 0x000fe20000000a00 */ /*0080*/ IMAD.SHL.U32 R8, R9, 0x2, RZ ; /* 0x0000000209087824 */ /* 0x001fe400078e00ff */ /*0090*/ ISETP.GE.AND P1, PT, R9, R0, PT ; /* 0x000000000900720c */ /* 0x000fe20003f26270 */ /*00a0*/ BSSY B0, 0x190 ; /* 0x000000e000007945 */ /* 0x000fe20003800000 */ /*00b0*/ ISETP.NE.AND P0, PT, R0, 0x1, PT ; /* 0x000000010000780c */ /* 0x000fd60003f05270 */ /*00c0*/ @P1 BRA 0x180 ; /* 0x000000b000001947 */ /* 0x000fea0003800000 */ /*00d0*/ IMAD R5, R8, UR4, RZ ; /* 0x0000000408057c24 */ /* 0x000fca000f8e02ff */ /*00e0*/ IADD3 R3, R5, UR4, RZ ; /* 0x0000000405037c10 */ /* 0x000fc8000fffe0ff */ /*00f0*/ ISETP.GE.AND P1, PT, R3, c[0x0][0x168], PT ; /* 0x00005a0003007a0c */ /* 0x000fda0003f26270 */ /*0100*/ @P1 BRA 0x180 ; /* 0x0000007000001947 */ /* 0x000fea0003800000 */ /*0110*/ HFMA2.MMA R4, -RZ, RZ, 0, 2.384185791015625e-07 ; /* 0x00000004ff047435 */ /* 0x000fd400000001ff */ /*0120*/ IMAD.WIDE R2, R3, R4, c[0x0][0x160] ; /* 0x0000580003027625 */ /* 0x000fc800078e0204 */ /*0130*/ IMAD.WIDE R4, R5, R4, c[0x0][0x160] ; /* 0x0000580005047625 */ /* 0x000fe400078e0204 */ /*0140*/ LDG.E R3, [R2.64] ; /* 0x0000000602037981 */ /* 0x000ea8000c1e1900 */ /*0150*/ LDG.E R6, [R4.64] ; /* 0x0000000604067981 */ /* 0x000ea4000c1e1900 */ /*0160*/ IMAD.IADD R7, R6, 0x1, R3 ; /* 0x0000000106077824 */ /* 0x004fca00078e0203 */ /*0170*/ STG.E [R4.64], R7 ; /* 0x0000000704007986 */ /* 0x0001e4000c101906 */ /*0180*/ BSYNC B0 ; /* 0x0000000000007941 */ /* 0x000fea0003800000 */ /*0190*/ @!P0 EXIT ; /* 0x000000000000894d */ /* 0x000fea0003800000 */ /*01a0*/ I2F R4, R0 ; /* 0x0000000000047306 */ /* 0x001e220000201400 */ /*01b0*/ BAR.SYNC.DEFER_BLOCKING 0x0 ; /* 0x0000000000007b1d */ /* 0x000fe20000010000 */ /*01c0*/ USHF.L.U32 UR4, UR4, 0x1, URZ ; /* 0x0000000104047899 */ /* 0x000fcc000800063f */ /*01d0*/ F2F.F64.F32 R2, R4 ; /* 0x0000000400027310 */ /* 0x001e240000201800 */ /*01e0*/ DMUL R2, R2, 0.5 ; /* 0x3fe0000002027828 */ /* 0x001e0c0000000000 */ /*01f0*/ F2I.F64.CEIL R0, R2 ; /* 0x0000000200007311 */ /* 0x001e240000309100 */ /*0200*/ ISETP.GT.AND P0, PT, R0, RZ, PT ; /* 0x000000ff0000720c */ /* 0x001fda0003f04270 */ /*0210*/ @P0 BRA 0x90 ; /* 0xfffffe7000000947 */ /* 0x000fea000383ffff */ /*0220*/ EXIT ; /* 0x000000000000794d */ /* 0x000fea0003800000 */ /*0230*/ BRA 0x230; /* 0xfffffff000007947 */ /* 0x000fc0000383ffff */ /*0240*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0250*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0260*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0270*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0280*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0290*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*02a0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*02b0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*02c0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*02d0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*02e0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*02f0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ ..........
.text .amdgcn_target "amdgcn-amd-amdhsa--gfx1100" .protected _Z3sumPii .globl _Z3sumPii .p2align 8 .type _Z3sumPii,@function _Z3sumPii: s_load_b32 s4, s[0:1], 0x1c s_waitcnt lgkmcnt(0) v_cmp_eq_u16_e64 s2, s4, 0 s_delay_alu instid0(VALU_DEP_1) s_and_b32 vcc_lo, exec_lo, s2 s_cbranch_vccnz .LBB0_7 s_clause 0x1 s_load_b64 s[2:3], s[0:1], 0x0 s_load_b32 s1, s[0:1], 0x8 s_and_b32 s0, 0xffff, s4 s_delay_alu instid0(SALU_CYCLE_1) v_dual_mov_b32 v6, s0 :: v_dual_lshlrev_b32 v5, 1, v0 s_mov_b32 s4, 1 s_branch .LBB0_3 .p2align 6 .LBB0_2: v_cvt_f32_i32_e32 v1, v6 s_lshl_b32 s4, s4, 1 s_waitcnt_vscnt null, 0x0 s_barrier buffer_gl0_inv v_cvt_f64_f32_e32 v[1:2], v1 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_mul_f64 v[1:2], v[1:2], 0.5 v_ceil_f64_e32 v[1:2], v[1:2] s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_cvt_i32_f64_e32 v6, v[1:2] v_cmp_gt_i32_e64 s0, 1, v6 s_delay_alu instid0(VALU_DEP_1) s_and_not1_b32 vcc_lo, exec_lo, s0 s_cbranch_vccz .LBB0_7 .LBB0_3: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2) v_mul_lo_u32 v1, v5, s4 v_cmp_lt_u32_e32 vcc_lo, v0, v6 s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_1) v_add_nc_u32_e32 v3, s4, v1 s_waitcnt lgkmcnt(0) v_cmp_gt_i32_e64 s0, s1, v3 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(SALU_CYCLE_1) s_and_b32 s5, vcc_lo, s0 s_and_saveexec_b32 s0, s5 s_cbranch_execz .LBB0_5 v_ashrrev_i32_e32 v4, 31, v3 v_ashrrev_i32_e32 v2, 31, v1 s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) v_lshlrev_b64 v[3:4], 2, v[3:4] v_lshlrev_b64 v[1:2], 2, v[1:2] s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3) v_add_co_u32 v3, vcc_lo, s2, v3 v_add_co_ci_u32_e32 v4, vcc_lo, s3, v4, vcc_lo s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_4) v_add_co_u32 v1, vcc_lo, s2, v1 v_add_co_ci_u32_e32 v2, vcc_lo, s3, v2, vcc_lo s_clause 0x1 global_load_b32 v3, v[3:4], off global_load_b32 v4, v[1:2], off s_waitcnt vmcnt(0) v_add_nc_u32_e32 v3, v4, v3 global_store_b32 v[1:2], v3, off .LBB0_5: s_or_b32 exec_lo, exec_lo, s0 v_cmp_eq_u32_e32 vcc_lo, 1, v6 s_cbranch_vccz .LBB0_2 .LBB0_7: s_nop 0 s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) s_endpgm .section .rodata,"a",@progbits .p2align 6, 0x0 .amdhsa_kernel _Z3sumPii .amdhsa_group_segment_fixed_size 0 .amdhsa_private_segment_fixed_size 0 .amdhsa_kernarg_size 272 .amdhsa_user_sgpr_count 15 .amdhsa_user_sgpr_dispatch_ptr 0 .amdhsa_user_sgpr_queue_ptr 0 .amdhsa_user_sgpr_kernarg_segment_ptr 1 .amdhsa_user_sgpr_dispatch_id 0 .amdhsa_user_sgpr_private_segment_size 0 .amdhsa_wavefront_size32 1 .amdhsa_uses_dynamic_stack 0 .amdhsa_enable_private_segment 0 .amdhsa_system_sgpr_workgroup_id_x 1 .amdhsa_system_sgpr_workgroup_id_y 0 .amdhsa_system_sgpr_workgroup_id_z 0 .amdhsa_system_sgpr_workgroup_info 0 .amdhsa_system_vgpr_workitem_id 0 .amdhsa_next_free_vgpr 7 .amdhsa_next_free_sgpr 6 .amdhsa_float_round_mode_32 0 .amdhsa_float_round_mode_16_64 0 .amdhsa_float_denorm_mode_32 3 .amdhsa_float_denorm_mode_16_64 3 .amdhsa_dx10_clamp 1 .amdhsa_ieee_mode 1 .amdhsa_fp16_overflow 0 .amdhsa_workgroup_processor_mode 1 .amdhsa_memory_ordered 1 .amdhsa_forward_progress 0 .amdhsa_shared_vgpr_count 0 .amdhsa_exception_fp_ieee_invalid_op 0 .amdhsa_exception_fp_denorm_src 0 .amdhsa_exception_fp_ieee_div_zero 0 .amdhsa_exception_fp_ieee_overflow 0 .amdhsa_exception_fp_ieee_underflow 0 .amdhsa_exception_fp_ieee_inexact 0 .amdhsa_exception_int_div_zero 0 .end_amdhsa_kernel .text .Lfunc_end0: .size _Z3sumPii, .Lfunc_end0-_Z3sumPii .section .AMDGPU.csdata,"",@progbits .text .p2alignl 7, 3214868480 .fill 96, 4, 3214868480 .type __hip_cuid_,@object .section .bss,"aw",@nobits .globl __hip_cuid_ __hip_cuid_: .byte 0 .size __hip_cuid_, 1 .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym __hip_cuid_ .amdgpu_metadata --- amdhsa.kernels: - .args: - .address_space: global .offset: 0 .size: 8 .value_kind: global_buffer - .offset: 8 .size: 4 .value_kind: by_value - .offset: 16 .size: 4 .value_kind: hidden_block_count_x - .offset: 20 .size: 4 .value_kind: hidden_block_count_y - .offset: 24 .size: 4 .value_kind: hidden_block_count_z - .offset: 28 .size: 2 .value_kind: hidden_group_size_x - .offset: 30 .size: 2 .value_kind: hidden_group_size_y - .offset: 32 .size: 2 .value_kind: hidden_group_size_z - .offset: 34 .size: 2 .value_kind: hidden_remainder_x - .offset: 36 .size: 2 .value_kind: hidden_remainder_y - .offset: 38 .size: 2 .value_kind: hidden_remainder_z - .offset: 56 .size: 8 .value_kind: hidden_global_offset_x - .offset: 64 .size: 8 .value_kind: hidden_global_offset_y - .offset: 72 .size: 8 .value_kind: hidden_global_offset_z - .offset: 80 .size: 2 .value_kind: hidden_grid_dims .group_segment_fixed_size: 0 .kernarg_segment_align: 8 .kernarg_segment_size: 272 .language: OpenCL C .language_version: - 2 - 0 .max_flat_workgroup_size: 1024 .name: _Z3sumPii .private_segment_fixed_size: 0 .sgpr_count: 8 .sgpr_spill_count: 0 .symbol: _Z3sumPii.kd .uniform_work_group_size: 1 .uses_dynamic_stack: false .vgpr_count: 7 .vgpr_spill_count: 0 .wavefront_size: 32 .workgroup_processor_mode: 1 amdhsa.target: amdgcn-amd-amdhsa--gfx1100 amdhsa.version: - 1 - 2 ... .end_amdgpu_metadata
You are a helpful coding assistant specialized in converting from CUDA host assembly to AMD host assembly.
.file "tmpxft_0002ac37_00000000-6_ParRedSum.cudafe1.cpp" .text #APP .globl _ZSt21ios_base_library_initv #NO_APP .type _ZL26__cudaUnregisterBinaryUtilv, @function _ZL26__cudaUnregisterBinaryUtilv: .LFB3711: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaUnregisterFatBinary@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE3711: .size _ZL26__cudaUnregisterBinaryUtilv, .-_ZL26__cudaUnregisterBinaryUtilv .globl _Z23__device_stub__Z3sumPiiPii .type _Z23__device_stub__Z3sumPiiPii, @function _Z23__device_stub__Z3sumPiiPii: .LFB3733: .cfi_startproc endbr64 subq $120, %rsp .cfi_def_cfa_offset 128 movq %rdi, 8(%rsp) movl %esi, 4(%rsp) movq %fs:40, %rax movq %rax, 104(%rsp) xorl %eax, %eax leaq 8(%rsp), %rax movq %rax, 80(%rsp) leaq 4(%rsp), %rax movq %rax, 88(%rsp) movl $1, 32(%rsp) movl $1, 36(%rsp) movl $1, 40(%rsp) movl $1, 44(%rsp) movl $1, 48(%rsp) movl $1, 52(%rsp) leaq 24(%rsp), %rcx leaq 16(%rsp), %rdx leaq 44(%rsp), %rsi leaq 32(%rsp), %rdi call __cudaPopCallConfiguration@PLT testl %eax, %eax je .L7 .L3: movq 104(%rsp), %rax subq %fs:40, %rax jne .L8 addq $120, %rsp .cfi_remember_state .cfi_def_cfa_offset 8 ret .L7: .cfi_restore_state pushq 24(%rsp) .cfi_def_cfa_offset 136 pushq 24(%rsp) .cfi_def_cfa_offset 144 leaq 96(%rsp), %r9 movq 60(%rsp), %rcx movl 68(%rsp), %r8d movq 48(%rsp), %rsi movl 56(%rsp), %edx leaq _Z3sumPii(%rip), %rdi call cudaLaunchKernel@PLT addq $16, %rsp .cfi_def_cfa_offset 128 jmp .L3 .L8: call __stack_chk_fail@PLT .cfi_endproc .LFE3733: .size _Z23__device_stub__Z3sumPiiPii, .-_Z23__device_stub__Z3sumPiiPii .globl _Z3sumPii .type _Z3sumPii, @function _Z3sumPii: .LFB3734: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 call _Z23__device_stub__Z3sumPiiPii addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE3734: .size _Z3sumPii, .-_Z3sumPii .section .rodata.str1.8,"aMS",@progbits,1 .align 8 .LC0: .string "\nEnter the number of elements : " .section .rodata.str1.1,"aMS",@progbits,1 .LC1: .string "\nEnter the elements : \n" .LC6: .string "Sum is " .text .globl main .type main, @function main: .LFB3708: .cfi_startproc endbr64 pushq %r14 .cfi_def_cfa_offset 16 .cfi_offset 14, -16 pushq %r13 .cfi_def_cfa_offset 24 .cfi_offset 13, -24 pushq %r12 .cfi_def_cfa_offset 32 .cfi_offset 12, -32 pushq %rbp .cfi_def_cfa_offset 40 .cfi_offset 6, -40 pushq %rbx .cfi_def_cfa_offset 48 .cfi_offset 3, -48 subq $48, %rsp .cfi_def_cfa_offset 96 movq %fs:40, %rax movq %rax, 40(%rsp) xorl %eax, %eax movl $0, 4(%rsp) leaq .LC0(%rip), %rsi leaq _ZSt4cout(%rip), %rdi call _ZStlsISt11char_traitsIcEERSt13basic_ostreamIcT_ES5_PKc@PLT leaq 4(%rsp), %rsi leaq _ZSt3cin(%rip), %rdi call _ZNSirsERi@PLT movl 4(%rsp), %ebx movslq %ebx, %rdi movabsq $2305843009213693950, %rax cmpq %rdi, %rax jb .L12 leal 0(,%rbx,4), %r14d salq $2, %rdi call _Znam@PLT movq %rax, %r13 leaq .LC1(%rip), %rsi leaq _ZSt4cout(%rip), %rdi call _ZStlsISt11char_traitsIcEERSt13basic_ostreamIcT_ES5_PKc@PLT movq %r13, %rbp movl $0, %ebx leaq _ZSt3cin(%rip), %r12 cmpl $0, 4(%rsp) jle .L14 .L16: movq %rbp, %rsi movq %r12, %rdi call _ZNSirsERi@PLT addl $1, %ebx addq $4, %rbp cmpl %ebx, 4(%rsp) jg .L16 .L14: movslq %r14d, %r14 leaq 8(%rsp), %rdi movq %r14, %rsi call cudaMalloc@PLT movl $1, %ecx movq %r14, %rdx movq %r13, %rsi movq 8(%rsp), %rdi call cudaMemcpy@PLT pxor %xmm0, %xmm0 cvtsi2ssl 4(%rsp), %xmm0 cvtss2sd %xmm0, %xmm0 mulsd .LC2(%rip), %xmm0 movapd %xmm0, %xmm3 movsd .LC7(%rip), %xmm2 movapd %xmm0, %xmm1 andpd %xmm2, %xmm1 movsd .LC3(%rip), %xmm4 ucomisd %xmm1, %xmm4 jbe .L17 cvttsd2siq %xmm0, %rax pxor %xmm1, %xmm1 cvtsi2sdq %rax, %xmm1 cmpnlesd %xmm1, %xmm3 movsd .LC5(%rip), %xmm4 andpd %xmm4, %xmm3 addsd %xmm1, %xmm3 andnpd %xmm0, %xmm2 orpd %xmm2, %xmm3 .L17: cvttsd2siq %xmm3, %rax movl %eax, 28(%rsp) movl $1, 32(%rsp) movl $1, 36(%rsp) movl $1, 16(%rsp) movl $1, 20(%rsp) movl $0, %r9d movl $0, %r8d movq 28(%rsp), %rdx movl $1, %ecx movq 16(%rsp), %rdi movl $1, %esi call __cudaPushCallConfiguration@PLT testl %eax, %eax je .L23 .L18: leaq 28(%rsp), %rdi movl $2, %ecx movl $4, %edx movq 8(%rsp), %rsi call cudaMemcpy@PLT leaq .LC6(%rip), %rsi leaq _ZSt4cout(%rip), %rdi call _ZStlsISt11char_traitsIcEERSt13basic_ostreamIcT_ES5_PKc@PLT movq %rax, %rdi movl 28(%rsp), %esi call _ZNSolsEi@PLT movq %rax, %rdi call _ZSt4endlIcSt11char_traitsIcEERSt13basic_ostreamIT_T0_ES6_@PLT movq stdin(%rip), %rdi call getc@PLT movq 8(%rsp), %rdi call cudaFree@PLT movq %r13, %rdi call _ZdaPv@PLT movq 40(%rsp), %rax subq %fs:40, %rax jne .L24 movl $0, %eax addq $48, %rsp .cfi_remember_state .cfi_def_cfa_offset 48 popq %rbx .cfi_def_cfa_offset 40 popq %rbp .cfi_def_cfa_offset 32 popq %r12 .cfi_def_cfa_offset 24 popq %r13 .cfi_def_cfa_offset 16 popq %r14 .cfi_def_cfa_offset 8 ret .L12: .cfi_restore_state movq 40(%rsp), %rax subq %fs:40, %rax je .L15 call __stack_chk_fail@PLT .L15: call __cxa_throw_bad_array_new_length@PLT .L23: movl 4(%rsp), %esi movq 8(%rsp), %rdi call _Z23__device_stub__Z3sumPiiPii jmp .L18 .L24: call __stack_chk_fail@PLT .cfi_endproc .LFE3708: .size main, .-main .section .rodata.str1.1 .LC8: .string "_Z3sumPii" .text .type _ZL24__sti____cudaRegisterAllv, @function _ZL24__sti____cudaRegisterAllv: .LFB3736: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 leaq _ZL15__fatDeviceText(%rip), %rdi call __cudaRegisterFatBinary@PLT movq %rax, %rdi movq %rax, _ZL20__cudaFatCubinHandle(%rip) pushq $0 .cfi_def_cfa_offset 24 pushq $0 .cfi_def_cfa_offset 32 pushq $0 .cfi_def_cfa_offset 40 pushq $0 .cfi_def_cfa_offset 48 movl $0, %r9d movl $-1, %r8d leaq .LC8(%rip), %rdx movq %rdx, %rcx leaq _Z3sumPii(%rip), %rsi call __cudaRegisterFunction@PLT addq $32, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaRegisterFatBinaryEnd@PLT leaq _ZL26__cudaUnregisterBinaryUtilv(%rip), %rdi call atexit@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE3736: .size _ZL24__sti____cudaRegisterAllv, .-_ZL24__sti____cudaRegisterAllv .section .init_array,"aw" .align 8 .quad _ZL24__sti____cudaRegisterAllv .section .nvFatBinSegment,"aw" .align 8 .type _ZL15__fatDeviceText, @object .size _ZL15__fatDeviceText, 24 _ZL15__fatDeviceText: .long 1180844977 .long 1 .quad fatbinData .quad 0 .local _ZL20__cudaFatCubinHandle .comm _ZL20__cudaFatCubinHandle,8,8 .section .rodata.cst8,"aM",@progbits,8 .align 8 .LC2: .long 0 .long 1071644672 .align 8 .LC3: .long 0 .long 1127219200 .align 8 .LC5: .long 0 .long 1072693248 .align 8 .LC7: .long -1 .long 2147483647 .ident "GCC: (Ubuntu 13.3.0-6ubuntu2~24.04) 13.3.0" .section .note.GNU-stack,"",@progbits .section .note.gnu.property,"a" .align 8 .long 1f - 0f .long 4f - 1f .long 5 0: .string "GNU" 1: .align 8 .long 0xc0000002 .long 3f - 2f 2: .long 0x3 3: .align 8 4:
.text .file "ParRedSum.hip" # Start of file scope inline assembly .globl _ZSt21ios_base_library_initv # End of file scope inline assembly .globl _Z18__device_stub__sumPii # -- Begin function _Z18__device_stub__sumPii .p2align 4, 0x90 .type _Z18__device_stub__sumPii,@function _Z18__device_stub__sumPii: # @_Z18__device_stub__sumPii .cfi_startproc # %bb.0: subq $88, %rsp .cfi_def_cfa_offset 96 movq %rdi, 56(%rsp) movl %esi, 4(%rsp) leaq 56(%rsp), %rax movq %rax, 64(%rsp) leaq 4(%rsp), %rax movq %rax, 72(%rsp) leaq 40(%rsp), %rdi leaq 24(%rsp), %rsi leaq 16(%rsp), %rdx leaq 8(%rsp), %rcx callq __hipPopCallConfiguration movq 40(%rsp), %rsi movl 48(%rsp), %edx movq 24(%rsp), %rcx movl 32(%rsp), %r8d leaq 64(%rsp), %r9 movl $_Z3sumPii, %edi pushq 8(%rsp) .cfi_adjust_cfa_offset 8 pushq 24(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $104, %rsp .cfi_adjust_cfa_offset -104 retq .Lfunc_end0: .size _Z18__device_stub__sumPii, .Lfunc_end0-_Z18__device_stub__sumPii .cfi_endproc # -- End function .section .rodata.cst8,"aM",@progbits,8 .p2align 3, 0x0 # -- Begin function main .LCPI1_0: .quad 0x3fe0000000000000 # double 0.5 .text .globl main .p2align 4, 0x90 .type main,@function main: # @main .cfi_startproc # %bb.0: pushq %rbp .cfi_def_cfa_offset 16 pushq %r15 .cfi_def_cfa_offset 24 pushq %r14 .cfi_def_cfa_offset 32 pushq %rbx .cfi_def_cfa_offset 40 subq $104, %rsp .cfi_def_cfa_offset 144 .cfi_offset %rbx, -40 .cfi_offset %r14, -32 .cfi_offset %r15, -24 .cfi_offset %rbp, -16 movl $0, 12(%rsp) movl $_ZSt4cout, %edi movl $.L.str, %esi movl $32, %edx callq _ZSt16__ostream_insertIcSt11char_traitsIcEERSt13basic_ostreamIT_T0_ES6_PKS3_l leaq 12(%rsp), %rsi movl $_ZSt3cin, %edi callq _ZNSirsERi movslq 12(%rsp), %rax leal (,%rax,4), %ebp leaq (,%rax,4), %rcx testq %rax, %rax movq $-1, %rdi cmovnsq %rcx, %rdi callq _Znam movq %rax, %rbx movl $_ZSt4cout, %edi movl $.L.str.1, %esi movl $23, %edx callq _ZSt16__ostream_insertIcSt11char_traitsIcEERSt13basic_ostreamIT_T0_ES6_PKS3_l cmpl $0, 12(%rsp) jle .LBB1_3 # %bb.1: # %.lr.ph.preheader xorl %r15d, %r15d movq %rbx, %r14 .p2align 4, 0x90 .LBB1_2: # %.lr.ph # =>This Inner Loop Header: Depth=1 movl $_ZSt3cin, %edi movq %r14, %rsi callq _ZNSirsERi incq %r15 movslq 12(%rsp), %rax addq $4, %r14 cmpq %rax, %r15 jl .LBB1_2 .LBB1_3: # %._crit_edge movslq %ebp, %r14 leaq 16(%rsp), %rdi movq %r14, %rsi callq hipMalloc movq 16(%rsp), %rdi movq %rbx, %rsi movq %r14, %rdx movl $1, %ecx callq hipMemcpy cvtsi2ssl 12(%rsp), %xmm0 cvtss2sd %xmm0, %xmm0 mulsd .LCPI1_0(%rip), %xmm0 callq ceil@PLT cvttsd2si %xmm0, %rax movl %eax, %edx movabsq $4294967296, %rdi # imm = 0x100000000 orq %rdi, %rdx orq $1, %rdi movl $1, %esi movl $1, %ecx xorl %r8d, %r8d xorl %r9d, %r9d callq __hipPushCallConfiguration testl %eax, %eax jne .LBB1_5 # %bb.4: movq 16(%rsp), %rax movl 12(%rsp), %ecx movq %rax, 96(%rsp) movl %ecx, 28(%rsp) leaq 96(%rsp), %rax movq %rax, 32(%rsp) leaq 28(%rsp), %rax movq %rax, 40(%rsp) leaq 80(%rsp), %rdi leaq 64(%rsp), %rsi leaq 56(%rsp), %rdx leaq 48(%rsp), %rcx callq __hipPopCallConfiguration movq 80(%rsp), %rsi movl 88(%rsp), %edx movq 64(%rsp), %rcx movl 72(%rsp), %r8d leaq 32(%rsp), %r9 movl $_Z3sumPii, %edi pushq 48(%rsp) .cfi_adjust_cfa_offset 8 pushq 64(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $16, %rsp .cfi_adjust_cfa_offset -16 .LBB1_5: movq 16(%rsp), %rsi leaq 32(%rsp), %rdi movl $4, %edx movl $2, %ecx callq hipMemcpy movl $_ZSt4cout, %edi movl $.L.str.2, %esi movl $7, %edx callq _ZSt16__ostream_insertIcSt11char_traitsIcEERSt13basic_ostreamIT_T0_ES6_PKS3_l movl 32(%rsp), %esi movl $_ZSt4cout, %edi callq _ZNSolsEi movq (%rax), %rcx movq -24(%rcx), %rcx movq 240(%rax,%rcx), %r14 testq %r14, %r14 je .LBB1_10 # %bb.6: # %_ZSt13__check_facetISt5ctypeIcEERKT_PS3_.exit.i.i cmpb $0, 56(%r14) je .LBB1_8 # %bb.7: movzbl 67(%r14), %ecx jmp .LBB1_9 .LBB1_8: movq %r14, %rdi movq %rax, %r15 callq _ZNKSt5ctypeIcE13_M_widen_initEv movq (%r14), %rax movq %r14, %rdi movl $10, %esi callq *48(%rax) movl %eax, %ecx movq %r15, %rax .LBB1_9: # %_ZSt4endlIcSt11char_traitsIcEERSt13basic_ostreamIT_T0_ES6_.exit movsbl %cl, %esi movq %rax, %rdi callq _ZNSo3putEc movq %rax, %rdi callq _ZNSo5flushEv movq stdin(%rip), %rdi callq getc movq 16(%rsp), %rdi callq hipFree movq %rbx, %rdi callq _ZdaPv xorl %eax, %eax addq $104, %rsp .cfi_def_cfa_offset 40 popq %rbx .cfi_def_cfa_offset 32 popq %r14 .cfi_def_cfa_offset 24 popq %r15 .cfi_def_cfa_offset 16 popq %rbp .cfi_def_cfa_offset 8 retq .LBB1_10: .cfi_def_cfa_offset 144 callq _ZSt16__throw_bad_castv .Lfunc_end1: .size main, .Lfunc_end1-main .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_ctor .type __hip_module_ctor,@function __hip_module_ctor: # @__hip_module_ctor .cfi_startproc # %bb.0: subq $40, %rsp .cfi_def_cfa_offset 48 cmpq $0, __hip_gpubin_handle(%rip) jne .LBB2_2 # %bb.1: movl $__hip_fatbin_wrapper, %edi callq __hipRegisterFatBinary movq %rax, __hip_gpubin_handle(%rip) .LBB2_2: movq __hip_gpubin_handle(%rip), %rdi xorps %xmm0, %xmm0 movups %xmm0, 16(%rsp) movups %xmm0, (%rsp) movl $_Z3sumPii, %esi movl $.L__unnamed_1, %edx movl $.L__unnamed_1, %ecx movl $-1, %r8d xorl %r9d, %r9d callq __hipRegisterFunction movl $__hip_module_dtor, %edi addq $40, %rsp .cfi_def_cfa_offset 8 jmp atexit # TAILCALL .Lfunc_end2: .size __hip_module_ctor, .Lfunc_end2-__hip_module_ctor .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_dtor .type __hip_module_dtor,@function __hip_module_dtor: # @__hip_module_dtor .cfi_startproc # %bb.0: movq __hip_gpubin_handle(%rip), %rdi testq %rdi, %rdi je .LBB3_2 # %bb.1: pushq %rax .cfi_def_cfa_offset 16 callq __hipUnregisterFatBinary movq $0, __hip_gpubin_handle(%rip) addq $8, %rsp .cfi_def_cfa_offset 8 .LBB3_2: retq .Lfunc_end3: .size __hip_module_dtor, .Lfunc_end3-__hip_module_dtor .cfi_endproc # -- End function .type _Z3sumPii,@object # @_Z3sumPii .section .rodata,"a",@progbits .globl _Z3sumPii .p2align 3, 0x0 _Z3sumPii: .quad _Z18__device_stub__sumPii .size _Z3sumPii, 8 .type .L.str,@object # @.str .section .rodata.str1.1,"aMS",@progbits,1 .L.str: .asciz "\nEnter the number of elements : " .size .L.str, 33 .type .L.str.1,@object # @.str.1 .L.str.1: .asciz "\nEnter the elements : \n" .size .L.str.1, 24 .type .L.str.2,@object # @.str.2 .L.str.2: .asciz "Sum is " .size .L.str.2, 8 .type .L__unnamed_1,@object # @0 .L__unnamed_1: .asciz "_Z3sumPii" .size .L__unnamed_1, 10 .type __hip_fatbin_wrapper,@object # @__hip_fatbin_wrapper .section .hipFatBinSegment,"a",@progbits .p2align 3, 0x0 __hip_fatbin_wrapper: .long 1212764230 # 0x48495046 .long 1 # 0x1 .quad __hip_fatbin .quad 0 .size __hip_fatbin_wrapper, 24 .type __hip_gpubin_handle,@object # @__hip_gpubin_handle .local __hip_gpubin_handle .comm __hip_gpubin_handle,8,8 .section .init_array,"aw",@init_array .p2align 3, 0x0 .quad __hip_module_ctor .type __hip_cuid_,@object # @__hip_cuid_ .bss .globl __hip_cuid_ __hip_cuid_: .byte 0 # 0x0 .size __hip_cuid_, 1 .section ".linker-options","e",@llvm_linker_options .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym _Z18__device_stub__sumPii .addrsig_sym __hip_module_ctor .addrsig_sym __hip_module_dtor .addrsig_sym _Z3sumPii .addrsig_sym _ZSt4cout .addrsig_sym _ZSt3cin .addrsig_sym __hip_fatbin .addrsig_sym __hip_fatbin_wrapper .addrsig_sym __hip_cuid_
You are a helpful coding assistant specialized in converting CUDA source code to CUDA host assembly.
#include <iostream> #include <cstdlib> #include <cstdio> #include <cuda_runtime.h> #include <sys/time.h> double get_time() { struct timeval tv; gettimeofday(&tv, nullptr); return tv.tv_sec + 1e-6 * tv.tv_usec; } constexpr int m = 256; inline __device__ int indirect(int *c, int i) { // return c[c[i & 127] & 127] + i; return int(exp(((((float(i))))) * 1e-18)) + i; // printf("%d\n", c[i % m] - i % m + i - i); // return i; } __constant__ int const_c[m]; __global__ void fd(float *a, float *b, int *c, int n) { __shared__ float b_s[m]; unsigned int i = blockIdx.x * blockDim.x + threadIdx.x; if (i < m) { b_s[i] = 0; } __syncthreads(); /* if (threadIdx.x < m) { b_s[threadIdx.x] = c[threadIdx.x]; } __syncthreads(); */ /* float sum = 0; if (i > 0) sum += a[indirect(c, i) - 1]; */ // sum += a[indirect(c, i)]; // sum += a[i + b_s[i & 127]]; /* if (i < n - 1) sum += a[indirect(c, i) + 1]; */ // b[i] = (i * 1e-18); // b[i] = i; // b[i] = c[c[c[i & 64]]]; // atomicAdd(b_s + ((unsigned)i * 34252345627) % m, 1.0f); // i = int(((((i * 1e-20f))))); // i = (i * 1e-10f); // i = i * i * i * i * i % m; // b_s[i % m] = 1; // #define C(x) i += (i >> x); // #define C(x) i += (i >> x); // for (int t = 0; t < 240; t++) // C(30); i += int(sin(i * 1e-20f)); i += int(sin(i * 1e-20f)); i += int(sin(i * 1e-20f)); i += int(sin(i * 1e-20f)); i += int(sin(i * 1e-20f)); i += int(sin(i * 1e-20f)); i += int(sin(i * 1e-20f)); i += int(sin(i * 1e-20f)); i += int(sin(i * 1e-20f)); i += int(sin(i * 1e-20f)); for (int j = 0; j < 27; j++) { atomicAdd(b_s + (unsigned int)(i / 4 + j * 431) % (m / 1), 1.0f); } __syncthreads(); if (i < m) { atomicAdd(&b[i], b_s[i]); } // atomicAdd(b + i % (m * m), 1); /* atomicAdd(&b_s[0], sqrt(sum)); if (threadIdx.x < m) { atomicAdd(b + threadIdx.x, b_s[threadIdx.x]); // b[threadIdx.x] += b_s[threadIdx.x]; } */ } int main() { int n = 128 * 1024 * 1024; float *a, *b; int *c; cudaMallocManaged(&a, n * sizeof(float)); cudaMallocManaged(&b, n * sizeof(float)); cudaMallocManaged(&c, m * sizeof(float)); for (int i = 0; i < n; i++) { a[i] = i * 1e-5f; } for (int i = 0; i < n; i++) { b[i] = i * 1e-5f; } for (int i = 0; i < m; i++) { c[i] = 0; } cudaMemcpyToSymbol(const_c, c, m * sizeof(float), 0, cudaMemcpyHostToDevice); for (auto bs : {256, 512, 1024}) { std::cout << "bs = " << bs << std::endl; for (int i = 0; i < 4; i++) { auto t = get_time(); fd<<<n / bs, bs>>>(a, b, c, n); cudaDeviceSynchronize(); t = get_time() - t; printf("%.2f ms bw %.3f GB/s\n", t * 1000, n * 2.0f * 4 / t / (1024 * 1024 * 1024.0f)); } std::cout << std::endl; } }
.file "tmpxft_000ef42a_00000000-6_bandwidth.cudafe1.cpp" .text #APP .globl _ZSt21ios_base_library_initv #NO_APP .type _ZL26__cudaUnregisterBinaryUtilv, @function _ZL26__cudaUnregisterBinaryUtilv: .LFB3674: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaUnregisterFatBinary@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE3674: .size _ZL26__cudaUnregisterBinaryUtilv, .-_ZL26__cudaUnregisterBinaryUtilv .globl _Z8get_timev .type _Z8get_timev, @function _Z8get_timev: .LFB3669: .cfi_startproc endbr64 subq $40, %rsp .cfi_def_cfa_offset 48 movq %fs:40, %rax movq %rax, 24(%rsp) xorl %eax, %eax movq %rsp, %rdi movl $0, %esi call gettimeofday@PLT pxor %xmm0, %xmm0 cvtsi2sdq 8(%rsp), %xmm0 mulsd .LC0(%rip), %xmm0 pxor %xmm1, %xmm1 cvtsi2sdq (%rsp), %xmm1 addsd %xmm1, %xmm0 movq 24(%rsp), %rax subq %fs:40, %rax jne .L6 addq $40, %rsp .cfi_remember_state .cfi_def_cfa_offset 8 ret .L6: .cfi_restore_state call __stack_chk_fail@PLT .cfi_endproc .LFE3669: .size _Z8get_timev, .-_Z8get_timev .globl _Z26__device_stub__Z2fdPfS_PiiPfS_Pii .type _Z26__device_stub__Z2fdPfS_PiiPfS_Pii, @function _Z26__device_stub__Z2fdPfS_PiiPfS_Pii: .LFB3696: .cfi_startproc endbr64 subq $152, %rsp .cfi_def_cfa_offset 160 movq %rdi, 24(%rsp) movq %rsi, 16(%rsp) movq %rdx, 8(%rsp) movl %ecx, 4(%rsp) movq %fs:40, %rax movq %rax, 136(%rsp) xorl %eax, %eax leaq 24(%rsp), %rax movq %rax, 96(%rsp) leaq 16(%rsp), %rax movq %rax, 104(%rsp) leaq 8(%rsp), %rax movq %rax, 112(%rsp) leaq 4(%rsp), %rax movq %rax, 120(%rsp) movl $1, 48(%rsp) movl $1, 52(%rsp) movl $1, 56(%rsp) movl $1, 60(%rsp) movl $1, 64(%rsp) movl $1, 68(%rsp) leaq 40(%rsp), %rcx leaq 32(%rsp), %rdx leaq 60(%rsp), %rsi leaq 48(%rsp), %rdi call __cudaPopCallConfiguration@PLT testl %eax, %eax je .L11 .L7: movq 136(%rsp), %rax subq %fs:40, %rax jne .L12 addq $152, %rsp .cfi_remember_state .cfi_def_cfa_offset 8 ret .L11: .cfi_restore_state pushq 40(%rsp) .cfi_def_cfa_offset 168 pushq 40(%rsp) .cfi_def_cfa_offset 176 leaq 112(%rsp), %r9 movq 76(%rsp), %rcx movl 84(%rsp), %r8d movq 64(%rsp), %rsi movl 72(%rsp), %edx leaq _Z2fdPfS_Pii(%rip), %rdi call cudaLaunchKernel@PLT addq $16, %rsp .cfi_def_cfa_offset 160 jmp .L7 .L12: call __stack_chk_fail@PLT .cfi_endproc .LFE3696: .size _Z26__device_stub__Z2fdPfS_PiiPfS_Pii, .-_Z26__device_stub__Z2fdPfS_PiiPfS_Pii .globl _Z2fdPfS_Pii .type _Z2fdPfS_Pii, @function _Z2fdPfS_Pii: .LFB3697: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 call _Z26__device_stub__Z2fdPfS_PiiPfS_Pii addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE3697: .size _Z2fdPfS_Pii, .-_Z2fdPfS_Pii .section .rodata.str1.1,"aMS",@progbits,1 .LC2: .string "bs = " .LC6: .string "%.2f ms bw %.3f GB/s\n" .text .globl main .type main, @function main: .LFB3671: .cfi_startproc endbr64 pushq %r15 .cfi_def_cfa_offset 16 .cfi_offset 15, -16 pushq %r14 .cfi_def_cfa_offset 24 .cfi_offset 14, -24 pushq %r13 .cfi_def_cfa_offset 32 .cfi_offset 13, -32 pushq %r12 .cfi_def_cfa_offset 40 .cfi_offset 12, -40 pushq %rbp .cfi_def_cfa_offset 48 .cfi_offset 6, -48 pushq %rbx .cfi_def_cfa_offset 56 .cfi_offset 3, -56 subq $104, %rsp .cfi_def_cfa_offset 160 movq %fs:40, %rax movq %rax, 88(%rsp) xorl %eax, %eax leaq 24(%rsp), %rdi movl $1, %edx movl $536870912, %esi call cudaMallocManaged@PLT leaq 32(%rsp), %rdi movl $1, %edx movl $536870912, %esi call cudaMallocManaged@PLT leaq 40(%rsp), %rdi movl $1, %edx movl $1024, %esi call cudaMallocManaged@PLT movl $0, %eax movss .LC1(%rip), %xmm1 .L16: pxor %xmm0, %xmm0 cvtsi2ssl %eax, %xmm0 mulss %xmm1, %xmm0 movq 24(%rsp), %rdx movss %xmm0, (%rdx,%rax,4) addq $1, %rax cmpq $134217728, %rax jne .L16 movl $0, %eax movss .LC1(%rip), %xmm1 .L17: pxor %xmm0, %xmm0 cvtsi2ssl %eax, %xmm0 mulss %xmm1, %xmm0 movq 32(%rsp), %rdx movss %xmm0, (%rdx,%rax,4) addq $1, %rax cmpq $134217728, %rax jne .L17 movl $0, %eax .L18: movq 40(%rsp), %rdx movl $0, (%rdx,%rax) addq $4, %rax cmpq $1024, %rax jne .L18 movl $1, %r8d movl $0, %ecx movl $1024, %edx movq 40(%rsp), %rsi leaq _ZL7const_c(%rip), %rdi call cudaMemcpyToSymbol@PLT movl $256, 76(%rsp) movl $512, 80(%rsp) movl $1024, 84(%rsp) leaq 76(%rsp), %r13 leaq 88(%rsp), %r15 leaq .LC2(%rip), %r14 jmp .L29 .L41: movq 88(%rsp), %rax subq %fs:40, %rax jne .L37 call _ZSt16__throw_bad_castv@PLT .L37: call __stack_chk_fail@PLT .L21: movq %r12, %rdi call _ZNKSt5ctypeIcE13_M_widen_initEv@PLT movq (%r12), %rax movl $10, %esi movq %r12, %rdi call *48(%rax) movl %eax, %esi jmp .L22 .L23: call cudaDeviceSynchronize@PLT call _Z8get_timev subsd 8(%rsp), %xmm0 movsd .LC3(%rip), %xmm1 divsd %xmm0, %xmm1 mulsd .LC5(%rip), %xmm0 mulsd .LC4(%rip), %xmm1 movq %r12, %rsi movl $2, %edi movl $2, %eax call __printf_chk@PLT subl $1, %ebp je .L38 .L24: call _Z8get_timev movsd %xmm0, 8(%rsp) movl %ebx, 64(%rsp) movl $1, 68(%rsp) movl $1, 72(%rsp) movl $134217728, %eax movl $0, %edx idivl %ebx movl %eax, 52(%rsp) movl $1, 56(%rsp) movl $1, 60(%rsp) movl $0, %r9d movl $0, %r8d movq 64(%rsp), %rdx movl $1, %ecx movq 52(%rsp), %rdi movl $1, %esi call __cudaPushCallConfiguration@PLT testl %eax, %eax jne .L23 movl $134217728, %ecx movq 40(%rsp), %rdx movq 32(%rsp), %rsi movq 24(%rsp), %rdi call _Z26__device_stub__Z2fdPfS_PiiPfS_Pii jmp .L23 .L38: movq _ZSt4cout(%rip), %rax movq -24(%rax), %rax leaq _ZSt4cout(%rip), %rdx movq 240(%rdx,%rax), %rbx testq %rbx, %rbx je .L39 cmpb $0, 56(%rbx) je .L27 movzbl 67(%rbx), %esi .L28: movsbl %sil, %esi leaq _ZSt4cout(%rip), %rdi call _ZNSo3putEc@PLT movq %rax, %rdi call _ZNSo5flushEv@PLT addq $4, %r13 cmpq %r15, %r13 je .L40 .L29: movl 0(%r13), %ebx movl $5, %edx movq %r14, %rsi leaq _ZSt4cout(%rip), %rbp movq %rbp, %rdi call _ZSt16__ostream_insertIcSt11char_traitsIcEERSt13basic_ostreamIT_T0_ES6_PKS3_l@PLT movl %ebx, %esi movq %rbp, %rdi call _ZNSolsEi@PLT movq %rax, %rbp movq (%rax), %rax movq -24(%rax), %rax movq 240(%rbp,%rax), %r12 testq %r12, %r12 je .L41 cmpb $0, 56(%r12) je .L21 movzbl 67(%r12), %esi .L22: movsbl %sil, %esi movq %rbp, %rdi call _ZNSo3putEc@PLT movq %rax, %rdi call _ZNSo5flushEv@PLT movl $4, %ebp leaq .LC6(%rip), %r12 jmp .L24 .L39: movq 88(%rsp), %rax subq %fs:40, %rax jne .L42 call _ZSt16__throw_bad_castv@PLT .L42: call __stack_chk_fail@PLT .L27: movq %rbx, %rdi call _ZNKSt5ctypeIcE13_M_widen_initEv@PLT movq (%rbx), %rax movl $10, %esi movq %rbx, %rdi call *48(%rax) movl %eax, %esi jmp .L28 .L40: movq 88(%rsp), %rax subq %fs:40, %rax jne .L43 movl $0, %eax addq $104, %rsp .cfi_remember_state .cfi_def_cfa_offset 56 popq %rbx .cfi_def_cfa_offset 48 popq %rbp .cfi_def_cfa_offset 40 popq %r12 .cfi_def_cfa_offset 32 popq %r13 .cfi_def_cfa_offset 24 popq %r14 .cfi_def_cfa_offset 16 popq %r15 .cfi_def_cfa_offset 8 ret .L43: .cfi_restore_state call __stack_chk_fail@PLT .cfi_endproc .LFE3671: .size main, .-main .section .rodata.str1.1 .LC7: .string "_Z2fdPfS_Pii" .LC8: .string "const_c" .text .type _ZL24__sti____cudaRegisterAllv, @function _ZL24__sti____cudaRegisterAllv: .LFB3699: .cfi_startproc endbr64 pushq %rbx .cfi_def_cfa_offset 16 .cfi_offset 3, -16 leaq _ZL15__fatDeviceText(%rip), %rdi call __cudaRegisterFatBinary@PLT movq %rax, %rbx movq %rax, _ZL20__cudaFatCubinHandle(%rip) pushq $0 .cfi_def_cfa_offset 24 pushq $0 .cfi_def_cfa_offset 32 pushq $0 .cfi_def_cfa_offset 40 pushq $0 .cfi_def_cfa_offset 48 movl $0, %r9d movl $-1, %r8d leaq .LC7(%rip), %rdx movq %rdx, %rcx leaq _Z2fdPfS_Pii(%rip), %rsi movq %rax, %rdi call __cudaRegisterFunction@PLT addq $32, %rsp .cfi_def_cfa_offset 16 pushq $0 .cfi_def_cfa_offset 24 pushq $1 .cfi_def_cfa_offset 32 movl $1024, %r9d movl $0, %r8d leaq .LC8(%rip), %rdx movq %rdx, %rcx leaq _ZL7const_c(%rip), %rsi movq %rbx, %rdi call __cudaRegisterVar@PLT addq $16, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaRegisterFatBinaryEnd@PLT leaq _ZL26__cudaUnregisterBinaryUtilv(%rip), %rdi call atexit@PLT popq %rbx .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE3699: .size _ZL24__sti____cudaRegisterAllv, .-_ZL24__sti____cudaRegisterAllv .section .init_array,"aw" .align 8 .quad _ZL24__sti____cudaRegisterAllv .section .nvFatBinSegment,"aw" .align 8 .type _ZL15__fatDeviceText, @object .size _ZL15__fatDeviceText, 24 _ZL15__fatDeviceText: .long 1180844977 .long 1 .quad fatbinData .quad 0 .local _ZL20__cudaFatCubinHandle .comm _ZL20__cudaFatCubinHandle,8,8 .local _ZL7const_c .comm _ZL7const_c,1024,32 .section .rodata.cst8,"aM",@progbits,8 .align 8 .LC0: .long -1598689907 .long 1051772663 .section .rodata.cst4,"aM",@progbits,4 .align 4 .LC1: .long 925353388 .section .rodata.cst8 .align 8 .LC3: .long 0 .long 1104150528 .align 8 .LC4: .long 0 .long 1041235968 .align 8 .LC5: .long 0 .long 1083129856 .ident "GCC: (Ubuntu 13.3.0-6ubuntu2~24.04) 13.3.0" .section .note.GNU-stack,"",@progbits .section .note.gnu.property,"a" .align 8 .long 1f - 0f .long 4f - 1f .long 5 0: .string "GNU" 1: .align 8 .long 0xc0000002 .long 3f - 2f 2: .long 0x3 3: .align 8 4:
You are a helpful coding assistant specialized in converting CUDA source code to HIP/AMD source code.
#include <iostream> #include <cstdlib> #include <cstdio> #include <cuda_runtime.h> #include <sys/time.h> double get_time() { struct timeval tv; gettimeofday(&tv, nullptr); return tv.tv_sec + 1e-6 * tv.tv_usec; } constexpr int m = 256; inline __device__ int indirect(int *c, int i) { // return c[c[i & 127] & 127] + i; return int(exp(((((float(i))))) * 1e-18)) + i; // printf("%d\n", c[i % m] - i % m + i - i); // return i; } __constant__ int const_c[m]; __global__ void fd(float *a, float *b, int *c, int n) { __shared__ float b_s[m]; unsigned int i = blockIdx.x * blockDim.x + threadIdx.x; if (i < m) { b_s[i] = 0; } __syncthreads(); /* if (threadIdx.x < m) { b_s[threadIdx.x] = c[threadIdx.x]; } __syncthreads(); */ /* float sum = 0; if (i > 0) sum += a[indirect(c, i) - 1]; */ // sum += a[indirect(c, i)]; // sum += a[i + b_s[i & 127]]; /* if (i < n - 1) sum += a[indirect(c, i) + 1]; */ // b[i] = (i * 1e-18); // b[i] = i; // b[i] = c[c[c[i & 64]]]; // atomicAdd(b_s + ((unsigned)i * 34252345627) % m, 1.0f); // i = int(((((i * 1e-20f))))); // i = (i * 1e-10f); // i = i * i * i * i * i % m; // b_s[i % m] = 1; // #define C(x) i += (i >> x); // #define C(x) i += (i >> x); // for (int t = 0; t < 240; t++) // C(30); i += int(sin(i * 1e-20f)); i += int(sin(i * 1e-20f)); i += int(sin(i * 1e-20f)); i += int(sin(i * 1e-20f)); i += int(sin(i * 1e-20f)); i += int(sin(i * 1e-20f)); i += int(sin(i * 1e-20f)); i += int(sin(i * 1e-20f)); i += int(sin(i * 1e-20f)); i += int(sin(i * 1e-20f)); for (int j = 0; j < 27; j++) { atomicAdd(b_s + (unsigned int)(i / 4 + j * 431) % (m / 1), 1.0f); } __syncthreads(); if (i < m) { atomicAdd(&b[i], b_s[i]); } // atomicAdd(b + i % (m * m), 1); /* atomicAdd(&b_s[0], sqrt(sum)); if (threadIdx.x < m) { atomicAdd(b + threadIdx.x, b_s[threadIdx.x]); // b[threadIdx.x] += b_s[threadIdx.x]; } */ } int main() { int n = 128 * 1024 * 1024; float *a, *b; int *c; cudaMallocManaged(&a, n * sizeof(float)); cudaMallocManaged(&b, n * sizeof(float)); cudaMallocManaged(&c, m * sizeof(float)); for (int i = 0; i < n; i++) { a[i] = i * 1e-5f; } for (int i = 0; i < n; i++) { b[i] = i * 1e-5f; } for (int i = 0; i < m; i++) { c[i] = 0; } cudaMemcpyToSymbol(const_c, c, m * sizeof(float), 0, cudaMemcpyHostToDevice); for (auto bs : {256, 512, 1024}) { std::cout << "bs = " << bs << std::endl; for (int i = 0; i < 4; i++) { auto t = get_time(); fd<<<n / bs, bs>>>(a, b, c, n); cudaDeviceSynchronize(); t = get_time() - t; printf("%.2f ms bw %.3f GB/s\n", t * 1000, n * 2.0f * 4 / t / (1024 * 1024 * 1024.0f)); } std::cout << std::endl; } }
#include <iostream> #include <cstdlib> #include <cstdio> #include <hip/hip_runtime.h> #include <sys/time.h> double get_time() { struct timeval tv; gettimeofday(&tv, nullptr); return tv.tv_sec + 1e-6 * tv.tv_usec; } constexpr int m = 256; inline __device__ int indirect(int *c, int i) { // return c[c[i & 127] & 127] + i; return int(exp(((((float(i))))) * 1e-18)) + i; // printf("%d\n", c[i % m] - i % m + i - i); // return i; } __constant__ int const_c[m]; __global__ void fd(float *a, float *b, int *c, int n) { __shared__ float b_s[m]; unsigned int i = blockIdx.x * blockDim.x + threadIdx.x; if (i < m) { b_s[i] = 0; } __syncthreads(); /* if (threadIdx.x < m) { b_s[threadIdx.x] = c[threadIdx.x]; } __syncthreads(); */ /* float sum = 0; if (i > 0) sum += a[indirect(c, i) - 1]; */ // sum += a[indirect(c, i)]; // sum += a[i + b_s[i & 127]]; /* if (i < n - 1) sum += a[indirect(c, i) + 1]; */ // b[i] = (i * 1e-18); // b[i] = i; // b[i] = c[c[c[i & 64]]]; // atomicAdd(b_s + ((unsigned)i * 34252345627) % m, 1.0f); // i = int(((((i * 1e-20f))))); // i = (i * 1e-10f); // i = i * i * i * i * i % m; // b_s[i % m] = 1; // #define C(x) i += (i >> x); // #define C(x) i += (i >> x); // for (int t = 0; t < 240; t++) // C(30); i += int(sin(i * 1e-20f)); i += int(sin(i * 1e-20f)); i += int(sin(i * 1e-20f)); i += int(sin(i * 1e-20f)); i += int(sin(i * 1e-20f)); i += int(sin(i * 1e-20f)); i += int(sin(i * 1e-20f)); i += int(sin(i * 1e-20f)); i += int(sin(i * 1e-20f)); i += int(sin(i * 1e-20f)); for (int j = 0; j < 27; j++) { atomicAdd(b_s + (unsigned int)(i / 4 + j * 431) % (m / 1), 1.0f); } __syncthreads(); if (i < m) { atomicAdd(&b[i], b_s[i]); } // atomicAdd(b + i % (m * m), 1); /* atomicAdd(&b_s[0], sqrt(sum)); if (threadIdx.x < m) { atomicAdd(b + threadIdx.x, b_s[threadIdx.x]); // b[threadIdx.x] += b_s[threadIdx.x]; } */ } int main() { int n = 128 * 1024 * 1024; float *a, *b; int *c; hipMallocManaged(&a, n * sizeof(float)); hipMallocManaged(&b, n * sizeof(float)); hipMallocManaged(&c, m * sizeof(float)); for (int i = 0; i < n; i++) { a[i] = i * 1e-5f; } for (int i = 0; i < n; i++) { b[i] = i * 1e-5f; } for (int i = 0; i < m; i++) { c[i] = 0; } hipMemcpyToSymbol(HIP_SYMBOL(const_c), c, m * sizeof(float), 0, hipMemcpyHostToDevice); for (auto bs : {256, 512, 1024}) { std::cout << "bs = " << bs << std::endl; for (int i = 0; i < 4; i++) { auto t = get_time(); fd<<<n / bs, bs>>>(a, b, c, n); hipDeviceSynchronize(); t = get_time() - t; printf("%.2f ms bw %.3f GB/s\n", t * 1000, n * 2.0f * 4 / t / (1024 * 1024 * 1024.0f)); } std::cout << std::endl; } }
You are a helpful coding assistant specialized in converting HIP source code to AMD host assembly.
#include <iostream> #include <cstdlib> #include <cstdio> #include <hip/hip_runtime.h> #include <sys/time.h> double get_time() { struct timeval tv; gettimeofday(&tv, nullptr); return tv.tv_sec + 1e-6 * tv.tv_usec; } constexpr int m = 256; inline __device__ int indirect(int *c, int i) { // return c[c[i & 127] & 127] + i; return int(exp(((((float(i))))) * 1e-18)) + i; // printf("%d\n", c[i % m] - i % m + i - i); // return i; } __constant__ int const_c[m]; __global__ void fd(float *a, float *b, int *c, int n) { __shared__ float b_s[m]; unsigned int i = blockIdx.x * blockDim.x + threadIdx.x; if (i < m) { b_s[i] = 0; } __syncthreads(); /* if (threadIdx.x < m) { b_s[threadIdx.x] = c[threadIdx.x]; } __syncthreads(); */ /* float sum = 0; if (i > 0) sum += a[indirect(c, i) - 1]; */ // sum += a[indirect(c, i)]; // sum += a[i + b_s[i & 127]]; /* if (i < n - 1) sum += a[indirect(c, i) + 1]; */ // b[i] = (i * 1e-18); // b[i] = i; // b[i] = c[c[c[i & 64]]]; // atomicAdd(b_s + ((unsigned)i * 34252345627) % m, 1.0f); // i = int(((((i * 1e-20f))))); // i = (i * 1e-10f); // i = i * i * i * i * i % m; // b_s[i % m] = 1; // #define C(x) i += (i >> x); // #define C(x) i += (i >> x); // for (int t = 0; t < 240; t++) // C(30); i += int(sin(i * 1e-20f)); i += int(sin(i * 1e-20f)); i += int(sin(i * 1e-20f)); i += int(sin(i * 1e-20f)); i += int(sin(i * 1e-20f)); i += int(sin(i * 1e-20f)); i += int(sin(i * 1e-20f)); i += int(sin(i * 1e-20f)); i += int(sin(i * 1e-20f)); i += int(sin(i * 1e-20f)); for (int j = 0; j < 27; j++) { atomicAdd(b_s + (unsigned int)(i / 4 + j * 431) % (m / 1), 1.0f); } __syncthreads(); if (i < m) { atomicAdd(&b[i], b_s[i]); } // atomicAdd(b + i % (m * m), 1); /* atomicAdd(&b_s[0], sqrt(sum)); if (threadIdx.x < m) { atomicAdd(b + threadIdx.x, b_s[threadIdx.x]); // b[threadIdx.x] += b_s[threadIdx.x]; } */ } int main() { int n = 128 * 1024 * 1024; float *a, *b; int *c; hipMallocManaged(&a, n * sizeof(float)); hipMallocManaged(&b, n * sizeof(float)); hipMallocManaged(&c, m * sizeof(float)); for (int i = 0; i < n; i++) { a[i] = i * 1e-5f; } for (int i = 0; i < n; i++) { b[i] = i * 1e-5f; } for (int i = 0; i < m; i++) { c[i] = 0; } hipMemcpyToSymbol(HIP_SYMBOL(const_c), c, m * sizeof(float), 0, hipMemcpyHostToDevice); for (auto bs : {256, 512, 1024}) { std::cout << "bs = " << bs << std::endl; for (int i = 0; i < 4; i++) { auto t = get_time(); fd<<<n / bs, bs>>>(a, b, c, n); hipDeviceSynchronize(); t = get_time() - t; printf("%.2f ms bw %.3f GB/s\n", t * 1000, n * 2.0f * 4 / t / (1024 * 1024 * 1024.0f)); } std::cout << std::endl; } }
.text .file "bandwidth.hip" # Start of file scope inline assembly .globl _ZSt21ios_base_library_initv # End of file scope inline assembly .section .rodata.cst8,"aM",@progbits,8 .p2align 3, 0x0 # -- Begin function _Z8get_timev .LCPI0_0: .quad 0x3eb0c6f7a0b5ed8d # double 9.9999999999999995E-7 .text .globl _Z8get_timev .p2align 4, 0x90 .type _Z8get_timev,@function _Z8get_timev: # @_Z8get_timev .cfi_startproc # %bb.0: subq $24, %rsp .cfi_def_cfa_offset 32 leaq 8(%rsp), %rdi xorl %esi, %esi callq gettimeofday cvtsi2sdq 8(%rsp), %xmm1 cvtsi2sdq 16(%rsp), %xmm0 mulsd .LCPI0_0(%rip), %xmm0 addsd %xmm1, %xmm0 addq $24, %rsp .cfi_def_cfa_offset 8 retq .Lfunc_end0: .size _Z8get_timev, .Lfunc_end0-_Z8get_timev .cfi_endproc # -- End function .globl _Z17__device_stub__fdPfS_Pii # -- Begin function _Z17__device_stub__fdPfS_Pii .p2align 4, 0x90 .type _Z17__device_stub__fdPfS_Pii,@function _Z17__device_stub__fdPfS_Pii: # @_Z17__device_stub__fdPfS_Pii .cfi_startproc # %bb.0: subq $120, %rsp .cfi_def_cfa_offset 128 movq %rdi, 72(%rsp) movq %rsi, 64(%rsp) movq %rdx, 56(%rsp) movl %ecx, 4(%rsp) leaq 72(%rsp), %rax movq %rax, 80(%rsp) leaq 64(%rsp), %rax movq %rax, 88(%rsp) leaq 56(%rsp), %rax movq %rax, 96(%rsp) leaq 4(%rsp), %rax movq %rax, 104(%rsp) leaq 40(%rsp), %rdi leaq 24(%rsp), %rsi leaq 16(%rsp), %rdx leaq 8(%rsp), %rcx callq __hipPopCallConfiguration movq 40(%rsp), %rsi movl 48(%rsp), %edx movq 24(%rsp), %rcx movl 32(%rsp), %r8d leaq 80(%rsp), %r9 movl $_Z2fdPfS_Pii, %edi pushq 8(%rsp) .cfi_adjust_cfa_offset 8 pushq 24(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $136, %rsp .cfi_adjust_cfa_offset -136 retq .Lfunc_end1: .size _Z17__device_stub__fdPfS_Pii, .Lfunc_end1-_Z17__device_stub__fdPfS_Pii .cfi_endproc # -- End function .section .rodata.cst4,"aM",@progbits,4 .p2align 2, 0x0 # -- Begin function main .LCPI2_0: .long 0x3727c5ac # float 9.99999974E-6 .section .rodata.cst8,"aM",@progbits,8 .p2align 3, 0x0 .LCPI2_1: .quad 0x3eb0c6f7a0b5ed8d # double 9.9999999999999995E-7 .LCPI2_2: .quad 0x408f400000000000 # double 1000 .LCPI2_3: .quad 0x41d0000000000000 # double 1073741824 .LCPI2_4: .quad 0x3e10000000000000 # double 9.3132257461547852E-10 .text .globl main .p2align 4, 0x90 .type main,@function main: # @main .cfi_startproc # %bb.0: pushq %rbp .cfi_def_cfa_offset 16 pushq %r15 .cfi_def_cfa_offset 24 pushq %r14 .cfi_def_cfa_offset 32 pushq %r13 .cfi_def_cfa_offset 40 pushq %r12 .cfi_def_cfa_offset 48 pushq %rbx .cfi_def_cfa_offset 56 subq $184, %rsp .cfi_def_cfa_offset 240 .cfi_offset %rbx, -56 .cfi_offset %r12, -48 .cfi_offset %r13, -40 .cfi_offset %r14, -32 .cfi_offset %r15, -24 .cfi_offset %rbp, -16 leaq 40(%rsp), %rdi movl $536870912, %esi # imm = 0x20000000 movl $1, %edx callq hipMallocManaged leaq 32(%rsp), %rdi movl $536870912, %esi # imm = 0x20000000 movl $1, %edx callq hipMallocManaged leaq 24(%rsp), %rdi movl $1024, %esi # imm = 0x400 movl $1, %edx callq hipMallocManaged xorl %eax, %eax movq 40(%rsp), %rcx movss .LCPI2_0(%rip), %xmm0 # xmm0 = mem[0],zero,zero,zero .p2align 4, 0x90 .LBB2_1: # =>This Inner Loop Header: Depth=1 xorps %xmm1, %xmm1 cvtsi2ss %eax, %xmm1 mulss %xmm0, %xmm1 movss %xmm1, (%rcx,%rax,4) incq %rax cmpq $134217728, %rax # imm = 0x8000000 jne .LBB2_1 # %bb.2: # %.preheader49 movq 32(%rsp), %rax xorl %ecx, %ecx .p2align 4, 0x90 .LBB2_3: # =>This Inner Loop Header: Depth=1 xorps %xmm1, %xmm1 cvtsi2ss %ecx, %xmm1 mulss %xmm0, %xmm1 movss %xmm1, (%rax,%rcx,4) incq %rcx cmpq $134217728, %rcx # imm = 0x8000000 jne .LBB2_3 # %bb.4: # %.preheader movabsq $4294967296, %r15 # imm = 0x100000000 movq 24(%rsp), %rbx xorl %r12d, %r12d movl $1024, %edx # imm = 0x400 movq %rbx, %rdi xorl %esi, %esi callq memset@PLT movl $const_c, %edi movl $1024, %edx # imm = 0x400 movq %rbx, %rsi xorl %ecx, %ecx movl $1, %r8d callq hipMemcpyToSymbol movabsq $2199023255808, %rax # imm = 0x20000000100 movq %rax, 172(%rsp) movl $1024, 180(%rsp) # imm = 0x400 leaq 48(%rsp), %rbx jmp .LBB2_5 .p2align 4, 0x90 .LBB2_16: # in Loop: Header=BB2_5 Depth=1 movq %r14, %rdi callq _ZNKSt5ctypeIcE13_M_widen_initEv movq (%r14), %rax movq %r14, %rdi movl $10, %esi callq *48(%rax) .LBB2_17: # %_ZSt4endlIcSt11char_traitsIcEERSt13basic_ostreamIT_T0_ES6_.exit43 # in Loop: Header=BB2_5 Depth=1 movsbl %al, %esi movl $_ZSt4cout, %edi callq _ZNSo3putEc movq %rax, %rdi callq _ZNSo5flushEv addq $4, %r12 cmpq $12, %r12 je .LBB2_18 .LBB2_5: # =>This Loop Header: Depth=1 # Child Loop BB2_10 Depth 2 movl 172(%rsp,%r12), %r14d movl $_ZSt4cout, %edi movl $.L.str, %esi movl $5, %edx callq _ZSt16__ostream_insertIcSt11char_traitsIcEERSt13basic_ostreamIT_T0_ES6_PKS3_l movl $_ZSt4cout, %edi movl %r14d, %esi callq _ZNSolsEi movq (%rax), %rcx movq -24(%rcx), %rcx movq 240(%rax,%rcx), %rbp testq %rbp, %rbp je .LBB2_19 # %bb.6: # %_ZSt13__check_facetISt5ctypeIcEERKT_PS3_.exit.i.i # in Loop: Header=BB2_5 Depth=1 cmpb $0, 56(%rbp) je .LBB2_8 # %bb.7: # in Loop: Header=BB2_5 Depth=1 movzbl 67(%rbp), %ecx jmp .LBB2_9 .p2align 4, 0x90 .LBB2_8: # in Loop: Header=BB2_5 Depth=1 movq %rbp, %rdi movq %rax, %r13 callq _ZNKSt5ctypeIcE13_M_widen_initEv movq (%rbp), %rax movq %rbp, %rdi movl $10, %esi callq *48(%rax) movl %eax, %ecx movq %r13, %rax .LBB2_9: # %_ZSt4endlIcSt11char_traitsIcEERSt13basic_ostreamIT_T0_ES6_.exit # in Loop: Header=BB2_5 Depth=1 movsbl %cl, %esi movq %rax, %rdi callq _ZNSo3putEc movq %rax, %rdi callq _ZNSo5flushEv movq %r14, %rbp orq %r15, %rbp movl $4, %r13d jmp .LBB2_10 .p2align 4, 0x90 .LBB2_12: # in Loop: Header=BB2_10 Depth=2 movsd 8(%rsp), %xmm0 # 8-byte Reload # xmm0 = mem[0],zero addsd 88(%rsp), %xmm0 # 8-byte Folded Reload movsd %xmm0, 8(%rsp) # 8-byte Spill callq hipDeviceSynchronize movq %rbx, %rdi xorl %esi, %esi callq gettimeofday xorps %xmm1, %xmm1 cvtsi2sdq 48(%rsp), %xmm1 xorps %xmm0, %xmm0 cvtsi2sdq 56(%rsp), %xmm0 mulsd .LCPI2_1(%rip), %xmm0 addsd %xmm1, %xmm0 subsd 8(%rsp), %xmm0 # 8-byte Folded Reload movsd .LCPI2_3(%rip), %xmm1 # xmm1 = mem[0],zero divsd %xmm0, %xmm1 mulsd .LCPI2_2(%rip), %xmm0 mulsd .LCPI2_4(%rip), %xmm1 movl $.L.str.1, %edi movb $2, %al callq printf decl %r13d je .LBB2_13 .LBB2_10: # Parent Loop BB2_5 Depth=1 # => This Inner Loop Header: Depth=2 movq %rbx, %rdi xorl %esi, %esi callq gettimeofday xorps %xmm0, %xmm0 cvtsi2sdq 48(%rsp), %xmm0 movsd %xmm0, 88(%rsp) # 8-byte Spill xorps %xmm0, %xmm0 cvtsi2sdq 56(%rsp), %xmm0 mulsd .LCPI2_1(%rip), %xmm0 movsd %xmm0, 8(%rsp) # 8-byte Spill movl $134217728, %eax # imm = 0x8000000 xorl %edx, %edx idivl %r14d # kill: def $eax killed $eax def $rax orq %r15, %rax movq %rax, %rdi movl $1, %esi movq %rbp, %rdx movl $1, %ecx xorl %r8d, %r8d xorl %r9d, %r9d callq __hipPushCallConfiguration testl %eax, %eax jne .LBB2_12 # %bb.11: # in Loop: Header=BB2_10 Depth=2 movq 40(%rsp), %rax movq 32(%rsp), %rcx movq 24(%rsp), %rdx movq %rax, 160(%rsp) movq %rcx, 152(%rsp) movq %rdx, 144(%rsp) movl $134217728, 20(%rsp) # imm = 0x8000000 leaq 160(%rsp), %rax movq %rax, 48(%rsp) leaq 152(%rsp), %rax movq %rax, 56(%rsp) leaq 144(%rsp), %rax movq %rax, 64(%rsp) leaq 20(%rsp), %rax movq %rax, 72(%rsp) leaq 128(%rsp), %rdi leaq 112(%rsp), %rsi leaq 104(%rsp), %rdx leaq 96(%rsp), %rcx callq __hipPopCallConfiguration movq 128(%rsp), %rsi movl 136(%rsp), %edx movq 112(%rsp), %rcx movl 120(%rsp), %r8d movl $_Z2fdPfS_Pii, %edi movq %rbx, %r9 pushq 96(%rsp) .cfi_adjust_cfa_offset 8 pushq 112(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $16, %rsp .cfi_adjust_cfa_offset -16 jmp .LBB2_12 .p2align 4, 0x90 .LBB2_13: # in Loop: Header=BB2_5 Depth=1 movq _ZSt4cout(%rip), %rax movq -24(%rax), %rax movq _ZSt4cout+240(%rax), %r14 testq %r14, %r14 je .LBB2_19 # %bb.14: # %_ZSt13__check_facetISt5ctypeIcEERKT_PS3_.exit.i.i40 # in Loop: Header=BB2_5 Depth=1 cmpb $0, 56(%r14) je .LBB2_16 # %bb.15: # in Loop: Header=BB2_5 Depth=1 movzbl 67(%r14), %eax jmp .LBB2_17 .LBB2_18: xorl %eax, %eax addq $184, %rsp .cfi_def_cfa_offset 56 popq %rbx .cfi_def_cfa_offset 48 popq %r12 .cfi_def_cfa_offset 40 popq %r13 .cfi_def_cfa_offset 32 popq %r14 .cfi_def_cfa_offset 24 popq %r15 .cfi_def_cfa_offset 16 popq %rbp .cfi_def_cfa_offset 8 retq .LBB2_19: .cfi_def_cfa_offset 240 callq _ZSt16__throw_bad_castv .Lfunc_end2: .size main, .Lfunc_end2-main .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_ctor .type __hip_module_ctor,@function __hip_module_ctor: # @__hip_module_ctor .cfi_startproc # %bb.0: pushq %rbx .cfi_def_cfa_offset 16 subq $32, %rsp .cfi_def_cfa_offset 48 .cfi_offset %rbx, -16 cmpq $0, __hip_gpubin_handle(%rip) jne .LBB3_2 # %bb.1: movl $__hip_fatbin_wrapper, %edi callq __hipRegisterFatBinary movq %rax, __hip_gpubin_handle(%rip) .LBB3_2: movq __hip_gpubin_handle(%rip), %rbx xorps %xmm0, %xmm0 movups %xmm0, 16(%rsp) movups %xmm0, (%rsp) movl $_Z2fdPfS_Pii, %esi movl $.L__unnamed_1, %edx movl $.L__unnamed_1, %ecx movq %rbx, %rdi movl $-1, %r8d xorl %r9d, %r9d callq __hipRegisterFunction movl $0, 8(%rsp) movl $1, (%rsp) movl $const_c, %esi movl $.L__unnamed_2, %edx movl $.L__unnamed_2, %ecx movl $1024, %r9d # imm = 0x400 movq %rbx, %rdi xorl %r8d, %r8d callq __hipRegisterVar movl $__hip_module_dtor, %edi addq $32, %rsp .cfi_def_cfa_offset 16 popq %rbx .cfi_def_cfa_offset 8 jmp atexit # TAILCALL .Lfunc_end3: .size __hip_module_ctor, .Lfunc_end3-__hip_module_ctor .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_dtor .type __hip_module_dtor,@function __hip_module_dtor: # @__hip_module_dtor .cfi_startproc # %bb.0: movq __hip_gpubin_handle(%rip), %rdi testq %rdi, %rdi je .LBB4_2 # %bb.1: pushq %rax .cfi_def_cfa_offset 16 callq __hipUnregisterFatBinary movq $0, __hip_gpubin_handle(%rip) addq $8, %rsp .cfi_def_cfa_offset 8 .LBB4_2: retq .Lfunc_end4: .size __hip_module_dtor, .Lfunc_end4-__hip_module_dtor .cfi_endproc # -- End function .type const_c,@object # @const_c .local const_c .comm const_c,1024,16 .type _Z2fdPfS_Pii,@object # @_Z2fdPfS_Pii .section .rodata,"a",@progbits .globl _Z2fdPfS_Pii .p2align 3, 0x0 _Z2fdPfS_Pii: .quad _Z17__device_stub__fdPfS_Pii .size _Z2fdPfS_Pii, 8 .type .L.str,@object # @.str .section .rodata.str1.1,"aMS",@progbits,1 .L.str: .asciz "bs = " .size .L.str, 6 .type .L.str.1,@object # @.str.1 .L.str.1: .asciz "%.2f ms bw %.3f GB/s\n" .size .L.str.1, 24 .type .L__unnamed_1,@object # @0 .L__unnamed_1: .asciz "_Z2fdPfS_Pii" .size .L__unnamed_1, 13 .type .L__unnamed_2,@object # @1 .L__unnamed_2: .asciz "const_c" .size .L__unnamed_2, 8 .type __hip_fatbin_wrapper,@object # @__hip_fatbin_wrapper .section .hipFatBinSegment,"a",@progbits .p2align 3, 0x0 __hip_fatbin_wrapper: .long 1212764230 # 0x48495046 .long 1 # 0x1 .quad __hip_fatbin .quad 0 .size __hip_fatbin_wrapper, 24 .type __hip_gpubin_handle,@object # @__hip_gpubin_handle .local __hip_gpubin_handle .comm __hip_gpubin_handle,8,8 .section .init_array,"aw",@init_array .p2align 3, 0x0 .quad __hip_module_ctor .type __hip_cuid_,@object # @__hip_cuid_ .bss .globl __hip_cuid_ __hip_cuid_: .byte 0 # 0x0 .size __hip_cuid_, 1 .section ".linker-options","e",@llvm_linker_options .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym _Z17__device_stub__fdPfS_Pii .addrsig_sym __hip_module_ctor .addrsig_sym __hip_module_dtor .addrsig_sym const_c .addrsig_sym _Z2fdPfS_Pii .addrsig_sym _ZSt4cout .addrsig_sym __hip_fatbin .addrsig_sym __hip_fatbin_wrapper .addrsig_sym __hip_cuid_
You are a helpful coding assistant specialized in converting from CUDA host assembly to AMD host assembly.
.file "tmpxft_000ef42a_00000000-6_bandwidth.cudafe1.cpp" .text #APP .globl _ZSt21ios_base_library_initv #NO_APP .type _ZL26__cudaUnregisterBinaryUtilv, @function _ZL26__cudaUnregisterBinaryUtilv: .LFB3674: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaUnregisterFatBinary@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE3674: .size _ZL26__cudaUnregisterBinaryUtilv, .-_ZL26__cudaUnregisterBinaryUtilv .globl _Z8get_timev .type _Z8get_timev, @function _Z8get_timev: .LFB3669: .cfi_startproc endbr64 subq $40, %rsp .cfi_def_cfa_offset 48 movq %fs:40, %rax movq %rax, 24(%rsp) xorl %eax, %eax movq %rsp, %rdi movl $0, %esi call gettimeofday@PLT pxor %xmm0, %xmm0 cvtsi2sdq 8(%rsp), %xmm0 mulsd .LC0(%rip), %xmm0 pxor %xmm1, %xmm1 cvtsi2sdq (%rsp), %xmm1 addsd %xmm1, %xmm0 movq 24(%rsp), %rax subq %fs:40, %rax jne .L6 addq $40, %rsp .cfi_remember_state .cfi_def_cfa_offset 8 ret .L6: .cfi_restore_state call __stack_chk_fail@PLT .cfi_endproc .LFE3669: .size _Z8get_timev, .-_Z8get_timev .globl _Z26__device_stub__Z2fdPfS_PiiPfS_Pii .type _Z26__device_stub__Z2fdPfS_PiiPfS_Pii, @function _Z26__device_stub__Z2fdPfS_PiiPfS_Pii: .LFB3696: .cfi_startproc endbr64 subq $152, %rsp .cfi_def_cfa_offset 160 movq %rdi, 24(%rsp) movq %rsi, 16(%rsp) movq %rdx, 8(%rsp) movl %ecx, 4(%rsp) movq %fs:40, %rax movq %rax, 136(%rsp) xorl %eax, %eax leaq 24(%rsp), %rax movq %rax, 96(%rsp) leaq 16(%rsp), %rax movq %rax, 104(%rsp) leaq 8(%rsp), %rax movq %rax, 112(%rsp) leaq 4(%rsp), %rax movq %rax, 120(%rsp) movl $1, 48(%rsp) movl $1, 52(%rsp) movl $1, 56(%rsp) movl $1, 60(%rsp) movl $1, 64(%rsp) movl $1, 68(%rsp) leaq 40(%rsp), %rcx leaq 32(%rsp), %rdx leaq 60(%rsp), %rsi leaq 48(%rsp), %rdi call __cudaPopCallConfiguration@PLT testl %eax, %eax je .L11 .L7: movq 136(%rsp), %rax subq %fs:40, %rax jne .L12 addq $152, %rsp .cfi_remember_state .cfi_def_cfa_offset 8 ret .L11: .cfi_restore_state pushq 40(%rsp) .cfi_def_cfa_offset 168 pushq 40(%rsp) .cfi_def_cfa_offset 176 leaq 112(%rsp), %r9 movq 76(%rsp), %rcx movl 84(%rsp), %r8d movq 64(%rsp), %rsi movl 72(%rsp), %edx leaq _Z2fdPfS_Pii(%rip), %rdi call cudaLaunchKernel@PLT addq $16, %rsp .cfi_def_cfa_offset 160 jmp .L7 .L12: call __stack_chk_fail@PLT .cfi_endproc .LFE3696: .size _Z26__device_stub__Z2fdPfS_PiiPfS_Pii, .-_Z26__device_stub__Z2fdPfS_PiiPfS_Pii .globl _Z2fdPfS_Pii .type _Z2fdPfS_Pii, @function _Z2fdPfS_Pii: .LFB3697: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 call _Z26__device_stub__Z2fdPfS_PiiPfS_Pii addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE3697: .size _Z2fdPfS_Pii, .-_Z2fdPfS_Pii .section .rodata.str1.1,"aMS",@progbits,1 .LC2: .string "bs = " .LC6: .string "%.2f ms bw %.3f GB/s\n" .text .globl main .type main, @function main: .LFB3671: .cfi_startproc endbr64 pushq %r15 .cfi_def_cfa_offset 16 .cfi_offset 15, -16 pushq %r14 .cfi_def_cfa_offset 24 .cfi_offset 14, -24 pushq %r13 .cfi_def_cfa_offset 32 .cfi_offset 13, -32 pushq %r12 .cfi_def_cfa_offset 40 .cfi_offset 12, -40 pushq %rbp .cfi_def_cfa_offset 48 .cfi_offset 6, -48 pushq %rbx .cfi_def_cfa_offset 56 .cfi_offset 3, -56 subq $104, %rsp .cfi_def_cfa_offset 160 movq %fs:40, %rax movq %rax, 88(%rsp) xorl %eax, %eax leaq 24(%rsp), %rdi movl $1, %edx movl $536870912, %esi call cudaMallocManaged@PLT leaq 32(%rsp), %rdi movl $1, %edx movl $536870912, %esi call cudaMallocManaged@PLT leaq 40(%rsp), %rdi movl $1, %edx movl $1024, %esi call cudaMallocManaged@PLT movl $0, %eax movss .LC1(%rip), %xmm1 .L16: pxor %xmm0, %xmm0 cvtsi2ssl %eax, %xmm0 mulss %xmm1, %xmm0 movq 24(%rsp), %rdx movss %xmm0, (%rdx,%rax,4) addq $1, %rax cmpq $134217728, %rax jne .L16 movl $0, %eax movss .LC1(%rip), %xmm1 .L17: pxor %xmm0, %xmm0 cvtsi2ssl %eax, %xmm0 mulss %xmm1, %xmm0 movq 32(%rsp), %rdx movss %xmm0, (%rdx,%rax,4) addq $1, %rax cmpq $134217728, %rax jne .L17 movl $0, %eax .L18: movq 40(%rsp), %rdx movl $0, (%rdx,%rax) addq $4, %rax cmpq $1024, %rax jne .L18 movl $1, %r8d movl $0, %ecx movl $1024, %edx movq 40(%rsp), %rsi leaq _ZL7const_c(%rip), %rdi call cudaMemcpyToSymbol@PLT movl $256, 76(%rsp) movl $512, 80(%rsp) movl $1024, 84(%rsp) leaq 76(%rsp), %r13 leaq 88(%rsp), %r15 leaq .LC2(%rip), %r14 jmp .L29 .L41: movq 88(%rsp), %rax subq %fs:40, %rax jne .L37 call _ZSt16__throw_bad_castv@PLT .L37: call __stack_chk_fail@PLT .L21: movq %r12, %rdi call _ZNKSt5ctypeIcE13_M_widen_initEv@PLT movq (%r12), %rax movl $10, %esi movq %r12, %rdi call *48(%rax) movl %eax, %esi jmp .L22 .L23: call cudaDeviceSynchronize@PLT call _Z8get_timev subsd 8(%rsp), %xmm0 movsd .LC3(%rip), %xmm1 divsd %xmm0, %xmm1 mulsd .LC5(%rip), %xmm0 mulsd .LC4(%rip), %xmm1 movq %r12, %rsi movl $2, %edi movl $2, %eax call __printf_chk@PLT subl $1, %ebp je .L38 .L24: call _Z8get_timev movsd %xmm0, 8(%rsp) movl %ebx, 64(%rsp) movl $1, 68(%rsp) movl $1, 72(%rsp) movl $134217728, %eax movl $0, %edx idivl %ebx movl %eax, 52(%rsp) movl $1, 56(%rsp) movl $1, 60(%rsp) movl $0, %r9d movl $0, %r8d movq 64(%rsp), %rdx movl $1, %ecx movq 52(%rsp), %rdi movl $1, %esi call __cudaPushCallConfiguration@PLT testl %eax, %eax jne .L23 movl $134217728, %ecx movq 40(%rsp), %rdx movq 32(%rsp), %rsi movq 24(%rsp), %rdi call _Z26__device_stub__Z2fdPfS_PiiPfS_Pii jmp .L23 .L38: movq _ZSt4cout(%rip), %rax movq -24(%rax), %rax leaq _ZSt4cout(%rip), %rdx movq 240(%rdx,%rax), %rbx testq %rbx, %rbx je .L39 cmpb $0, 56(%rbx) je .L27 movzbl 67(%rbx), %esi .L28: movsbl %sil, %esi leaq _ZSt4cout(%rip), %rdi call _ZNSo3putEc@PLT movq %rax, %rdi call _ZNSo5flushEv@PLT addq $4, %r13 cmpq %r15, %r13 je .L40 .L29: movl 0(%r13), %ebx movl $5, %edx movq %r14, %rsi leaq _ZSt4cout(%rip), %rbp movq %rbp, %rdi call _ZSt16__ostream_insertIcSt11char_traitsIcEERSt13basic_ostreamIT_T0_ES6_PKS3_l@PLT movl %ebx, %esi movq %rbp, %rdi call _ZNSolsEi@PLT movq %rax, %rbp movq (%rax), %rax movq -24(%rax), %rax movq 240(%rbp,%rax), %r12 testq %r12, %r12 je .L41 cmpb $0, 56(%r12) je .L21 movzbl 67(%r12), %esi .L22: movsbl %sil, %esi movq %rbp, %rdi call _ZNSo3putEc@PLT movq %rax, %rdi call _ZNSo5flushEv@PLT movl $4, %ebp leaq .LC6(%rip), %r12 jmp .L24 .L39: movq 88(%rsp), %rax subq %fs:40, %rax jne .L42 call _ZSt16__throw_bad_castv@PLT .L42: call __stack_chk_fail@PLT .L27: movq %rbx, %rdi call _ZNKSt5ctypeIcE13_M_widen_initEv@PLT movq (%rbx), %rax movl $10, %esi movq %rbx, %rdi call *48(%rax) movl %eax, %esi jmp .L28 .L40: movq 88(%rsp), %rax subq %fs:40, %rax jne .L43 movl $0, %eax addq $104, %rsp .cfi_remember_state .cfi_def_cfa_offset 56 popq %rbx .cfi_def_cfa_offset 48 popq %rbp .cfi_def_cfa_offset 40 popq %r12 .cfi_def_cfa_offset 32 popq %r13 .cfi_def_cfa_offset 24 popq %r14 .cfi_def_cfa_offset 16 popq %r15 .cfi_def_cfa_offset 8 ret .L43: .cfi_restore_state call __stack_chk_fail@PLT .cfi_endproc .LFE3671: .size main, .-main .section .rodata.str1.1 .LC7: .string "_Z2fdPfS_Pii" .LC8: .string "const_c" .text .type _ZL24__sti____cudaRegisterAllv, @function _ZL24__sti____cudaRegisterAllv: .LFB3699: .cfi_startproc endbr64 pushq %rbx .cfi_def_cfa_offset 16 .cfi_offset 3, -16 leaq _ZL15__fatDeviceText(%rip), %rdi call __cudaRegisterFatBinary@PLT movq %rax, %rbx movq %rax, _ZL20__cudaFatCubinHandle(%rip) pushq $0 .cfi_def_cfa_offset 24 pushq $0 .cfi_def_cfa_offset 32 pushq $0 .cfi_def_cfa_offset 40 pushq $0 .cfi_def_cfa_offset 48 movl $0, %r9d movl $-1, %r8d leaq .LC7(%rip), %rdx movq %rdx, %rcx leaq _Z2fdPfS_Pii(%rip), %rsi movq %rax, %rdi call __cudaRegisterFunction@PLT addq $32, %rsp .cfi_def_cfa_offset 16 pushq $0 .cfi_def_cfa_offset 24 pushq $1 .cfi_def_cfa_offset 32 movl $1024, %r9d movl $0, %r8d leaq .LC8(%rip), %rdx movq %rdx, %rcx leaq _ZL7const_c(%rip), %rsi movq %rbx, %rdi call __cudaRegisterVar@PLT addq $16, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaRegisterFatBinaryEnd@PLT leaq _ZL26__cudaUnregisterBinaryUtilv(%rip), %rdi call atexit@PLT popq %rbx .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE3699: .size _ZL24__sti____cudaRegisterAllv, .-_ZL24__sti____cudaRegisterAllv .section .init_array,"aw" .align 8 .quad _ZL24__sti____cudaRegisterAllv .section .nvFatBinSegment,"aw" .align 8 .type _ZL15__fatDeviceText, @object .size _ZL15__fatDeviceText, 24 _ZL15__fatDeviceText: .long 1180844977 .long 1 .quad fatbinData .quad 0 .local _ZL20__cudaFatCubinHandle .comm _ZL20__cudaFatCubinHandle,8,8 .local _ZL7const_c .comm _ZL7const_c,1024,32 .section .rodata.cst8,"aM",@progbits,8 .align 8 .LC0: .long -1598689907 .long 1051772663 .section .rodata.cst4,"aM",@progbits,4 .align 4 .LC1: .long 925353388 .section .rodata.cst8 .align 8 .LC3: .long 0 .long 1104150528 .align 8 .LC4: .long 0 .long 1041235968 .align 8 .LC5: .long 0 .long 1083129856 .ident "GCC: (Ubuntu 13.3.0-6ubuntu2~24.04) 13.3.0" .section .note.GNU-stack,"",@progbits .section .note.gnu.property,"a" .align 8 .long 1f - 0f .long 4f - 1f .long 5 0: .string "GNU" 1: .align 8 .long 0xc0000002 .long 3f - 2f 2: .long 0x3 3: .align 8 4:
.text .file "bandwidth.hip" # Start of file scope inline assembly .globl _ZSt21ios_base_library_initv # End of file scope inline assembly .section .rodata.cst8,"aM",@progbits,8 .p2align 3, 0x0 # -- Begin function _Z8get_timev .LCPI0_0: .quad 0x3eb0c6f7a0b5ed8d # double 9.9999999999999995E-7 .text .globl _Z8get_timev .p2align 4, 0x90 .type _Z8get_timev,@function _Z8get_timev: # @_Z8get_timev .cfi_startproc # %bb.0: subq $24, %rsp .cfi_def_cfa_offset 32 leaq 8(%rsp), %rdi xorl %esi, %esi callq gettimeofday cvtsi2sdq 8(%rsp), %xmm1 cvtsi2sdq 16(%rsp), %xmm0 mulsd .LCPI0_0(%rip), %xmm0 addsd %xmm1, %xmm0 addq $24, %rsp .cfi_def_cfa_offset 8 retq .Lfunc_end0: .size _Z8get_timev, .Lfunc_end0-_Z8get_timev .cfi_endproc # -- End function .globl _Z17__device_stub__fdPfS_Pii # -- Begin function _Z17__device_stub__fdPfS_Pii .p2align 4, 0x90 .type _Z17__device_stub__fdPfS_Pii,@function _Z17__device_stub__fdPfS_Pii: # @_Z17__device_stub__fdPfS_Pii .cfi_startproc # %bb.0: subq $120, %rsp .cfi_def_cfa_offset 128 movq %rdi, 72(%rsp) movq %rsi, 64(%rsp) movq %rdx, 56(%rsp) movl %ecx, 4(%rsp) leaq 72(%rsp), %rax movq %rax, 80(%rsp) leaq 64(%rsp), %rax movq %rax, 88(%rsp) leaq 56(%rsp), %rax movq %rax, 96(%rsp) leaq 4(%rsp), %rax movq %rax, 104(%rsp) leaq 40(%rsp), %rdi leaq 24(%rsp), %rsi leaq 16(%rsp), %rdx leaq 8(%rsp), %rcx callq __hipPopCallConfiguration movq 40(%rsp), %rsi movl 48(%rsp), %edx movq 24(%rsp), %rcx movl 32(%rsp), %r8d leaq 80(%rsp), %r9 movl $_Z2fdPfS_Pii, %edi pushq 8(%rsp) .cfi_adjust_cfa_offset 8 pushq 24(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $136, %rsp .cfi_adjust_cfa_offset -136 retq .Lfunc_end1: .size _Z17__device_stub__fdPfS_Pii, .Lfunc_end1-_Z17__device_stub__fdPfS_Pii .cfi_endproc # -- End function .section .rodata.cst4,"aM",@progbits,4 .p2align 2, 0x0 # -- Begin function main .LCPI2_0: .long 0x3727c5ac # float 9.99999974E-6 .section .rodata.cst8,"aM",@progbits,8 .p2align 3, 0x0 .LCPI2_1: .quad 0x3eb0c6f7a0b5ed8d # double 9.9999999999999995E-7 .LCPI2_2: .quad 0x408f400000000000 # double 1000 .LCPI2_3: .quad 0x41d0000000000000 # double 1073741824 .LCPI2_4: .quad 0x3e10000000000000 # double 9.3132257461547852E-10 .text .globl main .p2align 4, 0x90 .type main,@function main: # @main .cfi_startproc # %bb.0: pushq %rbp .cfi_def_cfa_offset 16 pushq %r15 .cfi_def_cfa_offset 24 pushq %r14 .cfi_def_cfa_offset 32 pushq %r13 .cfi_def_cfa_offset 40 pushq %r12 .cfi_def_cfa_offset 48 pushq %rbx .cfi_def_cfa_offset 56 subq $184, %rsp .cfi_def_cfa_offset 240 .cfi_offset %rbx, -56 .cfi_offset %r12, -48 .cfi_offset %r13, -40 .cfi_offset %r14, -32 .cfi_offset %r15, -24 .cfi_offset %rbp, -16 leaq 40(%rsp), %rdi movl $536870912, %esi # imm = 0x20000000 movl $1, %edx callq hipMallocManaged leaq 32(%rsp), %rdi movl $536870912, %esi # imm = 0x20000000 movl $1, %edx callq hipMallocManaged leaq 24(%rsp), %rdi movl $1024, %esi # imm = 0x400 movl $1, %edx callq hipMallocManaged xorl %eax, %eax movq 40(%rsp), %rcx movss .LCPI2_0(%rip), %xmm0 # xmm0 = mem[0],zero,zero,zero .p2align 4, 0x90 .LBB2_1: # =>This Inner Loop Header: Depth=1 xorps %xmm1, %xmm1 cvtsi2ss %eax, %xmm1 mulss %xmm0, %xmm1 movss %xmm1, (%rcx,%rax,4) incq %rax cmpq $134217728, %rax # imm = 0x8000000 jne .LBB2_1 # %bb.2: # %.preheader49 movq 32(%rsp), %rax xorl %ecx, %ecx .p2align 4, 0x90 .LBB2_3: # =>This Inner Loop Header: Depth=1 xorps %xmm1, %xmm1 cvtsi2ss %ecx, %xmm1 mulss %xmm0, %xmm1 movss %xmm1, (%rax,%rcx,4) incq %rcx cmpq $134217728, %rcx # imm = 0x8000000 jne .LBB2_3 # %bb.4: # %.preheader movabsq $4294967296, %r15 # imm = 0x100000000 movq 24(%rsp), %rbx xorl %r12d, %r12d movl $1024, %edx # imm = 0x400 movq %rbx, %rdi xorl %esi, %esi callq memset@PLT movl $const_c, %edi movl $1024, %edx # imm = 0x400 movq %rbx, %rsi xorl %ecx, %ecx movl $1, %r8d callq hipMemcpyToSymbol movabsq $2199023255808, %rax # imm = 0x20000000100 movq %rax, 172(%rsp) movl $1024, 180(%rsp) # imm = 0x400 leaq 48(%rsp), %rbx jmp .LBB2_5 .p2align 4, 0x90 .LBB2_16: # in Loop: Header=BB2_5 Depth=1 movq %r14, %rdi callq _ZNKSt5ctypeIcE13_M_widen_initEv movq (%r14), %rax movq %r14, %rdi movl $10, %esi callq *48(%rax) .LBB2_17: # %_ZSt4endlIcSt11char_traitsIcEERSt13basic_ostreamIT_T0_ES6_.exit43 # in Loop: Header=BB2_5 Depth=1 movsbl %al, %esi movl $_ZSt4cout, %edi callq _ZNSo3putEc movq %rax, %rdi callq _ZNSo5flushEv addq $4, %r12 cmpq $12, %r12 je .LBB2_18 .LBB2_5: # =>This Loop Header: Depth=1 # Child Loop BB2_10 Depth 2 movl 172(%rsp,%r12), %r14d movl $_ZSt4cout, %edi movl $.L.str, %esi movl $5, %edx callq _ZSt16__ostream_insertIcSt11char_traitsIcEERSt13basic_ostreamIT_T0_ES6_PKS3_l movl $_ZSt4cout, %edi movl %r14d, %esi callq _ZNSolsEi movq (%rax), %rcx movq -24(%rcx), %rcx movq 240(%rax,%rcx), %rbp testq %rbp, %rbp je .LBB2_19 # %bb.6: # %_ZSt13__check_facetISt5ctypeIcEERKT_PS3_.exit.i.i # in Loop: Header=BB2_5 Depth=1 cmpb $0, 56(%rbp) je .LBB2_8 # %bb.7: # in Loop: Header=BB2_5 Depth=1 movzbl 67(%rbp), %ecx jmp .LBB2_9 .p2align 4, 0x90 .LBB2_8: # in Loop: Header=BB2_5 Depth=1 movq %rbp, %rdi movq %rax, %r13 callq _ZNKSt5ctypeIcE13_M_widen_initEv movq (%rbp), %rax movq %rbp, %rdi movl $10, %esi callq *48(%rax) movl %eax, %ecx movq %r13, %rax .LBB2_9: # %_ZSt4endlIcSt11char_traitsIcEERSt13basic_ostreamIT_T0_ES6_.exit # in Loop: Header=BB2_5 Depth=1 movsbl %cl, %esi movq %rax, %rdi callq _ZNSo3putEc movq %rax, %rdi callq _ZNSo5flushEv movq %r14, %rbp orq %r15, %rbp movl $4, %r13d jmp .LBB2_10 .p2align 4, 0x90 .LBB2_12: # in Loop: Header=BB2_10 Depth=2 movsd 8(%rsp), %xmm0 # 8-byte Reload # xmm0 = mem[0],zero addsd 88(%rsp), %xmm0 # 8-byte Folded Reload movsd %xmm0, 8(%rsp) # 8-byte Spill callq hipDeviceSynchronize movq %rbx, %rdi xorl %esi, %esi callq gettimeofday xorps %xmm1, %xmm1 cvtsi2sdq 48(%rsp), %xmm1 xorps %xmm0, %xmm0 cvtsi2sdq 56(%rsp), %xmm0 mulsd .LCPI2_1(%rip), %xmm0 addsd %xmm1, %xmm0 subsd 8(%rsp), %xmm0 # 8-byte Folded Reload movsd .LCPI2_3(%rip), %xmm1 # xmm1 = mem[0],zero divsd %xmm0, %xmm1 mulsd .LCPI2_2(%rip), %xmm0 mulsd .LCPI2_4(%rip), %xmm1 movl $.L.str.1, %edi movb $2, %al callq printf decl %r13d je .LBB2_13 .LBB2_10: # Parent Loop BB2_5 Depth=1 # => This Inner Loop Header: Depth=2 movq %rbx, %rdi xorl %esi, %esi callq gettimeofday xorps %xmm0, %xmm0 cvtsi2sdq 48(%rsp), %xmm0 movsd %xmm0, 88(%rsp) # 8-byte Spill xorps %xmm0, %xmm0 cvtsi2sdq 56(%rsp), %xmm0 mulsd .LCPI2_1(%rip), %xmm0 movsd %xmm0, 8(%rsp) # 8-byte Spill movl $134217728, %eax # imm = 0x8000000 xorl %edx, %edx idivl %r14d # kill: def $eax killed $eax def $rax orq %r15, %rax movq %rax, %rdi movl $1, %esi movq %rbp, %rdx movl $1, %ecx xorl %r8d, %r8d xorl %r9d, %r9d callq __hipPushCallConfiguration testl %eax, %eax jne .LBB2_12 # %bb.11: # in Loop: Header=BB2_10 Depth=2 movq 40(%rsp), %rax movq 32(%rsp), %rcx movq 24(%rsp), %rdx movq %rax, 160(%rsp) movq %rcx, 152(%rsp) movq %rdx, 144(%rsp) movl $134217728, 20(%rsp) # imm = 0x8000000 leaq 160(%rsp), %rax movq %rax, 48(%rsp) leaq 152(%rsp), %rax movq %rax, 56(%rsp) leaq 144(%rsp), %rax movq %rax, 64(%rsp) leaq 20(%rsp), %rax movq %rax, 72(%rsp) leaq 128(%rsp), %rdi leaq 112(%rsp), %rsi leaq 104(%rsp), %rdx leaq 96(%rsp), %rcx callq __hipPopCallConfiguration movq 128(%rsp), %rsi movl 136(%rsp), %edx movq 112(%rsp), %rcx movl 120(%rsp), %r8d movl $_Z2fdPfS_Pii, %edi movq %rbx, %r9 pushq 96(%rsp) .cfi_adjust_cfa_offset 8 pushq 112(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $16, %rsp .cfi_adjust_cfa_offset -16 jmp .LBB2_12 .p2align 4, 0x90 .LBB2_13: # in Loop: Header=BB2_5 Depth=1 movq _ZSt4cout(%rip), %rax movq -24(%rax), %rax movq _ZSt4cout+240(%rax), %r14 testq %r14, %r14 je .LBB2_19 # %bb.14: # %_ZSt13__check_facetISt5ctypeIcEERKT_PS3_.exit.i.i40 # in Loop: Header=BB2_5 Depth=1 cmpb $0, 56(%r14) je .LBB2_16 # %bb.15: # in Loop: Header=BB2_5 Depth=1 movzbl 67(%r14), %eax jmp .LBB2_17 .LBB2_18: xorl %eax, %eax addq $184, %rsp .cfi_def_cfa_offset 56 popq %rbx .cfi_def_cfa_offset 48 popq %r12 .cfi_def_cfa_offset 40 popq %r13 .cfi_def_cfa_offset 32 popq %r14 .cfi_def_cfa_offset 24 popq %r15 .cfi_def_cfa_offset 16 popq %rbp .cfi_def_cfa_offset 8 retq .LBB2_19: .cfi_def_cfa_offset 240 callq _ZSt16__throw_bad_castv .Lfunc_end2: .size main, .Lfunc_end2-main .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_ctor .type __hip_module_ctor,@function __hip_module_ctor: # @__hip_module_ctor .cfi_startproc # %bb.0: pushq %rbx .cfi_def_cfa_offset 16 subq $32, %rsp .cfi_def_cfa_offset 48 .cfi_offset %rbx, -16 cmpq $0, __hip_gpubin_handle(%rip) jne .LBB3_2 # %bb.1: movl $__hip_fatbin_wrapper, %edi callq __hipRegisterFatBinary movq %rax, __hip_gpubin_handle(%rip) .LBB3_2: movq __hip_gpubin_handle(%rip), %rbx xorps %xmm0, %xmm0 movups %xmm0, 16(%rsp) movups %xmm0, (%rsp) movl $_Z2fdPfS_Pii, %esi movl $.L__unnamed_1, %edx movl $.L__unnamed_1, %ecx movq %rbx, %rdi movl $-1, %r8d xorl %r9d, %r9d callq __hipRegisterFunction movl $0, 8(%rsp) movl $1, (%rsp) movl $const_c, %esi movl $.L__unnamed_2, %edx movl $.L__unnamed_2, %ecx movl $1024, %r9d # imm = 0x400 movq %rbx, %rdi xorl %r8d, %r8d callq __hipRegisterVar movl $__hip_module_dtor, %edi addq $32, %rsp .cfi_def_cfa_offset 16 popq %rbx .cfi_def_cfa_offset 8 jmp atexit # TAILCALL .Lfunc_end3: .size __hip_module_ctor, .Lfunc_end3-__hip_module_ctor .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_dtor .type __hip_module_dtor,@function __hip_module_dtor: # @__hip_module_dtor .cfi_startproc # %bb.0: movq __hip_gpubin_handle(%rip), %rdi testq %rdi, %rdi je .LBB4_2 # %bb.1: pushq %rax .cfi_def_cfa_offset 16 callq __hipUnregisterFatBinary movq $0, __hip_gpubin_handle(%rip) addq $8, %rsp .cfi_def_cfa_offset 8 .LBB4_2: retq .Lfunc_end4: .size __hip_module_dtor, .Lfunc_end4-__hip_module_dtor .cfi_endproc # -- End function .type const_c,@object # @const_c .local const_c .comm const_c,1024,16 .type _Z2fdPfS_Pii,@object # @_Z2fdPfS_Pii .section .rodata,"a",@progbits .globl _Z2fdPfS_Pii .p2align 3, 0x0 _Z2fdPfS_Pii: .quad _Z17__device_stub__fdPfS_Pii .size _Z2fdPfS_Pii, 8 .type .L.str,@object # @.str .section .rodata.str1.1,"aMS",@progbits,1 .L.str: .asciz "bs = " .size .L.str, 6 .type .L.str.1,@object # @.str.1 .L.str.1: .asciz "%.2f ms bw %.3f GB/s\n" .size .L.str.1, 24 .type .L__unnamed_1,@object # @0 .L__unnamed_1: .asciz "_Z2fdPfS_Pii" .size .L__unnamed_1, 13 .type .L__unnamed_2,@object # @1 .L__unnamed_2: .asciz "const_c" .size .L__unnamed_2, 8 .type __hip_fatbin_wrapper,@object # @__hip_fatbin_wrapper .section .hipFatBinSegment,"a",@progbits .p2align 3, 0x0 __hip_fatbin_wrapper: .long 1212764230 # 0x48495046 .long 1 # 0x1 .quad __hip_fatbin .quad 0 .size __hip_fatbin_wrapper, 24 .type __hip_gpubin_handle,@object # @__hip_gpubin_handle .local __hip_gpubin_handle .comm __hip_gpubin_handle,8,8 .section .init_array,"aw",@init_array .p2align 3, 0x0 .quad __hip_module_ctor .type __hip_cuid_,@object # @__hip_cuid_ .bss .globl __hip_cuid_ __hip_cuid_: .byte 0 # 0x0 .size __hip_cuid_, 1 .section ".linker-options","e",@llvm_linker_options .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym _Z17__device_stub__fdPfS_Pii .addrsig_sym __hip_module_ctor .addrsig_sym __hip_module_dtor .addrsig_sym const_c .addrsig_sym _Z2fdPfS_Pii .addrsig_sym _ZSt4cout .addrsig_sym __hip_fatbin .addrsig_sym __hip_fatbin_wrapper .addrsig_sym __hip_cuid_
You are a helpful coding assistant specialized in converting CUDA source code to CUDA device assembly.
#include "includes.h" __global__ void cuda_gray(unsigned char *input, int offset, int streamSize, unsigned char* gray, int size) { int gray_idx = (offset/3) + (blockIdx.x * blockDim.x + threadIdx.x); int rgb_idx = (offset) + ((blockIdx.x * blockDim.x + threadIdx.x) * 3); if (((blockIdx.x * blockDim.x + threadIdx.x)*3)>=streamSize || gray_idx>=size) { return; } gray[gray_idx] = (gray_value[0] * input[rgb_idx]) + (gray_value[1] * input[rgb_idx + 1]) + (gray_value[2] * input[rgb_idx + 2]); }
code for sm_80 Function : _Z9cuda_grayPhiiS_i .headerflags @"EF_CUDA_TEXMODE_UNIFIED EF_CUDA_64BIT_ADDRESS EF_CUDA_SM80 EF_CUDA_VIRTUAL_SM(EF_CUDA_SM80)" /*0000*/ IMAD.MOV.U32 R1, RZ, RZ, c[0x0][0x28] ; /* 0x00000a00ff017624 */ /* 0x000fe400078e00ff */ /*0010*/ S2R R0, SR_CTAID.X ; /* 0x0000000000007919 */ /* 0x000e220000002500 */ /*0020*/ ULDC UR4, c[0x0][0x168] ; /* 0x00005a0000047ab9 */ /* 0x000fe40000000800 */ /*0030*/ UIMAD.WIDE UR4, UR4, 0x55555556, URZ ; /* 0x55555556040478a5 */ /* 0x000fe2000f8e023f */ /*0040*/ S2R R3, SR_TID.X ; /* 0x0000000000037919 */ /* 0x000e260000002100 */ /*0050*/ ULEA.HI UR5, UR5, UR5, URZ, 0x1 ; /* 0x0000000505057291 */ /* 0x000fe2000f8f083f */ /*0060*/ IMAD R0, R0, c[0x0][0x0], R3 ; /* 0x0000000000007a24 */ /* 0x001fca00078e0203 */ /*0070*/ IADD3 R10, R0.reuse, UR5, RZ ; /* 0x00000005000a7c10 */ /* 0x040fe2000fffe0ff */ /*0080*/ IMAD R0, R0, 0x3, RZ ; /* 0x0000000300007824 */ /* 0x000fc600078e02ff */ /*0090*/ ISETP.GE.AND P0, PT, R10, c[0x0][0x178], PT ; /* 0x00005e000a007a0c */ /* 0x000fc80003f06270 */ /*00a0*/ ISETP.GE.U32.OR P0, PT, R0, c[0x0][0x16c], P0 ; /* 0x00005b0000007a0c */ /* 0x000fda0000706470 */ /*00b0*/ @P0 EXIT ; /* 0x000000000000094d */ /* 0x000fea0003800000 */ /*00c0*/ IADD3 R0, R0, c[0x0][0x168], RZ ; /* 0x00005a0000007a10 */ /* 0x000fe20007ffe0ff */ /*00d0*/ ULDC.64 UR4, c[0x0][0x118] ; /* 0x0000460000047ab9 */ /* 0x000fc60000000a00 */ /*00e0*/ IADD3 R2, P0, R0, c[0x0][0x160], RZ ; /* 0x0000580000027a10 */ /* 0x000fc80007f1e0ff */ /*00f0*/ LEA.HI.X.SX32 R3, R0, c[0x0][0x164], 0x1, P0 ; /* 0x0000590000037a11 */ /* 0x000fca00000f0eff */ /*0100*/ LDG.E.U8 R6, [R2.64+0x1] ; /* 0x0000010402067981 */ /* 0x000ea8000c1e1100 */ /*0110*/ LDG.E.U8 R0, [R2.64] ; /* 0x0000000402007981 */ /* 0x000ee8000c1e1100 */ /*0120*/ LDG.E.U8 R7, [R2.64+0x2] ; /* 0x0000020402077981 */ /* 0x000f22000c1e1100 */ /*0130*/ I2F.U16 R6, R6 ; /* 0x0000000600067306 */ /* 0x004e300000101000 */ /*0140*/ I2F.U16 R0, R0 ; /* 0x0000000000007306 */ /* 0x008e700000101000 */ /*0150*/ I2F.U16 R4, R7 ; /* 0x0000000700047306 */ /* 0x010ea20000101000 */ /*0160*/ FMUL R5, R6, c[0x3][0x4] ; /* 0x00c0010006057a20 */ /* 0x001fc80000400000 */ /*0170*/ FFMA R5, R0, c[0x3][0x0], R5 ; /* 0x00c0000000057a23 */ /* 0x002fc80000000005 */ /*0180*/ FFMA R8, R4, c[0x3][0x8], R5 ; /* 0x00c0020004087a23 */ /* 0x004fe20000000005 */ /*0190*/ IADD3 R4, P0, R10, c[0x0][0x170], RZ ; /* 0x00005c000a047a10 */ /* 0x000fc60007f1e0ff */ /*01a0*/ F2I.U32.TRUNC.NTZ R9, R8 ; /* 0x0000000800097305 */ /* 0x000e22000020f000 */ /*01b0*/ LEA.HI.X.SX32 R5, R10, c[0x0][0x174], 0x1, P0 ; /* 0x00005d000a057a11 */ /* 0x000fca00000f0eff */ /*01c0*/ STG.E.U8 [R4.64], R9 ; /* 0x0000000904007986 */ /* 0x001fe2000c101104 */ /*01d0*/ EXIT ; /* 0x000000000000794d */ /* 0x000fea0003800000 */ /*01e0*/ BRA 0x1e0; /* 0xfffffff000007947 */ /* 0x000fc0000383ffff */ /*01f0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0200*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0210*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0220*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0230*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0240*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0250*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0260*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0270*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ ..........
You are a helpful coding assistant specialized in converting CUDA source code to CUDA host assembly.
#include "includes.h" __global__ void cuda_gray(unsigned char *input, int offset, int streamSize, unsigned char* gray, int size) { int gray_idx = (offset/3) + (blockIdx.x * blockDim.x + threadIdx.x); int rgb_idx = (offset) + ((blockIdx.x * blockDim.x + threadIdx.x) * 3); if (((blockIdx.x * blockDim.x + threadIdx.x)*3)>=streamSize || gray_idx>=size) { return; } gray[gray_idx] = (gray_value[0] * input[rgb_idx]) + (gray_value[1] * input[rgb_idx + 1]) + (gray_value[2] * input[rgb_idx + 2]); }
.file "tmpxft_000f499b_00000000-6_cuda_gray.cudafe1.cpp" .text #APP #NO_APP .type _ZL26__cudaUnregisterBinaryUtilv, @function _ZL26__cudaUnregisterBinaryUtilv: .LFB2029: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaUnregisterFatBinary@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2029: .size _ZL26__cudaUnregisterBinaryUtilv, .-_ZL26__cudaUnregisterBinaryUtilv .globl _Z33__device_stub__Z9cuda_grayPhiiS_iPhiiS_i .type _Z33__device_stub__Z9cuda_grayPhiiS_iPhiiS_i, @function _Z33__device_stub__Z9cuda_grayPhiiS_iPhiiS_i: .LFB2051: .cfi_startproc endbr64 subq $152, %rsp .cfi_def_cfa_offset 160 movq %rdi, 24(%rsp) movl %esi, 20(%rsp) movl %edx, 16(%rsp) movq %rcx, 8(%rsp) movl %r8d, 4(%rsp) movq %fs:40, %rax movq %rax, 136(%rsp) xorl %eax, %eax leaq 24(%rsp), %rax movq %rax, 96(%rsp) leaq 20(%rsp), %rax movq %rax, 104(%rsp) leaq 16(%rsp), %rax movq %rax, 112(%rsp) leaq 8(%rsp), %rax movq %rax, 120(%rsp) leaq 4(%rsp), %rax movq %rax, 128(%rsp) movl $1, 48(%rsp) movl $1, 52(%rsp) movl $1, 56(%rsp) movl $1, 60(%rsp) movl $1, 64(%rsp) movl $1, 68(%rsp) leaq 40(%rsp), %rcx leaq 32(%rsp), %rdx leaq 60(%rsp), %rsi leaq 48(%rsp), %rdi call __cudaPopCallConfiguration@PLT testl %eax, %eax je .L7 .L3: movq 136(%rsp), %rax subq %fs:40, %rax jne .L8 addq $152, %rsp .cfi_remember_state .cfi_def_cfa_offset 8 ret .L7: .cfi_restore_state pushq 40(%rsp) .cfi_def_cfa_offset 168 pushq 40(%rsp) .cfi_def_cfa_offset 176 leaq 112(%rsp), %r9 movq 76(%rsp), %rcx movl 84(%rsp), %r8d movq 64(%rsp), %rsi movl 72(%rsp), %edx leaq _Z9cuda_grayPhiiS_i(%rip), %rdi call cudaLaunchKernel@PLT addq $16, %rsp .cfi_def_cfa_offset 160 jmp .L3 .L8: call __stack_chk_fail@PLT .cfi_endproc .LFE2051: .size _Z33__device_stub__Z9cuda_grayPhiiS_iPhiiS_i, .-_Z33__device_stub__Z9cuda_grayPhiiS_iPhiiS_i .globl _Z9cuda_grayPhiiS_i .type _Z9cuda_grayPhiiS_i, @function _Z9cuda_grayPhiiS_i: .LFB2052: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 call _Z33__device_stub__Z9cuda_grayPhiiS_iPhiiS_i addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2052: .size _Z9cuda_grayPhiiS_i, .-_Z9cuda_grayPhiiS_i .section .rodata.str1.1,"aMS",@progbits,1 .LC0: .string "_Z9cuda_grayPhiiS_i" .LC1: .string "gray_value" .text .type _ZL24__sti____cudaRegisterAllv, @function _ZL24__sti____cudaRegisterAllv: .LFB2054: .cfi_startproc endbr64 pushq %rbx .cfi_def_cfa_offset 16 .cfi_offset 3, -16 leaq _ZL15__fatDeviceText(%rip), %rdi call __cudaRegisterFatBinary@PLT movq %rax, %rbx movq %rax, _ZL20__cudaFatCubinHandle(%rip) pushq $0 .cfi_def_cfa_offset 24 pushq $0 .cfi_def_cfa_offset 32 pushq $0 .cfi_def_cfa_offset 40 pushq $0 .cfi_def_cfa_offset 48 movl $0, %r9d movl $-1, %r8d leaq .LC0(%rip), %rdx movq %rdx, %rcx leaq _Z9cuda_grayPhiiS_i(%rip), %rsi movq %rax, %rdi call __cudaRegisterFunction@PLT addq $32, %rsp .cfi_def_cfa_offset 16 pushq $0 .cfi_def_cfa_offset 24 pushq $1 .cfi_def_cfa_offset 32 movl $12, %r9d movl $0, %r8d leaq .LC1(%rip), %rdx movq %rdx, %rcx leaq _ZL10gray_value(%rip), %rsi movq %rbx, %rdi call __cudaRegisterVar@PLT addq $16, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaRegisterFatBinaryEnd@PLT leaq _ZL26__cudaUnregisterBinaryUtilv(%rip), %rdi call atexit@PLT popq %rbx .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2054: .size _ZL24__sti____cudaRegisterAllv, .-_ZL24__sti____cudaRegisterAllv .section .init_array,"aw" .align 8 .quad _ZL24__sti____cudaRegisterAllv .section .nvFatBinSegment,"aw" .align 8 .type _ZL15__fatDeviceText, @object .size _ZL15__fatDeviceText, 24 _ZL15__fatDeviceText: .long 1180844977 .long 1 .quad fatbinData .quad 0 .local _ZL20__cudaFatCubinHandle .comm _ZL20__cudaFatCubinHandle,8,8 .local _ZL10gray_value .comm _ZL10gray_value,12,8 .ident "GCC: (Ubuntu 13.3.0-6ubuntu2~24.04) 13.3.0" .section .note.GNU-stack,"",@progbits .section .note.gnu.property,"a" .align 8 .long 1f - 0f .long 4f - 1f .long 5 0: .string "GNU" 1: .align 8 .long 0xc0000002 .long 3f - 2f 2: .long 0x3 3: .align 8 4:
You are a helpful coding assistant specialized in converting CUDA source code to HIP/AMD source code.
#include "includes.h" __global__ void cuda_gray(unsigned char *input, int offset, int streamSize, unsigned char* gray, int size) { int gray_idx = (offset/3) + (blockIdx.x * blockDim.x + threadIdx.x); int rgb_idx = (offset) + ((blockIdx.x * blockDim.x + threadIdx.x) * 3); if (((blockIdx.x * blockDim.x + threadIdx.x)*3)>=streamSize || gray_idx>=size) { return; } gray[gray_idx] = (gray_value[0] * input[rgb_idx]) + (gray_value[1] * input[rgb_idx + 1]) + (gray_value[2] * input[rgb_idx + 2]); }
#include <hip/hip_runtime.h> #include "includes.h" __global__ void cuda_gray(unsigned char *input, int offset, int streamSize, unsigned char* gray, int size) { int gray_idx = (offset/3) + (blockIdx.x * blockDim.x + threadIdx.x); int rgb_idx = (offset) + ((blockIdx.x * blockDim.x + threadIdx.x) * 3); if (((blockIdx.x * blockDim.x + threadIdx.x)*3)>=streamSize || gray_idx>=size) { return; } gray[gray_idx] = (gray_value[0] * input[rgb_idx]) + (gray_value[1] * input[rgb_idx + 1]) + (gray_value[2] * input[rgb_idx + 2]); }
You are a helpful coding assistant specialized in converting HIP source code to AMD device assembly.
#include <hip/hip_runtime.h> #include "includes.h" __global__ void cuda_gray(unsigned char *input, int offset, int streamSize, unsigned char* gray, int size) { int gray_idx = (offset/3) + (blockIdx.x * blockDim.x + threadIdx.x); int rgb_idx = (offset) + ((blockIdx.x * blockDim.x + threadIdx.x) * 3); if (((blockIdx.x * blockDim.x + threadIdx.x)*3)>=streamSize || gray_idx>=size) { return; } gray[gray_idx] = (gray_value[0] * input[rgb_idx]) + (gray_value[1] * input[rgb_idx + 1]) + (gray_value[2] * input[rgb_idx + 2]); }
.text .amdgcn_target "amdgcn-amd-amdhsa--gfx1100" .protected _Z9cuda_grayPhiiS_i .globl _Z9cuda_grayPhiiS_i .p2align 8 .type _Z9cuda_grayPhiiS_i,@function _Z9cuda_grayPhiiS_i: s_clause 0x2 s_load_b32 s2, s[0:1], 0x2c s_load_b64 s[4:5], s[0:1], 0x8 s_load_b32 s3, s[0:1], 0x18 s_waitcnt lgkmcnt(0) s_and_b32 s2, s2, 0xffff s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1) v_mad_u64_u32 v[2:3], null, s15, s2, v[0:1] s_mul_hi_i32 s2, s4, 0x55555556 s_lshr_b32 s6, s2, 31 s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2) v_lshl_add_u32 v1, v2, 1, v2 v_add3_u32 v0, s2, s6, v2 v_cmp_gt_u32_e32 vcc_lo, s5, v1 s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1) v_cmp_gt_i32_e64 s2, s3, v0 s_and_b32 s2, vcc_lo, s2 s_delay_alu instid0(SALU_CYCLE_1) s_and_saveexec_b32 s3, s2 s_cbranch_execz .LBB0_2 s_load_b64 s[2:3], s[0:1], 0x0 v_add_nc_u32_e32 v1, s4, v1 s_load_b64 s[0:1], s[0:1], 0x10 s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_2) v_ashrrev_i32_e32 v2, 31, v1 s_waitcnt lgkmcnt(0) v_add_co_u32 v1, vcc_lo, s2, v1 v_add_co_ci_u32_e32 v2, vcc_lo, s3, v2, vcc_lo s_getpc_b64 s[2:3] s_add_u32 s2, s2, gray_value@rel32@lo+8 s_addc_u32 s3, s3, gray_value@rel32@hi+16 s_getpc_b64 s[4:5] s_add_u32 s4, s4, gray_value@rel32@lo+4 s_addc_u32 s5, s5, gray_value@rel32@hi+12 s_load_b32 s6, s[2:3], 0x0 s_clause 0x2 global_load_u8 v3, v[1:2], off offset:1 global_load_u8 v4, v[1:2], off global_load_u8 v1, v[1:2], off offset:2 s_getpc_b64 s[2:3] s_add_u32 s2, s2, gray_value@rel32@lo+12 s_addc_u32 s3, s3, gray_value@rel32@hi+20 s_clause 0x1 s_load_b32 s4, s[4:5], 0x0 s_load_b32 s2, s[2:3], 0x0 s_waitcnt vmcnt(2) v_cvt_f32_ubyte0_e32 v2, v3 s_waitcnt vmcnt(1) v_cvt_f32_ubyte0_e32 v3, v4 s_waitcnt vmcnt(0) v_cvt_f32_ubyte0_e32 v1, v1 s_waitcnt lgkmcnt(0) v_mul_f32_e32 v2, s6, v2 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_fmac_f32_e32 v2, s4, v3 v_fmac_f32_e32 v2, s2, v1 v_ashrrev_i32_e32 v1, 31, v0 v_add_co_u32 v0, vcc_lo, s0, v0 s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3) v_cvt_i32_f32_e32 v2, v2 v_add_co_ci_u32_e32 v1, vcc_lo, s1, v1, vcc_lo global_store_b8 v[0:1], v2, off .LBB0_2: s_nop 0 s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) s_endpgm .section .rodata,"a",@progbits .p2align 6, 0x0 .amdhsa_kernel _Z9cuda_grayPhiiS_i .amdhsa_group_segment_fixed_size 0 .amdhsa_private_segment_fixed_size 0 .amdhsa_kernarg_size 288 .amdhsa_user_sgpr_count 15 .amdhsa_user_sgpr_dispatch_ptr 0 .amdhsa_user_sgpr_queue_ptr 0 .amdhsa_user_sgpr_kernarg_segment_ptr 1 .amdhsa_user_sgpr_dispatch_id 0 .amdhsa_user_sgpr_private_segment_size 0 .amdhsa_wavefront_size32 1 .amdhsa_uses_dynamic_stack 0 .amdhsa_enable_private_segment 0 .amdhsa_system_sgpr_workgroup_id_x 1 .amdhsa_system_sgpr_workgroup_id_y 0 .amdhsa_system_sgpr_workgroup_id_z 0 .amdhsa_system_sgpr_workgroup_info 0 .amdhsa_system_vgpr_workitem_id 0 .amdhsa_next_free_vgpr 5 .amdhsa_next_free_sgpr 16 .amdhsa_float_round_mode_32 0 .amdhsa_float_round_mode_16_64 0 .amdhsa_float_denorm_mode_32 3 .amdhsa_float_denorm_mode_16_64 3 .amdhsa_dx10_clamp 1 .amdhsa_ieee_mode 1 .amdhsa_fp16_overflow 0 .amdhsa_workgroup_processor_mode 1 .amdhsa_memory_ordered 1 .amdhsa_forward_progress 0 .amdhsa_shared_vgpr_count 0 .amdhsa_exception_fp_ieee_invalid_op 0 .amdhsa_exception_fp_denorm_src 0 .amdhsa_exception_fp_ieee_div_zero 0 .amdhsa_exception_fp_ieee_overflow 0 .amdhsa_exception_fp_ieee_underflow 0 .amdhsa_exception_fp_ieee_inexact 0 .amdhsa_exception_int_div_zero 0 .end_amdhsa_kernel .text .Lfunc_end0: .size _Z9cuda_grayPhiiS_i, .Lfunc_end0-_Z9cuda_grayPhiiS_i .section .AMDGPU.csdata,"",@progbits .text .p2alignl 7, 3214868480 .fill 96, 4, 3214868480 .protected gray_value .type gray_value,@object .data .globl gray_value .p2align 2, 0x0 gray_value: .long 0x3e99999a .long 0x3f147ae1 .long 0x3de147ae .size gray_value, 12 .type __hip_cuid_,@object .section .bss,"aw",@nobits .globl __hip_cuid_ __hip_cuid_: .byte 0 .size __hip_cuid_, 1 .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym gray_value .addrsig_sym __hip_cuid_ .amdgpu_metadata --- amdhsa.kernels: - .args: - .address_space: global .offset: 0 .size: 8 .value_kind: global_buffer - .offset: 8 .size: 4 .value_kind: by_value - .offset: 12 .size: 4 .value_kind: by_value - .address_space: global .offset: 16 .size: 8 .value_kind: global_buffer - .offset: 24 .size: 4 .value_kind: by_value - .offset: 32 .size: 4 .value_kind: hidden_block_count_x - .offset: 36 .size: 4 .value_kind: hidden_block_count_y - .offset: 40 .size: 4 .value_kind: hidden_block_count_z - .offset: 44 .size: 2 .value_kind: hidden_group_size_x - .offset: 46 .size: 2 .value_kind: hidden_group_size_y - .offset: 48 .size: 2 .value_kind: hidden_group_size_z - .offset: 50 .size: 2 .value_kind: hidden_remainder_x - .offset: 52 .size: 2 .value_kind: hidden_remainder_y - .offset: 54 .size: 2 .value_kind: hidden_remainder_z - .offset: 72 .size: 8 .value_kind: hidden_global_offset_x - .offset: 80 .size: 8 .value_kind: hidden_global_offset_y - .offset: 88 .size: 8 .value_kind: hidden_global_offset_z - .offset: 96 .size: 2 .value_kind: hidden_grid_dims .group_segment_fixed_size: 0 .kernarg_segment_align: 8 .kernarg_segment_size: 288 .language: OpenCL C .language_version: - 2 - 0 .max_flat_workgroup_size: 1024 .name: _Z9cuda_grayPhiiS_i .private_segment_fixed_size: 0 .sgpr_count: 18 .sgpr_spill_count: 0 .symbol: _Z9cuda_grayPhiiS_i.kd .uniform_work_group_size: 1 .uses_dynamic_stack: false .vgpr_count: 5 .vgpr_spill_count: 0 .wavefront_size: 32 .workgroup_processor_mode: 1 amdhsa.target: amdgcn-amd-amdhsa--gfx1100 amdhsa.version: - 1 - 2 ... .end_amdgpu_metadata
You are a helpful coding assistant specialized in converting HIP source code to AMD host assembly.
#include <hip/hip_runtime.h> #include "includes.h" __global__ void cuda_gray(unsigned char *input, int offset, int streamSize, unsigned char* gray, int size) { int gray_idx = (offset/3) + (blockIdx.x * blockDim.x + threadIdx.x); int rgb_idx = (offset) + ((blockIdx.x * blockDim.x + threadIdx.x) * 3); if (((blockIdx.x * blockDim.x + threadIdx.x)*3)>=streamSize || gray_idx>=size) { return; } gray[gray_idx] = (gray_value[0] * input[rgb_idx]) + (gray_value[1] * input[rgb_idx + 1]) + (gray_value[2] * input[rgb_idx + 2]); }
.text .file "cuda_gray.hip" .globl _Z24__device_stub__cuda_grayPhiiS_i # -- Begin function _Z24__device_stub__cuda_grayPhiiS_i .p2align 4, 0x90 .type _Z24__device_stub__cuda_grayPhiiS_i,@function _Z24__device_stub__cuda_grayPhiiS_i: # @_Z24__device_stub__cuda_grayPhiiS_i .cfi_startproc # %bb.0: subq $120, %rsp .cfi_def_cfa_offset 128 movq %rdi, 72(%rsp) movl %esi, 12(%rsp) movl %edx, 8(%rsp) movq %rcx, 64(%rsp) movl %r8d, 4(%rsp) leaq 72(%rsp), %rax movq %rax, 80(%rsp) leaq 12(%rsp), %rax movq %rax, 88(%rsp) leaq 8(%rsp), %rax movq %rax, 96(%rsp) leaq 64(%rsp), %rax movq %rax, 104(%rsp) leaq 4(%rsp), %rax movq %rax, 112(%rsp) leaq 48(%rsp), %rdi leaq 32(%rsp), %rsi leaq 24(%rsp), %rdx leaq 16(%rsp), %rcx callq __hipPopCallConfiguration movq 48(%rsp), %rsi movl 56(%rsp), %edx movq 32(%rsp), %rcx movl 40(%rsp), %r8d leaq 80(%rsp), %r9 movl $_Z9cuda_grayPhiiS_i, %edi pushq 16(%rsp) .cfi_adjust_cfa_offset 8 pushq 32(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $136, %rsp .cfi_adjust_cfa_offset -136 retq .Lfunc_end0: .size _Z24__device_stub__cuda_grayPhiiS_i, .Lfunc_end0-_Z24__device_stub__cuda_grayPhiiS_i .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_ctor .type __hip_module_ctor,@function __hip_module_ctor: # @__hip_module_ctor .cfi_startproc # %bb.0: pushq %rbx .cfi_def_cfa_offset 16 subq $32, %rsp .cfi_def_cfa_offset 48 .cfi_offset %rbx, -16 cmpq $0, __hip_gpubin_handle(%rip) jne .LBB1_2 # %bb.1: movl $__hip_fatbin_wrapper, %edi callq __hipRegisterFatBinary movq %rax, __hip_gpubin_handle(%rip) .LBB1_2: movq __hip_gpubin_handle(%rip), %rbx xorps %xmm0, %xmm0 movups %xmm0, 16(%rsp) movups %xmm0, (%rsp) movl $_Z9cuda_grayPhiiS_i, %esi movl $.L__unnamed_1, %edx movl $.L__unnamed_1, %ecx movq %rbx, %rdi movl $-1, %r8d xorl %r9d, %r9d callq __hipRegisterFunction movl $0, 8(%rsp) movl $1, (%rsp) movl $gray_value, %esi movl $.L__unnamed_2, %edx movl $.L__unnamed_2, %ecx movl $12, %r9d movq %rbx, %rdi xorl %r8d, %r8d callq __hipRegisterVar movl $__hip_module_dtor, %edi addq $32, %rsp .cfi_def_cfa_offset 16 popq %rbx .cfi_def_cfa_offset 8 jmp atexit # TAILCALL .Lfunc_end1: .size __hip_module_ctor, .Lfunc_end1-__hip_module_ctor .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_dtor .type __hip_module_dtor,@function __hip_module_dtor: # @__hip_module_dtor .cfi_startproc # %bb.0: movq __hip_gpubin_handle(%rip), %rdi testq %rdi, %rdi je .LBB2_2 # %bb.1: pushq %rax .cfi_def_cfa_offset 16 callq __hipUnregisterFatBinary movq $0, __hip_gpubin_handle(%rip) addq $8, %rsp .cfi_def_cfa_offset 8 .LBB2_2: retq .Lfunc_end2: .size __hip_module_dtor, .Lfunc_end2-__hip_module_dtor .cfi_endproc # -- End function .type gray_value,@object # @gray_value .local gray_value .comm gray_value,12,4 .type _Z9cuda_grayPhiiS_i,@object # @_Z9cuda_grayPhiiS_i .section .rodata,"a",@progbits .globl _Z9cuda_grayPhiiS_i .p2align 3, 0x0 _Z9cuda_grayPhiiS_i: .quad _Z24__device_stub__cuda_grayPhiiS_i .size _Z9cuda_grayPhiiS_i, 8 .type .L__unnamed_1,@object # @0 .section .rodata.str1.1,"aMS",@progbits,1 .L__unnamed_1: .asciz "_Z9cuda_grayPhiiS_i" .size .L__unnamed_1, 20 .type .L__unnamed_2,@object # @1 .L__unnamed_2: .asciz "gray_value" .size .L__unnamed_2, 11 .type __hip_fatbin_wrapper,@object # @__hip_fatbin_wrapper .section .hipFatBinSegment,"a",@progbits .p2align 3, 0x0 __hip_fatbin_wrapper: .long 1212764230 # 0x48495046 .long 1 # 0x1 .quad __hip_fatbin .quad 0 .size __hip_fatbin_wrapper, 24 .type __hip_gpubin_handle,@object # @__hip_gpubin_handle .local __hip_gpubin_handle .comm __hip_gpubin_handle,8,8 .section .init_array,"aw",@init_array .p2align 3, 0x0 .quad __hip_module_ctor .type __hip_cuid_,@object # @__hip_cuid_ .bss .globl __hip_cuid_ __hip_cuid_: .byte 0 # 0x0 .size __hip_cuid_, 1 .section ".linker-options","e",@llvm_linker_options .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym _Z24__device_stub__cuda_grayPhiiS_i .addrsig_sym __hip_module_ctor .addrsig_sym __hip_module_dtor .addrsig_sym gray_value .addrsig_sym _Z9cuda_grayPhiiS_i .addrsig_sym __hip_fatbin .addrsig_sym __hip_fatbin_wrapper .addrsig_sym __hip_cuid_
You are a helpful coding assistant specialized in converting from CUDA device assembly to AMD device assembly.
code for sm_80 Function : _Z9cuda_grayPhiiS_i .headerflags @"EF_CUDA_TEXMODE_UNIFIED EF_CUDA_64BIT_ADDRESS EF_CUDA_SM80 EF_CUDA_VIRTUAL_SM(EF_CUDA_SM80)" /*0000*/ IMAD.MOV.U32 R1, RZ, RZ, c[0x0][0x28] ; /* 0x00000a00ff017624 */ /* 0x000fe400078e00ff */ /*0010*/ S2R R0, SR_CTAID.X ; /* 0x0000000000007919 */ /* 0x000e220000002500 */ /*0020*/ ULDC UR4, c[0x0][0x168] ; /* 0x00005a0000047ab9 */ /* 0x000fe40000000800 */ /*0030*/ UIMAD.WIDE UR4, UR4, 0x55555556, URZ ; /* 0x55555556040478a5 */ /* 0x000fe2000f8e023f */ /*0040*/ S2R R3, SR_TID.X ; /* 0x0000000000037919 */ /* 0x000e260000002100 */ /*0050*/ ULEA.HI UR5, UR5, UR5, URZ, 0x1 ; /* 0x0000000505057291 */ /* 0x000fe2000f8f083f */ /*0060*/ IMAD R0, R0, c[0x0][0x0], R3 ; /* 0x0000000000007a24 */ /* 0x001fca00078e0203 */ /*0070*/ IADD3 R10, R0.reuse, UR5, RZ ; /* 0x00000005000a7c10 */ /* 0x040fe2000fffe0ff */ /*0080*/ IMAD R0, R0, 0x3, RZ ; /* 0x0000000300007824 */ /* 0x000fc600078e02ff */ /*0090*/ ISETP.GE.AND P0, PT, R10, c[0x0][0x178], PT ; /* 0x00005e000a007a0c */ /* 0x000fc80003f06270 */ /*00a0*/ ISETP.GE.U32.OR P0, PT, R0, c[0x0][0x16c], P0 ; /* 0x00005b0000007a0c */ /* 0x000fda0000706470 */ /*00b0*/ @P0 EXIT ; /* 0x000000000000094d */ /* 0x000fea0003800000 */ /*00c0*/ IADD3 R0, R0, c[0x0][0x168], RZ ; /* 0x00005a0000007a10 */ /* 0x000fe20007ffe0ff */ /*00d0*/ ULDC.64 UR4, c[0x0][0x118] ; /* 0x0000460000047ab9 */ /* 0x000fc60000000a00 */ /*00e0*/ IADD3 R2, P0, R0, c[0x0][0x160], RZ ; /* 0x0000580000027a10 */ /* 0x000fc80007f1e0ff */ /*00f0*/ LEA.HI.X.SX32 R3, R0, c[0x0][0x164], 0x1, P0 ; /* 0x0000590000037a11 */ /* 0x000fca00000f0eff */ /*0100*/ LDG.E.U8 R6, [R2.64+0x1] ; /* 0x0000010402067981 */ /* 0x000ea8000c1e1100 */ /*0110*/ LDG.E.U8 R0, [R2.64] ; /* 0x0000000402007981 */ /* 0x000ee8000c1e1100 */ /*0120*/ LDG.E.U8 R7, [R2.64+0x2] ; /* 0x0000020402077981 */ /* 0x000f22000c1e1100 */ /*0130*/ I2F.U16 R6, R6 ; /* 0x0000000600067306 */ /* 0x004e300000101000 */ /*0140*/ I2F.U16 R0, R0 ; /* 0x0000000000007306 */ /* 0x008e700000101000 */ /*0150*/ I2F.U16 R4, R7 ; /* 0x0000000700047306 */ /* 0x010ea20000101000 */ /*0160*/ FMUL R5, R6, c[0x3][0x4] ; /* 0x00c0010006057a20 */ /* 0x001fc80000400000 */ /*0170*/ FFMA R5, R0, c[0x3][0x0], R5 ; /* 0x00c0000000057a23 */ /* 0x002fc80000000005 */ /*0180*/ FFMA R8, R4, c[0x3][0x8], R5 ; /* 0x00c0020004087a23 */ /* 0x004fe20000000005 */ /*0190*/ IADD3 R4, P0, R10, c[0x0][0x170], RZ ; /* 0x00005c000a047a10 */ /* 0x000fc60007f1e0ff */ /*01a0*/ F2I.U32.TRUNC.NTZ R9, R8 ; /* 0x0000000800097305 */ /* 0x000e22000020f000 */ /*01b0*/ LEA.HI.X.SX32 R5, R10, c[0x0][0x174], 0x1, P0 ; /* 0x00005d000a057a11 */ /* 0x000fca00000f0eff */ /*01c0*/ STG.E.U8 [R4.64], R9 ; /* 0x0000000904007986 */ /* 0x001fe2000c101104 */ /*01d0*/ EXIT ; /* 0x000000000000794d */ /* 0x000fea0003800000 */ /*01e0*/ BRA 0x1e0; /* 0xfffffff000007947 */ /* 0x000fc0000383ffff */ /*01f0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0200*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0210*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0220*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0230*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0240*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0250*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0260*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0270*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ ..........
.text .amdgcn_target "amdgcn-amd-amdhsa--gfx1100" .protected _Z9cuda_grayPhiiS_i .globl _Z9cuda_grayPhiiS_i .p2align 8 .type _Z9cuda_grayPhiiS_i,@function _Z9cuda_grayPhiiS_i: s_clause 0x2 s_load_b32 s2, s[0:1], 0x2c s_load_b64 s[4:5], s[0:1], 0x8 s_load_b32 s3, s[0:1], 0x18 s_waitcnt lgkmcnt(0) s_and_b32 s2, s2, 0xffff s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1) v_mad_u64_u32 v[2:3], null, s15, s2, v[0:1] s_mul_hi_i32 s2, s4, 0x55555556 s_lshr_b32 s6, s2, 31 s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2) v_lshl_add_u32 v1, v2, 1, v2 v_add3_u32 v0, s2, s6, v2 v_cmp_gt_u32_e32 vcc_lo, s5, v1 s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1) v_cmp_gt_i32_e64 s2, s3, v0 s_and_b32 s2, vcc_lo, s2 s_delay_alu instid0(SALU_CYCLE_1) s_and_saveexec_b32 s3, s2 s_cbranch_execz .LBB0_2 s_load_b64 s[2:3], s[0:1], 0x0 v_add_nc_u32_e32 v1, s4, v1 s_load_b64 s[0:1], s[0:1], 0x10 s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_2) v_ashrrev_i32_e32 v2, 31, v1 s_waitcnt lgkmcnt(0) v_add_co_u32 v1, vcc_lo, s2, v1 v_add_co_ci_u32_e32 v2, vcc_lo, s3, v2, vcc_lo s_getpc_b64 s[2:3] s_add_u32 s2, s2, gray_value@rel32@lo+8 s_addc_u32 s3, s3, gray_value@rel32@hi+16 s_getpc_b64 s[4:5] s_add_u32 s4, s4, gray_value@rel32@lo+4 s_addc_u32 s5, s5, gray_value@rel32@hi+12 s_load_b32 s6, s[2:3], 0x0 s_clause 0x2 global_load_u8 v3, v[1:2], off offset:1 global_load_u8 v4, v[1:2], off global_load_u8 v1, v[1:2], off offset:2 s_getpc_b64 s[2:3] s_add_u32 s2, s2, gray_value@rel32@lo+12 s_addc_u32 s3, s3, gray_value@rel32@hi+20 s_clause 0x1 s_load_b32 s4, s[4:5], 0x0 s_load_b32 s2, s[2:3], 0x0 s_waitcnt vmcnt(2) v_cvt_f32_ubyte0_e32 v2, v3 s_waitcnt vmcnt(1) v_cvt_f32_ubyte0_e32 v3, v4 s_waitcnt vmcnt(0) v_cvt_f32_ubyte0_e32 v1, v1 s_waitcnt lgkmcnt(0) v_mul_f32_e32 v2, s6, v2 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_fmac_f32_e32 v2, s4, v3 v_fmac_f32_e32 v2, s2, v1 v_ashrrev_i32_e32 v1, 31, v0 v_add_co_u32 v0, vcc_lo, s0, v0 s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3) v_cvt_i32_f32_e32 v2, v2 v_add_co_ci_u32_e32 v1, vcc_lo, s1, v1, vcc_lo global_store_b8 v[0:1], v2, off .LBB0_2: s_nop 0 s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) s_endpgm .section .rodata,"a",@progbits .p2align 6, 0x0 .amdhsa_kernel _Z9cuda_grayPhiiS_i .amdhsa_group_segment_fixed_size 0 .amdhsa_private_segment_fixed_size 0 .amdhsa_kernarg_size 288 .amdhsa_user_sgpr_count 15 .amdhsa_user_sgpr_dispatch_ptr 0 .amdhsa_user_sgpr_queue_ptr 0 .amdhsa_user_sgpr_kernarg_segment_ptr 1 .amdhsa_user_sgpr_dispatch_id 0 .amdhsa_user_sgpr_private_segment_size 0 .amdhsa_wavefront_size32 1 .amdhsa_uses_dynamic_stack 0 .amdhsa_enable_private_segment 0 .amdhsa_system_sgpr_workgroup_id_x 1 .amdhsa_system_sgpr_workgroup_id_y 0 .amdhsa_system_sgpr_workgroup_id_z 0 .amdhsa_system_sgpr_workgroup_info 0 .amdhsa_system_vgpr_workitem_id 0 .amdhsa_next_free_vgpr 5 .amdhsa_next_free_sgpr 16 .amdhsa_float_round_mode_32 0 .amdhsa_float_round_mode_16_64 0 .amdhsa_float_denorm_mode_32 3 .amdhsa_float_denorm_mode_16_64 3 .amdhsa_dx10_clamp 1 .amdhsa_ieee_mode 1 .amdhsa_fp16_overflow 0 .amdhsa_workgroup_processor_mode 1 .amdhsa_memory_ordered 1 .amdhsa_forward_progress 0 .amdhsa_shared_vgpr_count 0 .amdhsa_exception_fp_ieee_invalid_op 0 .amdhsa_exception_fp_denorm_src 0 .amdhsa_exception_fp_ieee_div_zero 0 .amdhsa_exception_fp_ieee_overflow 0 .amdhsa_exception_fp_ieee_underflow 0 .amdhsa_exception_fp_ieee_inexact 0 .amdhsa_exception_int_div_zero 0 .end_amdhsa_kernel .text .Lfunc_end0: .size _Z9cuda_grayPhiiS_i, .Lfunc_end0-_Z9cuda_grayPhiiS_i .section .AMDGPU.csdata,"",@progbits .text .p2alignl 7, 3214868480 .fill 96, 4, 3214868480 .protected gray_value .type gray_value,@object .data .globl gray_value .p2align 2, 0x0 gray_value: .long 0x3e99999a .long 0x3f147ae1 .long 0x3de147ae .size gray_value, 12 .type __hip_cuid_,@object .section .bss,"aw",@nobits .globl __hip_cuid_ __hip_cuid_: .byte 0 .size __hip_cuid_, 1 .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym gray_value .addrsig_sym __hip_cuid_ .amdgpu_metadata --- amdhsa.kernels: - .args: - .address_space: global .offset: 0 .size: 8 .value_kind: global_buffer - .offset: 8 .size: 4 .value_kind: by_value - .offset: 12 .size: 4 .value_kind: by_value - .address_space: global .offset: 16 .size: 8 .value_kind: global_buffer - .offset: 24 .size: 4 .value_kind: by_value - .offset: 32 .size: 4 .value_kind: hidden_block_count_x - .offset: 36 .size: 4 .value_kind: hidden_block_count_y - .offset: 40 .size: 4 .value_kind: hidden_block_count_z - .offset: 44 .size: 2 .value_kind: hidden_group_size_x - .offset: 46 .size: 2 .value_kind: hidden_group_size_y - .offset: 48 .size: 2 .value_kind: hidden_group_size_z - .offset: 50 .size: 2 .value_kind: hidden_remainder_x - .offset: 52 .size: 2 .value_kind: hidden_remainder_y - .offset: 54 .size: 2 .value_kind: hidden_remainder_z - .offset: 72 .size: 8 .value_kind: hidden_global_offset_x - .offset: 80 .size: 8 .value_kind: hidden_global_offset_y - .offset: 88 .size: 8 .value_kind: hidden_global_offset_z - .offset: 96 .size: 2 .value_kind: hidden_grid_dims .group_segment_fixed_size: 0 .kernarg_segment_align: 8 .kernarg_segment_size: 288 .language: OpenCL C .language_version: - 2 - 0 .max_flat_workgroup_size: 1024 .name: _Z9cuda_grayPhiiS_i .private_segment_fixed_size: 0 .sgpr_count: 18 .sgpr_spill_count: 0 .symbol: _Z9cuda_grayPhiiS_i.kd .uniform_work_group_size: 1 .uses_dynamic_stack: false .vgpr_count: 5 .vgpr_spill_count: 0 .wavefront_size: 32 .workgroup_processor_mode: 1 amdhsa.target: amdgcn-amd-amdhsa--gfx1100 amdhsa.version: - 1 - 2 ... .end_amdgpu_metadata
You are a helpful coding assistant specialized in converting from CUDA host assembly to AMD host assembly.
.file "tmpxft_000f499b_00000000-6_cuda_gray.cudafe1.cpp" .text #APP #NO_APP .type _ZL26__cudaUnregisterBinaryUtilv, @function _ZL26__cudaUnregisterBinaryUtilv: .LFB2029: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaUnregisterFatBinary@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2029: .size _ZL26__cudaUnregisterBinaryUtilv, .-_ZL26__cudaUnregisterBinaryUtilv .globl _Z33__device_stub__Z9cuda_grayPhiiS_iPhiiS_i .type _Z33__device_stub__Z9cuda_grayPhiiS_iPhiiS_i, @function _Z33__device_stub__Z9cuda_grayPhiiS_iPhiiS_i: .LFB2051: .cfi_startproc endbr64 subq $152, %rsp .cfi_def_cfa_offset 160 movq %rdi, 24(%rsp) movl %esi, 20(%rsp) movl %edx, 16(%rsp) movq %rcx, 8(%rsp) movl %r8d, 4(%rsp) movq %fs:40, %rax movq %rax, 136(%rsp) xorl %eax, %eax leaq 24(%rsp), %rax movq %rax, 96(%rsp) leaq 20(%rsp), %rax movq %rax, 104(%rsp) leaq 16(%rsp), %rax movq %rax, 112(%rsp) leaq 8(%rsp), %rax movq %rax, 120(%rsp) leaq 4(%rsp), %rax movq %rax, 128(%rsp) movl $1, 48(%rsp) movl $1, 52(%rsp) movl $1, 56(%rsp) movl $1, 60(%rsp) movl $1, 64(%rsp) movl $1, 68(%rsp) leaq 40(%rsp), %rcx leaq 32(%rsp), %rdx leaq 60(%rsp), %rsi leaq 48(%rsp), %rdi call __cudaPopCallConfiguration@PLT testl %eax, %eax je .L7 .L3: movq 136(%rsp), %rax subq %fs:40, %rax jne .L8 addq $152, %rsp .cfi_remember_state .cfi_def_cfa_offset 8 ret .L7: .cfi_restore_state pushq 40(%rsp) .cfi_def_cfa_offset 168 pushq 40(%rsp) .cfi_def_cfa_offset 176 leaq 112(%rsp), %r9 movq 76(%rsp), %rcx movl 84(%rsp), %r8d movq 64(%rsp), %rsi movl 72(%rsp), %edx leaq _Z9cuda_grayPhiiS_i(%rip), %rdi call cudaLaunchKernel@PLT addq $16, %rsp .cfi_def_cfa_offset 160 jmp .L3 .L8: call __stack_chk_fail@PLT .cfi_endproc .LFE2051: .size _Z33__device_stub__Z9cuda_grayPhiiS_iPhiiS_i, .-_Z33__device_stub__Z9cuda_grayPhiiS_iPhiiS_i .globl _Z9cuda_grayPhiiS_i .type _Z9cuda_grayPhiiS_i, @function _Z9cuda_grayPhiiS_i: .LFB2052: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 call _Z33__device_stub__Z9cuda_grayPhiiS_iPhiiS_i addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2052: .size _Z9cuda_grayPhiiS_i, .-_Z9cuda_grayPhiiS_i .section .rodata.str1.1,"aMS",@progbits,1 .LC0: .string "_Z9cuda_grayPhiiS_i" .LC1: .string "gray_value" .text .type _ZL24__sti____cudaRegisterAllv, @function _ZL24__sti____cudaRegisterAllv: .LFB2054: .cfi_startproc endbr64 pushq %rbx .cfi_def_cfa_offset 16 .cfi_offset 3, -16 leaq _ZL15__fatDeviceText(%rip), %rdi call __cudaRegisterFatBinary@PLT movq %rax, %rbx movq %rax, _ZL20__cudaFatCubinHandle(%rip) pushq $0 .cfi_def_cfa_offset 24 pushq $0 .cfi_def_cfa_offset 32 pushq $0 .cfi_def_cfa_offset 40 pushq $0 .cfi_def_cfa_offset 48 movl $0, %r9d movl $-1, %r8d leaq .LC0(%rip), %rdx movq %rdx, %rcx leaq _Z9cuda_grayPhiiS_i(%rip), %rsi movq %rax, %rdi call __cudaRegisterFunction@PLT addq $32, %rsp .cfi_def_cfa_offset 16 pushq $0 .cfi_def_cfa_offset 24 pushq $1 .cfi_def_cfa_offset 32 movl $12, %r9d movl $0, %r8d leaq .LC1(%rip), %rdx movq %rdx, %rcx leaq _ZL10gray_value(%rip), %rsi movq %rbx, %rdi call __cudaRegisterVar@PLT addq $16, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaRegisterFatBinaryEnd@PLT leaq _ZL26__cudaUnregisterBinaryUtilv(%rip), %rdi call atexit@PLT popq %rbx .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2054: .size _ZL24__sti____cudaRegisterAllv, .-_ZL24__sti____cudaRegisterAllv .section .init_array,"aw" .align 8 .quad _ZL24__sti____cudaRegisterAllv .section .nvFatBinSegment,"aw" .align 8 .type _ZL15__fatDeviceText, @object .size _ZL15__fatDeviceText, 24 _ZL15__fatDeviceText: .long 1180844977 .long 1 .quad fatbinData .quad 0 .local _ZL20__cudaFatCubinHandle .comm _ZL20__cudaFatCubinHandle,8,8 .local _ZL10gray_value .comm _ZL10gray_value,12,8 .ident "GCC: (Ubuntu 13.3.0-6ubuntu2~24.04) 13.3.0" .section .note.GNU-stack,"",@progbits .section .note.gnu.property,"a" .align 8 .long 1f - 0f .long 4f - 1f .long 5 0: .string "GNU" 1: .align 8 .long 0xc0000002 .long 3f - 2f 2: .long 0x3 3: .align 8 4:
.text .file "cuda_gray.hip" .globl _Z24__device_stub__cuda_grayPhiiS_i # -- Begin function _Z24__device_stub__cuda_grayPhiiS_i .p2align 4, 0x90 .type _Z24__device_stub__cuda_grayPhiiS_i,@function _Z24__device_stub__cuda_grayPhiiS_i: # @_Z24__device_stub__cuda_grayPhiiS_i .cfi_startproc # %bb.0: subq $120, %rsp .cfi_def_cfa_offset 128 movq %rdi, 72(%rsp) movl %esi, 12(%rsp) movl %edx, 8(%rsp) movq %rcx, 64(%rsp) movl %r8d, 4(%rsp) leaq 72(%rsp), %rax movq %rax, 80(%rsp) leaq 12(%rsp), %rax movq %rax, 88(%rsp) leaq 8(%rsp), %rax movq %rax, 96(%rsp) leaq 64(%rsp), %rax movq %rax, 104(%rsp) leaq 4(%rsp), %rax movq %rax, 112(%rsp) leaq 48(%rsp), %rdi leaq 32(%rsp), %rsi leaq 24(%rsp), %rdx leaq 16(%rsp), %rcx callq __hipPopCallConfiguration movq 48(%rsp), %rsi movl 56(%rsp), %edx movq 32(%rsp), %rcx movl 40(%rsp), %r8d leaq 80(%rsp), %r9 movl $_Z9cuda_grayPhiiS_i, %edi pushq 16(%rsp) .cfi_adjust_cfa_offset 8 pushq 32(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $136, %rsp .cfi_adjust_cfa_offset -136 retq .Lfunc_end0: .size _Z24__device_stub__cuda_grayPhiiS_i, .Lfunc_end0-_Z24__device_stub__cuda_grayPhiiS_i .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_ctor .type __hip_module_ctor,@function __hip_module_ctor: # @__hip_module_ctor .cfi_startproc # %bb.0: pushq %rbx .cfi_def_cfa_offset 16 subq $32, %rsp .cfi_def_cfa_offset 48 .cfi_offset %rbx, -16 cmpq $0, __hip_gpubin_handle(%rip) jne .LBB1_2 # %bb.1: movl $__hip_fatbin_wrapper, %edi callq __hipRegisterFatBinary movq %rax, __hip_gpubin_handle(%rip) .LBB1_2: movq __hip_gpubin_handle(%rip), %rbx xorps %xmm0, %xmm0 movups %xmm0, 16(%rsp) movups %xmm0, (%rsp) movl $_Z9cuda_grayPhiiS_i, %esi movl $.L__unnamed_1, %edx movl $.L__unnamed_1, %ecx movq %rbx, %rdi movl $-1, %r8d xorl %r9d, %r9d callq __hipRegisterFunction movl $0, 8(%rsp) movl $1, (%rsp) movl $gray_value, %esi movl $.L__unnamed_2, %edx movl $.L__unnamed_2, %ecx movl $12, %r9d movq %rbx, %rdi xorl %r8d, %r8d callq __hipRegisterVar movl $__hip_module_dtor, %edi addq $32, %rsp .cfi_def_cfa_offset 16 popq %rbx .cfi_def_cfa_offset 8 jmp atexit # TAILCALL .Lfunc_end1: .size __hip_module_ctor, .Lfunc_end1-__hip_module_ctor .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_dtor .type __hip_module_dtor,@function __hip_module_dtor: # @__hip_module_dtor .cfi_startproc # %bb.0: movq __hip_gpubin_handle(%rip), %rdi testq %rdi, %rdi je .LBB2_2 # %bb.1: pushq %rax .cfi_def_cfa_offset 16 callq __hipUnregisterFatBinary movq $0, __hip_gpubin_handle(%rip) addq $8, %rsp .cfi_def_cfa_offset 8 .LBB2_2: retq .Lfunc_end2: .size __hip_module_dtor, .Lfunc_end2-__hip_module_dtor .cfi_endproc # -- End function .type gray_value,@object # @gray_value .local gray_value .comm gray_value,12,4 .type _Z9cuda_grayPhiiS_i,@object # @_Z9cuda_grayPhiiS_i .section .rodata,"a",@progbits .globl _Z9cuda_grayPhiiS_i .p2align 3, 0x0 _Z9cuda_grayPhiiS_i: .quad _Z24__device_stub__cuda_grayPhiiS_i .size _Z9cuda_grayPhiiS_i, 8 .type .L__unnamed_1,@object # @0 .section .rodata.str1.1,"aMS",@progbits,1 .L__unnamed_1: .asciz "_Z9cuda_grayPhiiS_i" .size .L__unnamed_1, 20 .type .L__unnamed_2,@object # @1 .L__unnamed_2: .asciz "gray_value" .size .L__unnamed_2, 11 .type __hip_fatbin_wrapper,@object # @__hip_fatbin_wrapper .section .hipFatBinSegment,"a",@progbits .p2align 3, 0x0 __hip_fatbin_wrapper: .long 1212764230 # 0x48495046 .long 1 # 0x1 .quad __hip_fatbin .quad 0 .size __hip_fatbin_wrapper, 24 .type __hip_gpubin_handle,@object # @__hip_gpubin_handle .local __hip_gpubin_handle .comm __hip_gpubin_handle,8,8 .section .init_array,"aw",@init_array .p2align 3, 0x0 .quad __hip_module_ctor .type __hip_cuid_,@object # @__hip_cuid_ .bss .globl __hip_cuid_ __hip_cuid_: .byte 0 # 0x0 .size __hip_cuid_, 1 .section ".linker-options","e",@llvm_linker_options .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym _Z24__device_stub__cuda_grayPhiiS_i .addrsig_sym __hip_module_ctor .addrsig_sym __hip_module_dtor .addrsig_sym gray_value .addrsig_sym _Z9cuda_grayPhiiS_i .addrsig_sym __hip_fatbin .addrsig_sym __hip_fatbin_wrapper .addrsig_sym __hip_cuid_
You are a helpful coding assistant specialized in converting CUDA source code to CUDA device assembly.
#include <stdio.h> int main() { int nDevices; cudaGetDeviceCount(&nDevices); for (int i = 0; i < nDevices; i++) { cudaDeviceProp prop; cudaGetDeviceProperties(&prop, i); printf("Device Number: %d\n", i); printf(" Device name: %s\n", prop.name); printf(" Version %d.%d\n", prop.major, prop.minor); printf(" Compute Mode: %d\n", prop.computeMode); printf(" Memory Clock Rate (KHz): %d\n", prop.memoryClockRate); printf(" Memory Bus Width (bits): %d\n", prop.memoryBusWidth); printf(" Peak Memory Bandwidth (GB/s): %f\n\n", 2.0*prop.memoryClockRate*(prop.memoryBusWidth/8)/1.0e6); printf(" Multi Processor Count: %d\n\n", prop.multiProcessorCount); printf(" TCC Driver: %d\n\n", prop.tccDriver); printf(" Total Global Mem: %d\n\n", prop.totalGlobalMem); printf(" Shared Mem Per Block: %d\n\n", prop.sharedMemPerBlock); printf(" Registers Per Block: %d\n\n", prop.regsPerBlock); printf(" Warpsize: %d\n\n", prop.warpSize); printf(" MemPitch: %d\n\n", prop.memPitch); printf(" MaxThreadsPerBlock: %d\n\n", prop.maxThreadsPerBlock); printf(" Can Map Host Memory: %d\n\n", prop.canMapHostMemory); } }
code for sm_80
You are a helpful coding assistant specialized in converting CUDA source code to CUDA host assembly.
#include <stdio.h> int main() { int nDevices; cudaGetDeviceCount(&nDevices); for (int i = 0; i < nDevices; i++) { cudaDeviceProp prop; cudaGetDeviceProperties(&prop, i); printf("Device Number: %d\n", i); printf(" Device name: %s\n", prop.name); printf(" Version %d.%d\n", prop.major, prop.minor); printf(" Compute Mode: %d\n", prop.computeMode); printf(" Memory Clock Rate (KHz): %d\n", prop.memoryClockRate); printf(" Memory Bus Width (bits): %d\n", prop.memoryBusWidth); printf(" Peak Memory Bandwidth (GB/s): %f\n\n", 2.0*prop.memoryClockRate*(prop.memoryBusWidth/8)/1.0e6); printf(" Multi Processor Count: %d\n\n", prop.multiProcessorCount); printf(" TCC Driver: %d\n\n", prop.tccDriver); printf(" Total Global Mem: %d\n\n", prop.totalGlobalMem); printf(" Shared Mem Per Block: %d\n\n", prop.sharedMemPerBlock); printf(" Registers Per Block: %d\n\n", prop.regsPerBlock); printf(" Warpsize: %d\n\n", prop.warpSize); printf(" MemPitch: %d\n\n", prop.memPitch); printf(" MaxThreadsPerBlock: %d\n\n", prop.maxThreadsPerBlock); printf(" Can Map Host Memory: %d\n\n", prop.canMapHostMemory); } }
.file "tmpxft_001631ba_00000000-6_cuda_properties.cudafe1.cpp" .text #APP #NO_APP .type _ZL26__cudaUnregisterBinaryUtilv, @function _ZL26__cudaUnregisterBinaryUtilv: .LFB2060: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaUnregisterFatBinary@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2060: .size _ZL26__cudaUnregisterBinaryUtilv, .-_ZL26__cudaUnregisterBinaryUtilv .section .rodata.str1.1,"aMS",@progbits,1 .LC0: .string "Device Number: %d\n" .LC1: .string " Device name: %s\n" .LC2: .string " Version %d.%d\n" .LC3: .string " Compute Mode: %d\n" .section .rodata.str1.8,"aMS",@progbits,1 .align 8 .LC4: .string " Memory Clock Rate (KHz): %d\n" .align 8 .LC5: .string " Memory Bus Width (bits): %d\n" .align 8 .LC7: .string " Peak Memory Bandwidth (GB/s): %f\n\n" .section .rodata.str1.1 .LC8: .string " Multi Processor Count: %d\n\n" .LC9: .string " TCC Driver: %d\n\n" .LC10: .string " Total Global Mem: %d\n\n" .LC11: .string " Shared Mem Per Block: %d\n\n" .LC12: .string " Registers Per Block: %d\n\n" .LC13: .string " Warpsize: %d\n\n" .LC14: .string " MemPitch: %d\n\n" .LC15: .string " MaxThreadsPerBlock: %d\n\n" .LC16: .string " Can Map Host Memory: %d\n\n" .text .globl main .type main, @function main: .LFB2057: .cfi_startproc endbr64 pushq %r14 .cfi_def_cfa_offset 16 .cfi_offset 14, -16 pushq %r13 .cfi_def_cfa_offset 24 .cfi_offset 13, -24 pushq %r12 .cfi_def_cfa_offset 32 .cfi_offset 12, -32 pushq %rbp .cfi_def_cfa_offset 40 .cfi_offset 6, -40 pushq %rbx .cfi_def_cfa_offset 48 .cfi_offset 3, -48 subq $1056, %rsp .cfi_def_cfa_offset 1104 movq %fs:40, %rax movq %rax, 1048(%rsp) xorl %eax, %eax leaq 12(%rsp), %rdi call cudaGetDeviceCount@PLT cmpl $0, 12(%rsp) jle .L4 movl $0, %ebx leaq .LC0(%rip), %r13 leaq .LC1(%rip), %r12 leaq .LC2(%rip), %rbp .L5: leaq 16(%rsp), %r14 movl %ebx, %esi movq %r14, %rdi call cudaGetDeviceProperties_v2@PLT movl %ebx, %edx movq %r13, %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT movq %r14, %rdx movq %r12, %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT movl 380(%rsp), %ecx movl 376(%rsp), %edx movq %rbp, %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT movl 420(%rsp), %edx leaq .LC3(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT movl 624(%rsp), %edx leaq .LC4(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT movl 628(%rsp), %edx leaq .LC5(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT pxor %xmm0, %xmm0 cvtsi2sdl 624(%rsp), %xmm0 addsd %xmm0, %xmm0 movl 628(%rsp), %edx leal 7(%rdx), %eax testl %edx, %edx cmovns %edx, %eax sarl $3, %eax pxor %xmm1, %xmm1 cvtsi2sdl %eax, %xmm1 mulsd %xmm1, %xmm0 divsd .LC6(%rip), %xmm0 leaq .LC7(%rip), %rsi movl $2, %edi movl $1, %eax call __printf_chk@PLT movl 404(%rsp), %edx leaq .LC8(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT movl 612(%rsp), %edx leaq .LC9(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT movq 304(%rsp), %rdx leaq .LC10(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT movq 312(%rsp), %rdx leaq .LC11(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT movl 320(%rsp), %edx leaq .LC12(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT movl 324(%rsp), %edx leaq .LC13(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT movq 328(%rsp), %rdx leaq .LC14(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT movl 336(%rsp), %edx leaq .LC15(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT movl 416(%rsp), %edx leaq .LC16(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT addl $1, %ebx cmpl %ebx, 12(%rsp) jg .L5 .L4: movq 1048(%rsp), %rax subq %fs:40, %rax jne .L9 movl $0, %eax addq $1056, %rsp .cfi_remember_state .cfi_def_cfa_offset 48 popq %rbx .cfi_def_cfa_offset 40 popq %rbp .cfi_def_cfa_offset 32 popq %r12 .cfi_def_cfa_offset 24 popq %r13 .cfi_def_cfa_offset 16 popq %r14 .cfi_def_cfa_offset 8 ret .L9: .cfi_restore_state call __stack_chk_fail@PLT .cfi_endproc .LFE2057: .size main, .-main .type _ZL24__sti____cudaRegisterAllv, @function _ZL24__sti____cudaRegisterAllv: .LFB2083: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 leaq _ZL15__fatDeviceText(%rip), %rdi call __cudaRegisterFatBinary@PLT movq %rax, %rdi movq %rax, _ZL20__cudaFatCubinHandle(%rip) call __cudaRegisterFatBinaryEnd@PLT leaq _ZL26__cudaUnregisterBinaryUtilv(%rip), %rdi call atexit@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2083: .size _ZL24__sti____cudaRegisterAllv, .-_ZL24__sti____cudaRegisterAllv .section .init_array,"aw" .align 8 .quad _ZL24__sti____cudaRegisterAllv .section .nvFatBinSegment,"aw" .align 8 .type _ZL15__fatDeviceText, @object .size _ZL15__fatDeviceText, 24 _ZL15__fatDeviceText: .long 1180844977 .long 1 .quad fatbinData .quad 0 .local _ZL20__cudaFatCubinHandle .comm _ZL20__cudaFatCubinHandle,8,8 .section .rodata.cst8,"aM",@progbits,8 .align 8 .LC6: .long 0 .long 1093567616 .ident "GCC: (Ubuntu 13.3.0-6ubuntu2~24.04) 13.3.0" .section .note.GNU-stack,"",@progbits .section .note.gnu.property,"a" .align 8 .long 1f - 0f .long 4f - 1f .long 5 0: .string "GNU" 1: .align 8 .long 0xc0000002 .long 3f - 2f 2: .long 0x3 3: .align 8 4:
You are a helpful coding assistant specialized in converting CUDA source code to HIP/AMD source code.
#include <stdio.h> int main() { int nDevices; cudaGetDeviceCount(&nDevices); for (int i = 0; i < nDevices; i++) { cudaDeviceProp prop; cudaGetDeviceProperties(&prop, i); printf("Device Number: %d\n", i); printf(" Device name: %s\n", prop.name); printf(" Version %d.%d\n", prop.major, prop.minor); printf(" Compute Mode: %d\n", prop.computeMode); printf(" Memory Clock Rate (KHz): %d\n", prop.memoryClockRate); printf(" Memory Bus Width (bits): %d\n", prop.memoryBusWidth); printf(" Peak Memory Bandwidth (GB/s): %f\n\n", 2.0*prop.memoryClockRate*(prop.memoryBusWidth/8)/1.0e6); printf(" Multi Processor Count: %d\n\n", prop.multiProcessorCount); printf(" TCC Driver: %d\n\n", prop.tccDriver); printf(" Total Global Mem: %d\n\n", prop.totalGlobalMem); printf(" Shared Mem Per Block: %d\n\n", prop.sharedMemPerBlock); printf(" Registers Per Block: %d\n\n", prop.regsPerBlock); printf(" Warpsize: %d\n\n", prop.warpSize); printf(" MemPitch: %d\n\n", prop.memPitch); printf(" MaxThreadsPerBlock: %d\n\n", prop.maxThreadsPerBlock); printf(" Can Map Host Memory: %d\n\n", prop.canMapHostMemory); } }
#include <hip/hip_runtime.h> #include <stdio.h> int main() { int nDevices; hipGetDeviceCount(&nDevices); for (int i = 0; i < nDevices; i++) { hipDeviceProp_t prop; hipGetDeviceProperties(&prop, i); printf("Device Number: %d\n", i); printf(" Device name: %s\n", prop.name); printf(" Version %d.%d\n", prop.major, prop.minor); printf(" Compute Mode: %d\n", prop.computeMode); printf(" Memory Clock Rate (KHz): %d\n", prop.memoryClockRate); printf(" Memory Bus Width (bits): %d\n", prop.memoryBusWidth); printf(" Peak Memory Bandwidth (GB/s): %f\n\n", 2.0*prop.memoryClockRate*(prop.memoryBusWidth/8)/1.0e6); printf(" Multi Processor Count: %d\n\n", prop.multiProcessorCount); printf(" TCC Driver: %d\n\n", prop.tccDriver); printf(" Total Global Mem: %d\n\n", prop.totalGlobalMem); printf(" Shared Mem Per Block: %d\n\n", prop.sharedMemPerBlock); printf(" Registers Per Block: %d\n\n", prop.regsPerBlock); printf(" Warpsize: %d\n\n", prop.warpSize); printf(" MemPitch: %d\n\n", prop.memPitch); printf(" MaxThreadsPerBlock: %d\n\n", prop.maxThreadsPerBlock); printf(" Can Map Host Memory: %d\n\n", prop.canMapHostMemory); } }
You are a helpful coding assistant specialized in converting HIP source code to AMD device assembly.
#include <hip/hip_runtime.h> #include <stdio.h> int main() { int nDevices; hipGetDeviceCount(&nDevices); for (int i = 0; i < nDevices; i++) { hipDeviceProp_t prop; hipGetDeviceProperties(&prop, i); printf("Device Number: %d\n", i); printf(" Device name: %s\n", prop.name); printf(" Version %d.%d\n", prop.major, prop.minor); printf(" Compute Mode: %d\n", prop.computeMode); printf(" Memory Clock Rate (KHz): %d\n", prop.memoryClockRate); printf(" Memory Bus Width (bits): %d\n", prop.memoryBusWidth); printf(" Peak Memory Bandwidth (GB/s): %f\n\n", 2.0*prop.memoryClockRate*(prop.memoryBusWidth/8)/1.0e6); printf(" Multi Processor Count: %d\n\n", prop.multiProcessorCount); printf(" TCC Driver: %d\n\n", prop.tccDriver); printf(" Total Global Mem: %d\n\n", prop.totalGlobalMem); printf(" Shared Mem Per Block: %d\n\n", prop.sharedMemPerBlock); printf(" Registers Per Block: %d\n\n", prop.regsPerBlock); printf(" Warpsize: %d\n\n", prop.warpSize); printf(" MemPitch: %d\n\n", prop.memPitch); printf(" MaxThreadsPerBlock: %d\n\n", prop.maxThreadsPerBlock); printf(" Can Map Host Memory: %d\n\n", prop.canMapHostMemory); } }
.text .p2alignl 7, 3214868480 .fill 96, 4, 3214868480 .type __hip_cuid_,@object .section .bss,"aw",@nobits .globl __hip_cuid_ __hip_cuid_: .byte 0 .size __hip_cuid_, 1 .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym __hip_cuid_ .amdgcn_target "amdgcn-amd-amdhsa--gfx1100" .amdgpu_metadata --- amdhsa.kernels: [] amdhsa.target: amdgcn-amd-amdhsa--gfx1100 amdhsa.version: - 1 - 2 ... .end_amdgpu_metadata
You are a helpful coding assistant specialized in converting HIP source code to AMD host assembly.
#include <hip/hip_runtime.h> #include <stdio.h> int main() { int nDevices; hipGetDeviceCount(&nDevices); for (int i = 0; i < nDevices; i++) { hipDeviceProp_t prop; hipGetDeviceProperties(&prop, i); printf("Device Number: %d\n", i); printf(" Device name: %s\n", prop.name); printf(" Version %d.%d\n", prop.major, prop.minor); printf(" Compute Mode: %d\n", prop.computeMode); printf(" Memory Clock Rate (KHz): %d\n", prop.memoryClockRate); printf(" Memory Bus Width (bits): %d\n", prop.memoryBusWidth); printf(" Peak Memory Bandwidth (GB/s): %f\n\n", 2.0*prop.memoryClockRate*(prop.memoryBusWidth/8)/1.0e6); printf(" Multi Processor Count: %d\n\n", prop.multiProcessorCount); printf(" TCC Driver: %d\n\n", prop.tccDriver); printf(" Total Global Mem: %d\n\n", prop.totalGlobalMem); printf(" Shared Mem Per Block: %d\n\n", prop.sharedMemPerBlock); printf(" Registers Per Block: %d\n\n", prop.regsPerBlock); printf(" Warpsize: %d\n\n", prop.warpSize); printf(" MemPitch: %d\n\n", prop.memPitch); printf(" MaxThreadsPerBlock: %d\n\n", prop.maxThreadsPerBlock); printf(" Can Map Host Memory: %d\n\n", prop.canMapHostMemory); } }
.text .file "cuda_properties.hip" .section .rodata.cst8,"aM",@progbits,8 .p2align 3, 0x0 # -- Begin function main .LCPI0_0: .quad 0x412e848000000000 # double 1.0E+6 .text .globl main .p2align 4, 0x90 .type main,@function main: # @main .cfi_startproc # %bb.0: pushq %rbp .cfi_def_cfa_offset 16 pushq %rbx .cfi_def_cfa_offset 24 subq $1480, %rsp # imm = 0x5C8 .cfi_def_cfa_offset 1504 .cfi_offset %rbx, -24 .cfi_offset %rbp, -16 leaq 4(%rsp), %rdi callq hipGetDeviceCount cmpl $0, 4(%rsp) jle .LBB0_3 # %bb.1: # %.lr.ph leaq 8(%rsp), %rbx xorl %ebp, %ebp .p2align 4, 0x90 .LBB0_2: # =>This Inner Loop Header: Depth=1 movq %rbx, %rdi movl %ebp, %esi callq hipGetDevicePropertiesR0600 movl $.L.str, %edi movl %ebp, %esi xorl %eax, %eax callq printf movl $.L.str.1, %edi movq %rbx, %rsi xorl %eax, %eax callq printf movl 368(%rsp), %esi movl 372(%rsp), %edx movl $.L.str.2, %edi xorl %eax, %eax callq printf movl 412(%rsp), %esi movl $.L.str.3, %edi xorl %eax, %eax callq printf movl 616(%rsp), %esi movl $.L.str.4, %edi xorl %eax, %eax callq printf movl 620(%rsp), %esi movl $.L.str.5, %edi xorl %eax, %eax callq printf cvtsi2sdl 616(%rsp), %xmm1 addsd %xmm1, %xmm1 movl 620(%rsp), %eax leal 7(%rax), %ecx testl %eax, %eax cmovnsl %eax, %ecx sarl $3, %ecx cvtsi2sd %ecx, %xmm0 mulsd %xmm1, %xmm0 divsd .LCPI0_0(%rip), %xmm0 movl $.L.str.6, %edi movb $1, %al callq printf movl 396(%rsp), %esi movl $.L.str.7, %edi xorl %eax, %eax callq printf movl 604(%rsp), %esi movl $.L.str.8, %edi xorl %eax, %eax callq printf movq 296(%rsp), %rsi movl $.L.str.9, %edi xorl %eax, %eax callq printf movq 304(%rsp), %rsi movl $.L.str.10, %edi xorl %eax, %eax callq printf movl 312(%rsp), %esi movl $.L.str.11, %edi xorl %eax, %eax callq printf movl 316(%rsp), %esi movl $.L.str.12, %edi xorl %eax, %eax callq printf movq 320(%rsp), %rsi movl $.L.str.13, %edi xorl %eax, %eax callq printf movl 328(%rsp), %esi movl $.L.str.14, %edi xorl %eax, %eax callq printf movl 408(%rsp), %esi movl $.L.str.15, %edi xorl %eax, %eax callq printf incl %ebp cmpl 4(%rsp), %ebp jl .LBB0_2 .LBB0_3: # %._crit_edge xorl %eax, %eax addq $1480, %rsp # imm = 0x5C8 .cfi_def_cfa_offset 24 popq %rbx .cfi_def_cfa_offset 16 popq %rbp .cfi_def_cfa_offset 8 retq .Lfunc_end0: .size main, .Lfunc_end0-main .cfi_endproc # -- End function .type .L.str,@object # @.str .section .rodata.str1.1,"aMS",@progbits,1 .L.str: .asciz "Device Number: %d\n" .size .L.str, 19 .type .L.str.1,@object # @.str.1 .L.str.1: .asciz " Device name: %s\n" .size .L.str.1, 19 .type .L.str.2,@object # @.str.2 .L.str.2: .asciz " Version %d.%d\n" .size .L.str.2, 17 .type .L.str.3,@object # @.str.3 .L.str.3: .asciz " Compute Mode: %d\n" .size .L.str.3, 20 .type .L.str.4,@object # @.str.4 .L.str.4: .asciz " Memory Clock Rate (KHz): %d\n" .size .L.str.4, 31 .type .L.str.5,@object # @.str.5 .L.str.5: .asciz " Memory Bus Width (bits): %d\n" .size .L.str.5, 31 .type .L.str.6,@object # @.str.6 .L.str.6: .asciz " Peak Memory Bandwidth (GB/s): %f\n\n" .size .L.str.6, 37 .type .L.str.7,@object # @.str.7 .L.str.7: .asciz " Multi Processor Count: %d\n\n" .size .L.str.7, 30 .type .L.str.8,@object # @.str.8 .L.str.8: .asciz " TCC Driver: %d\n\n" .size .L.str.8, 19 .type .L.str.9,@object # @.str.9 .L.str.9: .asciz " Total Global Mem: %d\n\n" .size .L.str.9, 25 .type .L.str.10,@object # @.str.10 .L.str.10: .asciz " Shared Mem Per Block: %d\n\n" .size .L.str.10, 29 .type .L.str.11,@object # @.str.11 .L.str.11: .asciz " Registers Per Block: %d\n\n" .size .L.str.11, 28 .type .L.str.12,@object # @.str.12 .L.str.12: .asciz " Warpsize: %d\n\n" .size .L.str.12, 17 .type .L.str.13,@object # @.str.13 .L.str.13: .asciz " MemPitch: %d\n\n" .size .L.str.13, 17 .type .L.str.14,@object # @.str.14 .L.str.14: .asciz " MaxThreadsPerBlock: %d\n\n" .size .L.str.14, 27 .type .L.str.15,@object # @.str.15 .L.str.15: .asciz " Can Map Host Memory: %d\n\n" .size .L.str.15, 28 .type __hip_cuid_,@object # @__hip_cuid_ .bss .globl __hip_cuid_ __hip_cuid_: .byte 0 # 0x0 .size __hip_cuid_, 1 .section ".linker-options","e",@llvm_linker_options .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym __hip_cuid_
You are a helpful coding assistant specialized in converting from CUDA device assembly to AMD device assembly.
code for sm_80
.text .p2alignl 7, 3214868480 .fill 96, 4, 3214868480 .type __hip_cuid_,@object .section .bss,"aw",@nobits .globl __hip_cuid_ __hip_cuid_: .byte 0 .size __hip_cuid_, 1 .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym __hip_cuid_ .amdgcn_target "amdgcn-amd-amdhsa--gfx1100" .amdgpu_metadata --- amdhsa.kernels: [] amdhsa.target: amdgcn-amd-amdhsa--gfx1100 amdhsa.version: - 1 - 2 ... .end_amdgpu_metadata
You are a helpful coding assistant specialized in converting from CUDA host assembly to AMD host assembly.
.file "tmpxft_001631ba_00000000-6_cuda_properties.cudafe1.cpp" .text #APP #NO_APP .type _ZL26__cudaUnregisterBinaryUtilv, @function _ZL26__cudaUnregisterBinaryUtilv: .LFB2060: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaUnregisterFatBinary@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2060: .size _ZL26__cudaUnregisterBinaryUtilv, .-_ZL26__cudaUnregisterBinaryUtilv .section .rodata.str1.1,"aMS",@progbits,1 .LC0: .string "Device Number: %d\n" .LC1: .string " Device name: %s\n" .LC2: .string " Version %d.%d\n" .LC3: .string " Compute Mode: %d\n" .section .rodata.str1.8,"aMS",@progbits,1 .align 8 .LC4: .string " Memory Clock Rate (KHz): %d\n" .align 8 .LC5: .string " Memory Bus Width (bits): %d\n" .align 8 .LC7: .string " Peak Memory Bandwidth (GB/s): %f\n\n" .section .rodata.str1.1 .LC8: .string " Multi Processor Count: %d\n\n" .LC9: .string " TCC Driver: %d\n\n" .LC10: .string " Total Global Mem: %d\n\n" .LC11: .string " Shared Mem Per Block: %d\n\n" .LC12: .string " Registers Per Block: %d\n\n" .LC13: .string " Warpsize: %d\n\n" .LC14: .string " MemPitch: %d\n\n" .LC15: .string " MaxThreadsPerBlock: %d\n\n" .LC16: .string " Can Map Host Memory: %d\n\n" .text .globl main .type main, @function main: .LFB2057: .cfi_startproc endbr64 pushq %r14 .cfi_def_cfa_offset 16 .cfi_offset 14, -16 pushq %r13 .cfi_def_cfa_offset 24 .cfi_offset 13, -24 pushq %r12 .cfi_def_cfa_offset 32 .cfi_offset 12, -32 pushq %rbp .cfi_def_cfa_offset 40 .cfi_offset 6, -40 pushq %rbx .cfi_def_cfa_offset 48 .cfi_offset 3, -48 subq $1056, %rsp .cfi_def_cfa_offset 1104 movq %fs:40, %rax movq %rax, 1048(%rsp) xorl %eax, %eax leaq 12(%rsp), %rdi call cudaGetDeviceCount@PLT cmpl $0, 12(%rsp) jle .L4 movl $0, %ebx leaq .LC0(%rip), %r13 leaq .LC1(%rip), %r12 leaq .LC2(%rip), %rbp .L5: leaq 16(%rsp), %r14 movl %ebx, %esi movq %r14, %rdi call cudaGetDeviceProperties_v2@PLT movl %ebx, %edx movq %r13, %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT movq %r14, %rdx movq %r12, %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT movl 380(%rsp), %ecx movl 376(%rsp), %edx movq %rbp, %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT movl 420(%rsp), %edx leaq .LC3(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT movl 624(%rsp), %edx leaq .LC4(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT movl 628(%rsp), %edx leaq .LC5(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT pxor %xmm0, %xmm0 cvtsi2sdl 624(%rsp), %xmm0 addsd %xmm0, %xmm0 movl 628(%rsp), %edx leal 7(%rdx), %eax testl %edx, %edx cmovns %edx, %eax sarl $3, %eax pxor %xmm1, %xmm1 cvtsi2sdl %eax, %xmm1 mulsd %xmm1, %xmm0 divsd .LC6(%rip), %xmm0 leaq .LC7(%rip), %rsi movl $2, %edi movl $1, %eax call __printf_chk@PLT movl 404(%rsp), %edx leaq .LC8(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT movl 612(%rsp), %edx leaq .LC9(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT movq 304(%rsp), %rdx leaq .LC10(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT movq 312(%rsp), %rdx leaq .LC11(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT movl 320(%rsp), %edx leaq .LC12(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT movl 324(%rsp), %edx leaq .LC13(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT movq 328(%rsp), %rdx leaq .LC14(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT movl 336(%rsp), %edx leaq .LC15(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT movl 416(%rsp), %edx leaq .LC16(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT addl $1, %ebx cmpl %ebx, 12(%rsp) jg .L5 .L4: movq 1048(%rsp), %rax subq %fs:40, %rax jne .L9 movl $0, %eax addq $1056, %rsp .cfi_remember_state .cfi_def_cfa_offset 48 popq %rbx .cfi_def_cfa_offset 40 popq %rbp .cfi_def_cfa_offset 32 popq %r12 .cfi_def_cfa_offset 24 popq %r13 .cfi_def_cfa_offset 16 popq %r14 .cfi_def_cfa_offset 8 ret .L9: .cfi_restore_state call __stack_chk_fail@PLT .cfi_endproc .LFE2057: .size main, .-main .type _ZL24__sti____cudaRegisterAllv, @function _ZL24__sti____cudaRegisterAllv: .LFB2083: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 leaq _ZL15__fatDeviceText(%rip), %rdi call __cudaRegisterFatBinary@PLT movq %rax, %rdi movq %rax, _ZL20__cudaFatCubinHandle(%rip) call __cudaRegisterFatBinaryEnd@PLT leaq _ZL26__cudaUnregisterBinaryUtilv(%rip), %rdi call atexit@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2083: .size _ZL24__sti____cudaRegisterAllv, .-_ZL24__sti____cudaRegisterAllv .section .init_array,"aw" .align 8 .quad _ZL24__sti____cudaRegisterAllv .section .nvFatBinSegment,"aw" .align 8 .type _ZL15__fatDeviceText, @object .size _ZL15__fatDeviceText, 24 _ZL15__fatDeviceText: .long 1180844977 .long 1 .quad fatbinData .quad 0 .local _ZL20__cudaFatCubinHandle .comm _ZL20__cudaFatCubinHandle,8,8 .section .rodata.cst8,"aM",@progbits,8 .align 8 .LC6: .long 0 .long 1093567616 .ident "GCC: (Ubuntu 13.3.0-6ubuntu2~24.04) 13.3.0" .section .note.GNU-stack,"",@progbits .section .note.gnu.property,"a" .align 8 .long 1f - 0f .long 4f - 1f .long 5 0: .string "GNU" 1: .align 8 .long 0xc0000002 .long 3f - 2f 2: .long 0x3 3: .align 8 4:
.text .file "cuda_properties.hip" .section .rodata.cst8,"aM",@progbits,8 .p2align 3, 0x0 # -- Begin function main .LCPI0_0: .quad 0x412e848000000000 # double 1.0E+6 .text .globl main .p2align 4, 0x90 .type main,@function main: # @main .cfi_startproc # %bb.0: pushq %rbp .cfi_def_cfa_offset 16 pushq %rbx .cfi_def_cfa_offset 24 subq $1480, %rsp # imm = 0x5C8 .cfi_def_cfa_offset 1504 .cfi_offset %rbx, -24 .cfi_offset %rbp, -16 leaq 4(%rsp), %rdi callq hipGetDeviceCount cmpl $0, 4(%rsp) jle .LBB0_3 # %bb.1: # %.lr.ph leaq 8(%rsp), %rbx xorl %ebp, %ebp .p2align 4, 0x90 .LBB0_2: # =>This Inner Loop Header: Depth=1 movq %rbx, %rdi movl %ebp, %esi callq hipGetDevicePropertiesR0600 movl $.L.str, %edi movl %ebp, %esi xorl %eax, %eax callq printf movl $.L.str.1, %edi movq %rbx, %rsi xorl %eax, %eax callq printf movl 368(%rsp), %esi movl 372(%rsp), %edx movl $.L.str.2, %edi xorl %eax, %eax callq printf movl 412(%rsp), %esi movl $.L.str.3, %edi xorl %eax, %eax callq printf movl 616(%rsp), %esi movl $.L.str.4, %edi xorl %eax, %eax callq printf movl 620(%rsp), %esi movl $.L.str.5, %edi xorl %eax, %eax callq printf cvtsi2sdl 616(%rsp), %xmm1 addsd %xmm1, %xmm1 movl 620(%rsp), %eax leal 7(%rax), %ecx testl %eax, %eax cmovnsl %eax, %ecx sarl $3, %ecx cvtsi2sd %ecx, %xmm0 mulsd %xmm1, %xmm0 divsd .LCPI0_0(%rip), %xmm0 movl $.L.str.6, %edi movb $1, %al callq printf movl 396(%rsp), %esi movl $.L.str.7, %edi xorl %eax, %eax callq printf movl 604(%rsp), %esi movl $.L.str.8, %edi xorl %eax, %eax callq printf movq 296(%rsp), %rsi movl $.L.str.9, %edi xorl %eax, %eax callq printf movq 304(%rsp), %rsi movl $.L.str.10, %edi xorl %eax, %eax callq printf movl 312(%rsp), %esi movl $.L.str.11, %edi xorl %eax, %eax callq printf movl 316(%rsp), %esi movl $.L.str.12, %edi xorl %eax, %eax callq printf movq 320(%rsp), %rsi movl $.L.str.13, %edi xorl %eax, %eax callq printf movl 328(%rsp), %esi movl $.L.str.14, %edi xorl %eax, %eax callq printf movl 408(%rsp), %esi movl $.L.str.15, %edi xorl %eax, %eax callq printf incl %ebp cmpl 4(%rsp), %ebp jl .LBB0_2 .LBB0_3: # %._crit_edge xorl %eax, %eax addq $1480, %rsp # imm = 0x5C8 .cfi_def_cfa_offset 24 popq %rbx .cfi_def_cfa_offset 16 popq %rbp .cfi_def_cfa_offset 8 retq .Lfunc_end0: .size main, .Lfunc_end0-main .cfi_endproc # -- End function .type .L.str,@object # @.str .section .rodata.str1.1,"aMS",@progbits,1 .L.str: .asciz "Device Number: %d\n" .size .L.str, 19 .type .L.str.1,@object # @.str.1 .L.str.1: .asciz " Device name: %s\n" .size .L.str.1, 19 .type .L.str.2,@object # @.str.2 .L.str.2: .asciz " Version %d.%d\n" .size .L.str.2, 17 .type .L.str.3,@object # @.str.3 .L.str.3: .asciz " Compute Mode: %d\n" .size .L.str.3, 20 .type .L.str.4,@object # @.str.4 .L.str.4: .asciz " Memory Clock Rate (KHz): %d\n" .size .L.str.4, 31 .type .L.str.5,@object # @.str.5 .L.str.5: .asciz " Memory Bus Width (bits): %d\n" .size .L.str.5, 31 .type .L.str.6,@object # @.str.6 .L.str.6: .asciz " Peak Memory Bandwidth (GB/s): %f\n\n" .size .L.str.6, 37 .type .L.str.7,@object # @.str.7 .L.str.7: .asciz " Multi Processor Count: %d\n\n" .size .L.str.7, 30 .type .L.str.8,@object # @.str.8 .L.str.8: .asciz " TCC Driver: %d\n\n" .size .L.str.8, 19 .type .L.str.9,@object # @.str.9 .L.str.9: .asciz " Total Global Mem: %d\n\n" .size .L.str.9, 25 .type .L.str.10,@object # @.str.10 .L.str.10: .asciz " Shared Mem Per Block: %d\n\n" .size .L.str.10, 29 .type .L.str.11,@object # @.str.11 .L.str.11: .asciz " Registers Per Block: %d\n\n" .size .L.str.11, 28 .type .L.str.12,@object # @.str.12 .L.str.12: .asciz " Warpsize: %d\n\n" .size .L.str.12, 17 .type .L.str.13,@object # @.str.13 .L.str.13: .asciz " MemPitch: %d\n\n" .size .L.str.13, 17 .type .L.str.14,@object # @.str.14 .L.str.14: .asciz " MaxThreadsPerBlock: %d\n\n" .size .L.str.14, 27 .type .L.str.15,@object # @.str.15 .L.str.15: .asciz " Can Map Host Memory: %d\n\n" .size .L.str.15, 28 .type __hip_cuid_,@object # @__hip_cuid_ .bss .globl __hip_cuid_ __hip_cuid_: .byte 0 # 0x0 .size __hip_cuid_, 1 .section ".linker-options","e",@llvm_linker_options .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym __hip_cuid_
You are a helpful coding assistant specialized in converting CUDA source code to CUDA host assembly.
// Copyright 2020 Christopher Khan // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. // You may obtain a copy of the license at // http://www.apache.org/licenses/LICENSE-2.0 // Unless required by applicable law or agreed to in writing, software // distributed under the License is distributed on an "AS IS" BASIS, // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. // See the License for the specific language governing permissions and // limitations under the License. // Description of GPU_kernels_single_precision.cu: // This file contains the CUDA code that allows for performing the computations // for GENRE on a GPU using single precision // Define the GPU kernel that performs predictor normalization __global__ void predictor_normalization(float * X_matrix_d, float * scaling_factors_d, double * X_matrix_thread_stride_d, double * B_thread_stride_d, double * num_observations_d, double * num_predictors_d, float * intercept_flag_d, int num_threads_per_block, int num_blocks, int num_threads_last_block) { // Obtain the index of the block int block_ind = blockIdx.x; // Obtain the thread index within one block int block_thread_ind = threadIdx.x; // Calculate the fit index int fit_ind = (block_ind * num_threads_per_block) + block_thread_ind; // Determine how many threads are in the block (accounts for the fact that the last block may contain less active threads than the other blocks) int num_threads_per_block_2 = num_threads_per_block; if (block_ind == (num_blocks - 1)) { num_threads_per_block_2 = num_threads_last_block; } // This if statement makes sure that extra threads aren't doing data processing if the last block has less fits to perform if (block_thread_ind < num_threads_per_block_2) { // Obtain the thread stride that is used to obtain the correct set of predictors for the fit int predictor_thread_stride = (int)B_thread_stride_d[fit_ind]; // Obtain the number of observations for the fit int num_observations = (int)num_observations_d[fit_ind]; // Obtain the number of predictors for the fit int num_predictors = (int)num_predictors_d[fit_ind]; // Obtain the thread stride that is used to obtain the correct model matrix for the fit int X_thread_stride = (int)X_matrix_thread_stride_d[fit_ind]; // Obtain the flag that determines whether the first predictor column is a column of ones for the intercept term or not int intercept_flag = (int)intercept_flag_d[fit_ind]; // Declare and initialize the variable that stores the number of the first predictor column to be normalized int start_ind = 0; // This if statement makes sure to not normalize the first predictor column if it corresponds to the intercept term if (intercept_flag == 1) { start_ind = 1; } // Normalize each predictor column so that the sum of the square of each predictor column is equal to 1 for (int predictor_column = start_ind; predictor_column < num_predictors; predictor_column++) { // Declare and initialize the variable that stores the sum of the square of the predictor column float sum_squared = 0.0f; // Calculate the sum of the square of the predictor column for (int observation_row = 0; observation_row < num_observations; observation_row++) { float X_value = X_matrix_d[X_thread_stride + (predictor_column * num_observations) + observation_row]; sum_squared = sum_squared + (X_value * X_value); } // Calculate the square root of the sum of the square of the predictor column float square_root_sum_squared = sqrtf(sum_squared); // Store the square root of the sum of the square of the predictor column scaling_factors_d[predictor_thread_stride + predictor_column] = square_root_sum_squared; // Normalize the predictor column by dividing each observation in the predictor column by the square root of the sum of the square of the predictor column for (int observation_row = 0; observation_row < num_observations; observation_row++) { X_matrix_d[X_thread_stride + (predictor_column * num_observations) + observation_row] = X_matrix_d[X_thread_stride + (predictor_column * num_observations) + observation_row] / square_root_sum_squared; } } // This if statement stores a scaling factor of 1 for the predictor column if it corresponds to an intercept term if (intercept_flag == 1) { scaling_factors_d[predictor_thread_stride] = 1.0f; } } } // Define the GPU kernel that performs predictor standardization __global__ void predictor_standardization(float * X_matrix_d, float * scaling_factors_d, float * mean_X_matrix_d, double * X_matrix_thread_stride_d, double * B_thread_stride_d, double * num_observations_d, double * num_predictors_d, float * intercept_flag_d, int num_threads_per_block, int num_blocks, int num_threads_last_block) { // Obtain the index of the block int block_ind = blockIdx.x; // Obtain the thread index within one block int block_thread_ind = threadIdx.x; // Calculate the fit index int fit_ind = (block_ind * num_threads_per_block) + block_thread_ind; // Determine how many threads are in the block (accounts for the fact that the last block may contain less active threads than the other blocks) int num_threads_per_block_2 = num_threads_per_block; if (block_ind == (num_blocks - 1)) { num_threads_per_block_2 = num_threads_last_block; } // This if statement makes sure that extra threads aren't doing data processing if the last block has less fits to perform if (block_thread_ind < num_threads_per_block_2) { // Obtain the thread stride that is used to obtain the correct set of predictors for the fit int predictor_thread_stride = (int)B_thread_stride_d[fit_ind]; // Obtain the number of observations for the fit int num_observations = (int)num_observations_d[fit_ind]; // Obtain the number of predictors for the fit int num_predictors = (int)num_predictors_d[fit_ind]; // Obtain the thread stride that is used to obtain the correct model matrix for the fit int X_thread_stride = (int)X_matrix_thread_stride_d[fit_ind]; // Obtain the flag that determines whether the first predictor column is a column of ones for the intercept term or not int intercept_flag = (int)intercept_flag_d[fit_ind]; // Declare and initialize the variable that stores the number of the first predictor column to be standardized int start_ind = 0; // This if statement makes sure to not standardize the first predictor column if it corresponds to the intercept term if (intercept_flag == 1) { start_ind = 1; } // Standardize each predictor column by subtracting the mean of the predictor column from each observation and diving each observation by the standard deviation of the predictor column for (int predictor_column = start_ind; predictor_column < num_predictors; predictor_column++) { // Declare and initialize the variable that stores the sum of the predictor column float sum_value = 0.0f; // Calculate the sum of the predictor column for (int observation_row = 0; observation_row < num_observations; observation_row++) { float X_value = X_matrix_d[X_thread_stride + (predictor_column * num_observations) + observation_row]; sum_value = sum_value + X_value; } // Calculate the mean of the predictor column float mean_value = sum_value / (float)num_observations; // Store the mean of the predictor column mean_X_matrix_d[predictor_thread_stride + predictor_column] = mean_value; // Declare and initialize the variable that stores the sum of the square of the demeaned predictor column float sum_squared = 0.0f; // Normalize the predictor column by dividing each observation in the predictor column by the square root of the sum of the square of the predictor column for (int observation_row = 0; observation_row < num_observations; observation_row++) { float X_value_demeaned = X_matrix_d[X_thread_stride + (predictor_column * num_observations) + observation_row] - mean_value; sum_squared = sum_squared + (X_value_demeaned * X_value_demeaned); } // Calculate the standard deviation of the demeaned predictor column float std = sqrtf(sum_squared / (float)num_observations); // Store the standard deviation of the demeaned predictor column scaling_factors_d[predictor_thread_stride + predictor_column] = std; // Standardize the predictor column by subtracting its mean and dividing by its standard deviation for (int observation_row = 0; observation_row < num_observations; observation_row++) { X_matrix_d[X_thread_stride + (predictor_column * num_observations) + observation_row] = (X_matrix_d[X_thread_stride + (predictor_column * num_observations) + observation_row] - mean_value) / std; } } // This if statement stores a scaling factor of 1 and a mean of 1 for the first column if it corresponds to an intercept term if (intercept_flag == 1) { scaling_factors_d[predictor_thread_stride] = 1.0f; mean_X_matrix_d[predictor_thread_stride] = 1.0f; } } } // Define the GPU kernel that calculates the standard deviations for each portion of the y_d array, standardizes the y_d array, and calculates the standardized lambda values __global__ void model_fit_preparation(float * y_d, float * residual_y_d, float * model_fit_flag_d, float * y_std_d, float * standardized_lambda_values_d, double * num_observations_d, double * observation_thread_stride_d, int num_threads_per_block, int num_blocks, int num_threads_last_block) { // Obtain the index of the block int block_ind = blockIdx.x; // Obtain the thread index within one block int block_thread_ind = threadIdx.x; // Calculate the fit index int fit_ind = (block_ind * num_threads_per_block) + block_thread_ind; // Determine how many threads are in the block (accounts for the fact that the last block may contain less active threads than the other blocks) int num_threads_per_block_2 = num_threads_per_block; if (block_ind == (num_blocks - 1)) { num_threads_per_block_2 = num_threads_last_block; } // This if statement makes sure that extra threads aren't doing data processing if the last block has less fits to perform if (block_thread_ind < num_threads_per_block_2) { // Obtain the number of observations for the fit int num_observations = (int)num_observations_d[fit_ind]; // Obtain the thread stride that is used to obtain the correct set of observations in the cropped_y_d array for the fit int observation_thread_stride = (int)observation_thread_stride_d[fit_ind]; // Declare and initialize the variable that stores the running sum of y for the fit float sum_value = 0.0f; // Calculate the running sums for sum_value for (int observation = 0; observation < num_observations; observation++) { float value = y_d[observation_thread_stride + observation]; sum_value += value; } // Calculate the mean of y for the fit float mean = sum_value / (float)num_observations; // Declare and initialize the variable that stores the standard deviation of y for the fit float std = 0.0f; // Calculate the standard deviation of y for the fit for (int observation = 0; observation < num_observations; observation++) { float value_2 = y_d[observation_thread_stride + observation]; std += ((value_2 - mean) * (value_2 - mean)); } std = sqrtf(std / (float)num_observations); // Store the standard deviation of y for the fit in the y_std_d array y_std_d[fit_ind] = std; // This if statement standardizes the lambda values and the y data if the standard deviation isn't 0 if (std != 0.0f) { // Set the model fit flag to 1 if the standard deviation is not 0 and a model fit should be performed model_fit_flag_d[fit_ind] = 1.0f; // Calculate the standardized lambda value and store it into the standardized_lambda_d array standardized_lambda_values_d[fit_ind] = standardized_lambda_values_d[fit_ind] / std; // Standardize y for the fit and store it into the y_d array and the residual_y_d array for (int observation = 0; observation < num_observations; observation++) { float standardized_value = y_d[observation_thread_stride + observation] / std; y_d[observation_thread_stride + observation] = standardized_value; residual_y_d[observation_thread_stride + observation] = standardized_value; } } } } // Define the GPU kernel that performs least-squares regression with elastic-net regularization using the cyclic coordinate descent optimization algorithm in order to fit the model matrices to the data __global__ void model_fit(float * B_d, double * B_thread_stride_d, float * model_fit_flag_d, float * X_matrix_d, double * X_matrix_thread_stride_d, double * observation_thread_stride_d, float * residual_y_d, float * y_std_d, float * standardized_lambda_values_d, double * num_observations_d, double * num_predictors_d, float * alpha_values_d, float * tolerance_values_d, float * max_iterations_values_d, float * intercept_flag_d, int transformation_flag, int num_threads_per_block, int num_blocks, int num_threads_last_block) { // Obtain the index of the block int block_ind = blockIdx.x; // Obtain the thread index within one block int block_thread_ind = threadIdx.x; // Calculate the fit index int fit_ind = (block_ind * num_threads_per_block) + block_thread_ind; // Determine how many threads are in the block (accounts for the fact that the last block may contain less active threads than the other blocks) int num_threads_per_block_2 = num_threads_per_block; if (block_ind == (num_blocks - 1)) { num_threads_per_block_2 = num_threads_last_block; } // This if statement makes sure that extra threads aren't doing data processing if the last block has less threads if (block_thread_ind < num_threads_per_block_2) { // Obtain the flag that determines whether to perform a model fit or not int model_fit_flag = (int)model_fit_flag_d[fit_ind]; // This if statement is to ensure that a model fit is performed only if the model fit flag is 1 if (model_fit_flag == 1) { // Obtain the thread stride that is used to obtain the correct set of predictors for the fit int predictor_thread_stride = (int)B_thread_stride_d[fit_ind]; // Obtain the thread stride that is used to obtain the correct model matrix for the fit int X_thread_stride = (int)X_matrix_thread_stride_d[fit_ind]; // Obtain the thread stride that is used to obtain the correct set of observations for the fit int observation_thread_stride = (int)observation_thread_stride_d[fit_ind]; // Obtain the alpha value for the fit float alpha = alpha_values_d[fit_ind]; // Obtain the standardized lambda value for the fit float lambda = standardized_lambda_values_d[fit_ind]; // Obtain the tolerance value for the fit float tolerance = tolerance_values_d[fit_ind]; // Obtain the max iterations value for the fit int max_iterations = (int)max_iterations_values_d[fit_ind]; // Obtain the number of observations for the fit int num_observations = (int)num_observations_d[fit_ind]; // Obtain the number of predictors for the fit int num_predictors = (int)num_predictors_d[fit_ind]; // Obtain the flag that determines whether the first predictor column is a column of ones for the intercept term or not int intercept_flag = (int)intercept_flag_d[fit_ind]; // Declare and initialize the variable that stores the maximum weighted (observation weights are all 1 in this case) sum of squares of the changes in the fitted values for one iteration of cyclic coordinate descent float global_max_change = 1E12; // Declare and initialize the variable that counts how many iterations of cyclic coordinate descent have been performed int iteration_count = 0; // Perform cyclic coordinate descent until either the maximum number of iterations is reached or the maximum weighted (observation weights are all 1 in this case) sum of squares of the changes in the fitted values becomes less than the tolerance while (global_max_change >= tolerance && iteration_count < max_iterations) { // Declare and initialize the variable that stores the maximum weighted (observation weights are all 1 in this case) sum of squares of the changes in the fitted values for one iteration of cyclic coordinate descent float max_change = 0.0f; // Declare and initialize the variable that stores the weighted (observation weights are all 1 in this case) sum of squares of the changes in the fitted values that are due to the current predictor coefficient value being updated using cyclic coordinate descent float change = 0.0f; // Cycle through all of the predictors for one iteration of cyclic coordinate descent for (int j = 0; j < num_predictors; j++) { // Obtain the predictor coefficient value for the current predictor float B_j = B_d[predictor_thread_stride + j]; // Store the predictor coefficent value before it's updated float previous_B_j = B_j; // Declare and initialize the variable that stores the correlation between the current predictor column and the residual values that are obtained leaving the current predictor out float p_j = 0.0f; // Calculate the residual values leaving the current predictor out (the predictor coefficients are initialized to zero, so the residual values are going to initially be y) // This if-else statement accounts for the fact that the contribution of the current predictor only needs to be removed from the residual values if the predictor coefficient is not zero // This is due to the fact that if the predictor coefficient is already zero, then the predictor contribution to the residual is zero if (B_j != 0.0f) { for (int observation_row = 0; observation_row < num_observations; observation_row++) { // Obtain the correct value from the model matrix for the current predictor float X_value = X_matrix_d[X_thread_stride + (j * num_observations) + observation_row]; // Remove the contribution of the current predictor from the current residual value float residual_y_value = residual_y_d[observation_thread_stride + observation_row] + (X_value * B_j); // Store the updated residual value back into the residual_y_d array residual_y_d[observation_thread_stride + observation_row] = residual_y_value; // Compute the correlation between the current predictor column and the residual values that are obtained leaving the current predictor out // The correlation is computed as a running sum p_j = p_j + (X_value * residual_y_value); } } else { for (int observation_row = 0; observation_row < num_observations; observation_row++) { // Obtain the correct value from the model matrix for the current predictor float X_value = X_matrix_d[X_thread_stride + (j * num_observations) + observation_row]; // Obtain the residual value (this is essentially the residual value leaving the current predictor out because the predictor coefficient value is zero) float residual_y_value = residual_y_d[observation_thread_stride + observation_row]; // Compute the correlation between the current predictor column and the residual values that are obtained leaving the current predictor out // The correlation is computed as a running sum p_j = p_j + (X_value * residual_y_value); } } // Divide the computed correlation by the total number of observations in y (also the total number of observations in one predictor column) p_j = (1.0f / (float)num_observations) * p_j; // Apply the soft-thresholding function that is associated with the L1-regularization component of elastic-net regularization float gamma = lambda * alpha; if (p_j > 0.0f && gamma < fabsf(p_j)) { B_j = p_j - gamma; } else if (p_j < 0.0f && gamma < fabsf(p_j)) { B_j = p_j + gamma; } else { B_j = 0.0f; } // Declare and initialize the mean of the square of the predictor column float mean_squared_predictor_value = 0.0f; // Obtain the mean of the square of the predictor column if (transformation_flag == 1 || transformation_flag == 3) { mean_squared_predictor_value = 1.0f; } else if (transformation_flag == 2 || transformation_flag == 4) { mean_squared_predictor_value = 1.0f / (float)num_observations; } // This if-else statemet accounts for the fact that regularization is not applied to the intercept term if one is included if (intercept_flag == 1 && j == 0) { // Use the computed correlation value as the updated predictor coefficient B_j = p_j; } else { // Calculate the updated predictor coefficient value by applying the component of elastic-net regularization that is associated with L2-regularization // The mean_squared_predictor_value term comes from the derivation of the coordinate descent update for a predictor coefficient B_j = B_j / (mean_squared_predictor_value + (lambda * (1.0f - alpha))); } // Store the updated predictor coefficient value into the B_d array B_d[predictor_thread_stride + j] = B_j; // Update the residual values to include the contribution of the current predictor using the updated predictor coefficient value // If the updated predictor coefficient value is 0, then its contribution to the residual values is zero if (B_j != 0.0f) { for (int observation_row = 0; observation_row < num_observations; observation_row++) { // Store the updated residual back into the residual_y_d array residual_y_d[observation_thread_stride + observation_row] = residual_y_d[observation_thread_stride + observation_row] - (X_matrix_d[X_thread_stride + (j * num_observations) + observation_row] * B_j); } } // Compute the weighted (observation weights are all 1 in this case) sum of squares of the changes in the fitted values (this is used for the tolerance convergence criterion) change = (previous_B_j - B_j) * (previous_B_j - B_j); if (transformation_flag == 2 || transformation_flag == 4) { if (intercept_flag == 1 && j > 0) { change = (1.0f / (float)num_observations) * change; } else if (intercept_flag == 0) { change = (1.0f / (float)num_observations) * change; } } if (change > max_change) { max_change = change; } } // Update the global_max_change variable global_max_change = max_change; // Update the iteration count variable iteration_count = iteration_count + 1; } // Account for the fact that the y in the model fit was divided by its standard deviation float std_y = y_std_d[fit_ind]; for (int j = 0; j < num_predictors; j++) { B_d[predictor_thread_stride + j] = B_d[predictor_thread_stride + j] * std_y; } } } } // Define the GPU kernel that performs least-squares regression with elastic-net regularization using the cyclic coordinate descent optimization algorithm in order to fit the model matrices to the data __global__ void model_fit_shared_memory(float * B_d, double * B_thread_stride_d, float * model_fit_flag_d, float * X_matrix_d, double * X_matrix_thread_stride_d, double * observation_thread_stride_d, float * residual_y_d, float * y_std_d, float * standardized_lambda_values_d, double * num_observations_d, double * num_predictors_d, float * alpha_values_d, float * tolerance_values_d, float * max_iterations_values_d, float * intercept_flag_d, int transformation_flag, int num_threads_per_block, int num_blocks, int num_threads_last_block) { // Define the shared memory array that stores the residual values of the model fits within one block (the amount of bytes is declared in the GPU kernel call) extern __shared__ float sdata[]; // Obtain the index of the block int block_ind = blockIdx.x; // Obtain the thread index within one block int block_thread_ind = threadIdx.x; // Calculate the fit index int fit_ind = (block_ind * num_threads_per_block) + block_thread_ind; // Determine how many threads are in the block (accounts for the fact that the last block may contain less active threads than the other blocks) int num_threads_per_block_2 = num_threads_per_block; if (block_ind == (num_blocks - 1)) { num_threads_per_block_2 = num_threads_last_block; } // This if statement makes sure that extra threads aren't doing data processing if the last block has less threads if (block_thread_ind < num_threads_per_block_2) { // Obtain the flag that determines whether to perform a model fit or not int model_fit_flag = (int)model_fit_flag_d[fit_ind]; // This if statement is to ensure that a model fit is performed only if the model fit flag is 1 if (model_fit_flag == 1) { // Obtain the thread stride that is used to obtain the correct set of predictors for the fit int predictor_thread_stride = (int)B_thread_stride_d[fit_ind]; // Obtain the thread stride that is used to obtain the correct model matrix for the fit int X_thread_stride = (int)X_matrix_thread_stride_d[fit_ind]; // Obtain the thread stride that is used to obtain the correct set of observations for the fit int observation_thread_stride = (int)observation_thread_stride_d[fit_ind]; // Obtain the alpha value for the fit float alpha = alpha_values_d[fit_ind]; // Obtain the standardized lambda value for the fit float lambda = standardized_lambda_values_d[fit_ind]; // Obtain the tolerance value for the fit float tolerance = tolerance_values_d[fit_ind]; // Obtain the max iterations value for the fit int max_iterations = (int)max_iterations_values_d[fit_ind]; // Obtain the number of observations for the fit int num_observations = (int)num_observations_d[fit_ind]; // Obtain the number of predictors for the fit int num_predictors = (int)num_predictors_d[fit_ind]; // Obtain the flag that determines whether the first predictor column is a column of ones for the intercept term or not int intercept_flag = (int)intercept_flag_d[fit_ind]; // Declare and initialize the variable that stores the maximum weighted (observation weights are all 1 in this case) sum of squares of the changes in the fitted values for one iteration of cyclic coordinate descent float global_max_change = 1E12; // Declare and initialize the variable that counts how many iterations of cyclic coordinate descent have been performed int iteration_count = 0; // Store the residual values for the fit into the shared memory array for (int observation_row = 0; observation_row < num_observations; observation_row++) { int store_ind = (observation_row * num_threads_per_block) + block_thread_ind; sdata[store_ind] = residual_y_d[observation_thread_stride + observation_row]; } // Perform cyclic coordinate descent until either the maximum number of iterations is reached or the maximum weighted (observation weights are all 1 in this case) sum of squares of the changes in the fitted values becomes less than the tolerance while (global_max_change >= tolerance && iteration_count < max_iterations) { // Declare and initialize the variable that stores the maximum weighted (observation weights are all 1 in this case) sum of squares of the changes in the fitted values for one iteration of cyclic coordinate descent float max_change = 0.0f; // Declare and initialize the variable that stores the weighted (observation weights are all 1 in this case) sum of squares of the changes in the fitted values that are due to the current predictor coefficient value being updated using cyclic coordinate descent float change = 0.0f; // Cycle through all of the predictors for one iteration of cyclic coordinate descent for (int j = 0; j < num_predictors; j++) { // Obtain the predictor coefficient value for the current predictor float B_j = B_d[predictor_thread_stride + j]; // Store the predictor coefficent value before it's updated float previous_B_j = B_j; // Declare and initialize the variable that stores the correlation between the current predictor column and the residual values that are obtained leaving the current predictor out float p_j = 0.0f; // Calculate the residual values leaving the current predictor out (the predictor coefficients are initialized to zero, so the residual values are going to initially be y) // This if-else statement accounts for the fact that the contribution of the current predictor only needs to be removed from the residual values if the predictor coefficient is not zero // This is due to the fact that if the predictor coefficient is already zero, then the predictor contribution to the residual is zero if (B_j != 0.0f) { for (int observation_row = 0; observation_row < num_observations; observation_row++) { // Obtain the correct value from the model matrix for the current predictor float X_value = X_matrix_d[X_thread_stride + (j * num_observations) + observation_row]; // Remove the contribution of the current predictor from the current residual value float residual_y_value = sdata[(observation_row * num_threads_per_block) + block_thread_ind] + (X_value * B_j); // Store the updated residual value back into the shared memory array sdata[(observation_row * num_threads_per_block) + block_thread_ind] = residual_y_value; // Compute the correlation between the current predictor column and the residual values that are obtained leaving the current predictor out // The correlation is computed as a running sum p_j = p_j + (X_value * residual_y_value); } } else { for (int observation_row = 0; observation_row < num_observations; observation_row++) { // Obtain the correct value from the model matrix for the current predictor float X_value = X_matrix_d[X_thread_stride + (j * num_observations) + observation_row]; // Obtain the residual value (this is essentially the residual value leaving the current predictor out because the predictor coefficient value is zero) float residual_y_value = sdata[(observation_row * num_threads_per_block) + block_thread_ind]; // Compute the correlation between the current predictor column and the residual values that are obtained leaving the current predictor out // The correlation is computed as a running sum p_j = p_j + (X_value * residual_y_value); } } // Divide the computed correlation by the total number of observations in y (also the total number of observations in one predictor column) p_j = (1.0f / (float)num_observations) * p_j; // Apply the soft-thresholding function that is associated with the L1-regularization component of elastic-net regularization float gamma = lambda * alpha; if (p_j > 0.0f && gamma < fabsf(p_j)) { B_j = p_j - gamma; } else if (p_j < 0.0f && gamma < fabsf(p_j)) { B_j = p_j + gamma; } else { B_j = 0.0f; } // Declare and initialize the mean of the square of the predictor column float mean_squared_predictor_value = 0.0f; // Obtain the mean of the square of the predictor column if (transformation_flag == 1 || transformation_flag == 3) { mean_squared_predictor_value = 1.0f; } else if (transformation_flag == 2 || transformation_flag == 4) { mean_squared_predictor_value = 1.0f / (float)num_observations; } // This if-else statemet accounts for the fact that regularization is not applied to the intercept term if one is included if (intercept_flag == 1 && j == 0) { // Use the computed correlation value as the updated predictor coefficient B_j = p_j; } else { // Calculate the updated predictor coefficient value by applying the component of elastic-net regularization that is associated with L2-regularization // The mean_squared_predictor_value term comes from the derivation of the coordinate descent update for a predictor coefficient B_j = B_j / (mean_squared_predictor_value + (lambda * (1.0f - alpha))); } // Store the updated predictor coefficient value into the B_d array B_d[predictor_thread_stride + j] = B_j; // Update the residual values to include the contribution of the current predictor using the updated predictor coefficient value // If the updated predictor coefficient value is 0, then its contribution to the residual values is zero if (B_j != 0.0f) { for (int observation_row = 0; observation_row < num_observations; observation_row++) { // Store the updated residual back into the shared memory array sdata[(observation_row * num_threads_per_block) + block_thread_ind] = sdata[(observation_row * num_threads_per_block) + block_thread_ind] - (X_matrix_d[X_thread_stride + (j * num_observations) + observation_row] * B_j); } } // Compute the weighted (observation weights are all 1 in this case) sum of squares of the changes in the fitted values (this is used for the tolerance convergence criterion) change = (previous_B_j - B_j) * (previous_B_j - B_j); if (transformation_flag == 2 || transformation_flag == 4) { if (intercept_flag == 1 && j > 0) { change = (1.0f / (float)num_observations) * change; } else if (intercept_flag == 0) { change = (1.0f / (float)num_observations) * change; } } if (change > max_change) { max_change = change; } } // Update the global_max_change variable global_max_change = max_change; // Update the iteration count variable iteration_count = iteration_count + 1; } // Account for the fact that the y in the model fit was divided by its standard deviation float std_y = y_std_d[fit_ind]; for (int j = 0; j < num_predictors; j++) { B_d[predictor_thread_stride + j] = B_d[predictor_thread_stride + j] * std_y; } } } } // Define the GPU kernel that performs predictor coefficient unnormalization __global__ void predictor_coefficient_unnormalization(float * B_d, double * B_thread_stride_d, float * model_fit_flag_d, float * X_matrix_d, double * X_matrix_thread_stride_d, float * scaling_factors_d, double * num_predictors_d, float * intercept_flag_d, int num_threads_per_block, int num_blocks, int num_threads_last_block) { // Obtain the index of the block int block_ind = blockIdx.x; // Obtain the thread index within one block int block_thread_ind = threadIdx.x; // Calculate the fit index int fit_ind = (block_ind * num_threads_per_block) + block_thread_ind; // Determine how many threads are in the block (accounts for the fact that the last block may contain less active threads than the other blocks) int num_threads_per_block_2 = num_threads_per_block; if (block_ind == (num_blocks - 1)) { num_threads_per_block_2 = num_threads_last_block; } // This if statement makes sure that extra threads aren't doing data processing if the last block has less fits to perform if (block_thread_ind < num_threads_per_block_2) { // Obtain the flag that determines whether a model fit was performed or not int model_fit_flag = (int)model_fit_flag_d[fit_ind]; // This if statement is to ensure that the coefficients are unnormalized only if a model fit was performed if (model_fit_flag == 1) { // Obtain the thread stride that is used to obtain the correct set of predictors for the fit int predictor_thread_stride = (int)B_thread_stride_d[fit_ind]; // Obtain the number of predictors for the fit int num_predictors = (int)num_predictors_d[fit_ind]; // Obtain the flag that determines whether the first predictor column is a column of ones for the intercept term or not int intercept_flag = (int)intercept_flag_d[fit_ind]; // Declare and initialize the variable that stores the number of the first predictor column to be standardized int start_ind = 0; // This if statement makes sure to not standardize the first predictor column if it corresponds to the intercept term if (intercept_flag == 1) { start_ind = 1; } // Unnormalize the predictor coefficients for (int predictor_column = start_ind; predictor_column < num_predictors; predictor_column++) { B_d[predictor_thread_stride + predictor_column] = B_d[predictor_thread_stride + predictor_column] / scaling_factors_d[predictor_thread_stride + predictor_column]; } } } } // Define the GPU kernel that performs predictor coefficient unstandardization __global__ void predictor_coefficient_unstandardization(float * B_d, double * B_thread_stride_d, float * model_fit_flag_d, float * X_matrix_d, double * X_matrix_thread_stride_d, float * scaling_factors_d, float * mean_X_matrix_d, double * num_predictors_d, float * intercept_flag_d, int num_threads_per_block, int num_blocks, int num_threads_last_block) { // Obtain the index of the block int block_ind = blockIdx.x; // Obtain the thread index within one block int block_thread_ind = threadIdx.x; // Calculate the fit index int fit_ind = (block_ind * num_threads_per_block) + block_thread_ind; // Determine how many threads are in the block (accounts for the fact that the last block may contain less active threads than the other blocks) int num_threads_per_block_2 = num_threads_per_block; if (block_ind == (num_blocks - 1)) { num_threads_per_block_2 = num_threads_last_block; } // This if statement makes sure that extra threads aren't doing data processing if the last block has less fits to perform if (block_thread_ind < num_threads_per_block_2) { // Obtain the flag that determines whether a model fit was performed or not int model_fit_flag = (int)model_fit_flag_d[fit_ind]; // This if statement is to ensure that the coefficients are unstandardized only if a model fit was performed if (model_fit_flag == 1) { // Obtain the thread stride that is used to obtain the correct set of predictors for the fit int predictor_thread_stride = (int)B_thread_stride_d[fit_ind]; // Obtain the number of predictors for the fit int num_predictors = (int)num_predictors_d[fit_ind]; // Obtain the flag that determines whether the first predictor column is a column of ones for the intercept term or not int intercept_flag = (int)intercept_flag_d[fit_ind]; // Declare and initialize the variable that stores the number of the first predictor column to be standardized int start_ind = 0; // This if statement makes sure to not standardize the first predictor column if it corresponds to the intercept term if (intercept_flag == 1) { start_ind = 1; } // Declare and initialize the variable that is used to adjust the intercept term if it is included float sum_value = 0.0f; // Perform predictor coefficient unstandardization for (int predictor_column = start_ind; predictor_column < num_predictors; predictor_column++) { float B_unstandardized = B_d[predictor_thread_stride + predictor_column] / scaling_factors_d[predictor_thread_stride + predictor_column]; B_d[predictor_thread_stride + predictor_column] = B_unstandardized; sum_value = sum_value + (B_unstandardized * mean_X_matrix_d[predictor_thread_stride + predictor_column]); } // Adjust the intercept term if it is included if (intercept_flag == 1) { B_d[predictor_thread_stride] = B_d[predictor_thread_stride] - sum_value; } } } }
.file "tmpxft_00190987_00000000-6_GPU_kernels_single_precision.cudafe1.cpp" .text #APP #NO_APP .type _ZL26__cudaUnregisterBinaryUtilv, @function _ZL26__cudaUnregisterBinaryUtilv: .LFB2029: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaUnregisterFatBinary@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2029: .size _ZL26__cudaUnregisterBinaryUtilv, .-_ZL26__cudaUnregisterBinaryUtilv .globl _Z61__device_stub__Z23predictor_normalizationPfS_PdS0_S0_S0_S_iiiPfS_PdS0_S0_S0_S_iii .type _Z61__device_stub__Z23predictor_normalizationPfS_PdS0_S0_S0_S_iiiPfS_PdS0_S0_S0_S_iii, @function _Z61__device_stub__Z23predictor_normalizationPfS_PdS0_S0_S0_S_iiiPfS_PdS0_S0_S0_S_iii: .LFB2051: .cfi_startproc endbr64 subq $232, %rsp .cfi_def_cfa_offset 240 movq %rdi, 56(%rsp) movq %rsi, 48(%rsp) movq %rdx, 40(%rsp) movq %rcx, 32(%rsp) movq %r8, 24(%rsp) movq %r9, 16(%rsp) movq 240(%rsp), %rax movq %rax, 8(%rsp) movq %fs:40, %rax movq %rax, 216(%rsp) xorl %eax, %eax leaq 56(%rsp), %rax movq %rax, 128(%rsp) leaq 48(%rsp), %rax movq %rax, 136(%rsp) leaq 40(%rsp), %rax movq %rax, 144(%rsp) leaq 32(%rsp), %rax movq %rax, 152(%rsp) leaq 24(%rsp), %rax movq %rax, 160(%rsp) leaq 16(%rsp), %rax movq %rax, 168(%rsp) leaq 8(%rsp), %rax movq %rax, 176(%rsp) leaq 248(%rsp), %rax movq %rax, 184(%rsp) leaq 256(%rsp), %rax movq %rax, 192(%rsp) leaq 264(%rsp), %rax movq %rax, 200(%rsp) movl $1, 80(%rsp) movl $1, 84(%rsp) movl $1, 88(%rsp) movl $1, 92(%rsp) movl $1, 96(%rsp) movl $1, 100(%rsp) leaq 72(%rsp), %rcx leaq 64(%rsp), %rdx leaq 92(%rsp), %rsi leaq 80(%rsp), %rdi call __cudaPopCallConfiguration@PLT testl %eax, %eax je .L7 .L3: movq 216(%rsp), %rax subq %fs:40, %rax jne .L8 addq $232, %rsp .cfi_remember_state .cfi_def_cfa_offset 8 ret .L7: .cfi_restore_state pushq 72(%rsp) .cfi_def_cfa_offset 248 pushq 72(%rsp) .cfi_def_cfa_offset 256 leaq 144(%rsp), %r9 movq 108(%rsp), %rcx movl 116(%rsp), %r8d movq 96(%rsp), %rsi movl 104(%rsp), %edx leaq _Z23predictor_normalizationPfS_PdS0_S0_S0_S_iii(%rip), %rdi call cudaLaunchKernel@PLT addq $16, %rsp .cfi_def_cfa_offset 240 jmp .L3 .L8: call __stack_chk_fail@PLT .cfi_endproc .LFE2051: .size _Z61__device_stub__Z23predictor_normalizationPfS_PdS0_S0_S0_S_iiiPfS_PdS0_S0_S0_S_iii, .-_Z61__device_stub__Z23predictor_normalizationPfS_PdS0_S0_S0_S_iiiPfS_PdS0_S0_S0_S_iii .globl _Z23predictor_normalizationPfS_PdS0_S0_S0_S_iii .type _Z23predictor_normalizationPfS_PdS0_S0_S0_S_iii, @function _Z23predictor_normalizationPfS_PdS0_S0_S0_S_iii: .LFB2052: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 movl 40(%rsp), %eax pushq %rax .cfi_def_cfa_offset 24 movl 40(%rsp), %eax pushq %rax .cfi_def_cfa_offset 32 movl 40(%rsp), %eax pushq %rax .cfi_def_cfa_offset 40 pushq 40(%rsp) .cfi_def_cfa_offset 48 call _Z61__device_stub__Z23predictor_normalizationPfS_PdS0_S0_S0_S_iiiPfS_PdS0_S0_S0_S_iii addq $40, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2052: .size _Z23predictor_normalizationPfS_PdS0_S0_S0_S_iii, .-_Z23predictor_normalizationPfS_PdS0_S0_S0_S_iii .globl _Z65__device_stub__Z25predictor_standardizationPfS_S_PdS0_S0_S0_S_iiiPfS_S_PdS0_S0_S0_S_iii .type _Z65__device_stub__Z25predictor_standardizationPfS_S_PdS0_S0_S0_S_iiiPfS_S_PdS0_S0_S0_S_iii, @function _Z65__device_stub__Z25predictor_standardizationPfS_S_PdS0_S0_S0_S_iiiPfS_S_PdS0_S0_S0_S_iii: .LFB2053: .cfi_startproc endbr64 subq $232, %rsp .cfi_def_cfa_offset 240 movq %rdi, 56(%rsp) movq %rsi, 48(%rsp) movq %rdx, 40(%rsp) movq %rcx, 32(%rsp) movq %r8, 24(%rsp) movq %r9, 16(%rsp) movq 240(%rsp), %rax movq %rax, 8(%rsp) movq 248(%rsp), %rax movq %rax, (%rsp) movq %fs:40, %rax movq %rax, 216(%rsp) xorl %eax, %eax leaq 56(%rsp), %rax movq %rax, 128(%rsp) leaq 48(%rsp), %rax movq %rax, 136(%rsp) leaq 40(%rsp), %rax movq %rax, 144(%rsp) leaq 32(%rsp), %rax movq %rax, 152(%rsp) leaq 24(%rsp), %rax movq %rax, 160(%rsp) leaq 16(%rsp), %rax movq %rax, 168(%rsp) leaq 8(%rsp), %rax movq %rax, 176(%rsp) movq %rsp, %rax movq %rax, 184(%rsp) leaq 256(%rsp), %rax movq %rax, 192(%rsp) leaq 264(%rsp), %rax movq %rax, 200(%rsp) leaq 272(%rsp), %rax movq %rax, 208(%rsp) movl $1, 80(%rsp) movl $1, 84(%rsp) movl $1, 88(%rsp) movl $1, 92(%rsp) movl $1, 96(%rsp) movl $1, 100(%rsp) leaq 72(%rsp), %rcx leaq 64(%rsp), %rdx leaq 92(%rsp), %rsi leaq 80(%rsp), %rdi call __cudaPopCallConfiguration@PLT testl %eax, %eax je .L15 .L11: movq 216(%rsp), %rax subq %fs:40, %rax jne .L16 addq $232, %rsp .cfi_remember_state .cfi_def_cfa_offset 8 ret .L15: .cfi_restore_state pushq 72(%rsp) .cfi_def_cfa_offset 248 pushq 72(%rsp) .cfi_def_cfa_offset 256 leaq 144(%rsp), %r9 movq 108(%rsp), %rcx movl 116(%rsp), %r8d movq 96(%rsp), %rsi movl 104(%rsp), %edx leaq _Z25predictor_standardizationPfS_S_PdS0_S0_S0_S_iii(%rip), %rdi call cudaLaunchKernel@PLT addq $16, %rsp .cfi_def_cfa_offset 240 jmp .L11 .L16: call __stack_chk_fail@PLT .cfi_endproc .LFE2053: .size _Z65__device_stub__Z25predictor_standardizationPfS_S_PdS0_S0_S0_S_iiiPfS_S_PdS0_S0_S0_S_iii, .-_Z65__device_stub__Z25predictor_standardizationPfS_S_PdS0_S0_S0_S_iiiPfS_S_PdS0_S0_S0_S_iii .globl _Z25predictor_standardizationPfS_S_PdS0_S0_S0_S_iii .type _Z25predictor_standardizationPfS_S_PdS0_S0_S0_S_iii, @function _Z25predictor_standardizationPfS_S_PdS0_S0_S0_S_iii: .LFB2054: .cfi_startproc endbr64 subq $16, %rsp .cfi_def_cfa_offset 24 movl 56(%rsp), %eax pushq %rax .cfi_def_cfa_offset 32 movl 56(%rsp), %eax pushq %rax .cfi_def_cfa_offset 40 movl 56(%rsp), %eax pushq %rax .cfi_def_cfa_offset 48 pushq 56(%rsp) .cfi_def_cfa_offset 56 pushq 56(%rsp) .cfi_def_cfa_offset 64 call _Z65__device_stub__Z25predictor_standardizationPfS_S_PdS0_S0_S0_S_iiiPfS_S_PdS0_S0_S0_S_iii addq $56, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2054: .size _Z25predictor_standardizationPfS_S_PdS0_S0_S0_S_iii, .-_Z25predictor_standardizationPfS_S_PdS0_S0_S0_S_iii .globl _Z57__device_stub__Z21model_fit_preparationPfS_S_S_S_PdS0_iiiPfS_S_S_S_PdS0_iii .type _Z57__device_stub__Z21model_fit_preparationPfS_S_S_S_PdS0_iiiPfS_S_S_S_PdS0_iii, @function _Z57__device_stub__Z21model_fit_preparationPfS_S_S_S_PdS0_iiiPfS_S_S_S_PdS0_iii: .LFB2055: .cfi_startproc endbr64 subq $232, %rsp .cfi_def_cfa_offset 240 movq %rdi, 56(%rsp) movq %rsi, 48(%rsp) movq %rdx, 40(%rsp) movq %rcx, 32(%rsp) movq %r8, 24(%rsp) movq %r9, 16(%rsp) movq 240(%rsp), %rax movq %rax, 8(%rsp) movq %fs:40, %rax movq %rax, 216(%rsp) xorl %eax, %eax leaq 56(%rsp), %rax movq %rax, 128(%rsp) leaq 48(%rsp), %rax movq %rax, 136(%rsp) leaq 40(%rsp), %rax movq %rax, 144(%rsp) leaq 32(%rsp), %rax movq %rax, 152(%rsp) leaq 24(%rsp), %rax movq %rax, 160(%rsp) leaq 16(%rsp), %rax movq %rax, 168(%rsp) leaq 8(%rsp), %rax movq %rax, 176(%rsp) leaq 248(%rsp), %rax movq %rax, 184(%rsp) leaq 256(%rsp), %rax movq %rax, 192(%rsp) leaq 264(%rsp), %rax movq %rax, 200(%rsp) movl $1, 80(%rsp) movl $1, 84(%rsp) movl $1, 88(%rsp) movl $1, 92(%rsp) movl $1, 96(%rsp) movl $1, 100(%rsp) leaq 72(%rsp), %rcx leaq 64(%rsp), %rdx leaq 92(%rsp), %rsi leaq 80(%rsp), %rdi call __cudaPopCallConfiguration@PLT testl %eax, %eax je .L23 .L19: movq 216(%rsp), %rax subq %fs:40, %rax jne .L24 addq $232, %rsp .cfi_remember_state .cfi_def_cfa_offset 8 ret .L23: .cfi_restore_state pushq 72(%rsp) .cfi_def_cfa_offset 248 pushq 72(%rsp) .cfi_def_cfa_offset 256 leaq 144(%rsp), %r9 movq 108(%rsp), %rcx movl 116(%rsp), %r8d movq 96(%rsp), %rsi movl 104(%rsp), %edx leaq _Z21model_fit_preparationPfS_S_S_S_PdS0_iii(%rip), %rdi call cudaLaunchKernel@PLT addq $16, %rsp .cfi_def_cfa_offset 240 jmp .L19 .L24: call __stack_chk_fail@PLT .cfi_endproc .LFE2055: .size _Z57__device_stub__Z21model_fit_preparationPfS_S_S_S_PdS0_iiiPfS_S_S_S_PdS0_iii, .-_Z57__device_stub__Z21model_fit_preparationPfS_S_S_S_PdS0_iiiPfS_S_S_S_PdS0_iii .globl _Z21model_fit_preparationPfS_S_S_S_PdS0_iii .type _Z21model_fit_preparationPfS_S_S_S_PdS0_iii, @function _Z21model_fit_preparationPfS_S_S_S_PdS0_iii: .LFB2056: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 movl 40(%rsp), %eax pushq %rax .cfi_def_cfa_offset 24 movl 40(%rsp), %eax pushq %rax .cfi_def_cfa_offset 32 movl 40(%rsp), %eax pushq %rax .cfi_def_cfa_offset 40 pushq 40(%rsp) .cfi_def_cfa_offset 48 call _Z57__device_stub__Z21model_fit_preparationPfS_S_S_S_PdS0_iiiPfS_S_S_S_PdS0_iii addq $40, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2056: .size _Z21model_fit_preparationPfS_S_S_S_PdS0_iii, .-_Z21model_fit_preparationPfS_S_S_S_PdS0_iii .globl _Z64__device_stub__Z9model_fitPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiiiPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii .type _Z64__device_stub__Z9model_fitPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiiiPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii, @function _Z64__device_stub__Z9model_fitPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiiiPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii: .LFB2057: .cfi_startproc endbr64 subq $360, %rsp .cfi_def_cfa_offset 368 movq %rdi, 120(%rsp) movq %rsi, 112(%rsp) movq %rdx, 104(%rsp) movq %rcx, 96(%rsp) movq %r8, 88(%rsp) movq %r9, 80(%rsp) movq 368(%rsp), %rax movq %rax, 72(%rsp) movq 376(%rsp), %rax movq %rax, 64(%rsp) movq 384(%rsp), %rax movq %rax, 56(%rsp) movq 392(%rsp), %rax movq %rax, 48(%rsp) movq 400(%rsp), %rax movq %rax, 40(%rsp) movq 408(%rsp), %rax movq %rax, 32(%rsp) movq 416(%rsp), %rax movq %rax, 24(%rsp) movq 424(%rsp), %rax movq %rax, 16(%rsp) movq 432(%rsp), %rax movq %rax, 8(%rsp) movq %fs:40, %rax movq %rax, 344(%rsp) xorl %eax, %eax leaq 120(%rsp), %rax movq %rax, 192(%rsp) leaq 112(%rsp), %rax movq %rax, 200(%rsp) leaq 104(%rsp), %rax movq %rax, 208(%rsp) leaq 96(%rsp), %rax movq %rax, 216(%rsp) leaq 88(%rsp), %rax movq %rax, 224(%rsp) leaq 80(%rsp), %rax movq %rax, 232(%rsp) leaq 72(%rsp), %rax movq %rax, 240(%rsp) leaq 64(%rsp), %rax movq %rax, 248(%rsp) leaq 56(%rsp), %rax movq %rax, 256(%rsp) leaq 48(%rsp), %rax movq %rax, 264(%rsp) leaq 40(%rsp), %rax movq %rax, 272(%rsp) leaq 32(%rsp), %rax movq %rax, 280(%rsp) leaq 24(%rsp), %rax movq %rax, 288(%rsp) leaq 16(%rsp), %rax movq %rax, 296(%rsp) leaq 8(%rsp), %rax movq %rax, 304(%rsp) leaq 440(%rsp), %rax movq %rax, 312(%rsp) leaq 448(%rsp), %rax movq %rax, 320(%rsp) leaq 456(%rsp), %rax movq %rax, 328(%rsp) leaq 464(%rsp), %rax movq %rax, 336(%rsp) movl $1, 144(%rsp) movl $1, 148(%rsp) movl $1, 152(%rsp) movl $1, 156(%rsp) movl $1, 160(%rsp) movl $1, 164(%rsp) leaq 136(%rsp), %rcx leaq 128(%rsp), %rdx leaq 156(%rsp), %rsi leaq 144(%rsp), %rdi call __cudaPopCallConfiguration@PLT testl %eax, %eax je .L31 .L27: movq 344(%rsp), %rax subq %fs:40, %rax jne .L32 addq $360, %rsp .cfi_remember_state .cfi_def_cfa_offset 8 ret .L31: .cfi_restore_state pushq 136(%rsp) .cfi_def_cfa_offset 376 pushq 136(%rsp) .cfi_def_cfa_offset 384 leaq 208(%rsp), %r9 movq 172(%rsp), %rcx movl 180(%rsp), %r8d movq 160(%rsp), %rsi movl 168(%rsp), %edx leaq _Z9model_fitPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii(%rip), %rdi call cudaLaunchKernel@PLT addq $16, %rsp .cfi_def_cfa_offset 368 jmp .L27 .L32: call __stack_chk_fail@PLT .cfi_endproc .LFE2057: .size _Z64__device_stub__Z9model_fitPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiiiPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii, .-_Z64__device_stub__Z9model_fitPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiiiPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii .globl _Z9model_fitPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii .type _Z9model_fitPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii, @function _Z9model_fitPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii: .LFB2058: .cfi_startproc endbr64 subq $16, %rsp .cfi_def_cfa_offset 24 movl 120(%rsp), %eax pushq %rax .cfi_def_cfa_offset 32 movl 120(%rsp), %eax pushq %rax .cfi_def_cfa_offset 40 movl 120(%rsp), %eax pushq %rax .cfi_def_cfa_offset 48 movl 120(%rsp), %eax pushq %rax .cfi_def_cfa_offset 56 pushq 120(%rsp) .cfi_def_cfa_offset 64 pushq 120(%rsp) .cfi_def_cfa_offset 72 pushq 120(%rsp) .cfi_def_cfa_offset 80 pushq 120(%rsp) .cfi_def_cfa_offset 88 pushq 120(%rsp) .cfi_def_cfa_offset 96 pushq 120(%rsp) .cfi_def_cfa_offset 104 pushq 120(%rsp) .cfi_def_cfa_offset 112 pushq 120(%rsp) .cfi_def_cfa_offset 120 pushq 120(%rsp) .cfi_def_cfa_offset 128 call _Z64__device_stub__Z9model_fitPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiiiPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii addq $120, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2058: .size _Z9model_fitPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii, .-_Z9model_fitPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii .globl _Z79__device_stub__Z23model_fit_shared_memoryPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiiiPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii .type _Z79__device_stub__Z23model_fit_shared_memoryPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiiiPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii, @function _Z79__device_stub__Z23model_fit_shared_memoryPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiiiPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii: .LFB2059: .cfi_startproc endbr64 subq $360, %rsp .cfi_def_cfa_offset 368 movq %rdi, 120(%rsp) movq %rsi, 112(%rsp) movq %rdx, 104(%rsp) movq %rcx, 96(%rsp) movq %r8, 88(%rsp) movq %r9, 80(%rsp) movq 368(%rsp), %rax movq %rax, 72(%rsp) movq 376(%rsp), %rax movq %rax, 64(%rsp) movq 384(%rsp), %rax movq %rax, 56(%rsp) movq 392(%rsp), %rax movq %rax, 48(%rsp) movq 400(%rsp), %rax movq %rax, 40(%rsp) movq 408(%rsp), %rax movq %rax, 32(%rsp) movq 416(%rsp), %rax movq %rax, 24(%rsp) movq 424(%rsp), %rax movq %rax, 16(%rsp) movq 432(%rsp), %rax movq %rax, 8(%rsp) movq %fs:40, %rax movq %rax, 344(%rsp) xorl %eax, %eax leaq 120(%rsp), %rax movq %rax, 192(%rsp) leaq 112(%rsp), %rax movq %rax, 200(%rsp) leaq 104(%rsp), %rax movq %rax, 208(%rsp) leaq 96(%rsp), %rax movq %rax, 216(%rsp) leaq 88(%rsp), %rax movq %rax, 224(%rsp) leaq 80(%rsp), %rax movq %rax, 232(%rsp) leaq 72(%rsp), %rax movq %rax, 240(%rsp) leaq 64(%rsp), %rax movq %rax, 248(%rsp) leaq 56(%rsp), %rax movq %rax, 256(%rsp) leaq 48(%rsp), %rax movq %rax, 264(%rsp) leaq 40(%rsp), %rax movq %rax, 272(%rsp) leaq 32(%rsp), %rax movq %rax, 280(%rsp) leaq 24(%rsp), %rax movq %rax, 288(%rsp) leaq 16(%rsp), %rax movq %rax, 296(%rsp) leaq 8(%rsp), %rax movq %rax, 304(%rsp) leaq 440(%rsp), %rax movq %rax, 312(%rsp) leaq 448(%rsp), %rax movq %rax, 320(%rsp) leaq 456(%rsp), %rax movq %rax, 328(%rsp) leaq 464(%rsp), %rax movq %rax, 336(%rsp) movl $1, 144(%rsp) movl $1, 148(%rsp) movl $1, 152(%rsp) movl $1, 156(%rsp) movl $1, 160(%rsp) movl $1, 164(%rsp) leaq 136(%rsp), %rcx leaq 128(%rsp), %rdx leaq 156(%rsp), %rsi leaq 144(%rsp), %rdi call __cudaPopCallConfiguration@PLT testl %eax, %eax je .L39 .L35: movq 344(%rsp), %rax subq %fs:40, %rax jne .L40 addq $360, %rsp .cfi_remember_state .cfi_def_cfa_offset 8 ret .L39: .cfi_restore_state pushq 136(%rsp) .cfi_def_cfa_offset 376 pushq 136(%rsp) .cfi_def_cfa_offset 384 leaq 208(%rsp), %r9 movq 172(%rsp), %rcx movl 180(%rsp), %r8d movq 160(%rsp), %rsi movl 168(%rsp), %edx leaq _Z23model_fit_shared_memoryPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii(%rip), %rdi call cudaLaunchKernel@PLT addq $16, %rsp .cfi_def_cfa_offset 368 jmp .L35 .L40: call __stack_chk_fail@PLT .cfi_endproc .LFE2059: .size _Z79__device_stub__Z23model_fit_shared_memoryPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiiiPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii, .-_Z79__device_stub__Z23model_fit_shared_memoryPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiiiPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii .globl _Z23model_fit_shared_memoryPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii .type _Z23model_fit_shared_memoryPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii, @function _Z23model_fit_shared_memoryPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii: .LFB2060: .cfi_startproc endbr64 subq $16, %rsp .cfi_def_cfa_offset 24 movl 120(%rsp), %eax pushq %rax .cfi_def_cfa_offset 32 movl 120(%rsp), %eax pushq %rax .cfi_def_cfa_offset 40 movl 120(%rsp), %eax pushq %rax .cfi_def_cfa_offset 48 movl 120(%rsp), %eax pushq %rax .cfi_def_cfa_offset 56 pushq 120(%rsp) .cfi_def_cfa_offset 64 pushq 120(%rsp) .cfi_def_cfa_offset 72 pushq 120(%rsp) .cfi_def_cfa_offset 80 pushq 120(%rsp) .cfi_def_cfa_offset 88 pushq 120(%rsp) .cfi_def_cfa_offset 96 pushq 120(%rsp) .cfi_def_cfa_offset 104 pushq 120(%rsp) .cfi_def_cfa_offset 112 pushq 120(%rsp) .cfi_def_cfa_offset 120 pushq 120(%rsp) .cfi_def_cfa_offset 128 call _Z79__device_stub__Z23model_fit_shared_memoryPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiiiPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii addq $120, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2060: .size _Z23model_fit_shared_memoryPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii, .-_Z23model_fit_shared_memoryPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii .globl _Z76__device_stub__Z37predictor_coefficient_unnormalizationPfPdS_S_S0_S_S0_S_iiiPfPdS_S_S0_S_S0_S_iii .type _Z76__device_stub__Z37predictor_coefficient_unnormalizationPfPdS_S_S0_S_S0_S_iiiPfPdS_S_S0_S_S0_S_iii, @function _Z76__device_stub__Z37predictor_coefficient_unnormalizationPfPdS_S_S0_S_S0_S_iiiPfPdS_S_S0_S_S0_S_iii: .LFB2061: .cfi_startproc endbr64 subq $232, %rsp .cfi_def_cfa_offset 240 movq %rdi, 56(%rsp) movq %rsi, 48(%rsp) movq %rdx, 40(%rsp) movq %rcx, 32(%rsp) movq %r8, 24(%rsp) movq %r9, 16(%rsp) movq 240(%rsp), %rax movq %rax, 8(%rsp) movq 248(%rsp), %rax movq %rax, (%rsp) movq %fs:40, %rax movq %rax, 216(%rsp) xorl %eax, %eax leaq 56(%rsp), %rax movq %rax, 128(%rsp) leaq 48(%rsp), %rax movq %rax, 136(%rsp) leaq 40(%rsp), %rax movq %rax, 144(%rsp) leaq 32(%rsp), %rax movq %rax, 152(%rsp) leaq 24(%rsp), %rax movq %rax, 160(%rsp) leaq 16(%rsp), %rax movq %rax, 168(%rsp) leaq 8(%rsp), %rax movq %rax, 176(%rsp) movq %rsp, %rax movq %rax, 184(%rsp) leaq 256(%rsp), %rax movq %rax, 192(%rsp) leaq 264(%rsp), %rax movq %rax, 200(%rsp) leaq 272(%rsp), %rax movq %rax, 208(%rsp) movl $1, 80(%rsp) movl $1, 84(%rsp) movl $1, 88(%rsp) movl $1, 92(%rsp) movl $1, 96(%rsp) movl $1, 100(%rsp) leaq 72(%rsp), %rcx leaq 64(%rsp), %rdx leaq 92(%rsp), %rsi leaq 80(%rsp), %rdi call __cudaPopCallConfiguration@PLT testl %eax, %eax je .L47 .L43: movq 216(%rsp), %rax subq %fs:40, %rax jne .L48 addq $232, %rsp .cfi_remember_state .cfi_def_cfa_offset 8 ret .L47: .cfi_restore_state pushq 72(%rsp) .cfi_def_cfa_offset 248 pushq 72(%rsp) .cfi_def_cfa_offset 256 leaq 144(%rsp), %r9 movq 108(%rsp), %rcx movl 116(%rsp), %r8d movq 96(%rsp), %rsi movl 104(%rsp), %edx leaq _Z37predictor_coefficient_unnormalizationPfPdS_S_S0_S_S0_S_iii(%rip), %rdi call cudaLaunchKernel@PLT addq $16, %rsp .cfi_def_cfa_offset 240 jmp .L43 .L48: call __stack_chk_fail@PLT .cfi_endproc .LFE2061: .size _Z76__device_stub__Z37predictor_coefficient_unnormalizationPfPdS_S_S0_S_S0_S_iiiPfPdS_S_S0_S_S0_S_iii, .-_Z76__device_stub__Z37predictor_coefficient_unnormalizationPfPdS_S_S0_S_S0_S_iiiPfPdS_S_S0_S_S0_S_iii .globl _Z37predictor_coefficient_unnormalizationPfPdS_S_S0_S_S0_S_iii .type _Z37predictor_coefficient_unnormalizationPfPdS_S_S0_S_S0_S_iii, @function _Z37predictor_coefficient_unnormalizationPfPdS_S_S0_S_S0_S_iii: .LFB2062: .cfi_startproc endbr64 subq $16, %rsp .cfi_def_cfa_offset 24 movl 56(%rsp), %eax pushq %rax .cfi_def_cfa_offset 32 movl 56(%rsp), %eax pushq %rax .cfi_def_cfa_offset 40 movl 56(%rsp), %eax pushq %rax .cfi_def_cfa_offset 48 pushq 56(%rsp) .cfi_def_cfa_offset 56 pushq 56(%rsp) .cfi_def_cfa_offset 64 call _Z76__device_stub__Z37predictor_coefficient_unnormalizationPfPdS_S_S0_S_S0_S_iiiPfPdS_S_S0_S_S0_S_iii addq $56, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2062: .size _Z37predictor_coefficient_unnormalizationPfPdS_S_S0_S_S0_S_iii, .-_Z37predictor_coefficient_unnormalizationPfPdS_S_S0_S_S0_S_iii .globl _Z80__device_stub__Z39predictor_coefficient_unstandardizationPfPdS_S_S0_S_S_S0_S_iiiPfPdS_S_S0_S_S_S0_S_iii .type _Z80__device_stub__Z39predictor_coefficient_unstandardizationPfPdS_S_S0_S_S_S0_S_iiiPfPdS_S_S0_S_S_S0_S_iii, @function _Z80__device_stub__Z39predictor_coefficient_unstandardizationPfPdS_S_S0_S_S_S0_S_iiiPfPdS_S_S0_S_S_S0_S_iii: .LFB2063: .cfi_startproc endbr64 subq $264, %rsp .cfi_def_cfa_offset 272 movq %rdi, 72(%rsp) movq %rsi, 64(%rsp) movq %rdx, 56(%rsp) movq %rcx, 48(%rsp) movq %r8, 40(%rsp) movq %r9, 32(%rsp) movq 272(%rsp), %rax movq %rax, 24(%rsp) movq 280(%rsp), %rax movq %rax, 16(%rsp) movq 288(%rsp), %rax movq %rax, 8(%rsp) movq %fs:40, %rax movq %rax, 248(%rsp) xorl %eax, %eax leaq 72(%rsp), %rax movq %rax, 144(%rsp) leaq 64(%rsp), %rax movq %rax, 152(%rsp) leaq 56(%rsp), %rax movq %rax, 160(%rsp) leaq 48(%rsp), %rax movq %rax, 168(%rsp) leaq 40(%rsp), %rax movq %rax, 176(%rsp) leaq 32(%rsp), %rax movq %rax, 184(%rsp) leaq 24(%rsp), %rax movq %rax, 192(%rsp) leaq 16(%rsp), %rax movq %rax, 200(%rsp) leaq 8(%rsp), %rax movq %rax, 208(%rsp) leaq 296(%rsp), %rax movq %rax, 216(%rsp) leaq 304(%rsp), %rax movq %rax, 224(%rsp) leaq 312(%rsp), %rax movq %rax, 232(%rsp) movl $1, 96(%rsp) movl $1, 100(%rsp) movl $1, 104(%rsp) movl $1, 108(%rsp) movl $1, 112(%rsp) movl $1, 116(%rsp) leaq 88(%rsp), %rcx leaq 80(%rsp), %rdx leaq 108(%rsp), %rsi leaq 96(%rsp), %rdi call __cudaPopCallConfiguration@PLT testl %eax, %eax je .L55 .L51: movq 248(%rsp), %rax subq %fs:40, %rax jne .L56 addq $264, %rsp .cfi_remember_state .cfi_def_cfa_offset 8 ret .L55: .cfi_restore_state pushq 88(%rsp) .cfi_def_cfa_offset 280 pushq 88(%rsp) .cfi_def_cfa_offset 288 leaq 160(%rsp), %r9 movq 124(%rsp), %rcx movl 132(%rsp), %r8d movq 112(%rsp), %rsi movl 120(%rsp), %edx leaq _Z39predictor_coefficient_unstandardizationPfPdS_S_S0_S_S_S0_S_iii(%rip), %rdi call cudaLaunchKernel@PLT addq $16, %rsp .cfi_def_cfa_offset 272 jmp .L51 .L56: call __stack_chk_fail@PLT .cfi_endproc .LFE2063: .size _Z80__device_stub__Z39predictor_coefficient_unstandardizationPfPdS_S_S0_S_S_S0_S_iiiPfPdS_S_S0_S_S_S0_S_iii, .-_Z80__device_stub__Z39predictor_coefficient_unstandardizationPfPdS_S_S0_S_S_S0_S_iiiPfPdS_S_S0_S_S_S0_S_iii .globl _Z39predictor_coefficient_unstandardizationPfPdS_S_S0_S_S_S0_S_iii .type _Z39predictor_coefficient_unstandardizationPfPdS_S_S0_S_S_S0_S_iii, @function _Z39predictor_coefficient_unstandardizationPfPdS_S_S0_S_S_S0_S_iii: .LFB2064: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 movl 56(%rsp), %eax pushq %rax .cfi_def_cfa_offset 24 movl 56(%rsp), %eax pushq %rax .cfi_def_cfa_offset 32 movl 56(%rsp), %eax pushq %rax .cfi_def_cfa_offset 40 pushq 56(%rsp) .cfi_def_cfa_offset 48 pushq 56(%rsp) .cfi_def_cfa_offset 56 pushq 56(%rsp) .cfi_def_cfa_offset 64 call _Z80__device_stub__Z39predictor_coefficient_unstandardizationPfPdS_S_S0_S_S_S0_S_iiiPfPdS_S_S0_S_S_S0_S_iii addq $56, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2064: .size _Z39predictor_coefficient_unstandardizationPfPdS_S_S0_S_S_S0_S_iii, .-_Z39predictor_coefficient_unstandardizationPfPdS_S_S0_S_S_S0_S_iii .section .rodata.str1.8,"aMS",@progbits,1 .align 8 .LC0: .string "_Z39predictor_coefficient_unstandardizationPfPdS_S_S0_S_S_S0_S_iii" .align 8 .LC1: .string "_Z37predictor_coefficient_unnormalizationPfPdS_S_S0_S_S0_S_iii" .align 8 .LC2: .string "_Z23model_fit_shared_memoryPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii" .align 8 .LC3: .string "_Z9model_fitPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii" .align 8 .LC4: .string "_Z21model_fit_preparationPfS_S_S_S_PdS0_iii" .align 8 .LC5: .string "_Z25predictor_standardizationPfS_S_PdS0_S0_S0_S_iii" .align 8 .LC6: .string "_Z23predictor_normalizationPfS_PdS0_S0_S0_S_iii" .text .type _ZL24__sti____cudaRegisterAllv, @function _ZL24__sti____cudaRegisterAllv: .LFB2066: .cfi_startproc endbr64 pushq %rbx .cfi_def_cfa_offset 16 .cfi_offset 3, -16 leaq _ZL15__fatDeviceText(%rip), %rdi call __cudaRegisterFatBinary@PLT movq %rax, %rbx movq %rax, _ZL20__cudaFatCubinHandle(%rip) pushq $0 .cfi_def_cfa_offset 24 pushq $0 .cfi_def_cfa_offset 32 pushq $0 .cfi_def_cfa_offset 40 pushq $0 .cfi_def_cfa_offset 48 movl $0, %r9d movl $-1, %r8d leaq .LC0(%rip), %rdx movq %rdx, %rcx leaq _Z39predictor_coefficient_unstandardizationPfPdS_S_S0_S_S_S0_S_iii(%rip), %rsi movq %rax, %rdi call __cudaRegisterFunction@PLT addq $32, %rsp .cfi_def_cfa_offset 16 pushq $0 .cfi_def_cfa_offset 24 pushq $0 .cfi_def_cfa_offset 32 pushq $0 .cfi_def_cfa_offset 40 pushq $0 .cfi_def_cfa_offset 48 movl $0, %r9d movl $-1, %r8d leaq .LC1(%rip), %rdx movq %rdx, %rcx leaq _Z37predictor_coefficient_unnormalizationPfPdS_S_S0_S_S0_S_iii(%rip), %rsi movq %rbx, %rdi call __cudaRegisterFunction@PLT addq $32, %rsp .cfi_def_cfa_offset 16 pushq $0 .cfi_def_cfa_offset 24 pushq $0 .cfi_def_cfa_offset 32 pushq $0 .cfi_def_cfa_offset 40 pushq $0 .cfi_def_cfa_offset 48 movl $0, %r9d movl $-1, %r8d leaq .LC2(%rip), %rdx movq %rdx, %rcx leaq _Z23model_fit_shared_memoryPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii(%rip), %rsi movq %rbx, %rdi call __cudaRegisterFunction@PLT addq $32, %rsp .cfi_def_cfa_offset 16 pushq $0 .cfi_def_cfa_offset 24 pushq $0 .cfi_def_cfa_offset 32 pushq $0 .cfi_def_cfa_offset 40 pushq $0 .cfi_def_cfa_offset 48 movl $0, %r9d movl $-1, %r8d leaq .LC3(%rip), %rdx movq %rdx, %rcx leaq _Z9model_fitPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii(%rip), %rsi movq %rbx, %rdi call __cudaRegisterFunction@PLT addq $32, %rsp .cfi_def_cfa_offset 16 pushq $0 .cfi_def_cfa_offset 24 pushq $0 .cfi_def_cfa_offset 32 pushq $0 .cfi_def_cfa_offset 40 pushq $0 .cfi_def_cfa_offset 48 movl $0, %r9d movl $-1, %r8d leaq .LC4(%rip), %rdx movq %rdx, %rcx leaq _Z21model_fit_preparationPfS_S_S_S_PdS0_iii(%rip), %rsi movq %rbx, %rdi call __cudaRegisterFunction@PLT addq $32, %rsp .cfi_def_cfa_offset 16 pushq $0 .cfi_def_cfa_offset 24 pushq $0 .cfi_def_cfa_offset 32 pushq $0 .cfi_def_cfa_offset 40 pushq $0 .cfi_def_cfa_offset 48 movl $0, %r9d movl $-1, %r8d leaq .LC5(%rip), %rdx movq %rdx, %rcx leaq _Z25predictor_standardizationPfS_S_PdS0_S0_S0_S_iii(%rip), %rsi movq %rbx, %rdi call __cudaRegisterFunction@PLT addq $32, %rsp .cfi_def_cfa_offset 16 pushq $0 .cfi_def_cfa_offset 24 pushq $0 .cfi_def_cfa_offset 32 pushq $0 .cfi_def_cfa_offset 40 pushq $0 .cfi_def_cfa_offset 48 movl $0, %r9d movl $-1, %r8d leaq .LC6(%rip), %rdx movq %rdx, %rcx leaq _Z23predictor_normalizationPfS_PdS0_S0_S0_S_iii(%rip), %rsi movq %rbx, %rdi call __cudaRegisterFunction@PLT addq $32, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaRegisterFatBinaryEnd@PLT leaq _ZL26__cudaUnregisterBinaryUtilv(%rip), %rdi call atexit@PLT popq %rbx .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2066: .size _ZL24__sti____cudaRegisterAllv, .-_ZL24__sti____cudaRegisterAllv .section .init_array,"aw" .align 8 .quad _ZL24__sti____cudaRegisterAllv .section .nvFatBinSegment,"aw" .align 8 .type _ZL15__fatDeviceText, @object .size _ZL15__fatDeviceText, 24 _ZL15__fatDeviceText: .long 1180844977 .long 1 .quad fatbinData .quad 0 .local _ZL20__cudaFatCubinHandle .comm _ZL20__cudaFatCubinHandle,8,8 .ident "GCC: (Ubuntu 13.3.0-6ubuntu2~24.04) 13.3.0" .section .note.GNU-stack,"",@progbits .section .note.gnu.property,"a" .align 8 .long 1f - 0f .long 4f - 1f .long 5 0: .string "GNU" 1: .align 8 .long 0xc0000002 .long 3f - 2f 2: .long 0x3 3: .align 8 4:
You are a helpful coding assistant specialized in converting CUDA source code to HIP/AMD source code.
// Copyright 2020 Christopher Khan // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. // You may obtain a copy of the license at // http://www.apache.org/licenses/LICENSE-2.0 // Unless required by applicable law or agreed to in writing, software // distributed under the License is distributed on an "AS IS" BASIS, // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. // See the License for the specific language governing permissions and // limitations under the License. // Description of GPU_kernels_single_precision.cu: // This file contains the CUDA code that allows for performing the computations // for GENRE on a GPU using single precision // Define the GPU kernel that performs predictor normalization __global__ void predictor_normalization(float * X_matrix_d, float * scaling_factors_d, double * X_matrix_thread_stride_d, double * B_thread_stride_d, double * num_observations_d, double * num_predictors_d, float * intercept_flag_d, int num_threads_per_block, int num_blocks, int num_threads_last_block) { // Obtain the index of the block int block_ind = blockIdx.x; // Obtain the thread index within one block int block_thread_ind = threadIdx.x; // Calculate the fit index int fit_ind = (block_ind * num_threads_per_block) + block_thread_ind; // Determine how many threads are in the block (accounts for the fact that the last block may contain less active threads than the other blocks) int num_threads_per_block_2 = num_threads_per_block; if (block_ind == (num_blocks - 1)) { num_threads_per_block_2 = num_threads_last_block; } // This if statement makes sure that extra threads aren't doing data processing if the last block has less fits to perform if (block_thread_ind < num_threads_per_block_2) { // Obtain the thread stride that is used to obtain the correct set of predictors for the fit int predictor_thread_stride = (int)B_thread_stride_d[fit_ind]; // Obtain the number of observations for the fit int num_observations = (int)num_observations_d[fit_ind]; // Obtain the number of predictors for the fit int num_predictors = (int)num_predictors_d[fit_ind]; // Obtain the thread stride that is used to obtain the correct model matrix for the fit int X_thread_stride = (int)X_matrix_thread_stride_d[fit_ind]; // Obtain the flag that determines whether the first predictor column is a column of ones for the intercept term or not int intercept_flag = (int)intercept_flag_d[fit_ind]; // Declare and initialize the variable that stores the number of the first predictor column to be normalized int start_ind = 0; // This if statement makes sure to not normalize the first predictor column if it corresponds to the intercept term if (intercept_flag == 1) { start_ind = 1; } // Normalize each predictor column so that the sum of the square of each predictor column is equal to 1 for (int predictor_column = start_ind; predictor_column < num_predictors; predictor_column++) { // Declare and initialize the variable that stores the sum of the square of the predictor column float sum_squared = 0.0f; // Calculate the sum of the square of the predictor column for (int observation_row = 0; observation_row < num_observations; observation_row++) { float X_value = X_matrix_d[X_thread_stride + (predictor_column * num_observations) + observation_row]; sum_squared = sum_squared + (X_value * X_value); } // Calculate the square root of the sum of the square of the predictor column float square_root_sum_squared = sqrtf(sum_squared); // Store the square root of the sum of the square of the predictor column scaling_factors_d[predictor_thread_stride + predictor_column] = square_root_sum_squared; // Normalize the predictor column by dividing each observation in the predictor column by the square root of the sum of the square of the predictor column for (int observation_row = 0; observation_row < num_observations; observation_row++) { X_matrix_d[X_thread_stride + (predictor_column * num_observations) + observation_row] = X_matrix_d[X_thread_stride + (predictor_column * num_observations) + observation_row] / square_root_sum_squared; } } // This if statement stores a scaling factor of 1 for the predictor column if it corresponds to an intercept term if (intercept_flag == 1) { scaling_factors_d[predictor_thread_stride] = 1.0f; } } } // Define the GPU kernel that performs predictor standardization __global__ void predictor_standardization(float * X_matrix_d, float * scaling_factors_d, float * mean_X_matrix_d, double * X_matrix_thread_stride_d, double * B_thread_stride_d, double * num_observations_d, double * num_predictors_d, float * intercept_flag_d, int num_threads_per_block, int num_blocks, int num_threads_last_block) { // Obtain the index of the block int block_ind = blockIdx.x; // Obtain the thread index within one block int block_thread_ind = threadIdx.x; // Calculate the fit index int fit_ind = (block_ind * num_threads_per_block) + block_thread_ind; // Determine how many threads are in the block (accounts for the fact that the last block may contain less active threads than the other blocks) int num_threads_per_block_2 = num_threads_per_block; if (block_ind == (num_blocks - 1)) { num_threads_per_block_2 = num_threads_last_block; } // This if statement makes sure that extra threads aren't doing data processing if the last block has less fits to perform if (block_thread_ind < num_threads_per_block_2) { // Obtain the thread stride that is used to obtain the correct set of predictors for the fit int predictor_thread_stride = (int)B_thread_stride_d[fit_ind]; // Obtain the number of observations for the fit int num_observations = (int)num_observations_d[fit_ind]; // Obtain the number of predictors for the fit int num_predictors = (int)num_predictors_d[fit_ind]; // Obtain the thread stride that is used to obtain the correct model matrix for the fit int X_thread_stride = (int)X_matrix_thread_stride_d[fit_ind]; // Obtain the flag that determines whether the first predictor column is a column of ones for the intercept term or not int intercept_flag = (int)intercept_flag_d[fit_ind]; // Declare and initialize the variable that stores the number of the first predictor column to be standardized int start_ind = 0; // This if statement makes sure to not standardize the first predictor column if it corresponds to the intercept term if (intercept_flag == 1) { start_ind = 1; } // Standardize each predictor column by subtracting the mean of the predictor column from each observation and diving each observation by the standard deviation of the predictor column for (int predictor_column = start_ind; predictor_column < num_predictors; predictor_column++) { // Declare and initialize the variable that stores the sum of the predictor column float sum_value = 0.0f; // Calculate the sum of the predictor column for (int observation_row = 0; observation_row < num_observations; observation_row++) { float X_value = X_matrix_d[X_thread_stride + (predictor_column * num_observations) + observation_row]; sum_value = sum_value + X_value; } // Calculate the mean of the predictor column float mean_value = sum_value / (float)num_observations; // Store the mean of the predictor column mean_X_matrix_d[predictor_thread_stride + predictor_column] = mean_value; // Declare and initialize the variable that stores the sum of the square of the demeaned predictor column float sum_squared = 0.0f; // Normalize the predictor column by dividing each observation in the predictor column by the square root of the sum of the square of the predictor column for (int observation_row = 0; observation_row < num_observations; observation_row++) { float X_value_demeaned = X_matrix_d[X_thread_stride + (predictor_column * num_observations) + observation_row] - mean_value; sum_squared = sum_squared + (X_value_demeaned * X_value_demeaned); } // Calculate the standard deviation of the demeaned predictor column float std = sqrtf(sum_squared / (float)num_observations); // Store the standard deviation of the demeaned predictor column scaling_factors_d[predictor_thread_stride + predictor_column] = std; // Standardize the predictor column by subtracting its mean and dividing by its standard deviation for (int observation_row = 0; observation_row < num_observations; observation_row++) { X_matrix_d[X_thread_stride + (predictor_column * num_observations) + observation_row] = (X_matrix_d[X_thread_stride + (predictor_column * num_observations) + observation_row] - mean_value) / std; } } // This if statement stores a scaling factor of 1 and a mean of 1 for the first column if it corresponds to an intercept term if (intercept_flag == 1) { scaling_factors_d[predictor_thread_stride] = 1.0f; mean_X_matrix_d[predictor_thread_stride] = 1.0f; } } } // Define the GPU kernel that calculates the standard deviations for each portion of the y_d array, standardizes the y_d array, and calculates the standardized lambda values __global__ void model_fit_preparation(float * y_d, float * residual_y_d, float * model_fit_flag_d, float * y_std_d, float * standardized_lambda_values_d, double * num_observations_d, double * observation_thread_stride_d, int num_threads_per_block, int num_blocks, int num_threads_last_block) { // Obtain the index of the block int block_ind = blockIdx.x; // Obtain the thread index within one block int block_thread_ind = threadIdx.x; // Calculate the fit index int fit_ind = (block_ind * num_threads_per_block) + block_thread_ind; // Determine how many threads are in the block (accounts for the fact that the last block may contain less active threads than the other blocks) int num_threads_per_block_2 = num_threads_per_block; if (block_ind == (num_blocks - 1)) { num_threads_per_block_2 = num_threads_last_block; } // This if statement makes sure that extra threads aren't doing data processing if the last block has less fits to perform if (block_thread_ind < num_threads_per_block_2) { // Obtain the number of observations for the fit int num_observations = (int)num_observations_d[fit_ind]; // Obtain the thread stride that is used to obtain the correct set of observations in the cropped_y_d array for the fit int observation_thread_stride = (int)observation_thread_stride_d[fit_ind]; // Declare and initialize the variable that stores the running sum of y for the fit float sum_value = 0.0f; // Calculate the running sums for sum_value for (int observation = 0; observation < num_observations; observation++) { float value = y_d[observation_thread_stride + observation]; sum_value += value; } // Calculate the mean of y for the fit float mean = sum_value / (float)num_observations; // Declare and initialize the variable that stores the standard deviation of y for the fit float std = 0.0f; // Calculate the standard deviation of y for the fit for (int observation = 0; observation < num_observations; observation++) { float value_2 = y_d[observation_thread_stride + observation]; std += ((value_2 - mean) * (value_2 - mean)); } std = sqrtf(std / (float)num_observations); // Store the standard deviation of y for the fit in the y_std_d array y_std_d[fit_ind] = std; // This if statement standardizes the lambda values and the y data if the standard deviation isn't 0 if (std != 0.0f) { // Set the model fit flag to 1 if the standard deviation is not 0 and a model fit should be performed model_fit_flag_d[fit_ind] = 1.0f; // Calculate the standardized lambda value and store it into the standardized_lambda_d array standardized_lambda_values_d[fit_ind] = standardized_lambda_values_d[fit_ind] / std; // Standardize y for the fit and store it into the y_d array and the residual_y_d array for (int observation = 0; observation < num_observations; observation++) { float standardized_value = y_d[observation_thread_stride + observation] / std; y_d[observation_thread_stride + observation] = standardized_value; residual_y_d[observation_thread_stride + observation] = standardized_value; } } } } // Define the GPU kernel that performs least-squares regression with elastic-net regularization using the cyclic coordinate descent optimization algorithm in order to fit the model matrices to the data __global__ void model_fit(float * B_d, double * B_thread_stride_d, float * model_fit_flag_d, float * X_matrix_d, double * X_matrix_thread_stride_d, double * observation_thread_stride_d, float * residual_y_d, float * y_std_d, float * standardized_lambda_values_d, double * num_observations_d, double * num_predictors_d, float * alpha_values_d, float * tolerance_values_d, float * max_iterations_values_d, float * intercept_flag_d, int transformation_flag, int num_threads_per_block, int num_blocks, int num_threads_last_block) { // Obtain the index of the block int block_ind = blockIdx.x; // Obtain the thread index within one block int block_thread_ind = threadIdx.x; // Calculate the fit index int fit_ind = (block_ind * num_threads_per_block) + block_thread_ind; // Determine how many threads are in the block (accounts for the fact that the last block may contain less active threads than the other blocks) int num_threads_per_block_2 = num_threads_per_block; if (block_ind == (num_blocks - 1)) { num_threads_per_block_2 = num_threads_last_block; } // This if statement makes sure that extra threads aren't doing data processing if the last block has less threads if (block_thread_ind < num_threads_per_block_2) { // Obtain the flag that determines whether to perform a model fit or not int model_fit_flag = (int)model_fit_flag_d[fit_ind]; // This if statement is to ensure that a model fit is performed only if the model fit flag is 1 if (model_fit_flag == 1) { // Obtain the thread stride that is used to obtain the correct set of predictors for the fit int predictor_thread_stride = (int)B_thread_stride_d[fit_ind]; // Obtain the thread stride that is used to obtain the correct model matrix for the fit int X_thread_stride = (int)X_matrix_thread_stride_d[fit_ind]; // Obtain the thread stride that is used to obtain the correct set of observations for the fit int observation_thread_stride = (int)observation_thread_stride_d[fit_ind]; // Obtain the alpha value for the fit float alpha = alpha_values_d[fit_ind]; // Obtain the standardized lambda value for the fit float lambda = standardized_lambda_values_d[fit_ind]; // Obtain the tolerance value for the fit float tolerance = tolerance_values_d[fit_ind]; // Obtain the max iterations value for the fit int max_iterations = (int)max_iterations_values_d[fit_ind]; // Obtain the number of observations for the fit int num_observations = (int)num_observations_d[fit_ind]; // Obtain the number of predictors for the fit int num_predictors = (int)num_predictors_d[fit_ind]; // Obtain the flag that determines whether the first predictor column is a column of ones for the intercept term or not int intercept_flag = (int)intercept_flag_d[fit_ind]; // Declare and initialize the variable that stores the maximum weighted (observation weights are all 1 in this case) sum of squares of the changes in the fitted values for one iteration of cyclic coordinate descent float global_max_change = 1E12; // Declare and initialize the variable that counts how many iterations of cyclic coordinate descent have been performed int iteration_count = 0; // Perform cyclic coordinate descent until either the maximum number of iterations is reached or the maximum weighted (observation weights are all 1 in this case) sum of squares of the changes in the fitted values becomes less than the tolerance while (global_max_change >= tolerance && iteration_count < max_iterations) { // Declare and initialize the variable that stores the maximum weighted (observation weights are all 1 in this case) sum of squares of the changes in the fitted values for one iteration of cyclic coordinate descent float max_change = 0.0f; // Declare and initialize the variable that stores the weighted (observation weights are all 1 in this case) sum of squares of the changes in the fitted values that are due to the current predictor coefficient value being updated using cyclic coordinate descent float change = 0.0f; // Cycle through all of the predictors for one iteration of cyclic coordinate descent for (int j = 0; j < num_predictors; j++) { // Obtain the predictor coefficient value for the current predictor float B_j = B_d[predictor_thread_stride + j]; // Store the predictor coefficent value before it's updated float previous_B_j = B_j; // Declare and initialize the variable that stores the correlation between the current predictor column and the residual values that are obtained leaving the current predictor out float p_j = 0.0f; // Calculate the residual values leaving the current predictor out (the predictor coefficients are initialized to zero, so the residual values are going to initially be y) // This if-else statement accounts for the fact that the contribution of the current predictor only needs to be removed from the residual values if the predictor coefficient is not zero // This is due to the fact that if the predictor coefficient is already zero, then the predictor contribution to the residual is zero if (B_j != 0.0f) { for (int observation_row = 0; observation_row < num_observations; observation_row++) { // Obtain the correct value from the model matrix for the current predictor float X_value = X_matrix_d[X_thread_stride + (j * num_observations) + observation_row]; // Remove the contribution of the current predictor from the current residual value float residual_y_value = residual_y_d[observation_thread_stride + observation_row] + (X_value * B_j); // Store the updated residual value back into the residual_y_d array residual_y_d[observation_thread_stride + observation_row] = residual_y_value; // Compute the correlation between the current predictor column and the residual values that are obtained leaving the current predictor out // The correlation is computed as a running sum p_j = p_j + (X_value * residual_y_value); } } else { for (int observation_row = 0; observation_row < num_observations; observation_row++) { // Obtain the correct value from the model matrix for the current predictor float X_value = X_matrix_d[X_thread_stride + (j * num_observations) + observation_row]; // Obtain the residual value (this is essentially the residual value leaving the current predictor out because the predictor coefficient value is zero) float residual_y_value = residual_y_d[observation_thread_stride + observation_row]; // Compute the correlation between the current predictor column and the residual values that are obtained leaving the current predictor out // The correlation is computed as a running sum p_j = p_j + (X_value * residual_y_value); } } // Divide the computed correlation by the total number of observations in y (also the total number of observations in one predictor column) p_j = (1.0f / (float)num_observations) * p_j; // Apply the soft-thresholding function that is associated with the L1-regularization component of elastic-net regularization float gamma = lambda * alpha; if (p_j > 0.0f && gamma < fabsf(p_j)) { B_j = p_j - gamma; } else if (p_j < 0.0f && gamma < fabsf(p_j)) { B_j = p_j + gamma; } else { B_j = 0.0f; } // Declare and initialize the mean of the square of the predictor column float mean_squared_predictor_value = 0.0f; // Obtain the mean of the square of the predictor column if (transformation_flag == 1 || transformation_flag == 3) { mean_squared_predictor_value = 1.0f; } else if (transformation_flag == 2 || transformation_flag == 4) { mean_squared_predictor_value = 1.0f / (float)num_observations; } // This if-else statemet accounts for the fact that regularization is not applied to the intercept term if one is included if (intercept_flag == 1 && j == 0) { // Use the computed correlation value as the updated predictor coefficient B_j = p_j; } else { // Calculate the updated predictor coefficient value by applying the component of elastic-net regularization that is associated with L2-regularization // The mean_squared_predictor_value term comes from the derivation of the coordinate descent update for a predictor coefficient B_j = B_j / (mean_squared_predictor_value + (lambda * (1.0f - alpha))); } // Store the updated predictor coefficient value into the B_d array B_d[predictor_thread_stride + j] = B_j; // Update the residual values to include the contribution of the current predictor using the updated predictor coefficient value // If the updated predictor coefficient value is 0, then its contribution to the residual values is zero if (B_j != 0.0f) { for (int observation_row = 0; observation_row < num_observations; observation_row++) { // Store the updated residual back into the residual_y_d array residual_y_d[observation_thread_stride + observation_row] = residual_y_d[observation_thread_stride + observation_row] - (X_matrix_d[X_thread_stride + (j * num_observations) + observation_row] * B_j); } } // Compute the weighted (observation weights are all 1 in this case) sum of squares of the changes in the fitted values (this is used for the tolerance convergence criterion) change = (previous_B_j - B_j) * (previous_B_j - B_j); if (transformation_flag == 2 || transformation_flag == 4) { if (intercept_flag == 1 && j > 0) { change = (1.0f / (float)num_observations) * change; } else if (intercept_flag == 0) { change = (1.0f / (float)num_observations) * change; } } if (change > max_change) { max_change = change; } } // Update the global_max_change variable global_max_change = max_change; // Update the iteration count variable iteration_count = iteration_count + 1; } // Account for the fact that the y in the model fit was divided by its standard deviation float std_y = y_std_d[fit_ind]; for (int j = 0; j < num_predictors; j++) { B_d[predictor_thread_stride + j] = B_d[predictor_thread_stride + j] * std_y; } } } } // Define the GPU kernel that performs least-squares regression with elastic-net regularization using the cyclic coordinate descent optimization algorithm in order to fit the model matrices to the data __global__ void model_fit_shared_memory(float * B_d, double * B_thread_stride_d, float * model_fit_flag_d, float * X_matrix_d, double * X_matrix_thread_stride_d, double * observation_thread_stride_d, float * residual_y_d, float * y_std_d, float * standardized_lambda_values_d, double * num_observations_d, double * num_predictors_d, float * alpha_values_d, float * tolerance_values_d, float * max_iterations_values_d, float * intercept_flag_d, int transformation_flag, int num_threads_per_block, int num_blocks, int num_threads_last_block) { // Define the shared memory array that stores the residual values of the model fits within one block (the amount of bytes is declared in the GPU kernel call) extern __shared__ float sdata[]; // Obtain the index of the block int block_ind = blockIdx.x; // Obtain the thread index within one block int block_thread_ind = threadIdx.x; // Calculate the fit index int fit_ind = (block_ind * num_threads_per_block) + block_thread_ind; // Determine how many threads are in the block (accounts for the fact that the last block may contain less active threads than the other blocks) int num_threads_per_block_2 = num_threads_per_block; if (block_ind == (num_blocks - 1)) { num_threads_per_block_2 = num_threads_last_block; } // This if statement makes sure that extra threads aren't doing data processing if the last block has less threads if (block_thread_ind < num_threads_per_block_2) { // Obtain the flag that determines whether to perform a model fit or not int model_fit_flag = (int)model_fit_flag_d[fit_ind]; // This if statement is to ensure that a model fit is performed only if the model fit flag is 1 if (model_fit_flag == 1) { // Obtain the thread stride that is used to obtain the correct set of predictors for the fit int predictor_thread_stride = (int)B_thread_stride_d[fit_ind]; // Obtain the thread stride that is used to obtain the correct model matrix for the fit int X_thread_stride = (int)X_matrix_thread_stride_d[fit_ind]; // Obtain the thread stride that is used to obtain the correct set of observations for the fit int observation_thread_stride = (int)observation_thread_stride_d[fit_ind]; // Obtain the alpha value for the fit float alpha = alpha_values_d[fit_ind]; // Obtain the standardized lambda value for the fit float lambda = standardized_lambda_values_d[fit_ind]; // Obtain the tolerance value for the fit float tolerance = tolerance_values_d[fit_ind]; // Obtain the max iterations value for the fit int max_iterations = (int)max_iterations_values_d[fit_ind]; // Obtain the number of observations for the fit int num_observations = (int)num_observations_d[fit_ind]; // Obtain the number of predictors for the fit int num_predictors = (int)num_predictors_d[fit_ind]; // Obtain the flag that determines whether the first predictor column is a column of ones for the intercept term or not int intercept_flag = (int)intercept_flag_d[fit_ind]; // Declare and initialize the variable that stores the maximum weighted (observation weights are all 1 in this case) sum of squares of the changes in the fitted values for one iteration of cyclic coordinate descent float global_max_change = 1E12; // Declare and initialize the variable that counts how many iterations of cyclic coordinate descent have been performed int iteration_count = 0; // Store the residual values for the fit into the shared memory array for (int observation_row = 0; observation_row < num_observations; observation_row++) { int store_ind = (observation_row * num_threads_per_block) + block_thread_ind; sdata[store_ind] = residual_y_d[observation_thread_stride + observation_row]; } // Perform cyclic coordinate descent until either the maximum number of iterations is reached or the maximum weighted (observation weights are all 1 in this case) sum of squares of the changes in the fitted values becomes less than the tolerance while (global_max_change >= tolerance && iteration_count < max_iterations) { // Declare and initialize the variable that stores the maximum weighted (observation weights are all 1 in this case) sum of squares of the changes in the fitted values for one iteration of cyclic coordinate descent float max_change = 0.0f; // Declare and initialize the variable that stores the weighted (observation weights are all 1 in this case) sum of squares of the changes in the fitted values that are due to the current predictor coefficient value being updated using cyclic coordinate descent float change = 0.0f; // Cycle through all of the predictors for one iteration of cyclic coordinate descent for (int j = 0; j < num_predictors; j++) { // Obtain the predictor coefficient value for the current predictor float B_j = B_d[predictor_thread_stride + j]; // Store the predictor coefficent value before it's updated float previous_B_j = B_j; // Declare and initialize the variable that stores the correlation between the current predictor column and the residual values that are obtained leaving the current predictor out float p_j = 0.0f; // Calculate the residual values leaving the current predictor out (the predictor coefficients are initialized to zero, so the residual values are going to initially be y) // This if-else statement accounts for the fact that the contribution of the current predictor only needs to be removed from the residual values if the predictor coefficient is not zero // This is due to the fact that if the predictor coefficient is already zero, then the predictor contribution to the residual is zero if (B_j != 0.0f) { for (int observation_row = 0; observation_row < num_observations; observation_row++) { // Obtain the correct value from the model matrix for the current predictor float X_value = X_matrix_d[X_thread_stride + (j * num_observations) + observation_row]; // Remove the contribution of the current predictor from the current residual value float residual_y_value = sdata[(observation_row * num_threads_per_block) + block_thread_ind] + (X_value * B_j); // Store the updated residual value back into the shared memory array sdata[(observation_row * num_threads_per_block) + block_thread_ind] = residual_y_value; // Compute the correlation between the current predictor column and the residual values that are obtained leaving the current predictor out // The correlation is computed as a running sum p_j = p_j + (X_value * residual_y_value); } } else { for (int observation_row = 0; observation_row < num_observations; observation_row++) { // Obtain the correct value from the model matrix for the current predictor float X_value = X_matrix_d[X_thread_stride + (j * num_observations) + observation_row]; // Obtain the residual value (this is essentially the residual value leaving the current predictor out because the predictor coefficient value is zero) float residual_y_value = sdata[(observation_row * num_threads_per_block) + block_thread_ind]; // Compute the correlation between the current predictor column and the residual values that are obtained leaving the current predictor out // The correlation is computed as a running sum p_j = p_j + (X_value * residual_y_value); } } // Divide the computed correlation by the total number of observations in y (also the total number of observations in one predictor column) p_j = (1.0f / (float)num_observations) * p_j; // Apply the soft-thresholding function that is associated with the L1-regularization component of elastic-net regularization float gamma = lambda * alpha; if (p_j > 0.0f && gamma < fabsf(p_j)) { B_j = p_j - gamma; } else if (p_j < 0.0f && gamma < fabsf(p_j)) { B_j = p_j + gamma; } else { B_j = 0.0f; } // Declare and initialize the mean of the square of the predictor column float mean_squared_predictor_value = 0.0f; // Obtain the mean of the square of the predictor column if (transformation_flag == 1 || transformation_flag == 3) { mean_squared_predictor_value = 1.0f; } else if (transformation_flag == 2 || transformation_flag == 4) { mean_squared_predictor_value = 1.0f / (float)num_observations; } // This if-else statemet accounts for the fact that regularization is not applied to the intercept term if one is included if (intercept_flag == 1 && j == 0) { // Use the computed correlation value as the updated predictor coefficient B_j = p_j; } else { // Calculate the updated predictor coefficient value by applying the component of elastic-net regularization that is associated with L2-regularization // The mean_squared_predictor_value term comes from the derivation of the coordinate descent update for a predictor coefficient B_j = B_j / (mean_squared_predictor_value + (lambda * (1.0f - alpha))); } // Store the updated predictor coefficient value into the B_d array B_d[predictor_thread_stride + j] = B_j; // Update the residual values to include the contribution of the current predictor using the updated predictor coefficient value // If the updated predictor coefficient value is 0, then its contribution to the residual values is zero if (B_j != 0.0f) { for (int observation_row = 0; observation_row < num_observations; observation_row++) { // Store the updated residual back into the shared memory array sdata[(observation_row * num_threads_per_block) + block_thread_ind] = sdata[(observation_row * num_threads_per_block) + block_thread_ind] - (X_matrix_d[X_thread_stride + (j * num_observations) + observation_row] * B_j); } } // Compute the weighted (observation weights are all 1 in this case) sum of squares of the changes in the fitted values (this is used for the tolerance convergence criterion) change = (previous_B_j - B_j) * (previous_B_j - B_j); if (transformation_flag == 2 || transformation_flag == 4) { if (intercept_flag == 1 && j > 0) { change = (1.0f / (float)num_observations) * change; } else if (intercept_flag == 0) { change = (1.0f / (float)num_observations) * change; } } if (change > max_change) { max_change = change; } } // Update the global_max_change variable global_max_change = max_change; // Update the iteration count variable iteration_count = iteration_count + 1; } // Account for the fact that the y in the model fit was divided by its standard deviation float std_y = y_std_d[fit_ind]; for (int j = 0; j < num_predictors; j++) { B_d[predictor_thread_stride + j] = B_d[predictor_thread_stride + j] * std_y; } } } } // Define the GPU kernel that performs predictor coefficient unnormalization __global__ void predictor_coefficient_unnormalization(float * B_d, double * B_thread_stride_d, float * model_fit_flag_d, float * X_matrix_d, double * X_matrix_thread_stride_d, float * scaling_factors_d, double * num_predictors_d, float * intercept_flag_d, int num_threads_per_block, int num_blocks, int num_threads_last_block) { // Obtain the index of the block int block_ind = blockIdx.x; // Obtain the thread index within one block int block_thread_ind = threadIdx.x; // Calculate the fit index int fit_ind = (block_ind * num_threads_per_block) + block_thread_ind; // Determine how many threads are in the block (accounts for the fact that the last block may contain less active threads than the other blocks) int num_threads_per_block_2 = num_threads_per_block; if (block_ind == (num_blocks - 1)) { num_threads_per_block_2 = num_threads_last_block; } // This if statement makes sure that extra threads aren't doing data processing if the last block has less fits to perform if (block_thread_ind < num_threads_per_block_2) { // Obtain the flag that determines whether a model fit was performed or not int model_fit_flag = (int)model_fit_flag_d[fit_ind]; // This if statement is to ensure that the coefficients are unnormalized only if a model fit was performed if (model_fit_flag == 1) { // Obtain the thread stride that is used to obtain the correct set of predictors for the fit int predictor_thread_stride = (int)B_thread_stride_d[fit_ind]; // Obtain the number of predictors for the fit int num_predictors = (int)num_predictors_d[fit_ind]; // Obtain the flag that determines whether the first predictor column is a column of ones for the intercept term or not int intercept_flag = (int)intercept_flag_d[fit_ind]; // Declare and initialize the variable that stores the number of the first predictor column to be standardized int start_ind = 0; // This if statement makes sure to not standardize the first predictor column if it corresponds to the intercept term if (intercept_flag == 1) { start_ind = 1; } // Unnormalize the predictor coefficients for (int predictor_column = start_ind; predictor_column < num_predictors; predictor_column++) { B_d[predictor_thread_stride + predictor_column] = B_d[predictor_thread_stride + predictor_column] / scaling_factors_d[predictor_thread_stride + predictor_column]; } } } } // Define the GPU kernel that performs predictor coefficient unstandardization __global__ void predictor_coefficient_unstandardization(float * B_d, double * B_thread_stride_d, float * model_fit_flag_d, float * X_matrix_d, double * X_matrix_thread_stride_d, float * scaling_factors_d, float * mean_X_matrix_d, double * num_predictors_d, float * intercept_flag_d, int num_threads_per_block, int num_blocks, int num_threads_last_block) { // Obtain the index of the block int block_ind = blockIdx.x; // Obtain the thread index within one block int block_thread_ind = threadIdx.x; // Calculate the fit index int fit_ind = (block_ind * num_threads_per_block) + block_thread_ind; // Determine how many threads are in the block (accounts for the fact that the last block may contain less active threads than the other blocks) int num_threads_per_block_2 = num_threads_per_block; if (block_ind == (num_blocks - 1)) { num_threads_per_block_2 = num_threads_last_block; } // This if statement makes sure that extra threads aren't doing data processing if the last block has less fits to perform if (block_thread_ind < num_threads_per_block_2) { // Obtain the flag that determines whether a model fit was performed or not int model_fit_flag = (int)model_fit_flag_d[fit_ind]; // This if statement is to ensure that the coefficients are unstandardized only if a model fit was performed if (model_fit_flag == 1) { // Obtain the thread stride that is used to obtain the correct set of predictors for the fit int predictor_thread_stride = (int)B_thread_stride_d[fit_ind]; // Obtain the number of predictors for the fit int num_predictors = (int)num_predictors_d[fit_ind]; // Obtain the flag that determines whether the first predictor column is a column of ones for the intercept term or not int intercept_flag = (int)intercept_flag_d[fit_ind]; // Declare and initialize the variable that stores the number of the first predictor column to be standardized int start_ind = 0; // This if statement makes sure to not standardize the first predictor column if it corresponds to the intercept term if (intercept_flag == 1) { start_ind = 1; } // Declare and initialize the variable that is used to adjust the intercept term if it is included float sum_value = 0.0f; // Perform predictor coefficient unstandardization for (int predictor_column = start_ind; predictor_column < num_predictors; predictor_column++) { float B_unstandardized = B_d[predictor_thread_stride + predictor_column] / scaling_factors_d[predictor_thread_stride + predictor_column]; B_d[predictor_thread_stride + predictor_column] = B_unstandardized; sum_value = sum_value + (B_unstandardized * mean_X_matrix_d[predictor_thread_stride + predictor_column]); } // Adjust the intercept term if it is included if (intercept_flag == 1) { B_d[predictor_thread_stride] = B_d[predictor_thread_stride] - sum_value; } } } }
#include <hip/hip_runtime.h> // Copyright 2020 Christopher Khan // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. // You may obtain a copy of the license at // http://www.apache.org/licenses/LICENSE-2.0 // Unless required by applicable law or agreed to in writing, software // distributed under the License is distributed on an "AS IS" BASIS, // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. // See the License for the specific language governing permissions and // limitations under the License. // Description of GPU_kernels_single_precision.cu: // This file contains the CUDA code that allows for performing the computations // for GENRE on a GPU using single precision // Define the GPU kernel that performs predictor normalization __global__ void predictor_normalization(float * X_matrix_d, float * scaling_factors_d, double * X_matrix_thread_stride_d, double * B_thread_stride_d, double * num_observations_d, double * num_predictors_d, float * intercept_flag_d, int num_threads_per_block, int num_blocks, int num_threads_last_block) { // Obtain the index of the block int block_ind = blockIdx.x; // Obtain the thread index within one block int block_thread_ind = threadIdx.x; // Calculate the fit index int fit_ind = (block_ind * num_threads_per_block) + block_thread_ind; // Determine how many threads are in the block (accounts for the fact that the last block may contain less active threads than the other blocks) int num_threads_per_block_2 = num_threads_per_block; if (block_ind == (num_blocks - 1)) { num_threads_per_block_2 = num_threads_last_block; } // This if statement makes sure that extra threads aren't doing data processing if the last block has less fits to perform if (block_thread_ind < num_threads_per_block_2) { // Obtain the thread stride that is used to obtain the correct set of predictors for the fit int predictor_thread_stride = (int)B_thread_stride_d[fit_ind]; // Obtain the number of observations for the fit int num_observations = (int)num_observations_d[fit_ind]; // Obtain the number of predictors for the fit int num_predictors = (int)num_predictors_d[fit_ind]; // Obtain the thread stride that is used to obtain the correct model matrix for the fit int X_thread_stride = (int)X_matrix_thread_stride_d[fit_ind]; // Obtain the flag that determines whether the first predictor column is a column of ones for the intercept term or not int intercept_flag = (int)intercept_flag_d[fit_ind]; // Declare and initialize the variable that stores the number of the first predictor column to be normalized int start_ind = 0; // This if statement makes sure to not normalize the first predictor column if it corresponds to the intercept term if (intercept_flag == 1) { start_ind = 1; } // Normalize each predictor column so that the sum of the square of each predictor column is equal to 1 for (int predictor_column = start_ind; predictor_column < num_predictors; predictor_column++) { // Declare and initialize the variable that stores the sum of the square of the predictor column float sum_squared = 0.0f; // Calculate the sum of the square of the predictor column for (int observation_row = 0; observation_row < num_observations; observation_row++) { float X_value = X_matrix_d[X_thread_stride + (predictor_column * num_observations) + observation_row]; sum_squared = sum_squared + (X_value * X_value); } // Calculate the square root of the sum of the square of the predictor column float square_root_sum_squared = sqrtf(sum_squared); // Store the square root of the sum of the square of the predictor column scaling_factors_d[predictor_thread_stride + predictor_column] = square_root_sum_squared; // Normalize the predictor column by dividing each observation in the predictor column by the square root of the sum of the square of the predictor column for (int observation_row = 0; observation_row < num_observations; observation_row++) { X_matrix_d[X_thread_stride + (predictor_column * num_observations) + observation_row] = X_matrix_d[X_thread_stride + (predictor_column * num_observations) + observation_row] / square_root_sum_squared; } } // This if statement stores a scaling factor of 1 for the predictor column if it corresponds to an intercept term if (intercept_flag == 1) { scaling_factors_d[predictor_thread_stride] = 1.0f; } } } // Define the GPU kernel that performs predictor standardization __global__ void predictor_standardization(float * X_matrix_d, float * scaling_factors_d, float * mean_X_matrix_d, double * X_matrix_thread_stride_d, double * B_thread_stride_d, double * num_observations_d, double * num_predictors_d, float * intercept_flag_d, int num_threads_per_block, int num_blocks, int num_threads_last_block) { // Obtain the index of the block int block_ind = blockIdx.x; // Obtain the thread index within one block int block_thread_ind = threadIdx.x; // Calculate the fit index int fit_ind = (block_ind * num_threads_per_block) + block_thread_ind; // Determine how many threads are in the block (accounts for the fact that the last block may contain less active threads than the other blocks) int num_threads_per_block_2 = num_threads_per_block; if (block_ind == (num_blocks - 1)) { num_threads_per_block_2 = num_threads_last_block; } // This if statement makes sure that extra threads aren't doing data processing if the last block has less fits to perform if (block_thread_ind < num_threads_per_block_2) { // Obtain the thread stride that is used to obtain the correct set of predictors for the fit int predictor_thread_stride = (int)B_thread_stride_d[fit_ind]; // Obtain the number of observations for the fit int num_observations = (int)num_observations_d[fit_ind]; // Obtain the number of predictors for the fit int num_predictors = (int)num_predictors_d[fit_ind]; // Obtain the thread stride that is used to obtain the correct model matrix for the fit int X_thread_stride = (int)X_matrix_thread_stride_d[fit_ind]; // Obtain the flag that determines whether the first predictor column is a column of ones for the intercept term or not int intercept_flag = (int)intercept_flag_d[fit_ind]; // Declare and initialize the variable that stores the number of the first predictor column to be standardized int start_ind = 0; // This if statement makes sure to not standardize the first predictor column if it corresponds to the intercept term if (intercept_flag == 1) { start_ind = 1; } // Standardize each predictor column by subtracting the mean of the predictor column from each observation and diving each observation by the standard deviation of the predictor column for (int predictor_column = start_ind; predictor_column < num_predictors; predictor_column++) { // Declare and initialize the variable that stores the sum of the predictor column float sum_value = 0.0f; // Calculate the sum of the predictor column for (int observation_row = 0; observation_row < num_observations; observation_row++) { float X_value = X_matrix_d[X_thread_stride + (predictor_column * num_observations) + observation_row]; sum_value = sum_value + X_value; } // Calculate the mean of the predictor column float mean_value = sum_value / (float)num_observations; // Store the mean of the predictor column mean_X_matrix_d[predictor_thread_stride + predictor_column] = mean_value; // Declare and initialize the variable that stores the sum of the square of the demeaned predictor column float sum_squared = 0.0f; // Normalize the predictor column by dividing each observation in the predictor column by the square root of the sum of the square of the predictor column for (int observation_row = 0; observation_row < num_observations; observation_row++) { float X_value_demeaned = X_matrix_d[X_thread_stride + (predictor_column * num_observations) + observation_row] - mean_value; sum_squared = sum_squared + (X_value_demeaned * X_value_demeaned); } // Calculate the standard deviation of the demeaned predictor column float std = sqrtf(sum_squared / (float)num_observations); // Store the standard deviation of the demeaned predictor column scaling_factors_d[predictor_thread_stride + predictor_column] = std; // Standardize the predictor column by subtracting its mean and dividing by its standard deviation for (int observation_row = 0; observation_row < num_observations; observation_row++) { X_matrix_d[X_thread_stride + (predictor_column * num_observations) + observation_row] = (X_matrix_d[X_thread_stride + (predictor_column * num_observations) + observation_row] - mean_value) / std; } } // This if statement stores a scaling factor of 1 and a mean of 1 for the first column if it corresponds to an intercept term if (intercept_flag == 1) { scaling_factors_d[predictor_thread_stride] = 1.0f; mean_X_matrix_d[predictor_thread_stride] = 1.0f; } } } // Define the GPU kernel that calculates the standard deviations for each portion of the y_d array, standardizes the y_d array, and calculates the standardized lambda values __global__ void model_fit_preparation(float * y_d, float * residual_y_d, float * model_fit_flag_d, float * y_std_d, float * standardized_lambda_values_d, double * num_observations_d, double * observation_thread_stride_d, int num_threads_per_block, int num_blocks, int num_threads_last_block) { // Obtain the index of the block int block_ind = blockIdx.x; // Obtain the thread index within one block int block_thread_ind = threadIdx.x; // Calculate the fit index int fit_ind = (block_ind * num_threads_per_block) + block_thread_ind; // Determine how many threads are in the block (accounts for the fact that the last block may contain less active threads than the other blocks) int num_threads_per_block_2 = num_threads_per_block; if (block_ind == (num_blocks - 1)) { num_threads_per_block_2 = num_threads_last_block; } // This if statement makes sure that extra threads aren't doing data processing if the last block has less fits to perform if (block_thread_ind < num_threads_per_block_2) { // Obtain the number of observations for the fit int num_observations = (int)num_observations_d[fit_ind]; // Obtain the thread stride that is used to obtain the correct set of observations in the cropped_y_d array for the fit int observation_thread_stride = (int)observation_thread_stride_d[fit_ind]; // Declare and initialize the variable that stores the running sum of y for the fit float sum_value = 0.0f; // Calculate the running sums for sum_value for (int observation = 0; observation < num_observations; observation++) { float value = y_d[observation_thread_stride + observation]; sum_value += value; } // Calculate the mean of y for the fit float mean = sum_value / (float)num_observations; // Declare and initialize the variable that stores the standard deviation of y for the fit float std = 0.0f; // Calculate the standard deviation of y for the fit for (int observation = 0; observation < num_observations; observation++) { float value_2 = y_d[observation_thread_stride + observation]; std += ((value_2 - mean) * (value_2 - mean)); } std = sqrtf(std / (float)num_observations); // Store the standard deviation of y for the fit in the y_std_d array y_std_d[fit_ind] = std; // This if statement standardizes the lambda values and the y data if the standard deviation isn't 0 if (std != 0.0f) { // Set the model fit flag to 1 if the standard deviation is not 0 and a model fit should be performed model_fit_flag_d[fit_ind] = 1.0f; // Calculate the standardized lambda value and store it into the standardized_lambda_d array standardized_lambda_values_d[fit_ind] = standardized_lambda_values_d[fit_ind] / std; // Standardize y for the fit and store it into the y_d array and the residual_y_d array for (int observation = 0; observation < num_observations; observation++) { float standardized_value = y_d[observation_thread_stride + observation] / std; y_d[observation_thread_stride + observation] = standardized_value; residual_y_d[observation_thread_stride + observation] = standardized_value; } } } } // Define the GPU kernel that performs least-squares regression with elastic-net regularization using the cyclic coordinate descent optimization algorithm in order to fit the model matrices to the data __global__ void model_fit(float * B_d, double * B_thread_stride_d, float * model_fit_flag_d, float * X_matrix_d, double * X_matrix_thread_stride_d, double * observation_thread_stride_d, float * residual_y_d, float * y_std_d, float * standardized_lambda_values_d, double * num_observations_d, double * num_predictors_d, float * alpha_values_d, float * tolerance_values_d, float * max_iterations_values_d, float * intercept_flag_d, int transformation_flag, int num_threads_per_block, int num_blocks, int num_threads_last_block) { // Obtain the index of the block int block_ind = blockIdx.x; // Obtain the thread index within one block int block_thread_ind = threadIdx.x; // Calculate the fit index int fit_ind = (block_ind * num_threads_per_block) + block_thread_ind; // Determine how many threads are in the block (accounts for the fact that the last block may contain less active threads than the other blocks) int num_threads_per_block_2 = num_threads_per_block; if (block_ind == (num_blocks - 1)) { num_threads_per_block_2 = num_threads_last_block; } // This if statement makes sure that extra threads aren't doing data processing if the last block has less threads if (block_thread_ind < num_threads_per_block_2) { // Obtain the flag that determines whether to perform a model fit or not int model_fit_flag = (int)model_fit_flag_d[fit_ind]; // This if statement is to ensure that a model fit is performed only if the model fit flag is 1 if (model_fit_flag == 1) { // Obtain the thread stride that is used to obtain the correct set of predictors for the fit int predictor_thread_stride = (int)B_thread_stride_d[fit_ind]; // Obtain the thread stride that is used to obtain the correct model matrix for the fit int X_thread_stride = (int)X_matrix_thread_stride_d[fit_ind]; // Obtain the thread stride that is used to obtain the correct set of observations for the fit int observation_thread_stride = (int)observation_thread_stride_d[fit_ind]; // Obtain the alpha value for the fit float alpha = alpha_values_d[fit_ind]; // Obtain the standardized lambda value for the fit float lambda = standardized_lambda_values_d[fit_ind]; // Obtain the tolerance value for the fit float tolerance = tolerance_values_d[fit_ind]; // Obtain the max iterations value for the fit int max_iterations = (int)max_iterations_values_d[fit_ind]; // Obtain the number of observations for the fit int num_observations = (int)num_observations_d[fit_ind]; // Obtain the number of predictors for the fit int num_predictors = (int)num_predictors_d[fit_ind]; // Obtain the flag that determines whether the first predictor column is a column of ones for the intercept term or not int intercept_flag = (int)intercept_flag_d[fit_ind]; // Declare and initialize the variable that stores the maximum weighted (observation weights are all 1 in this case) sum of squares of the changes in the fitted values for one iteration of cyclic coordinate descent float global_max_change = 1E12; // Declare and initialize the variable that counts how many iterations of cyclic coordinate descent have been performed int iteration_count = 0; // Perform cyclic coordinate descent until either the maximum number of iterations is reached or the maximum weighted (observation weights are all 1 in this case) sum of squares of the changes in the fitted values becomes less than the tolerance while (global_max_change >= tolerance && iteration_count < max_iterations) { // Declare and initialize the variable that stores the maximum weighted (observation weights are all 1 in this case) sum of squares of the changes in the fitted values for one iteration of cyclic coordinate descent float max_change = 0.0f; // Declare and initialize the variable that stores the weighted (observation weights are all 1 in this case) sum of squares of the changes in the fitted values that are due to the current predictor coefficient value being updated using cyclic coordinate descent float change = 0.0f; // Cycle through all of the predictors for one iteration of cyclic coordinate descent for (int j = 0; j < num_predictors; j++) { // Obtain the predictor coefficient value for the current predictor float B_j = B_d[predictor_thread_stride + j]; // Store the predictor coefficent value before it's updated float previous_B_j = B_j; // Declare and initialize the variable that stores the correlation between the current predictor column and the residual values that are obtained leaving the current predictor out float p_j = 0.0f; // Calculate the residual values leaving the current predictor out (the predictor coefficients are initialized to zero, so the residual values are going to initially be y) // This if-else statement accounts for the fact that the contribution of the current predictor only needs to be removed from the residual values if the predictor coefficient is not zero // This is due to the fact that if the predictor coefficient is already zero, then the predictor contribution to the residual is zero if (B_j != 0.0f) { for (int observation_row = 0; observation_row < num_observations; observation_row++) { // Obtain the correct value from the model matrix for the current predictor float X_value = X_matrix_d[X_thread_stride + (j * num_observations) + observation_row]; // Remove the contribution of the current predictor from the current residual value float residual_y_value = residual_y_d[observation_thread_stride + observation_row] + (X_value * B_j); // Store the updated residual value back into the residual_y_d array residual_y_d[observation_thread_stride + observation_row] = residual_y_value; // Compute the correlation between the current predictor column and the residual values that are obtained leaving the current predictor out // The correlation is computed as a running sum p_j = p_j + (X_value * residual_y_value); } } else { for (int observation_row = 0; observation_row < num_observations; observation_row++) { // Obtain the correct value from the model matrix for the current predictor float X_value = X_matrix_d[X_thread_stride + (j * num_observations) + observation_row]; // Obtain the residual value (this is essentially the residual value leaving the current predictor out because the predictor coefficient value is zero) float residual_y_value = residual_y_d[observation_thread_stride + observation_row]; // Compute the correlation between the current predictor column and the residual values that are obtained leaving the current predictor out // The correlation is computed as a running sum p_j = p_j + (X_value * residual_y_value); } } // Divide the computed correlation by the total number of observations in y (also the total number of observations in one predictor column) p_j = (1.0f / (float)num_observations) * p_j; // Apply the soft-thresholding function that is associated with the L1-regularization component of elastic-net regularization float gamma = lambda * alpha; if (p_j > 0.0f && gamma < fabsf(p_j)) { B_j = p_j - gamma; } else if (p_j < 0.0f && gamma < fabsf(p_j)) { B_j = p_j + gamma; } else { B_j = 0.0f; } // Declare and initialize the mean of the square of the predictor column float mean_squared_predictor_value = 0.0f; // Obtain the mean of the square of the predictor column if (transformation_flag == 1 || transformation_flag == 3) { mean_squared_predictor_value = 1.0f; } else if (transformation_flag == 2 || transformation_flag == 4) { mean_squared_predictor_value = 1.0f / (float)num_observations; } // This if-else statemet accounts for the fact that regularization is not applied to the intercept term if one is included if (intercept_flag == 1 && j == 0) { // Use the computed correlation value as the updated predictor coefficient B_j = p_j; } else { // Calculate the updated predictor coefficient value by applying the component of elastic-net regularization that is associated with L2-regularization // The mean_squared_predictor_value term comes from the derivation of the coordinate descent update for a predictor coefficient B_j = B_j / (mean_squared_predictor_value + (lambda * (1.0f - alpha))); } // Store the updated predictor coefficient value into the B_d array B_d[predictor_thread_stride + j] = B_j; // Update the residual values to include the contribution of the current predictor using the updated predictor coefficient value // If the updated predictor coefficient value is 0, then its contribution to the residual values is zero if (B_j != 0.0f) { for (int observation_row = 0; observation_row < num_observations; observation_row++) { // Store the updated residual back into the residual_y_d array residual_y_d[observation_thread_stride + observation_row] = residual_y_d[observation_thread_stride + observation_row] - (X_matrix_d[X_thread_stride + (j * num_observations) + observation_row] * B_j); } } // Compute the weighted (observation weights are all 1 in this case) sum of squares of the changes in the fitted values (this is used for the tolerance convergence criterion) change = (previous_B_j - B_j) * (previous_B_j - B_j); if (transformation_flag == 2 || transformation_flag == 4) { if (intercept_flag == 1 && j > 0) { change = (1.0f / (float)num_observations) * change; } else if (intercept_flag == 0) { change = (1.0f / (float)num_observations) * change; } } if (change > max_change) { max_change = change; } } // Update the global_max_change variable global_max_change = max_change; // Update the iteration count variable iteration_count = iteration_count + 1; } // Account for the fact that the y in the model fit was divided by its standard deviation float std_y = y_std_d[fit_ind]; for (int j = 0; j < num_predictors; j++) { B_d[predictor_thread_stride + j] = B_d[predictor_thread_stride + j] * std_y; } } } } // Define the GPU kernel that performs least-squares regression with elastic-net regularization using the cyclic coordinate descent optimization algorithm in order to fit the model matrices to the data __global__ void model_fit_shared_memory(float * B_d, double * B_thread_stride_d, float * model_fit_flag_d, float * X_matrix_d, double * X_matrix_thread_stride_d, double * observation_thread_stride_d, float * residual_y_d, float * y_std_d, float * standardized_lambda_values_d, double * num_observations_d, double * num_predictors_d, float * alpha_values_d, float * tolerance_values_d, float * max_iterations_values_d, float * intercept_flag_d, int transformation_flag, int num_threads_per_block, int num_blocks, int num_threads_last_block) { // Define the shared memory array that stores the residual values of the model fits within one block (the amount of bytes is declared in the GPU kernel call) extern __shared__ float sdata[]; // Obtain the index of the block int block_ind = blockIdx.x; // Obtain the thread index within one block int block_thread_ind = threadIdx.x; // Calculate the fit index int fit_ind = (block_ind * num_threads_per_block) + block_thread_ind; // Determine how many threads are in the block (accounts for the fact that the last block may contain less active threads than the other blocks) int num_threads_per_block_2 = num_threads_per_block; if (block_ind == (num_blocks - 1)) { num_threads_per_block_2 = num_threads_last_block; } // This if statement makes sure that extra threads aren't doing data processing if the last block has less threads if (block_thread_ind < num_threads_per_block_2) { // Obtain the flag that determines whether to perform a model fit or not int model_fit_flag = (int)model_fit_flag_d[fit_ind]; // This if statement is to ensure that a model fit is performed only if the model fit flag is 1 if (model_fit_flag == 1) { // Obtain the thread stride that is used to obtain the correct set of predictors for the fit int predictor_thread_stride = (int)B_thread_stride_d[fit_ind]; // Obtain the thread stride that is used to obtain the correct model matrix for the fit int X_thread_stride = (int)X_matrix_thread_stride_d[fit_ind]; // Obtain the thread stride that is used to obtain the correct set of observations for the fit int observation_thread_stride = (int)observation_thread_stride_d[fit_ind]; // Obtain the alpha value for the fit float alpha = alpha_values_d[fit_ind]; // Obtain the standardized lambda value for the fit float lambda = standardized_lambda_values_d[fit_ind]; // Obtain the tolerance value for the fit float tolerance = tolerance_values_d[fit_ind]; // Obtain the max iterations value for the fit int max_iterations = (int)max_iterations_values_d[fit_ind]; // Obtain the number of observations for the fit int num_observations = (int)num_observations_d[fit_ind]; // Obtain the number of predictors for the fit int num_predictors = (int)num_predictors_d[fit_ind]; // Obtain the flag that determines whether the first predictor column is a column of ones for the intercept term or not int intercept_flag = (int)intercept_flag_d[fit_ind]; // Declare and initialize the variable that stores the maximum weighted (observation weights are all 1 in this case) sum of squares of the changes in the fitted values for one iteration of cyclic coordinate descent float global_max_change = 1E12; // Declare and initialize the variable that counts how many iterations of cyclic coordinate descent have been performed int iteration_count = 0; // Store the residual values for the fit into the shared memory array for (int observation_row = 0; observation_row < num_observations; observation_row++) { int store_ind = (observation_row * num_threads_per_block) + block_thread_ind; sdata[store_ind] = residual_y_d[observation_thread_stride + observation_row]; } // Perform cyclic coordinate descent until either the maximum number of iterations is reached or the maximum weighted (observation weights are all 1 in this case) sum of squares of the changes in the fitted values becomes less than the tolerance while (global_max_change >= tolerance && iteration_count < max_iterations) { // Declare and initialize the variable that stores the maximum weighted (observation weights are all 1 in this case) sum of squares of the changes in the fitted values for one iteration of cyclic coordinate descent float max_change = 0.0f; // Declare and initialize the variable that stores the weighted (observation weights are all 1 in this case) sum of squares of the changes in the fitted values that are due to the current predictor coefficient value being updated using cyclic coordinate descent float change = 0.0f; // Cycle through all of the predictors for one iteration of cyclic coordinate descent for (int j = 0; j < num_predictors; j++) { // Obtain the predictor coefficient value for the current predictor float B_j = B_d[predictor_thread_stride + j]; // Store the predictor coefficent value before it's updated float previous_B_j = B_j; // Declare and initialize the variable that stores the correlation between the current predictor column and the residual values that are obtained leaving the current predictor out float p_j = 0.0f; // Calculate the residual values leaving the current predictor out (the predictor coefficients are initialized to zero, so the residual values are going to initially be y) // This if-else statement accounts for the fact that the contribution of the current predictor only needs to be removed from the residual values if the predictor coefficient is not zero // This is due to the fact that if the predictor coefficient is already zero, then the predictor contribution to the residual is zero if (B_j != 0.0f) { for (int observation_row = 0; observation_row < num_observations; observation_row++) { // Obtain the correct value from the model matrix for the current predictor float X_value = X_matrix_d[X_thread_stride + (j * num_observations) + observation_row]; // Remove the contribution of the current predictor from the current residual value float residual_y_value = sdata[(observation_row * num_threads_per_block) + block_thread_ind] + (X_value * B_j); // Store the updated residual value back into the shared memory array sdata[(observation_row * num_threads_per_block) + block_thread_ind] = residual_y_value; // Compute the correlation between the current predictor column and the residual values that are obtained leaving the current predictor out // The correlation is computed as a running sum p_j = p_j + (X_value * residual_y_value); } } else { for (int observation_row = 0; observation_row < num_observations; observation_row++) { // Obtain the correct value from the model matrix for the current predictor float X_value = X_matrix_d[X_thread_stride + (j * num_observations) + observation_row]; // Obtain the residual value (this is essentially the residual value leaving the current predictor out because the predictor coefficient value is zero) float residual_y_value = sdata[(observation_row * num_threads_per_block) + block_thread_ind]; // Compute the correlation between the current predictor column and the residual values that are obtained leaving the current predictor out // The correlation is computed as a running sum p_j = p_j + (X_value * residual_y_value); } } // Divide the computed correlation by the total number of observations in y (also the total number of observations in one predictor column) p_j = (1.0f / (float)num_observations) * p_j; // Apply the soft-thresholding function that is associated with the L1-regularization component of elastic-net regularization float gamma = lambda * alpha; if (p_j > 0.0f && gamma < fabsf(p_j)) { B_j = p_j - gamma; } else if (p_j < 0.0f && gamma < fabsf(p_j)) { B_j = p_j + gamma; } else { B_j = 0.0f; } // Declare and initialize the mean of the square of the predictor column float mean_squared_predictor_value = 0.0f; // Obtain the mean of the square of the predictor column if (transformation_flag == 1 || transformation_flag == 3) { mean_squared_predictor_value = 1.0f; } else if (transformation_flag == 2 || transformation_flag == 4) { mean_squared_predictor_value = 1.0f / (float)num_observations; } // This if-else statemet accounts for the fact that regularization is not applied to the intercept term if one is included if (intercept_flag == 1 && j == 0) { // Use the computed correlation value as the updated predictor coefficient B_j = p_j; } else { // Calculate the updated predictor coefficient value by applying the component of elastic-net regularization that is associated with L2-regularization // The mean_squared_predictor_value term comes from the derivation of the coordinate descent update for a predictor coefficient B_j = B_j / (mean_squared_predictor_value + (lambda * (1.0f - alpha))); } // Store the updated predictor coefficient value into the B_d array B_d[predictor_thread_stride + j] = B_j; // Update the residual values to include the contribution of the current predictor using the updated predictor coefficient value // If the updated predictor coefficient value is 0, then its contribution to the residual values is zero if (B_j != 0.0f) { for (int observation_row = 0; observation_row < num_observations; observation_row++) { // Store the updated residual back into the shared memory array sdata[(observation_row * num_threads_per_block) + block_thread_ind] = sdata[(observation_row * num_threads_per_block) + block_thread_ind] - (X_matrix_d[X_thread_stride + (j * num_observations) + observation_row] * B_j); } } // Compute the weighted (observation weights are all 1 in this case) sum of squares of the changes in the fitted values (this is used for the tolerance convergence criterion) change = (previous_B_j - B_j) * (previous_B_j - B_j); if (transformation_flag == 2 || transformation_flag == 4) { if (intercept_flag == 1 && j > 0) { change = (1.0f / (float)num_observations) * change; } else if (intercept_flag == 0) { change = (1.0f / (float)num_observations) * change; } } if (change > max_change) { max_change = change; } } // Update the global_max_change variable global_max_change = max_change; // Update the iteration count variable iteration_count = iteration_count + 1; } // Account for the fact that the y in the model fit was divided by its standard deviation float std_y = y_std_d[fit_ind]; for (int j = 0; j < num_predictors; j++) { B_d[predictor_thread_stride + j] = B_d[predictor_thread_stride + j] * std_y; } } } } // Define the GPU kernel that performs predictor coefficient unnormalization __global__ void predictor_coefficient_unnormalization(float * B_d, double * B_thread_stride_d, float * model_fit_flag_d, float * X_matrix_d, double * X_matrix_thread_stride_d, float * scaling_factors_d, double * num_predictors_d, float * intercept_flag_d, int num_threads_per_block, int num_blocks, int num_threads_last_block) { // Obtain the index of the block int block_ind = blockIdx.x; // Obtain the thread index within one block int block_thread_ind = threadIdx.x; // Calculate the fit index int fit_ind = (block_ind * num_threads_per_block) + block_thread_ind; // Determine how many threads are in the block (accounts for the fact that the last block may contain less active threads than the other blocks) int num_threads_per_block_2 = num_threads_per_block; if (block_ind == (num_blocks - 1)) { num_threads_per_block_2 = num_threads_last_block; } // This if statement makes sure that extra threads aren't doing data processing if the last block has less fits to perform if (block_thread_ind < num_threads_per_block_2) { // Obtain the flag that determines whether a model fit was performed or not int model_fit_flag = (int)model_fit_flag_d[fit_ind]; // This if statement is to ensure that the coefficients are unnormalized only if a model fit was performed if (model_fit_flag == 1) { // Obtain the thread stride that is used to obtain the correct set of predictors for the fit int predictor_thread_stride = (int)B_thread_stride_d[fit_ind]; // Obtain the number of predictors for the fit int num_predictors = (int)num_predictors_d[fit_ind]; // Obtain the flag that determines whether the first predictor column is a column of ones for the intercept term or not int intercept_flag = (int)intercept_flag_d[fit_ind]; // Declare and initialize the variable that stores the number of the first predictor column to be standardized int start_ind = 0; // This if statement makes sure to not standardize the first predictor column if it corresponds to the intercept term if (intercept_flag == 1) { start_ind = 1; } // Unnormalize the predictor coefficients for (int predictor_column = start_ind; predictor_column < num_predictors; predictor_column++) { B_d[predictor_thread_stride + predictor_column] = B_d[predictor_thread_stride + predictor_column] / scaling_factors_d[predictor_thread_stride + predictor_column]; } } } } // Define the GPU kernel that performs predictor coefficient unstandardization __global__ void predictor_coefficient_unstandardization(float * B_d, double * B_thread_stride_d, float * model_fit_flag_d, float * X_matrix_d, double * X_matrix_thread_stride_d, float * scaling_factors_d, float * mean_X_matrix_d, double * num_predictors_d, float * intercept_flag_d, int num_threads_per_block, int num_blocks, int num_threads_last_block) { // Obtain the index of the block int block_ind = blockIdx.x; // Obtain the thread index within one block int block_thread_ind = threadIdx.x; // Calculate the fit index int fit_ind = (block_ind * num_threads_per_block) + block_thread_ind; // Determine how many threads are in the block (accounts for the fact that the last block may contain less active threads than the other blocks) int num_threads_per_block_2 = num_threads_per_block; if (block_ind == (num_blocks - 1)) { num_threads_per_block_2 = num_threads_last_block; } // This if statement makes sure that extra threads aren't doing data processing if the last block has less fits to perform if (block_thread_ind < num_threads_per_block_2) { // Obtain the flag that determines whether a model fit was performed or not int model_fit_flag = (int)model_fit_flag_d[fit_ind]; // This if statement is to ensure that the coefficients are unstandardized only if a model fit was performed if (model_fit_flag == 1) { // Obtain the thread stride that is used to obtain the correct set of predictors for the fit int predictor_thread_stride = (int)B_thread_stride_d[fit_ind]; // Obtain the number of predictors for the fit int num_predictors = (int)num_predictors_d[fit_ind]; // Obtain the flag that determines whether the first predictor column is a column of ones for the intercept term or not int intercept_flag = (int)intercept_flag_d[fit_ind]; // Declare and initialize the variable that stores the number of the first predictor column to be standardized int start_ind = 0; // This if statement makes sure to not standardize the first predictor column if it corresponds to the intercept term if (intercept_flag == 1) { start_ind = 1; } // Declare and initialize the variable that is used to adjust the intercept term if it is included float sum_value = 0.0f; // Perform predictor coefficient unstandardization for (int predictor_column = start_ind; predictor_column < num_predictors; predictor_column++) { float B_unstandardized = B_d[predictor_thread_stride + predictor_column] / scaling_factors_d[predictor_thread_stride + predictor_column]; B_d[predictor_thread_stride + predictor_column] = B_unstandardized; sum_value = sum_value + (B_unstandardized * mean_X_matrix_d[predictor_thread_stride + predictor_column]); } // Adjust the intercept term if it is included if (intercept_flag == 1) { B_d[predictor_thread_stride] = B_d[predictor_thread_stride] - sum_value; } } } }
You are a helpful coding assistant specialized in converting HIP source code to AMD host assembly.
#include <hip/hip_runtime.h> // Copyright 2020 Christopher Khan // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. // You may obtain a copy of the license at // http://www.apache.org/licenses/LICENSE-2.0 // Unless required by applicable law or agreed to in writing, software // distributed under the License is distributed on an "AS IS" BASIS, // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. // See the License for the specific language governing permissions and // limitations under the License. // Description of GPU_kernels_single_precision.cu: // This file contains the CUDA code that allows for performing the computations // for GENRE on a GPU using single precision // Define the GPU kernel that performs predictor normalization __global__ void predictor_normalization(float * X_matrix_d, float * scaling_factors_d, double * X_matrix_thread_stride_d, double * B_thread_stride_d, double * num_observations_d, double * num_predictors_d, float * intercept_flag_d, int num_threads_per_block, int num_blocks, int num_threads_last_block) { // Obtain the index of the block int block_ind = blockIdx.x; // Obtain the thread index within one block int block_thread_ind = threadIdx.x; // Calculate the fit index int fit_ind = (block_ind * num_threads_per_block) + block_thread_ind; // Determine how many threads are in the block (accounts for the fact that the last block may contain less active threads than the other blocks) int num_threads_per_block_2 = num_threads_per_block; if (block_ind == (num_blocks - 1)) { num_threads_per_block_2 = num_threads_last_block; } // This if statement makes sure that extra threads aren't doing data processing if the last block has less fits to perform if (block_thread_ind < num_threads_per_block_2) { // Obtain the thread stride that is used to obtain the correct set of predictors for the fit int predictor_thread_stride = (int)B_thread_stride_d[fit_ind]; // Obtain the number of observations for the fit int num_observations = (int)num_observations_d[fit_ind]; // Obtain the number of predictors for the fit int num_predictors = (int)num_predictors_d[fit_ind]; // Obtain the thread stride that is used to obtain the correct model matrix for the fit int X_thread_stride = (int)X_matrix_thread_stride_d[fit_ind]; // Obtain the flag that determines whether the first predictor column is a column of ones for the intercept term or not int intercept_flag = (int)intercept_flag_d[fit_ind]; // Declare and initialize the variable that stores the number of the first predictor column to be normalized int start_ind = 0; // This if statement makes sure to not normalize the first predictor column if it corresponds to the intercept term if (intercept_flag == 1) { start_ind = 1; } // Normalize each predictor column so that the sum of the square of each predictor column is equal to 1 for (int predictor_column = start_ind; predictor_column < num_predictors; predictor_column++) { // Declare and initialize the variable that stores the sum of the square of the predictor column float sum_squared = 0.0f; // Calculate the sum of the square of the predictor column for (int observation_row = 0; observation_row < num_observations; observation_row++) { float X_value = X_matrix_d[X_thread_stride + (predictor_column * num_observations) + observation_row]; sum_squared = sum_squared + (X_value * X_value); } // Calculate the square root of the sum of the square of the predictor column float square_root_sum_squared = sqrtf(sum_squared); // Store the square root of the sum of the square of the predictor column scaling_factors_d[predictor_thread_stride + predictor_column] = square_root_sum_squared; // Normalize the predictor column by dividing each observation in the predictor column by the square root of the sum of the square of the predictor column for (int observation_row = 0; observation_row < num_observations; observation_row++) { X_matrix_d[X_thread_stride + (predictor_column * num_observations) + observation_row] = X_matrix_d[X_thread_stride + (predictor_column * num_observations) + observation_row] / square_root_sum_squared; } } // This if statement stores a scaling factor of 1 for the predictor column if it corresponds to an intercept term if (intercept_flag == 1) { scaling_factors_d[predictor_thread_stride] = 1.0f; } } } // Define the GPU kernel that performs predictor standardization __global__ void predictor_standardization(float * X_matrix_d, float * scaling_factors_d, float * mean_X_matrix_d, double * X_matrix_thread_stride_d, double * B_thread_stride_d, double * num_observations_d, double * num_predictors_d, float * intercept_flag_d, int num_threads_per_block, int num_blocks, int num_threads_last_block) { // Obtain the index of the block int block_ind = blockIdx.x; // Obtain the thread index within one block int block_thread_ind = threadIdx.x; // Calculate the fit index int fit_ind = (block_ind * num_threads_per_block) + block_thread_ind; // Determine how many threads are in the block (accounts for the fact that the last block may contain less active threads than the other blocks) int num_threads_per_block_2 = num_threads_per_block; if (block_ind == (num_blocks - 1)) { num_threads_per_block_2 = num_threads_last_block; } // This if statement makes sure that extra threads aren't doing data processing if the last block has less fits to perform if (block_thread_ind < num_threads_per_block_2) { // Obtain the thread stride that is used to obtain the correct set of predictors for the fit int predictor_thread_stride = (int)B_thread_stride_d[fit_ind]; // Obtain the number of observations for the fit int num_observations = (int)num_observations_d[fit_ind]; // Obtain the number of predictors for the fit int num_predictors = (int)num_predictors_d[fit_ind]; // Obtain the thread stride that is used to obtain the correct model matrix for the fit int X_thread_stride = (int)X_matrix_thread_stride_d[fit_ind]; // Obtain the flag that determines whether the first predictor column is a column of ones for the intercept term or not int intercept_flag = (int)intercept_flag_d[fit_ind]; // Declare and initialize the variable that stores the number of the first predictor column to be standardized int start_ind = 0; // This if statement makes sure to not standardize the first predictor column if it corresponds to the intercept term if (intercept_flag == 1) { start_ind = 1; } // Standardize each predictor column by subtracting the mean of the predictor column from each observation and diving each observation by the standard deviation of the predictor column for (int predictor_column = start_ind; predictor_column < num_predictors; predictor_column++) { // Declare and initialize the variable that stores the sum of the predictor column float sum_value = 0.0f; // Calculate the sum of the predictor column for (int observation_row = 0; observation_row < num_observations; observation_row++) { float X_value = X_matrix_d[X_thread_stride + (predictor_column * num_observations) + observation_row]; sum_value = sum_value + X_value; } // Calculate the mean of the predictor column float mean_value = sum_value / (float)num_observations; // Store the mean of the predictor column mean_X_matrix_d[predictor_thread_stride + predictor_column] = mean_value; // Declare and initialize the variable that stores the sum of the square of the demeaned predictor column float sum_squared = 0.0f; // Normalize the predictor column by dividing each observation in the predictor column by the square root of the sum of the square of the predictor column for (int observation_row = 0; observation_row < num_observations; observation_row++) { float X_value_demeaned = X_matrix_d[X_thread_stride + (predictor_column * num_observations) + observation_row] - mean_value; sum_squared = sum_squared + (X_value_demeaned * X_value_demeaned); } // Calculate the standard deviation of the demeaned predictor column float std = sqrtf(sum_squared / (float)num_observations); // Store the standard deviation of the demeaned predictor column scaling_factors_d[predictor_thread_stride + predictor_column] = std; // Standardize the predictor column by subtracting its mean and dividing by its standard deviation for (int observation_row = 0; observation_row < num_observations; observation_row++) { X_matrix_d[X_thread_stride + (predictor_column * num_observations) + observation_row] = (X_matrix_d[X_thread_stride + (predictor_column * num_observations) + observation_row] - mean_value) / std; } } // This if statement stores a scaling factor of 1 and a mean of 1 for the first column if it corresponds to an intercept term if (intercept_flag == 1) { scaling_factors_d[predictor_thread_stride] = 1.0f; mean_X_matrix_d[predictor_thread_stride] = 1.0f; } } } // Define the GPU kernel that calculates the standard deviations for each portion of the y_d array, standardizes the y_d array, and calculates the standardized lambda values __global__ void model_fit_preparation(float * y_d, float * residual_y_d, float * model_fit_flag_d, float * y_std_d, float * standardized_lambda_values_d, double * num_observations_d, double * observation_thread_stride_d, int num_threads_per_block, int num_blocks, int num_threads_last_block) { // Obtain the index of the block int block_ind = blockIdx.x; // Obtain the thread index within one block int block_thread_ind = threadIdx.x; // Calculate the fit index int fit_ind = (block_ind * num_threads_per_block) + block_thread_ind; // Determine how many threads are in the block (accounts for the fact that the last block may contain less active threads than the other blocks) int num_threads_per_block_2 = num_threads_per_block; if (block_ind == (num_blocks - 1)) { num_threads_per_block_2 = num_threads_last_block; } // This if statement makes sure that extra threads aren't doing data processing if the last block has less fits to perform if (block_thread_ind < num_threads_per_block_2) { // Obtain the number of observations for the fit int num_observations = (int)num_observations_d[fit_ind]; // Obtain the thread stride that is used to obtain the correct set of observations in the cropped_y_d array for the fit int observation_thread_stride = (int)observation_thread_stride_d[fit_ind]; // Declare and initialize the variable that stores the running sum of y for the fit float sum_value = 0.0f; // Calculate the running sums for sum_value for (int observation = 0; observation < num_observations; observation++) { float value = y_d[observation_thread_stride + observation]; sum_value += value; } // Calculate the mean of y for the fit float mean = sum_value / (float)num_observations; // Declare and initialize the variable that stores the standard deviation of y for the fit float std = 0.0f; // Calculate the standard deviation of y for the fit for (int observation = 0; observation < num_observations; observation++) { float value_2 = y_d[observation_thread_stride + observation]; std += ((value_2 - mean) * (value_2 - mean)); } std = sqrtf(std / (float)num_observations); // Store the standard deviation of y for the fit in the y_std_d array y_std_d[fit_ind] = std; // This if statement standardizes the lambda values and the y data if the standard deviation isn't 0 if (std != 0.0f) { // Set the model fit flag to 1 if the standard deviation is not 0 and a model fit should be performed model_fit_flag_d[fit_ind] = 1.0f; // Calculate the standardized lambda value and store it into the standardized_lambda_d array standardized_lambda_values_d[fit_ind] = standardized_lambda_values_d[fit_ind] / std; // Standardize y for the fit and store it into the y_d array and the residual_y_d array for (int observation = 0; observation < num_observations; observation++) { float standardized_value = y_d[observation_thread_stride + observation] / std; y_d[observation_thread_stride + observation] = standardized_value; residual_y_d[observation_thread_stride + observation] = standardized_value; } } } } // Define the GPU kernel that performs least-squares regression with elastic-net regularization using the cyclic coordinate descent optimization algorithm in order to fit the model matrices to the data __global__ void model_fit(float * B_d, double * B_thread_stride_d, float * model_fit_flag_d, float * X_matrix_d, double * X_matrix_thread_stride_d, double * observation_thread_stride_d, float * residual_y_d, float * y_std_d, float * standardized_lambda_values_d, double * num_observations_d, double * num_predictors_d, float * alpha_values_d, float * tolerance_values_d, float * max_iterations_values_d, float * intercept_flag_d, int transformation_flag, int num_threads_per_block, int num_blocks, int num_threads_last_block) { // Obtain the index of the block int block_ind = blockIdx.x; // Obtain the thread index within one block int block_thread_ind = threadIdx.x; // Calculate the fit index int fit_ind = (block_ind * num_threads_per_block) + block_thread_ind; // Determine how many threads are in the block (accounts for the fact that the last block may contain less active threads than the other blocks) int num_threads_per_block_2 = num_threads_per_block; if (block_ind == (num_blocks - 1)) { num_threads_per_block_2 = num_threads_last_block; } // This if statement makes sure that extra threads aren't doing data processing if the last block has less threads if (block_thread_ind < num_threads_per_block_2) { // Obtain the flag that determines whether to perform a model fit or not int model_fit_flag = (int)model_fit_flag_d[fit_ind]; // This if statement is to ensure that a model fit is performed only if the model fit flag is 1 if (model_fit_flag == 1) { // Obtain the thread stride that is used to obtain the correct set of predictors for the fit int predictor_thread_stride = (int)B_thread_stride_d[fit_ind]; // Obtain the thread stride that is used to obtain the correct model matrix for the fit int X_thread_stride = (int)X_matrix_thread_stride_d[fit_ind]; // Obtain the thread stride that is used to obtain the correct set of observations for the fit int observation_thread_stride = (int)observation_thread_stride_d[fit_ind]; // Obtain the alpha value for the fit float alpha = alpha_values_d[fit_ind]; // Obtain the standardized lambda value for the fit float lambda = standardized_lambda_values_d[fit_ind]; // Obtain the tolerance value for the fit float tolerance = tolerance_values_d[fit_ind]; // Obtain the max iterations value for the fit int max_iterations = (int)max_iterations_values_d[fit_ind]; // Obtain the number of observations for the fit int num_observations = (int)num_observations_d[fit_ind]; // Obtain the number of predictors for the fit int num_predictors = (int)num_predictors_d[fit_ind]; // Obtain the flag that determines whether the first predictor column is a column of ones for the intercept term or not int intercept_flag = (int)intercept_flag_d[fit_ind]; // Declare and initialize the variable that stores the maximum weighted (observation weights are all 1 in this case) sum of squares of the changes in the fitted values for one iteration of cyclic coordinate descent float global_max_change = 1E12; // Declare and initialize the variable that counts how many iterations of cyclic coordinate descent have been performed int iteration_count = 0; // Perform cyclic coordinate descent until either the maximum number of iterations is reached or the maximum weighted (observation weights are all 1 in this case) sum of squares of the changes in the fitted values becomes less than the tolerance while (global_max_change >= tolerance && iteration_count < max_iterations) { // Declare and initialize the variable that stores the maximum weighted (observation weights are all 1 in this case) sum of squares of the changes in the fitted values for one iteration of cyclic coordinate descent float max_change = 0.0f; // Declare and initialize the variable that stores the weighted (observation weights are all 1 in this case) sum of squares of the changes in the fitted values that are due to the current predictor coefficient value being updated using cyclic coordinate descent float change = 0.0f; // Cycle through all of the predictors for one iteration of cyclic coordinate descent for (int j = 0; j < num_predictors; j++) { // Obtain the predictor coefficient value for the current predictor float B_j = B_d[predictor_thread_stride + j]; // Store the predictor coefficent value before it's updated float previous_B_j = B_j; // Declare and initialize the variable that stores the correlation between the current predictor column and the residual values that are obtained leaving the current predictor out float p_j = 0.0f; // Calculate the residual values leaving the current predictor out (the predictor coefficients are initialized to zero, so the residual values are going to initially be y) // This if-else statement accounts for the fact that the contribution of the current predictor only needs to be removed from the residual values if the predictor coefficient is not zero // This is due to the fact that if the predictor coefficient is already zero, then the predictor contribution to the residual is zero if (B_j != 0.0f) { for (int observation_row = 0; observation_row < num_observations; observation_row++) { // Obtain the correct value from the model matrix for the current predictor float X_value = X_matrix_d[X_thread_stride + (j * num_observations) + observation_row]; // Remove the contribution of the current predictor from the current residual value float residual_y_value = residual_y_d[observation_thread_stride + observation_row] + (X_value * B_j); // Store the updated residual value back into the residual_y_d array residual_y_d[observation_thread_stride + observation_row] = residual_y_value; // Compute the correlation between the current predictor column and the residual values that are obtained leaving the current predictor out // The correlation is computed as a running sum p_j = p_j + (X_value * residual_y_value); } } else { for (int observation_row = 0; observation_row < num_observations; observation_row++) { // Obtain the correct value from the model matrix for the current predictor float X_value = X_matrix_d[X_thread_stride + (j * num_observations) + observation_row]; // Obtain the residual value (this is essentially the residual value leaving the current predictor out because the predictor coefficient value is zero) float residual_y_value = residual_y_d[observation_thread_stride + observation_row]; // Compute the correlation between the current predictor column and the residual values that are obtained leaving the current predictor out // The correlation is computed as a running sum p_j = p_j + (X_value * residual_y_value); } } // Divide the computed correlation by the total number of observations in y (also the total number of observations in one predictor column) p_j = (1.0f / (float)num_observations) * p_j; // Apply the soft-thresholding function that is associated with the L1-regularization component of elastic-net regularization float gamma = lambda * alpha; if (p_j > 0.0f && gamma < fabsf(p_j)) { B_j = p_j - gamma; } else if (p_j < 0.0f && gamma < fabsf(p_j)) { B_j = p_j + gamma; } else { B_j = 0.0f; } // Declare and initialize the mean of the square of the predictor column float mean_squared_predictor_value = 0.0f; // Obtain the mean of the square of the predictor column if (transformation_flag == 1 || transformation_flag == 3) { mean_squared_predictor_value = 1.0f; } else if (transformation_flag == 2 || transformation_flag == 4) { mean_squared_predictor_value = 1.0f / (float)num_observations; } // This if-else statemet accounts for the fact that regularization is not applied to the intercept term if one is included if (intercept_flag == 1 && j == 0) { // Use the computed correlation value as the updated predictor coefficient B_j = p_j; } else { // Calculate the updated predictor coefficient value by applying the component of elastic-net regularization that is associated with L2-regularization // The mean_squared_predictor_value term comes from the derivation of the coordinate descent update for a predictor coefficient B_j = B_j / (mean_squared_predictor_value + (lambda * (1.0f - alpha))); } // Store the updated predictor coefficient value into the B_d array B_d[predictor_thread_stride + j] = B_j; // Update the residual values to include the contribution of the current predictor using the updated predictor coefficient value // If the updated predictor coefficient value is 0, then its contribution to the residual values is zero if (B_j != 0.0f) { for (int observation_row = 0; observation_row < num_observations; observation_row++) { // Store the updated residual back into the residual_y_d array residual_y_d[observation_thread_stride + observation_row] = residual_y_d[observation_thread_stride + observation_row] - (X_matrix_d[X_thread_stride + (j * num_observations) + observation_row] * B_j); } } // Compute the weighted (observation weights are all 1 in this case) sum of squares of the changes in the fitted values (this is used for the tolerance convergence criterion) change = (previous_B_j - B_j) * (previous_B_j - B_j); if (transformation_flag == 2 || transformation_flag == 4) { if (intercept_flag == 1 && j > 0) { change = (1.0f / (float)num_observations) * change; } else if (intercept_flag == 0) { change = (1.0f / (float)num_observations) * change; } } if (change > max_change) { max_change = change; } } // Update the global_max_change variable global_max_change = max_change; // Update the iteration count variable iteration_count = iteration_count + 1; } // Account for the fact that the y in the model fit was divided by its standard deviation float std_y = y_std_d[fit_ind]; for (int j = 0; j < num_predictors; j++) { B_d[predictor_thread_stride + j] = B_d[predictor_thread_stride + j] * std_y; } } } } // Define the GPU kernel that performs least-squares regression with elastic-net regularization using the cyclic coordinate descent optimization algorithm in order to fit the model matrices to the data __global__ void model_fit_shared_memory(float * B_d, double * B_thread_stride_d, float * model_fit_flag_d, float * X_matrix_d, double * X_matrix_thread_stride_d, double * observation_thread_stride_d, float * residual_y_d, float * y_std_d, float * standardized_lambda_values_d, double * num_observations_d, double * num_predictors_d, float * alpha_values_d, float * tolerance_values_d, float * max_iterations_values_d, float * intercept_flag_d, int transformation_flag, int num_threads_per_block, int num_blocks, int num_threads_last_block) { // Define the shared memory array that stores the residual values of the model fits within one block (the amount of bytes is declared in the GPU kernel call) extern __shared__ float sdata[]; // Obtain the index of the block int block_ind = blockIdx.x; // Obtain the thread index within one block int block_thread_ind = threadIdx.x; // Calculate the fit index int fit_ind = (block_ind * num_threads_per_block) + block_thread_ind; // Determine how many threads are in the block (accounts for the fact that the last block may contain less active threads than the other blocks) int num_threads_per_block_2 = num_threads_per_block; if (block_ind == (num_blocks - 1)) { num_threads_per_block_2 = num_threads_last_block; } // This if statement makes sure that extra threads aren't doing data processing if the last block has less threads if (block_thread_ind < num_threads_per_block_2) { // Obtain the flag that determines whether to perform a model fit or not int model_fit_flag = (int)model_fit_flag_d[fit_ind]; // This if statement is to ensure that a model fit is performed only if the model fit flag is 1 if (model_fit_flag == 1) { // Obtain the thread stride that is used to obtain the correct set of predictors for the fit int predictor_thread_stride = (int)B_thread_stride_d[fit_ind]; // Obtain the thread stride that is used to obtain the correct model matrix for the fit int X_thread_stride = (int)X_matrix_thread_stride_d[fit_ind]; // Obtain the thread stride that is used to obtain the correct set of observations for the fit int observation_thread_stride = (int)observation_thread_stride_d[fit_ind]; // Obtain the alpha value for the fit float alpha = alpha_values_d[fit_ind]; // Obtain the standardized lambda value for the fit float lambda = standardized_lambda_values_d[fit_ind]; // Obtain the tolerance value for the fit float tolerance = tolerance_values_d[fit_ind]; // Obtain the max iterations value for the fit int max_iterations = (int)max_iterations_values_d[fit_ind]; // Obtain the number of observations for the fit int num_observations = (int)num_observations_d[fit_ind]; // Obtain the number of predictors for the fit int num_predictors = (int)num_predictors_d[fit_ind]; // Obtain the flag that determines whether the first predictor column is a column of ones for the intercept term or not int intercept_flag = (int)intercept_flag_d[fit_ind]; // Declare and initialize the variable that stores the maximum weighted (observation weights are all 1 in this case) sum of squares of the changes in the fitted values for one iteration of cyclic coordinate descent float global_max_change = 1E12; // Declare and initialize the variable that counts how many iterations of cyclic coordinate descent have been performed int iteration_count = 0; // Store the residual values for the fit into the shared memory array for (int observation_row = 0; observation_row < num_observations; observation_row++) { int store_ind = (observation_row * num_threads_per_block) + block_thread_ind; sdata[store_ind] = residual_y_d[observation_thread_stride + observation_row]; } // Perform cyclic coordinate descent until either the maximum number of iterations is reached or the maximum weighted (observation weights are all 1 in this case) sum of squares of the changes in the fitted values becomes less than the tolerance while (global_max_change >= tolerance && iteration_count < max_iterations) { // Declare and initialize the variable that stores the maximum weighted (observation weights are all 1 in this case) sum of squares of the changes in the fitted values for one iteration of cyclic coordinate descent float max_change = 0.0f; // Declare and initialize the variable that stores the weighted (observation weights are all 1 in this case) sum of squares of the changes in the fitted values that are due to the current predictor coefficient value being updated using cyclic coordinate descent float change = 0.0f; // Cycle through all of the predictors for one iteration of cyclic coordinate descent for (int j = 0; j < num_predictors; j++) { // Obtain the predictor coefficient value for the current predictor float B_j = B_d[predictor_thread_stride + j]; // Store the predictor coefficent value before it's updated float previous_B_j = B_j; // Declare and initialize the variable that stores the correlation between the current predictor column and the residual values that are obtained leaving the current predictor out float p_j = 0.0f; // Calculate the residual values leaving the current predictor out (the predictor coefficients are initialized to zero, so the residual values are going to initially be y) // This if-else statement accounts for the fact that the contribution of the current predictor only needs to be removed from the residual values if the predictor coefficient is not zero // This is due to the fact that if the predictor coefficient is already zero, then the predictor contribution to the residual is zero if (B_j != 0.0f) { for (int observation_row = 0; observation_row < num_observations; observation_row++) { // Obtain the correct value from the model matrix for the current predictor float X_value = X_matrix_d[X_thread_stride + (j * num_observations) + observation_row]; // Remove the contribution of the current predictor from the current residual value float residual_y_value = sdata[(observation_row * num_threads_per_block) + block_thread_ind] + (X_value * B_j); // Store the updated residual value back into the shared memory array sdata[(observation_row * num_threads_per_block) + block_thread_ind] = residual_y_value; // Compute the correlation between the current predictor column and the residual values that are obtained leaving the current predictor out // The correlation is computed as a running sum p_j = p_j + (X_value * residual_y_value); } } else { for (int observation_row = 0; observation_row < num_observations; observation_row++) { // Obtain the correct value from the model matrix for the current predictor float X_value = X_matrix_d[X_thread_stride + (j * num_observations) + observation_row]; // Obtain the residual value (this is essentially the residual value leaving the current predictor out because the predictor coefficient value is zero) float residual_y_value = sdata[(observation_row * num_threads_per_block) + block_thread_ind]; // Compute the correlation between the current predictor column and the residual values that are obtained leaving the current predictor out // The correlation is computed as a running sum p_j = p_j + (X_value * residual_y_value); } } // Divide the computed correlation by the total number of observations in y (also the total number of observations in one predictor column) p_j = (1.0f / (float)num_observations) * p_j; // Apply the soft-thresholding function that is associated with the L1-regularization component of elastic-net regularization float gamma = lambda * alpha; if (p_j > 0.0f && gamma < fabsf(p_j)) { B_j = p_j - gamma; } else if (p_j < 0.0f && gamma < fabsf(p_j)) { B_j = p_j + gamma; } else { B_j = 0.0f; } // Declare and initialize the mean of the square of the predictor column float mean_squared_predictor_value = 0.0f; // Obtain the mean of the square of the predictor column if (transformation_flag == 1 || transformation_flag == 3) { mean_squared_predictor_value = 1.0f; } else if (transformation_flag == 2 || transformation_flag == 4) { mean_squared_predictor_value = 1.0f / (float)num_observations; } // This if-else statemet accounts for the fact that regularization is not applied to the intercept term if one is included if (intercept_flag == 1 && j == 0) { // Use the computed correlation value as the updated predictor coefficient B_j = p_j; } else { // Calculate the updated predictor coefficient value by applying the component of elastic-net regularization that is associated with L2-regularization // The mean_squared_predictor_value term comes from the derivation of the coordinate descent update for a predictor coefficient B_j = B_j / (mean_squared_predictor_value + (lambda * (1.0f - alpha))); } // Store the updated predictor coefficient value into the B_d array B_d[predictor_thread_stride + j] = B_j; // Update the residual values to include the contribution of the current predictor using the updated predictor coefficient value // If the updated predictor coefficient value is 0, then its contribution to the residual values is zero if (B_j != 0.0f) { for (int observation_row = 0; observation_row < num_observations; observation_row++) { // Store the updated residual back into the shared memory array sdata[(observation_row * num_threads_per_block) + block_thread_ind] = sdata[(observation_row * num_threads_per_block) + block_thread_ind] - (X_matrix_d[X_thread_stride + (j * num_observations) + observation_row] * B_j); } } // Compute the weighted (observation weights are all 1 in this case) sum of squares of the changes in the fitted values (this is used for the tolerance convergence criterion) change = (previous_B_j - B_j) * (previous_B_j - B_j); if (transformation_flag == 2 || transformation_flag == 4) { if (intercept_flag == 1 && j > 0) { change = (1.0f / (float)num_observations) * change; } else if (intercept_flag == 0) { change = (1.0f / (float)num_observations) * change; } } if (change > max_change) { max_change = change; } } // Update the global_max_change variable global_max_change = max_change; // Update the iteration count variable iteration_count = iteration_count + 1; } // Account for the fact that the y in the model fit was divided by its standard deviation float std_y = y_std_d[fit_ind]; for (int j = 0; j < num_predictors; j++) { B_d[predictor_thread_stride + j] = B_d[predictor_thread_stride + j] * std_y; } } } } // Define the GPU kernel that performs predictor coefficient unnormalization __global__ void predictor_coefficient_unnormalization(float * B_d, double * B_thread_stride_d, float * model_fit_flag_d, float * X_matrix_d, double * X_matrix_thread_stride_d, float * scaling_factors_d, double * num_predictors_d, float * intercept_flag_d, int num_threads_per_block, int num_blocks, int num_threads_last_block) { // Obtain the index of the block int block_ind = blockIdx.x; // Obtain the thread index within one block int block_thread_ind = threadIdx.x; // Calculate the fit index int fit_ind = (block_ind * num_threads_per_block) + block_thread_ind; // Determine how many threads are in the block (accounts for the fact that the last block may contain less active threads than the other blocks) int num_threads_per_block_2 = num_threads_per_block; if (block_ind == (num_blocks - 1)) { num_threads_per_block_2 = num_threads_last_block; } // This if statement makes sure that extra threads aren't doing data processing if the last block has less fits to perform if (block_thread_ind < num_threads_per_block_2) { // Obtain the flag that determines whether a model fit was performed or not int model_fit_flag = (int)model_fit_flag_d[fit_ind]; // This if statement is to ensure that the coefficients are unnormalized only if a model fit was performed if (model_fit_flag == 1) { // Obtain the thread stride that is used to obtain the correct set of predictors for the fit int predictor_thread_stride = (int)B_thread_stride_d[fit_ind]; // Obtain the number of predictors for the fit int num_predictors = (int)num_predictors_d[fit_ind]; // Obtain the flag that determines whether the first predictor column is a column of ones for the intercept term or not int intercept_flag = (int)intercept_flag_d[fit_ind]; // Declare and initialize the variable that stores the number of the first predictor column to be standardized int start_ind = 0; // This if statement makes sure to not standardize the first predictor column if it corresponds to the intercept term if (intercept_flag == 1) { start_ind = 1; } // Unnormalize the predictor coefficients for (int predictor_column = start_ind; predictor_column < num_predictors; predictor_column++) { B_d[predictor_thread_stride + predictor_column] = B_d[predictor_thread_stride + predictor_column] / scaling_factors_d[predictor_thread_stride + predictor_column]; } } } } // Define the GPU kernel that performs predictor coefficient unstandardization __global__ void predictor_coefficient_unstandardization(float * B_d, double * B_thread_stride_d, float * model_fit_flag_d, float * X_matrix_d, double * X_matrix_thread_stride_d, float * scaling_factors_d, float * mean_X_matrix_d, double * num_predictors_d, float * intercept_flag_d, int num_threads_per_block, int num_blocks, int num_threads_last_block) { // Obtain the index of the block int block_ind = blockIdx.x; // Obtain the thread index within one block int block_thread_ind = threadIdx.x; // Calculate the fit index int fit_ind = (block_ind * num_threads_per_block) + block_thread_ind; // Determine how many threads are in the block (accounts for the fact that the last block may contain less active threads than the other blocks) int num_threads_per_block_2 = num_threads_per_block; if (block_ind == (num_blocks - 1)) { num_threads_per_block_2 = num_threads_last_block; } // This if statement makes sure that extra threads aren't doing data processing if the last block has less fits to perform if (block_thread_ind < num_threads_per_block_2) { // Obtain the flag that determines whether a model fit was performed or not int model_fit_flag = (int)model_fit_flag_d[fit_ind]; // This if statement is to ensure that the coefficients are unstandardized only if a model fit was performed if (model_fit_flag == 1) { // Obtain the thread stride that is used to obtain the correct set of predictors for the fit int predictor_thread_stride = (int)B_thread_stride_d[fit_ind]; // Obtain the number of predictors for the fit int num_predictors = (int)num_predictors_d[fit_ind]; // Obtain the flag that determines whether the first predictor column is a column of ones for the intercept term or not int intercept_flag = (int)intercept_flag_d[fit_ind]; // Declare and initialize the variable that stores the number of the first predictor column to be standardized int start_ind = 0; // This if statement makes sure to not standardize the first predictor column if it corresponds to the intercept term if (intercept_flag == 1) { start_ind = 1; } // Declare and initialize the variable that is used to adjust the intercept term if it is included float sum_value = 0.0f; // Perform predictor coefficient unstandardization for (int predictor_column = start_ind; predictor_column < num_predictors; predictor_column++) { float B_unstandardized = B_d[predictor_thread_stride + predictor_column] / scaling_factors_d[predictor_thread_stride + predictor_column]; B_d[predictor_thread_stride + predictor_column] = B_unstandardized; sum_value = sum_value + (B_unstandardized * mean_X_matrix_d[predictor_thread_stride + predictor_column]); } // Adjust the intercept term if it is included if (intercept_flag == 1) { B_d[predictor_thread_stride] = B_d[predictor_thread_stride] - sum_value; } } } }
.text .file "GPU_kernels_single_precision.hip" .globl _Z38__device_stub__predictor_normalizationPfS_PdS0_S0_S0_S_iii # -- Begin function _Z38__device_stub__predictor_normalizationPfS_PdS0_S0_S0_S_iii .p2align 4, 0x90 .type _Z38__device_stub__predictor_normalizationPfS_PdS0_S0_S0_S_iii,@function _Z38__device_stub__predictor_normalizationPfS_PdS0_S0_S0_S_iii: # @_Z38__device_stub__predictor_normalizationPfS_PdS0_S0_S0_S_iii .cfi_startproc # %bb.0: subq $184, %rsp .cfi_def_cfa_offset 192 movq %rdi, 88(%rsp) movq %rsi, 80(%rsp) movq %rdx, 72(%rsp) movq %rcx, 64(%rsp) movq %r8, 56(%rsp) movq %r9, 48(%rsp) leaq 88(%rsp), %rax movq %rax, 96(%rsp) leaq 80(%rsp), %rax movq %rax, 104(%rsp) leaq 72(%rsp), %rax movq %rax, 112(%rsp) leaq 64(%rsp), %rax movq %rax, 120(%rsp) leaq 56(%rsp), %rax movq %rax, 128(%rsp) leaq 48(%rsp), %rax movq %rax, 136(%rsp) leaq 192(%rsp), %rax movq %rax, 144(%rsp) leaq 200(%rsp), %rax movq %rax, 152(%rsp) leaq 208(%rsp), %rax movq %rax, 160(%rsp) leaq 216(%rsp), %rax movq %rax, 168(%rsp) leaq 32(%rsp), %rdi leaq 16(%rsp), %rsi leaq 8(%rsp), %rdx movq %rsp, %rcx callq __hipPopCallConfiguration movq 32(%rsp), %rsi movl 40(%rsp), %edx movq 16(%rsp), %rcx movl 24(%rsp), %r8d leaq 96(%rsp), %r9 movl $_Z23predictor_normalizationPfS_PdS0_S0_S0_S_iii, %edi pushq (%rsp) .cfi_adjust_cfa_offset 8 pushq 16(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $200, %rsp .cfi_adjust_cfa_offset -200 retq .Lfunc_end0: .size _Z38__device_stub__predictor_normalizationPfS_PdS0_S0_S0_S_iii, .Lfunc_end0-_Z38__device_stub__predictor_normalizationPfS_PdS0_S0_S0_S_iii .cfi_endproc # -- End function .globl _Z40__device_stub__predictor_standardizationPfS_S_PdS0_S0_S0_S_iii # -- Begin function _Z40__device_stub__predictor_standardizationPfS_S_PdS0_S0_S0_S_iii .p2align 4, 0x90 .type _Z40__device_stub__predictor_standardizationPfS_S_PdS0_S0_S0_S_iii,@function _Z40__device_stub__predictor_standardizationPfS_S_PdS0_S0_S0_S_iii: # @_Z40__device_stub__predictor_standardizationPfS_S_PdS0_S0_S0_S_iii .cfi_startproc # %bb.0: subq $184, %rsp .cfi_def_cfa_offset 192 movq %rdi, 88(%rsp) movq %rsi, 80(%rsp) movq %rdx, 72(%rsp) movq %rcx, 64(%rsp) movq %r8, 56(%rsp) movq %r9, 48(%rsp) leaq 88(%rsp), %rax movq %rax, 96(%rsp) leaq 80(%rsp), %rax movq %rax, 104(%rsp) leaq 72(%rsp), %rax movq %rax, 112(%rsp) leaq 64(%rsp), %rax movq %rax, 120(%rsp) leaq 56(%rsp), %rax movq %rax, 128(%rsp) leaq 48(%rsp), %rax movq %rax, 136(%rsp) leaq 192(%rsp), %rax movq %rax, 144(%rsp) leaq 200(%rsp), %rax movq %rax, 152(%rsp) leaq 208(%rsp), %rax movq %rax, 160(%rsp) leaq 216(%rsp), %rax movq %rax, 168(%rsp) leaq 224(%rsp), %rax movq %rax, 176(%rsp) leaq 32(%rsp), %rdi leaq 16(%rsp), %rsi leaq 8(%rsp), %rdx movq %rsp, %rcx callq __hipPopCallConfiguration movq 32(%rsp), %rsi movl 40(%rsp), %edx movq 16(%rsp), %rcx movl 24(%rsp), %r8d leaq 96(%rsp), %r9 movl $_Z25predictor_standardizationPfS_S_PdS0_S0_S0_S_iii, %edi pushq (%rsp) .cfi_adjust_cfa_offset 8 pushq 16(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $200, %rsp .cfi_adjust_cfa_offset -200 retq .Lfunc_end1: .size _Z40__device_stub__predictor_standardizationPfS_S_PdS0_S0_S0_S_iii, .Lfunc_end1-_Z40__device_stub__predictor_standardizationPfS_S_PdS0_S0_S0_S_iii .cfi_endproc # -- End function .globl _Z36__device_stub__model_fit_preparationPfS_S_S_S_PdS0_iii # -- Begin function _Z36__device_stub__model_fit_preparationPfS_S_S_S_PdS0_iii .p2align 4, 0x90 .type _Z36__device_stub__model_fit_preparationPfS_S_S_S_PdS0_iii,@function _Z36__device_stub__model_fit_preparationPfS_S_S_S_PdS0_iii: # @_Z36__device_stub__model_fit_preparationPfS_S_S_S_PdS0_iii .cfi_startproc # %bb.0: subq $184, %rsp .cfi_def_cfa_offset 192 movq %rdi, 88(%rsp) movq %rsi, 80(%rsp) movq %rdx, 72(%rsp) movq %rcx, 64(%rsp) movq %r8, 56(%rsp) movq %r9, 48(%rsp) leaq 88(%rsp), %rax movq %rax, 96(%rsp) leaq 80(%rsp), %rax movq %rax, 104(%rsp) leaq 72(%rsp), %rax movq %rax, 112(%rsp) leaq 64(%rsp), %rax movq %rax, 120(%rsp) leaq 56(%rsp), %rax movq %rax, 128(%rsp) leaq 48(%rsp), %rax movq %rax, 136(%rsp) leaq 192(%rsp), %rax movq %rax, 144(%rsp) leaq 200(%rsp), %rax movq %rax, 152(%rsp) leaq 208(%rsp), %rax movq %rax, 160(%rsp) leaq 216(%rsp), %rax movq %rax, 168(%rsp) leaq 32(%rsp), %rdi leaq 16(%rsp), %rsi leaq 8(%rsp), %rdx movq %rsp, %rcx callq __hipPopCallConfiguration movq 32(%rsp), %rsi movl 40(%rsp), %edx movq 16(%rsp), %rcx movl 24(%rsp), %r8d leaq 96(%rsp), %r9 movl $_Z21model_fit_preparationPfS_S_S_S_PdS0_iii, %edi pushq (%rsp) .cfi_adjust_cfa_offset 8 pushq 16(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $200, %rsp .cfi_adjust_cfa_offset -200 retq .Lfunc_end2: .size _Z36__device_stub__model_fit_preparationPfS_S_S_S_PdS0_iii, .Lfunc_end2-_Z36__device_stub__model_fit_preparationPfS_S_S_S_PdS0_iii .cfi_endproc # -- End function .globl _Z24__device_stub__model_fitPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii # -- Begin function _Z24__device_stub__model_fitPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii .p2align 4, 0x90 .type _Z24__device_stub__model_fitPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii,@function _Z24__device_stub__model_fitPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii: # @_Z24__device_stub__model_fitPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii .cfi_startproc # %bb.0: subq $248, %rsp .cfi_def_cfa_offset 256 movq %rdi, 88(%rsp) movq %rsi, 80(%rsp) movq %rdx, 72(%rsp) movq %rcx, 64(%rsp) movq %r8, 56(%rsp) movq %r9, 48(%rsp) leaq 88(%rsp), %rax movq %rax, 96(%rsp) leaq 80(%rsp), %rax movq %rax, 104(%rsp) leaq 72(%rsp), %rax movq %rax, 112(%rsp) leaq 64(%rsp), %rax movq %rax, 120(%rsp) leaq 56(%rsp), %rax movq %rax, 128(%rsp) leaq 48(%rsp), %rax movq %rax, 136(%rsp) leaq 256(%rsp), %rax movq %rax, 144(%rsp) leaq 264(%rsp), %rax movq %rax, 152(%rsp) leaq 272(%rsp), %rax movq %rax, 160(%rsp) leaq 280(%rsp), %rax movq %rax, 168(%rsp) leaq 288(%rsp), %rax movq %rax, 176(%rsp) leaq 296(%rsp), %rax movq %rax, 184(%rsp) leaq 304(%rsp), %rax movq %rax, 192(%rsp) leaq 312(%rsp), %rax movq %rax, 200(%rsp) leaq 320(%rsp), %rax movq %rax, 208(%rsp) leaq 328(%rsp), %rax movq %rax, 216(%rsp) leaq 336(%rsp), %rax movq %rax, 224(%rsp) leaq 344(%rsp), %rax movq %rax, 232(%rsp) leaq 352(%rsp), %rax movq %rax, 240(%rsp) leaq 32(%rsp), %rdi leaq 16(%rsp), %rsi leaq 8(%rsp), %rdx movq %rsp, %rcx callq __hipPopCallConfiguration movq 32(%rsp), %rsi movl 40(%rsp), %edx movq 16(%rsp), %rcx movl 24(%rsp), %r8d leaq 96(%rsp), %r9 movl $_Z9model_fitPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii, %edi pushq (%rsp) .cfi_adjust_cfa_offset 8 pushq 16(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $264, %rsp # imm = 0x108 .cfi_adjust_cfa_offset -264 retq .Lfunc_end3: .size _Z24__device_stub__model_fitPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii, .Lfunc_end3-_Z24__device_stub__model_fitPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii .cfi_endproc # -- End function .globl _Z38__device_stub__model_fit_shared_memoryPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii # -- Begin function _Z38__device_stub__model_fit_shared_memoryPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii .p2align 4, 0x90 .type _Z38__device_stub__model_fit_shared_memoryPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii,@function _Z38__device_stub__model_fit_shared_memoryPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii: # @_Z38__device_stub__model_fit_shared_memoryPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii .cfi_startproc # %bb.0: subq $248, %rsp .cfi_def_cfa_offset 256 movq %rdi, 88(%rsp) movq %rsi, 80(%rsp) movq %rdx, 72(%rsp) movq %rcx, 64(%rsp) movq %r8, 56(%rsp) movq %r9, 48(%rsp) leaq 88(%rsp), %rax movq %rax, 96(%rsp) leaq 80(%rsp), %rax movq %rax, 104(%rsp) leaq 72(%rsp), %rax movq %rax, 112(%rsp) leaq 64(%rsp), %rax movq %rax, 120(%rsp) leaq 56(%rsp), %rax movq %rax, 128(%rsp) leaq 48(%rsp), %rax movq %rax, 136(%rsp) leaq 256(%rsp), %rax movq %rax, 144(%rsp) leaq 264(%rsp), %rax movq %rax, 152(%rsp) leaq 272(%rsp), %rax movq %rax, 160(%rsp) leaq 280(%rsp), %rax movq %rax, 168(%rsp) leaq 288(%rsp), %rax movq %rax, 176(%rsp) leaq 296(%rsp), %rax movq %rax, 184(%rsp) leaq 304(%rsp), %rax movq %rax, 192(%rsp) leaq 312(%rsp), %rax movq %rax, 200(%rsp) leaq 320(%rsp), %rax movq %rax, 208(%rsp) leaq 328(%rsp), %rax movq %rax, 216(%rsp) leaq 336(%rsp), %rax movq %rax, 224(%rsp) leaq 344(%rsp), %rax movq %rax, 232(%rsp) leaq 352(%rsp), %rax movq %rax, 240(%rsp) leaq 32(%rsp), %rdi leaq 16(%rsp), %rsi leaq 8(%rsp), %rdx movq %rsp, %rcx callq __hipPopCallConfiguration movq 32(%rsp), %rsi movl 40(%rsp), %edx movq 16(%rsp), %rcx movl 24(%rsp), %r8d leaq 96(%rsp), %r9 movl $_Z23model_fit_shared_memoryPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii, %edi pushq (%rsp) .cfi_adjust_cfa_offset 8 pushq 16(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $264, %rsp # imm = 0x108 .cfi_adjust_cfa_offset -264 retq .Lfunc_end4: .size _Z38__device_stub__model_fit_shared_memoryPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii, .Lfunc_end4-_Z38__device_stub__model_fit_shared_memoryPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii .cfi_endproc # -- End function .globl _Z52__device_stub__predictor_coefficient_unnormalizationPfPdS_S_S0_S_S0_S_iii # -- Begin function _Z52__device_stub__predictor_coefficient_unnormalizationPfPdS_S_S0_S_S0_S_iii .p2align 4, 0x90 .type _Z52__device_stub__predictor_coefficient_unnormalizationPfPdS_S_S0_S_S0_S_iii,@function _Z52__device_stub__predictor_coefficient_unnormalizationPfPdS_S_S0_S_S0_S_iii: # @_Z52__device_stub__predictor_coefficient_unnormalizationPfPdS_S_S0_S_S0_S_iii .cfi_startproc # %bb.0: subq $184, %rsp .cfi_def_cfa_offset 192 movq %rdi, 88(%rsp) movq %rsi, 80(%rsp) movq %rdx, 72(%rsp) movq %rcx, 64(%rsp) movq %r8, 56(%rsp) movq %r9, 48(%rsp) leaq 88(%rsp), %rax movq %rax, 96(%rsp) leaq 80(%rsp), %rax movq %rax, 104(%rsp) leaq 72(%rsp), %rax movq %rax, 112(%rsp) leaq 64(%rsp), %rax movq %rax, 120(%rsp) leaq 56(%rsp), %rax movq %rax, 128(%rsp) leaq 48(%rsp), %rax movq %rax, 136(%rsp) leaq 192(%rsp), %rax movq %rax, 144(%rsp) leaq 200(%rsp), %rax movq %rax, 152(%rsp) leaq 208(%rsp), %rax movq %rax, 160(%rsp) leaq 216(%rsp), %rax movq %rax, 168(%rsp) leaq 224(%rsp), %rax movq %rax, 176(%rsp) leaq 32(%rsp), %rdi leaq 16(%rsp), %rsi leaq 8(%rsp), %rdx movq %rsp, %rcx callq __hipPopCallConfiguration movq 32(%rsp), %rsi movl 40(%rsp), %edx movq 16(%rsp), %rcx movl 24(%rsp), %r8d leaq 96(%rsp), %r9 movl $_Z37predictor_coefficient_unnormalizationPfPdS_S_S0_S_S0_S_iii, %edi pushq (%rsp) .cfi_adjust_cfa_offset 8 pushq 16(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $200, %rsp .cfi_adjust_cfa_offset -200 retq .Lfunc_end5: .size _Z52__device_stub__predictor_coefficient_unnormalizationPfPdS_S_S0_S_S0_S_iii, .Lfunc_end5-_Z52__device_stub__predictor_coefficient_unnormalizationPfPdS_S_S0_S_S0_S_iii .cfi_endproc # -- End function .globl _Z54__device_stub__predictor_coefficient_unstandardizationPfPdS_S_S0_S_S_S0_S_iii # -- Begin function _Z54__device_stub__predictor_coefficient_unstandardizationPfPdS_S_S0_S_S_S0_S_iii .p2align 4, 0x90 .type _Z54__device_stub__predictor_coefficient_unstandardizationPfPdS_S_S0_S_S_S0_S_iii,@function _Z54__device_stub__predictor_coefficient_unstandardizationPfPdS_S_S0_S_S_S0_S_iii: # @_Z54__device_stub__predictor_coefficient_unstandardizationPfPdS_S_S0_S_S_S0_S_iii .cfi_startproc # %bb.0: subq $200, %rsp .cfi_def_cfa_offset 208 movq %rdi, 88(%rsp) movq %rsi, 80(%rsp) movq %rdx, 72(%rsp) movq %rcx, 64(%rsp) movq %r8, 56(%rsp) movq %r9, 48(%rsp) leaq 88(%rsp), %rax movq %rax, 96(%rsp) leaq 80(%rsp), %rax movq %rax, 104(%rsp) leaq 72(%rsp), %rax movq %rax, 112(%rsp) leaq 64(%rsp), %rax movq %rax, 120(%rsp) leaq 56(%rsp), %rax movq %rax, 128(%rsp) leaq 48(%rsp), %rax movq %rax, 136(%rsp) leaq 208(%rsp), %rax movq %rax, 144(%rsp) leaq 216(%rsp), %rax movq %rax, 152(%rsp) leaq 224(%rsp), %rax movq %rax, 160(%rsp) leaq 232(%rsp), %rax movq %rax, 168(%rsp) leaq 240(%rsp), %rax movq %rax, 176(%rsp) leaq 248(%rsp), %rax movq %rax, 184(%rsp) leaq 32(%rsp), %rdi leaq 16(%rsp), %rsi leaq 8(%rsp), %rdx movq %rsp, %rcx callq __hipPopCallConfiguration movq 32(%rsp), %rsi movl 40(%rsp), %edx movq 16(%rsp), %rcx movl 24(%rsp), %r8d leaq 96(%rsp), %r9 movl $_Z39predictor_coefficient_unstandardizationPfPdS_S_S0_S_S_S0_S_iii, %edi pushq (%rsp) .cfi_adjust_cfa_offset 8 pushq 16(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $216, %rsp .cfi_adjust_cfa_offset -216 retq .Lfunc_end6: .size _Z54__device_stub__predictor_coefficient_unstandardizationPfPdS_S_S0_S_S_S0_S_iii, .Lfunc_end6-_Z54__device_stub__predictor_coefficient_unstandardizationPfPdS_S_S0_S_S_S0_S_iii .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_ctor .type __hip_module_ctor,@function __hip_module_ctor: # @__hip_module_ctor .cfi_startproc # %bb.0: pushq %rbx .cfi_def_cfa_offset 16 subq $32, %rsp .cfi_def_cfa_offset 48 .cfi_offset %rbx, -16 cmpq $0, __hip_gpubin_handle(%rip) jne .LBB7_2 # %bb.1: movl $__hip_fatbin_wrapper, %edi callq __hipRegisterFatBinary movq %rax, __hip_gpubin_handle(%rip) .LBB7_2: movq __hip_gpubin_handle(%rip), %rbx xorps %xmm0, %xmm0 movups %xmm0, 16(%rsp) movups %xmm0, (%rsp) movl $_Z23predictor_normalizationPfS_PdS0_S0_S0_S_iii, %esi movl $.L__unnamed_1, %edx movl $.L__unnamed_1, %ecx movq %rbx, %rdi movl $-1, %r8d xorl %r9d, %r9d callq __hipRegisterFunction xorps %xmm0, %xmm0 movups %xmm0, 16(%rsp) movups %xmm0, (%rsp) movl $_Z25predictor_standardizationPfS_S_PdS0_S0_S0_S_iii, %esi movl $.L__unnamed_2, %edx movl $.L__unnamed_2, %ecx movq %rbx, %rdi movl $-1, %r8d xorl %r9d, %r9d callq __hipRegisterFunction xorps %xmm0, %xmm0 movups %xmm0, 16(%rsp) movups %xmm0, (%rsp) movl $_Z21model_fit_preparationPfS_S_S_S_PdS0_iii, %esi movl $.L__unnamed_3, %edx movl $.L__unnamed_3, %ecx movq %rbx, %rdi movl $-1, %r8d xorl %r9d, %r9d callq __hipRegisterFunction xorps %xmm0, %xmm0 movups %xmm0, 16(%rsp) movups %xmm0, (%rsp) movl $_Z9model_fitPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii, %esi movl $.L__unnamed_4, %edx movl $.L__unnamed_4, %ecx movq %rbx, %rdi movl $-1, %r8d xorl %r9d, %r9d callq __hipRegisterFunction xorps %xmm0, %xmm0 movups %xmm0, 16(%rsp) movups %xmm0, (%rsp) movl $_Z23model_fit_shared_memoryPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii, %esi movl $.L__unnamed_5, %edx movl $.L__unnamed_5, %ecx movq %rbx, %rdi movl $-1, %r8d xorl %r9d, %r9d callq __hipRegisterFunction xorps %xmm0, %xmm0 movups %xmm0, 16(%rsp) movups %xmm0, (%rsp) movl $_Z37predictor_coefficient_unnormalizationPfPdS_S_S0_S_S0_S_iii, %esi movl $.L__unnamed_6, %edx movl $.L__unnamed_6, %ecx movq %rbx, %rdi movl $-1, %r8d xorl %r9d, %r9d callq __hipRegisterFunction xorps %xmm0, %xmm0 movups %xmm0, 16(%rsp) movups %xmm0, (%rsp) movl $_Z39predictor_coefficient_unstandardizationPfPdS_S_S0_S_S_S0_S_iii, %esi movl $.L__unnamed_7, %edx movl $.L__unnamed_7, %ecx movq %rbx, %rdi movl $-1, %r8d xorl %r9d, %r9d callq __hipRegisterFunction movl $__hip_module_dtor, %edi addq $32, %rsp .cfi_def_cfa_offset 16 popq %rbx .cfi_def_cfa_offset 8 jmp atexit # TAILCALL .Lfunc_end7: .size __hip_module_ctor, .Lfunc_end7-__hip_module_ctor .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_dtor .type __hip_module_dtor,@function __hip_module_dtor: # @__hip_module_dtor .cfi_startproc # %bb.0: movq __hip_gpubin_handle(%rip), %rdi testq %rdi, %rdi je .LBB8_2 # %bb.1: pushq %rax .cfi_def_cfa_offset 16 callq __hipUnregisterFatBinary movq $0, __hip_gpubin_handle(%rip) addq $8, %rsp .cfi_def_cfa_offset 8 .LBB8_2: retq .Lfunc_end8: .size __hip_module_dtor, .Lfunc_end8-__hip_module_dtor .cfi_endproc # -- End function .type _Z23predictor_normalizationPfS_PdS0_S0_S0_S_iii,@object # @_Z23predictor_normalizationPfS_PdS0_S0_S0_S_iii .section .rodata,"a",@progbits .globl _Z23predictor_normalizationPfS_PdS0_S0_S0_S_iii .p2align 3, 0x0 _Z23predictor_normalizationPfS_PdS0_S0_S0_S_iii: .quad _Z38__device_stub__predictor_normalizationPfS_PdS0_S0_S0_S_iii .size _Z23predictor_normalizationPfS_PdS0_S0_S0_S_iii, 8 .type _Z25predictor_standardizationPfS_S_PdS0_S0_S0_S_iii,@object # @_Z25predictor_standardizationPfS_S_PdS0_S0_S0_S_iii .globl _Z25predictor_standardizationPfS_S_PdS0_S0_S0_S_iii .p2align 3, 0x0 _Z25predictor_standardizationPfS_S_PdS0_S0_S0_S_iii: .quad _Z40__device_stub__predictor_standardizationPfS_S_PdS0_S0_S0_S_iii .size _Z25predictor_standardizationPfS_S_PdS0_S0_S0_S_iii, 8 .type _Z21model_fit_preparationPfS_S_S_S_PdS0_iii,@object # @_Z21model_fit_preparationPfS_S_S_S_PdS0_iii .globl _Z21model_fit_preparationPfS_S_S_S_PdS0_iii .p2align 3, 0x0 _Z21model_fit_preparationPfS_S_S_S_PdS0_iii: .quad _Z36__device_stub__model_fit_preparationPfS_S_S_S_PdS0_iii .size _Z21model_fit_preparationPfS_S_S_S_PdS0_iii, 8 .type _Z9model_fitPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii,@object # @_Z9model_fitPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii .globl _Z9model_fitPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii .p2align 3, 0x0 _Z9model_fitPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii: .quad _Z24__device_stub__model_fitPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii .size _Z9model_fitPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii, 8 .type _Z23model_fit_shared_memoryPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii,@object # @_Z23model_fit_shared_memoryPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii .globl _Z23model_fit_shared_memoryPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii .p2align 3, 0x0 _Z23model_fit_shared_memoryPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii: .quad _Z38__device_stub__model_fit_shared_memoryPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii .size _Z23model_fit_shared_memoryPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii, 8 .type _Z37predictor_coefficient_unnormalizationPfPdS_S_S0_S_S0_S_iii,@object # @_Z37predictor_coefficient_unnormalizationPfPdS_S_S0_S_S0_S_iii .globl _Z37predictor_coefficient_unnormalizationPfPdS_S_S0_S_S0_S_iii .p2align 3, 0x0 _Z37predictor_coefficient_unnormalizationPfPdS_S_S0_S_S0_S_iii: .quad _Z52__device_stub__predictor_coefficient_unnormalizationPfPdS_S_S0_S_S0_S_iii .size _Z37predictor_coefficient_unnormalizationPfPdS_S_S0_S_S0_S_iii, 8 .type _Z39predictor_coefficient_unstandardizationPfPdS_S_S0_S_S_S0_S_iii,@object # @_Z39predictor_coefficient_unstandardizationPfPdS_S_S0_S_S_S0_S_iii .globl _Z39predictor_coefficient_unstandardizationPfPdS_S_S0_S_S_S0_S_iii .p2align 3, 0x0 _Z39predictor_coefficient_unstandardizationPfPdS_S_S0_S_S_S0_S_iii: .quad _Z54__device_stub__predictor_coefficient_unstandardizationPfPdS_S_S0_S_S_S0_S_iii .size _Z39predictor_coefficient_unstandardizationPfPdS_S_S0_S_S_S0_S_iii, 8 .type .L__unnamed_1,@object # @0 .section .rodata.str1.1,"aMS",@progbits,1 .L__unnamed_1: .asciz "_Z23predictor_normalizationPfS_PdS0_S0_S0_S_iii" .size .L__unnamed_1, 48 .type .L__unnamed_2,@object # @1 .L__unnamed_2: .asciz "_Z25predictor_standardizationPfS_S_PdS0_S0_S0_S_iii" .size .L__unnamed_2, 52 .type .L__unnamed_3,@object # @2 .L__unnamed_3: .asciz "_Z21model_fit_preparationPfS_S_S_S_PdS0_iii" .size .L__unnamed_3, 44 .type .L__unnamed_4,@object # @3 .L__unnamed_4: .asciz "_Z9model_fitPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii" .size .L__unnamed_4, 51 .type .L__unnamed_5,@object # @4 .L__unnamed_5: .asciz "_Z23model_fit_shared_memoryPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii" .size .L__unnamed_5, 66 .type .L__unnamed_6,@object # @5 .L__unnamed_6: .asciz "_Z37predictor_coefficient_unnormalizationPfPdS_S_S0_S_S0_S_iii" .size .L__unnamed_6, 63 .type .L__unnamed_7,@object # @6 .L__unnamed_7: .asciz "_Z39predictor_coefficient_unstandardizationPfPdS_S_S0_S_S_S0_S_iii" .size .L__unnamed_7, 67 .type __hip_fatbin_wrapper,@object # @__hip_fatbin_wrapper .section .hipFatBinSegment,"a",@progbits .p2align 3, 0x0 __hip_fatbin_wrapper: .long 1212764230 # 0x48495046 .long 1 # 0x1 .quad __hip_fatbin .quad 0 .size __hip_fatbin_wrapper, 24 .type __hip_gpubin_handle,@object # @__hip_gpubin_handle .local __hip_gpubin_handle .comm __hip_gpubin_handle,8,8 .section .init_array,"aw",@init_array .p2align 3, 0x0 .quad __hip_module_ctor .type __hip_cuid_,@object # @__hip_cuid_ .bss .globl __hip_cuid_ __hip_cuid_: .byte 0 # 0x0 .size __hip_cuid_, 1 .section ".linker-options","e",@llvm_linker_options .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym _Z38__device_stub__predictor_normalizationPfS_PdS0_S0_S0_S_iii .addrsig_sym _Z40__device_stub__predictor_standardizationPfS_S_PdS0_S0_S0_S_iii .addrsig_sym _Z36__device_stub__model_fit_preparationPfS_S_S_S_PdS0_iii .addrsig_sym _Z24__device_stub__model_fitPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii .addrsig_sym _Z38__device_stub__model_fit_shared_memoryPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii .addrsig_sym _Z52__device_stub__predictor_coefficient_unnormalizationPfPdS_S_S0_S_S0_S_iii .addrsig_sym _Z54__device_stub__predictor_coefficient_unstandardizationPfPdS_S_S0_S_S_S0_S_iii .addrsig_sym __hip_module_ctor .addrsig_sym __hip_module_dtor .addrsig_sym _Z23predictor_normalizationPfS_PdS0_S0_S0_S_iii .addrsig_sym _Z25predictor_standardizationPfS_S_PdS0_S0_S0_S_iii .addrsig_sym _Z21model_fit_preparationPfS_S_S_S_PdS0_iii .addrsig_sym _Z9model_fitPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii .addrsig_sym _Z23model_fit_shared_memoryPfPdS_S_S0_S0_S_S_S_S0_S0_S_S_S_S_iiii .addrsig_sym _Z37predictor_coefficient_unnormalizationPfPdS_S_S0_S_S0_S_iii .addrsig_sym _Z39predictor_coefficient_unstandardizationPfPdS_S_S0_S_S_S0_S_iii .addrsig_sym __hip_fatbin .addrsig_sym __hip_fatbin_wrapper .addrsig_sym __hip_cuid_
You are a helpful coding assistant specialized in converting CUDA source code to CUDA device assembly.
#include "includes.h" __global__ void MatrixMultiplication__CudaKernel(int* in_tabA, int* in_tabB, int* out_tabC, int outTabWidth) { int row = blockIdx.x * blockDim.x + threadIdx.x; int col = blockIdx.y * blockDim.y + threadIdx.y; //making sure that extra threads will do not any work if (row < outTabWidth && col < outTabWidth) { int tmp_sum = 0; //#pragma unroll for (int i = 0; i < outTabWidth; i++) { tmp_sum += in_tabA[row * outTabWidth + i] * in_tabB[i * outTabWidth + col]; } out_tabC[row * outTabWidth + col] = tmp_sum; } }
code for sm_80 Function : _Z32MatrixMultiplication__CudaKernelPiS_S_i .headerflags @"EF_CUDA_TEXMODE_UNIFIED EF_CUDA_64BIT_ADDRESS EF_CUDA_SM80 EF_CUDA_VIRTUAL_SM(EF_CUDA_SM80)" /*0000*/ MOV R1, c[0x0][0x28] ; /* 0x00000a0000017a02 */ /* 0x000fe40000000f00 */ /*0010*/ S2R R0, SR_CTAID.Y ; /* 0x0000000000007919 */ /* 0x000e280000002600 */ /*0020*/ S2R R5, SR_TID.Y ; /* 0x0000000000057919 */ /* 0x000e280000002200 */ /*0030*/ S2R R3, SR_CTAID.X ; /* 0x0000000000037919 */ /* 0x000e680000002500 */ /*0040*/ S2R R2, SR_TID.X ; /* 0x0000000000027919 */ /* 0x000e620000002100 */ /*0050*/ IMAD R0, R0, c[0x0][0x4], R5 ; /* 0x0000010000007a24 */ /* 0x001fca00078e0205 */ /*0060*/ ISETP.GE.AND P0, PT, R0, c[0x0][0x178], PT ; /* 0x00005e0000007a0c */ /* 0x000fe20003f06270 */ /*0070*/ IMAD R3, R3, c[0x0][0x0], R2 ; /* 0x0000000003037a24 */ /* 0x002fca00078e0202 */ /*0080*/ ISETP.GE.OR P0, PT, R3, c[0x0][0x178], P0 ; /* 0x00005e0003007a0c */ /* 0x000fda0000706670 */ /*0090*/ @P0 EXIT ; /* 0x000000000000094d */ /* 0x000fea0003800000 */ /*00a0*/ MOV R2, c[0x0][0x178] ; /* 0x00005e0000027a02 */ /* 0x000fe20000000f00 */ /*00b0*/ ULDC.64 UR4, c[0x0][0x118] ; /* 0x0000460000047ab9 */ /* 0x000fe20000000a00 */ /*00c0*/ HFMA2.MMA R28, -RZ, RZ, 0, 0 ; /* 0x00000000ff1c7435 */ /* 0x000fe200000001ff */ /*00d0*/ IMAD R3, R3, c[0x0][0x178], RZ ; /* 0x00005e0003037a24 */ /* 0x000fe200078e02ff */ /*00e0*/ ISETP.GE.AND P0, PT, R2, 0x1, PT ; /* 0x000000010200780c */ /* 0x000fda0003f06270 */ /*00f0*/ @!P0 BRA 0xbf0 ; /* 0x00000af000008947 */ /* 0x000fea0003800000 */ /*0100*/ IADD3 R4, R2.reuse, -0x1, RZ ; /* 0xffffffff02047810 */ /* 0x040fe40007ffe0ff */ /*0110*/ LOP3.LUT R5, R2, 0x3, RZ, 0xc0, !PT ; /* 0x0000000302057812 */ /* 0x000fe400078ec0ff */ /*0120*/ ISETP.GE.U32.AND P0, PT, R4, 0x3, PT ; /* 0x000000030400780c */ /* 0x000fe40003f06070 */ /*0130*/ MOV R4, RZ ; /* 0x000000ff00047202 */ /* 0x000fe40000000f00 */ /*0140*/ MOV R28, RZ ; /* 0x000000ff001c7202 */ /* 0x000fd20000000f00 */ /*0150*/ @!P0 BRA 0xaf0 ; /* 0x0000099000008947 */ /* 0x000fea0003800000 */ /*0160*/ IADD3 R6, -R5, c[0x0][0x178], RZ ; /* 0x00005e0005067a10 */ /* 0x000fe20007ffe1ff */ /*0170*/ HFMA2.MMA R25, -RZ, RZ, 0, 2.384185791015625e-07 ; /* 0x00000004ff197435 */ /* 0x000fe200000001ff */ /*0180*/ ULDC.64 UR6, c[0x0][0x160] ; /* 0x0000580000067ab9 */ /* 0x000fe20000000a00 */ /*0190*/ MOV R4, RZ ; /* 0x000000ff00047202 */ /* 0x000fe40000000f00 */ /*01a0*/ ISETP.GT.AND P0, PT, R6, RZ, PT ; /* 0x000000ff0600720c */ /* 0x000fcc0003f04270 */ /*01b0*/ IMAD.WIDE R24, R0, R25, c[0x0][0x168] ; /* 0x00005a0000187625 */ /* 0x000fce00078e0219 */ /*01c0*/ @!P0 BRA 0x960 ; /* 0x0000079000008947 */ /* 0x000fea0003800000 */ /*01d0*/ ISETP.GT.AND P1, PT, R6, 0xc, PT ; /* 0x0000000c0600780c */ /* 0x000fe40003f24270 */ /*01e0*/ PLOP3.LUT P0, PT, PT, PT, PT, 0x80, 0x0 ; /* 0x000000000000781c */ /* 0x000fd60003f0f070 */ /*01f0*/ @!P1 BRA 0x6a0 ; /* 0x000004a000009947 */ /* 0x000fea0003800000 */ /*0200*/ PLOP3.LUT P0, PT, PT, PT, PT, 0x8, 0x0 ; /* 0x000000000000781c */ /* 0x000fe40003f0e170 */ /*0210*/ MOV R12, UR6 ; /* 0x00000006000c7c02 */ /* 0x000fe20008000f00 */ /*0220*/ LDG.E R29, [R24.64] ; /* 0x00000004181d7981 */ /* 0x0000a2000c1e1900 */ /*0230*/ MOV R13, UR7 ; /* 0x00000007000d7c02 */ /* 0x000fca0008000f00 */ /*0240*/ IMAD.WIDE R12, R3, 0x4, R12 ; /* 0x00000004030c7825 */ /* 0x000fca00078e020c */ /*0250*/ LDG.E R27, [R12.64] ; /* 0x000000040c1b7981 */ /* 0x000ea2000c1e1900 */ /*0260*/ IMAD.WIDE R10, R2, 0x4, R24 ; /* 0x00000004020a7825 */ /* 0x000fc600078e0218 */ /*0270*/ LDG.E R17, [R12.64+0x4] ; /* 0x000004040c117981 */ /* 0x000ee6000c1e1900 */ /*0280*/ IMAD.WIDE R18, R2.reuse, 0x4, R10 ; /* 0x0000000402127825 */ /* 0x040fe200078e020a */ /*0290*/ LDG.E R16, [R10.64] ; /* 0x000000040a107981 */ /* 0x0002e8000c1e1900 */ /*02a0*/ LDG.E R7, [R12.64+0xc] ; /* 0x00000c040c077981 */ /* 0x000f22000c1e1900 */ /*02b0*/ IMAD.WIDE R14, R2, 0x4, R18 ; /* 0x00000004020e7825 */ /* 0x000fc600078e0212 */ /*02c0*/ LDG.E R18, [R18.64] ; /* 0x0000000412127981 */ /* 0x000b26000c1e1900 */ /*02d0*/ IMAD.WIDE R20, R2.reuse, 0x4, R14 ; /* 0x0000000402147825 */ /* 0x040fe200078e020e */ /*02e0*/ LDG.E R26, [R14.64] ; /* 0x000000040e1a7981 */ /* 0x000128000c1e1900 */ /*02f0*/ LDG.E R9, [R12.64+0x10] ; /* 0x000010040c097981 */ /* 0x000f28000c1e1900 */ /*0300*/ LDG.E R19, [R12.64+0x8] ; /* 0x000008040c137981 */ /* 0x020f22000c1e1900 */ /*0310*/ IMAD.WIDE R14, R2, 0x4, R20 ; /* 0x00000004020e7825 */ /* 0x001fc600078e0214 */ /*0320*/ LDG.E R20, [R20.64] ; /* 0x0000000414147981 */ /* 0x000166000c1e1900 */ /*0330*/ IMAD.WIDE R22, R2.reuse, 0x4, R14 ; /* 0x0000000402167825 */ /* 0x040fe200078e020e */ /*0340*/ LDG.E R8, [R14.64] ; /* 0x000000040e087981 */ /* 0x000168000c1e1900 */ /*0350*/ LDG.E R11, [R12.64+0x14] ; /* 0x000014040c0b7981 */ /* 0x002f62000c1e1900 */ /*0360*/ IMAD.WIDE R24, R2, 0x4, R22 ; /* 0x0000000402187825 */ /* 0x000fc600078e0216 */ /*0370*/ LDG.E R10, [R22.64] ; /* 0x00000004160a7981 */ /* 0x000368000c1e1900 */ /*0380*/ LDG.E R21, [R12.64+0x18] ; /* 0x000018040c157981 */ /* 0x001f62000c1e1900 */ /*0390*/ IMAD R29, R29, R27, R28 ; /* 0x0000001b1d1d7224 */ /* 0x004fc600078e021c */ /*03a0*/ LDG.E R27, [R12.64+0x1c] ; /* 0x00001c040c1b7981 */ /* 0x000ea8000c1e1900 */ /*03b0*/ LDG.E R28, [R24.64] ; /* 0x00000004181c7981 */ /* 0x0000a2000c1e1900 */ /*03c0*/ IMAD.WIDE R14, R2, 0x4, R24 ; /* 0x00000004020e7825 */ /* 0x000fc800078e0218 */ /*03d0*/ IMAD R29, R16, R17, R29 ; /* 0x00000011101d7224 */ /* 0x008fe400078e021d */ /*03e0*/ IMAD.WIDE R16, R2, 0x4, R14 ; /* 0x0000000402107825 */ /* 0x000fe400078e020e */ /*03f0*/ LDG.E R14, [R14.64] ; /* 0x000000040e0e7981 */ /* 0x0006a4000c1e1900 */ /*0400*/ IMAD R29, R18, R19, R29 ; /* 0x00000013121d7224 */ /* 0x010fe400078e021d */ /*0410*/ IMAD.WIDE R18, R2, 0x4, R16 ; /* 0x0000000402127825 */ /* 0x000fe400078e0210 */ /*0420*/ LDG.E R16, [R16.64] ; /* 0x0000000410107981 */ /* 0x0008a4000c1e1900 */ /*0430*/ IMAD R26, R26, R7, R29 ; /* 0x000000071a1a7224 */ /* 0x000fc400078e021d */ /*0440*/ IMAD.WIDE R22, R2.reuse, 0x4, R18 ; /* 0x0000000402167825 */ /* 0x042fe200078e0212 */ /*0450*/ LDG.E R7, [R12.64+0x20] ; /* 0x000020040c077981 */ /* 0x000ea8000c1e1900 */ /*0460*/ LDG.E R29, [R12.64+0x24] ; /* 0x000024040c1d7981 */ /* 0x000ea2000c1e1900 */ /*0470*/ IMAD.WIDE R24, R2, 0x4, R22 ; /* 0x0000000402187825 */ /* 0x001fc600078e0216 */ /*0480*/ LDG.E R18, [R18.64] ; /* 0x0000000412127981 */ /* 0x0000a2000c1e1900 */ /*0490*/ IMAD R9, R20, R9, R26 ; /* 0x0000000914097224 */ /* 0x020fc600078e021a */ /*04a0*/ LDG.E R26, [R12.64+0x28] ; /* 0x000028040c1a7981 */ /* 0x000f62000c1e1900 */ /*04b0*/ IMAD R11, R8, R11, R9 ; /* 0x0000000b080b7224 */ /* 0x000fe400078e0209 */ /*04c0*/ IMAD.WIDE R8, R2, 0x4, R24 ; /* 0x0000000402087825 */ /* 0x000fe200078e0218 */ /*04d0*/ LDG.E R22, [R22.64] ; /* 0x0000000416167981 */ /* 0x000368000c1e1900 */ /*04e0*/ LDG.E R17, [R12.64+0x2c] ; /* 0x00002c040c117981 */ /* 0x010f22000c1e1900 */ /*04f0*/ IMAD R21, R10, R21, R11 ; /* 0x000000150a157224 */ /* 0x000fc600078e020b */ /*0500*/ LDG.E R15, [R24.64] ; /* 0x00000004180f7981 */ /* 0x008722000c1e1900 */ /*0510*/ IMAD.WIDE R10, R2, 0x4, R8 ; /* 0x00000004020a7825 */ /* 0x000fc600078e0208 */ /*0520*/ LDG.E R19, [R8.64] ; /* 0x0000000408137981 */ /* 0x001128000c1e1900 */ /*0530*/ LDG.E R23, [R10.64] ; /* 0x000000040a177981 */ /* 0x002f28000c1e1900 */ /*0540*/ LDG.E R24, [R12.64+0x30] ; /* 0x000030040c187981 */ /* 0x008ee8000c1e1900 */ /*0550*/ LDG.E R25, [R12.64+0x38] ; /* 0x000038040c197981 */ /* 0x000ee8000c1e1900 */ /*0560*/ LDG.E R8, [R12.64+0x3c] ; /* 0x00003c040c087981 */ /* 0x001ee2000c1e1900 */ /*0570*/ IMAD R9, R28, R27, R21 ; /* 0x0000001b1c097224 */ /* 0x004fc600078e0215 */ /*0580*/ LDG.E R28, [R12.64+0x34] ; /* 0x000034040c1c7981 */ /* 0x000ea2000c1e1900 */ /*0590*/ IMAD.WIDE R20, R2, 0x4, R10 ; /* 0x0000000402147825 */ /* 0x000fca00078e020a */ /*05a0*/ LDG.E R27, [R20.64] ; /* 0x00000004141b7981 */ /* 0x000ea2000c1e1900 */ /*05b0*/ IADD3 R6, R6, -0x10, RZ ; /* 0xfffffff006067810 */ /* 0x000fc80007ffe0ff */ /*05c0*/ ISETP.GT.AND P1, PT, R6, 0xc, PT ; /* 0x0000000c0600780c */ /* 0x000fe20003f24270 */ /*05d0*/ IMAD R7, R14, R7, R9 ; /* 0x000000070e077224 */ /* 0x000fc800078e0209 */ /*05e0*/ IMAD R7, R16, R29, R7 ; /* 0x0000001d10077224 */ /* 0x000fc800078e0207 */ /*05f0*/ IMAD R7, R18, R26, R7 ; /* 0x0000001a12077224 */ /* 0x020fc800078e0207 */ /*0600*/ IMAD R7, R22, R17, R7 ; /* 0x0000001116077224 */ /* 0x010fe200078e0207 */ /*0610*/ UIADD3 UR6, UP0, UR6, 0x40, URZ ; /* 0x0000004006067890 */ /* 0x000fe2000ff1e03f */ /*0620*/ IADD3 R4, R4, 0x10, RZ ; /* 0x0000001004047810 */ /* 0x000fc60007ffe0ff */ /*0630*/ UIADD3.X UR7, URZ, UR7, URZ, UP0, !UPT ; /* 0x000000073f077290 */ /* 0x000fe200087fe43f */ /*0640*/ IMAD R7, R15, R24, R7 ; /* 0x000000180f077224 */ /* 0x008fc800078e0207 */ /*0650*/ IMAD R28, R19, R28, R7 ; /* 0x0000001c131c7224 */ /* 0x004fc800078e0207 */ /*0660*/ IMAD R28, R23, R25, R28 ; /* 0x00000019171c7224 */ /* 0x000fe400078e021c */ /*0670*/ IMAD.WIDE R24, R2, 0x4, R20 ; /* 0x0000000402187825 */ /* 0x000fc800078e0214 */ /*0680*/ IMAD R28, R27, R8, R28 ; /* 0x000000081b1c7224 */ /* 0x000fe200078e021c */ /*0690*/ @P1 BRA 0x210 ; /* 0xfffffb7000001947 */ /* 0x000fea000383ffff */ /*06a0*/ ISETP.GT.AND P1, PT, R6, 0x4, PT ; /* 0x000000040600780c */ /* 0x000fda0003f24270 */ /*06b0*/ @!P1 BRA 0x940 ; /* 0x0000028000009947 */ /* 0x000fea0003800000 */ /*06c0*/ IMAD.WIDE R16, R2, 0x4, R24 ; /* 0x0000000402107825 */ /* 0x000fe200078e0218 */ /*06d0*/ MOV R8, UR6 ; /* 0x0000000600087c02 */ /* 0x000fe20008000f00 */ /*06e0*/ LDG.E R7, [R24.64] ; /* 0x0000000418077981 */ /* 0x0000a2000c1e1900 */ /*06f0*/ MOV R9, UR7 ; /* 0x0000000700097c02 */ /* 0x000fc60008000f00 */ /*0700*/ IMAD.WIDE R12, R2.reuse, 0x4, R16 ; /* 0x00000004020c7825 */ /* 0x040fe200078e0210 */ /*0710*/ LDG.E R21, [R16.64] ; /* 0x0000000410157981 */ /* 0x0002e6000c1e1900 */ /*0720*/ IMAD.WIDE R8, R3, 0x4, R8 ; /* 0x0000000403087825 */ /* 0x000fe200078e0208 */ /*0730*/ LDG.E R23, [R12.64] ; /* 0x000000040c177981 */ /* 0x000966000c1e1900 */ /*0740*/ IMAD.WIDE R14, R2.reuse, 0x4, R12 ; /* 0x00000004020e7825 */ /* 0x040fe200078e020c */ /*0750*/ LDG.E R20, [R8.64] ; /* 0x0000000408147981 */ /* 0x000ea8000c1e1900 */ /*0760*/ LDG.E R22, [R8.64+0x4] ; /* 0x0000040408167981 */ /* 0x000ee2000c1e1900 */ /*0770*/ IMAD.WIDE R10, R2, 0x4, R14 ; /* 0x00000004020a7825 */ /* 0x000fc600078e020e */ /*0780*/ LDG.E R26, [R8.64+0x8] ; /* 0x00000804081a7981 */ /* 0x000f66000c1e1900 */ /*0790*/ IMAD.WIDE R16, R2.reuse, 0x4, R10 ; /* 0x0000000402107825 */ /* 0x042fe200078e020a */ /*07a0*/ LDG.E R14, [R14.64] ; /* 0x000000040e0e7981 */ /* 0x000368000c1e1900 */ /*07b0*/ LDG.E R27, [R8.64+0xc] ; /* 0x00000c04081b7981 */ /* 0x000f62000c1e1900 */ /*07c0*/ IMAD.WIDE R18, R2, 0x4, R16 ; /* 0x0000000402127825 */ /* 0x000fc600078e0210 */ /*07d0*/ LDG.E R10, [R10.64] ; /* 0x000000040a0a7981 */ /* 0x000368000c1e1900 */ /*07e0*/ LDG.E R25, [R8.64+0x10] ; /* 0x0000100408197981 */ /* 0x001f62000c1e1900 */ /*07f0*/ IMAD.WIDE R12, R2, 0x4, R18 ; /* 0x00000004020c7825 */ /* 0x010fc600078e0212 */ /*0800*/ LDG.E R16, [R16.64] ; /* 0x0000000410107981 */ /* 0x000f28000c1e1900 */ /*0810*/ LDG.E R29, [R8.64+0x14] ; /* 0x00001404081d7981 */ /* 0x000f28000c1e1900 */ /*0820*/ LDG.E R24, [R18.64] ; /* 0x0000000412187981 */ /* 0x000128000c1e1900 */ /*0830*/ LDG.E R11, [R8.64+0x18] ; /* 0x00001804080b7981 */ /* 0x002f28000c1e1900 */ /*0840*/ LDG.E R15, [R12.64] ; /* 0x000000040c0f7981 */ /* 0x000f28000c1e1900 */ /*0850*/ LDG.E R18, [R8.64+0x1c] ; /* 0x00001c0408127981 */ /* 0x001f22000c1e1900 */ /*0860*/ UIADD3 UR6, UP0, UR6, 0x20, URZ ; /* 0x0000002006067890 */ /* 0x000fe2000ff1e03f */ /*0870*/ PLOP3.LUT P0, PT, PT, PT, PT, 0x8, 0x0 ; /* 0x000000000000781c */ /* 0x000fc40003f0e170 */ /*0880*/ IADD3 R4, R4, 0x8, RZ ; /* 0x0000000804047810 */ /* 0x000fe40007ffe0ff */ /*0890*/ IADD3 R6, R6, -0x8, RZ ; /* 0xfffffff806067810 */ /* 0x000fe20007ffe0ff */ /*08a0*/ UIADD3.X UR7, URZ, UR7, URZ, UP0, !UPT ; /* 0x000000073f077290 */ /* 0x000fe200087fe43f */ /*08b0*/ IMAD R7, R7, R20, R28 ; /* 0x0000001407077224 */ /* 0x004fc800078e021c */ /*08c0*/ IMAD R7, R21, R22, R7 ; /* 0x0000001615077224 */ /* 0x008fc800078e0207 */ /*08d0*/ IMAD R7, R23, R26, R7 ; /* 0x0000001a17077224 */ /* 0x020fc800078e0207 */ /*08e0*/ IMAD R7, R14, R27, R7 ; /* 0x0000001b0e077224 */ /* 0x000fc800078e0207 */ /*08f0*/ IMAD R7, R10, R25, R7 ; /* 0x000000190a077224 */ /* 0x000fc800078e0207 */ /*0900*/ IMAD R7, R16, R29, R7 ; /* 0x0000001d10077224 */ /* 0x010fc800078e0207 */ /*0910*/ IMAD R7, R24, R11, R7 ; /* 0x0000000b18077224 */ /* 0x000fe400078e0207 */ /*0920*/ IMAD.WIDE R24, R2, 0x4, R12 ; /* 0x0000000402187825 */ /* 0x000fc800078e020c */ /*0930*/ IMAD R28, R15, R18, R7 ; /* 0x000000120f1c7224 */ /* 0x000fe400078e0207 */ /*0940*/ ISETP.NE.OR P0, PT, R6, RZ, P0 ; /* 0x000000ff0600720c */ /* 0x000fda0000705670 */ /*0950*/ @!P0 BRA 0xaf0 ; /* 0x0000019000008947 */ /* 0x000fea0003800000 */ /*0960*/ MOV R8, UR6 ; /* 0x0000000600087c02 */ /* 0x000fe20008000f00 */ /*0970*/ IMAD.WIDE R14, R2, 0x4, R24 ; /* 0x00000004020e7825 */ /* 0x000fe200078e0218 */ /*0980*/ MOV R9, UR7 ; /* 0x0000000700097c02 */ /* 0x000fe20008000f00 */ /*0990*/ LDG.E R25, [R24.64] ; /* 0x0000000418197981 */ /* 0x000ea8000c1e1900 */ /*09a0*/ IMAD.WIDE R8, R3, 0x4, R8 ; /* 0x0000000403087825 */ /* 0x000fc800078e0208 */ /*09b0*/ IMAD.WIDE R12, R2.reuse, 0x4, R14 ; /* 0x00000004020c7825 */ /* 0x040fe200078e020e */ /*09c0*/ LDG.E R7, [R8.64] ; /* 0x0000000408077981 */ /* 0x000ea8000c1e1900 */ /*09d0*/ LDG.E R14, [R14.64] ; /* 0x000000040e0e7981 */ /* 0x000ee2000c1e1900 */ /*09e0*/ IMAD.WIDE R10, R2, 0x4, R12 ; /* 0x00000004020a7825 */ /* 0x000fc600078e020c */ /*09f0*/ LDG.E R16, [R8.64+0x4] ; /* 0x0000040408107981 */ /* 0x000ee8000c1e1900 */ /*0a00*/ LDG.E R18, [R12.64] ; /* 0x000000040c127981 */ /* 0x000f28000c1e1900 */ /*0a10*/ LDG.E R17, [R8.64+0x8] ; /* 0x0000080408117981 */ /* 0x000f28000c1e1900 */ /*0a20*/ LDG.E R19, [R8.64+0xc] ; /* 0x00000c0408137981 */ /* 0x000f68000c1e1900 */ /*0a30*/ LDG.E R20, [R10.64] ; /* 0x000000040a147981 */ /* 0x000f62000c1e1900 */ /*0a40*/ IADD3 R6, R6, -0x4, RZ ; /* 0xfffffffc06067810 */ /* 0x000fc80007ffe0ff */ /*0a50*/ ISETP.NE.AND P0, PT, R6, RZ, PT ; /* 0x000000ff0600720c */ /* 0x000fe20003f05270 */ /*0a60*/ UIADD3 UR6, UP0, UR6, 0x10, URZ ; /* 0x0000001006067890 */ /* 0x000fe2000ff1e03f */ /*0a70*/ IADD3 R4, R4, 0x4, RZ ; /* 0x0000000404047810 */ /* 0x000fc60007ffe0ff */ /*0a80*/ UIADD3.X UR7, URZ, UR7, URZ, UP0, !UPT ; /* 0x000000073f077290 */ /* 0x000fe200087fe43f */ /*0a90*/ IMAD R7, R25, R7, R28 ; /* 0x0000000719077224 */ /* 0x004fc800078e021c */ /*0aa0*/ IMAD R7, R14, R16, R7 ; /* 0x000000100e077224 */ /* 0x008fe400078e0207 */ /*0ab0*/ IMAD.WIDE R24, R2, 0x4, R10 ; /* 0x0000000402187825 */ /* 0x000fc800078e020a */ /*0ac0*/ IMAD R7, R18, R17, R7 ; /* 0x0000001112077224 */ /* 0x010fc800078e0207 */ /*0ad0*/ IMAD R28, R20, R19, R7 ; /* 0x00000013141c7224 */ /* 0x020fe200078e0207 */ /*0ae0*/ @P0 BRA 0x960 ; /* 0xfffffe7000000947 */ /* 0x000fea000383ffff */ /*0af0*/ ISETP.NE.AND P0, PT, R5, RZ, PT ; /* 0x000000ff0500720c */ /* 0x000fda0003f05270 */ /*0b00*/ @!P0 BRA 0xbf0 ; /* 0x000000e000008947 */ /* 0x000fea0003800000 */ /*0b10*/ HFMA2.MMA R9, -RZ, RZ, 0, 2.384185791015625e-07 ; /* 0x00000004ff097435 */ /* 0x000fe200000001ff */ /*0b20*/ IADD3 R6, R3, R4, RZ ; /* 0x0000000403067210 */ /* 0x000fe20007ffe0ff */ /*0b30*/ IMAD R4, R4, c[0x0][0x178], R0 ; /* 0x00005e0004047a24 */ /* 0x000fd000078e0200 */ /*0b40*/ IMAD.WIDE R6, R6, R9, c[0x0][0x160] ; /* 0x0000580006067625 */ /* 0x000fc800078e0209 */ /*0b50*/ IMAD.WIDE R8, R4, R9, c[0x0][0x168] ; /* 0x00005a0004087625 */ /* 0x000fca00078e0209 */ /*0b60*/ LDG.E R11, [R8.64] ; /* 0x00000004080b7981 */ /* 0x0000a8000c1e1900 */ /*0b70*/ LDG.E R4, [R6.64] ; /* 0x0000000406047981 */ /* 0x0002a2000c1e1900 */ /*0b80*/ IADD3 R5, R5, -0x1, RZ ; /* 0xffffffff05057810 */ /* 0x000fc80007ffe0ff */ /*0b90*/ ISETP.NE.AND P0, PT, R5, RZ, PT ; /* 0x000000ff0500720c */ /* 0x000fe20003f05270 */ /*0ba0*/ IMAD.WIDE R8, R2, 0x4, R8 ; /* 0x0000000402087825 */ /* 0x001fe200078e0208 */ /*0bb0*/ IADD3 R6, P1, R6, 0x4, RZ ; /* 0x0000000406067810 */ /* 0x002fc80007f3e0ff */ /*0bc0*/ IADD3.X R7, RZ, R7, RZ, P1, !PT ; /* 0x00000007ff077210 */ /* 0x000fe20000ffe4ff */ /*0bd0*/ IMAD R28, R11, R4, R28 ; /* 0x000000040b1c7224 */ /* 0x004fcc00078e021c */ /*0be0*/ @P0 BRA 0xb60 ; /* 0xffffff7000000947 */ /* 0x000fea000383ffff */ /*0bf0*/ IADD3 R3, R0, R3, RZ ; /* 0x0000000300037210 */ /* 0x000fe40007ffe0ff */ /*0c00*/ MOV R2, 0x4 ; /* 0x0000000400027802 */ /* 0x000fca0000000f00 */ /*0c10*/ IMAD.WIDE R2, R3, R2, c[0x0][0x170] ; /* 0x00005c0003027625 */ /* 0x000fca00078e0202 */ /*0c20*/ STG.E [R2.64], R28 ; /* 0x0000001c02007986 */ /* 0x000fe2000c101904 */ /*0c30*/ EXIT ; /* 0x000000000000794d */ /* 0x000fea0003800000 */ /*0c40*/ BRA 0xc40; /* 0xfffffff000007947 */ /* 0x000fc0000383ffff */ /*0c50*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0c60*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0c70*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0c80*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0c90*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0ca0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0cb0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0cc0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0cd0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0ce0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0cf0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ ..........
You are a helpful coding assistant specialized in converting CUDA source code to CUDA host assembly.
#include "includes.h" __global__ void MatrixMultiplication__CudaKernel(int* in_tabA, int* in_tabB, int* out_tabC, int outTabWidth) { int row = blockIdx.x * blockDim.x + threadIdx.x; int col = blockIdx.y * blockDim.y + threadIdx.y; //making sure that extra threads will do not any work if (row < outTabWidth && col < outTabWidth) { int tmp_sum = 0; //#pragma unroll for (int i = 0; i < outTabWidth; i++) { tmp_sum += in_tabA[row * outTabWidth + i] * in_tabB[i * outTabWidth + col]; } out_tabC[row * outTabWidth + col] = tmp_sum; } }
.file "tmpxft_0011c356_00000000-6_MatrixMultiplication__CudaKernel.cudafe1.cpp" .text #APP #NO_APP .type _ZL26__cudaUnregisterBinaryUtilv, @function _ZL26__cudaUnregisterBinaryUtilv: .LFB2029: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaUnregisterFatBinary@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2029: .size _ZL26__cudaUnregisterBinaryUtilv, .-_ZL26__cudaUnregisterBinaryUtilv .globl _Z57__device_stub__Z32MatrixMultiplication__CudaKernelPiS_S_iPiS_S_i .type _Z57__device_stub__Z32MatrixMultiplication__CudaKernelPiS_S_iPiS_S_i, @function _Z57__device_stub__Z32MatrixMultiplication__CudaKernelPiS_S_iPiS_S_i: .LFB2051: .cfi_startproc endbr64 subq $152, %rsp .cfi_def_cfa_offset 160 movq %rdi, 24(%rsp) movq %rsi, 16(%rsp) movq %rdx, 8(%rsp) movl %ecx, 4(%rsp) movq %fs:40, %rax movq %rax, 136(%rsp) xorl %eax, %eax leaq 24(%rsp), %rax movq %rax, 96(%rsp) leaq 16(%rsp), %rax movq %rax, 104(%rsp) leaq 8(%rsp), %rax movq %rax, 112(%rsp) leaq 4(%rsp), %rax movq %rax, 120(%rsp) movl $1, 48(%rsp) movl $1, 52(%rsp) movl $1, 56(%rsp) movl $1, 60(%rsp) movl $1, 64(%rsp) movl $1, 68(%rsp) leaq 40(%rsp), %rcx leaq 32(%rsp), %rdx leaq 60(%rsp), %rsi leaq 48(%rsp), %rdi call __cudaPopCallConfiguration@PLT testl %eax, %eax je .L7 .L3: movq 136(%rsp), %rax subq %fs:40, %rax jne .L8 addq $152, %rsp .cfi_remember_state .cfi_def_cfa_offset 8 ret .L7: .cfi_restore_state pushq 40(%rsp) .cfi_def_cfa_offset 168 pushq 40(%rsp) .cfi_def_cfa_offset 176 leaq 112(%rsp), %r9 movq 76(%rsp), %rcx movl 84(%rsp), %r8d movq 64(%rsp), %rsi movl 72(%rsp), %edx leaq _Z32MatrixMultiplication__CudaKernelPiS_S_i(%rip), %rdi call cudaLaunchKernel@PLT addq $16, %rsp .cfi_def_cfa_offset 160 jmp .L3 .L8: call __stack_chk_fail@PLT .cfi_endproc .LFE2051: .size _Z57__device_stub__Z32MatrixMultiplication__CudaKernelPiS_S_iPiS_S_i, .-_Z57__device_stub__Z32MatrixMultiplication__CudaKernelPiS_S_iPiS_S_i .globl _Z32MatrixMultiplication__CudaKernelPiS_S_i .type _Z32MatrixMultiplication__CudaKernelPiS_S_i, @function _Z32MatrixMultiplication__CudaKernelPiS_S_i: .LFB2052: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 call _Z57__device_stub__Z32MatrixMultiplication__CudaKernelPiS_S_iPiS_S_i addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2052: .size _Z32MatrixMultiplication__CudaKernelPiS_S_i, .-_Z32MatrixMultiplication__CudaKernelPiS_S_i .section .rodata.str1.8,"aMS",@progbits,1 .align 8 .LC0: .string "_Z32MatrixMultiplication__CudaKernelPiS_S_i" .text .type _ZL24__sti____cudaRegisterAllv, @function _ZL24__sti____cudaRegisterAllv: .LFB2054: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 leaq _ZL15__fatDeviceText(%rip), %rdi call __cudaRegisterFatBinary@PLT movq %rax, %rdi movq %rax, _ZL20__cudaFatCubinHandle(%rip) pushq $0 .cfi_def_cfa_offset 24 pushq $0 .cfi_def_cfa_offset 32 pushq $0 .cfi_def_cfa_offset 40 pushq $0 .cfi_def_cfa_offset 48 movl $0, %r9d movl $-1, %r8d leaq .LC0(%rip), %rdx movq %rdx, %rcx leaq _Z32MatrixMultiplication__CudaKernelPiS_S_i(%rip), %rsi call __cudaRegisterFunction@PLT addq $32, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaRegisterFatBinaryEnd@PLT leaq _ZL26__cudaUnregisterBinaryUtilv(%rip), %rdi call atexit@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2054: .size _ZL24__sti____cudaRegisterAllv, .-_ZL24__sti____cudaRegisterAllv .section .init_array,"aw" .align 8 .quad _ZL24__sti____cudaRegisterAllv .section .nvFatBinSegment,"aw" .align 8 .type _ZL15__fatDeviceText, @object .size _ZL15__fatDeviceText, 24 _ZL15__fatDeviceText: .long 1180844977 .long 1 .quad fatbinData .quad 0 .local _ZL20__cudaFatCubinHandle .comm _ZL20__cudaFatCubinHandle,8,8 .ident "GCC: (Ubuntu 13.3.0-6ubuntu2~24.04) 13.3.0" .section .note.GNU-stack,"",@progbits .section .note.gnu.property,"a" .align 8 .long 1f - 0f .long 4f - 1f .long 5 0: .string "GNU" 1: .align 8 .long 0xc0000002 .long 3f - 2f 2: .long 0x3 3: .align 8 4:
You are a helpful coding assistant specialized in converting CUDA source code to HIP/AMD source code.
#include "includes.h" __global__ void MatrixMultiplication__CudaKernel(int* in_tabA, int* in_tabB, int* out_tabC, int outTabWidth) { int row = blockIdx.x * blockDim.x + threadIdx.x; int col = blockIdx.y * blockDim.y + threadIdx.y; //making sure that extra threads will do not any work if (row < outTabWidth && col < outTabWidth) { int tmp_sum = 0; //#pragma unroll for (int i = 0; i < outTabWidth; i++) { tmp_sum += in_tabA[row * outTabWidth + i] * in_tabB[i * outTabWidth + col]; } out_tabC[row * outTabWidth + col] = tmp_sum; } }
#include <hip/hip_runtime.h> #include "includes.h" __global__ void MatrixMultiplication__CudaKernel(int* in_tabA, int* in_tabB, int* out_tabC, int outTabWidth) { int row = blockIdx.x * blockDim.x + threadIdx.x; int col = blockIdx.y * blockDim.y + threadIdx.y; //making sure that extra threads will do not any work if (row < outTabWidth && col < outTabWidth) { int tmp_sum = 0; //#pragma unroll for (int i = 0; i < outTabWidth; i++) { tmp_sum += in_tabA[row * outTabWidth + i] * in_tabB[i * outTabWidth + col]; } out_tabC[row * outTabWidth + col] = tmp_sum; } }
You are a helpful coding assistant specialized in converting HIP source code to AMD device assembly.
#include <hip/hip_runtime.h> #include "includes.h" __global__ void MatrixMultiplication__CudaKernel(int* in_tabA, int* in_tabB, int* out_tabC, int outTabWidth) { int row = blockIdx.x * blockDim.x + threadIdx.x; int col = blockIdx.y * blockDim.y + threadIdx.y; //making sure that extra threads will do not any work if (row < outTabWidth && col < outTabWidth) { int tmp_sum = 0; //#pragma unroll for (int i = 0; i < outTabWidth; i++) { tmp_sum += in_tabA[row * outTabWidth + i] * in_tabB[i * outTabWidth + col]; } out_tabC[row * outTabWidth + col] = tmp_sum; } }
.text .amdgcn_target "amdgcn-amd-amdhsa--gfx1100" .protected _Z32MatrixMultiplication__CudaKernelPiS_S_i .globl _Z32MatrixMultiplication__CudaKernelPiS_S_i .p2align 8 .type _Z32MatrixMultiplication__CudaKernelPiS_S_i,@function _Z32MatrixMultiplication__CudaKernelPiS_S_i: s_clause 0x1 s_load_b32 s3, s[0:1], 0x2c s_load_b32 s2, s[0:1], 0x18 v_and_b32_e32 v2, 0x3ff, v0 v_bfe_u32 v3, v0, 10, 10 s_waitcnt lgkmcnt(0) s_and_b32 s4, s3, 0xffff s_lshr_b32 s3, s3, 16 s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1) v_mad_u64_u32 v[0:1], null, s14, s4, v[2:3] v_mad_u64_u32 v[1:2], null, s15, s3, v[3:4] s_mov_b32 s3, exec_lo v_max_i32_e32 v2, v0, v1 s_delay_alu instid0(VALU_DEP_1) v_cmpx_gt_i32_e64 s2, v2 s_cbranch_execz .LBB0_6 s_cmp_lt_i32 s2, 1 s_cbranch_scc1 .LBB0_4 s_load_b128 s[4:7], s[0:1], 0x0 v_mul_lo_u32 v2, v0, s2 s_mov_b32 s3, s2 v_mov_b32_e32 v5, v1 s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1) v_ashrrev_i32_e32 v3, 31, v2 v_lshlrev_b64 v[3:4], 2, v[2:3] v_mov_b32_e32 v2, 0 s_waitcnt lgkmcnt(0) s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3) v_add_co_u32 v3, vcc_lo, s4, v3 v_add_co_ci_u32_e32 v4, vcc_lo, s5, v4, vcc_lo .p2align 6 .LBB0_3: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1) v_ashrrev_i32_e32 v6, 31, v5 s_add_i32 s3, s3, -1 s_cmp_eq_u32 s3, 0 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_lshlrev_b64 v[6:7], 2, v[5:6] v_add_co_u32 v6, vcc_lo, s6, v6 s_delay_alu instid0(VALU_DEP_2) v_add_co_ci_u32_e32 v7, vcc_lo, s7, v7, vcc_lo global_load_b32 v8, v[3:4], off global_load_b32 v9, v[6:7], off s_waitcnt vmcnt(0) v_mad_u64_u32 v[6:7], null, v9, v8, v[2:3] v_add_co_u32 v3, vcc_lo, v3, 4 v_add_co_ci_u32_e32 v4, vcc_lo, 0, v4, vcc_lo s_delay_alu instid0(VALU_DEP_3) v_dual_mov_b32 v2, v6 :: v_dual_add_nc_u32 v5, s2, v5 s_cbranch_scc0 .LBB0_3 s_branch .LBB0_5 .LBB0_4: v_mov_b32_e32 v2, 0 .LBB0_5: s_load_b64 s[0:1], s[0:1], 0x10 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_mad_u64_u32 v[3:4], null, v0, s2, v[1:2] v_ashrrev_i32_e32 v4, 31, v3 s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1) v_lshlrev_b64 v[0:1], 2, v[3:4] s_waitcnt lgkmcnt(0) v_add_co_u32 v0, vcc_lo, s0, v0 s_delay_alu instid0(VALU_DEP_2) v_add_co_ci_u32_e32 v1, vcc_lo, s1, v1, vcc_lo global_store_b32 v[0:1], v2, off .LBB0_6: s_nop 0 s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) s_endpgm .section .rodata,"a",@progbits .p2align 6, 0x0 .amdhsa_kernel _Z32MatrixMultiplication__CudaKernelPiS_S_i .amdhsa_group_segment_fixed_size 0 .amdhsa_private_segment_fixed_size 0 .amdhsa_kernarg_size 288 .amdhsa_user_sgpr_count 14 .amdhsa_user_sgpr_dispatch_ptr 0 .amdhsa_user_sgpr_queue_ptr 0 .amdhsa_user_sgpr_kernarg_segment_ptr 1 .amdhsa_user_sgpr_dispatch_id 0 .amdhsa_user_sgpr_private_segment_size 0 .amdhsa_wavefront_size32 1 .amdhsa_uses_dynamic_stack 0 .amdhsa_enable_private_segment 0 .amdhsa_system_sgpr_workgroup_id_x 1 .amdhsa_system_sgpr_workgroup_id_y 1 .amdhsa_system_sgpr_workgroup_id_z 0 .amdhsa_system_sgpr_workgroup_info 0 .amdhsa_system_vgpr_workitem_id 1 .amdhsa_next_free_vgpr 10 .amdhsa_next_free_sgpr 16 .amdhsa_float_round_mode_32 0 .amdhsa_float_round_mode_16_64 0 .amdhsa_float_denorm_mode_32 3 .amdhsa_float_denorm_mode_16_64 3 .amdhsa_dx10_clamp 1 .amdhsa_ieee_mode 1 .amdhsa_fp16_overflow 0 .amdhsa_workgroup_processor_mode 1 .amdhsa_memory_ordered 1 .amdhsa_forward_progress 0 .amdhsa_shared_vgpr_count 0 .amdhsa_exception_fp_ieee_invalid_op 0 .amdhsa_exception_fp_denorm_src 0 .amdhsa_exception_fp_ieee_div_zero 0 .amdhsa_exception_fp_ieee_overflow 0 .amdhsa_exception_fp_ieee_underflow 0 .amdhsa_exception_fp_ieee_inexact 0 .amdhsa_exception_int_div_zero 0 .end_amdhsa_kernel .text .Lfunc_end0: .size _Z32MatrixMultiplication__CudaKernelPiS_S_i, .Lfunc_end0-_Z32MatrixMultiplication__CudaKernelPiS_S_i .section .AMDGPU.csdata,"",@progbits .text .p2alignl 7, 3214868480 .fill 96, 4, 3214868480 .type __hip_cuid_,@object .section .bss,"aw",@nobits .globl __hip_cuid_ __hip_cuid_: .byte 0 .size __hip_cuid_, 1 .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym __hip_cuid_ .amdgpu_metadata --- amdhsa.kernels: - .args: - .address_space: global .offset: 0 .size: 8 .value_kind: global_buffer - .address_space: global .offset: 8 .size: 8 .value_kind: global_buffer - .address_space: global .offset: 16 .size: 8 .value_kind: global_buffer - .offset: 24 .size: 4 .value_kind: by_value - .offset: 32 .size: 4 .value_kind: hidden_block_count_x - .offset: 36 .size: 4 .value_kind: hidden_block_count_y - .offset: 40 .size: 4 .value_kind: hidden_block_count_z - .offset: 44 .size: 2 .value_kind: hidden_group_size_x - .offset: 46 .size: 2 .value_kind: hidden_group_size_y - .offset: 48 .size: 2 .value_kind: hidden_group_size_z - .offset: 50 .size: 2 .value_kind: hidden_remainder_x - .offset: 52 .size: 2 .value_kind: hidden_remainder_y - .offset: 54 .size: 2 .value_kind: hidden_remainder_z - .offset: 72 .size: 8 .value_kind: hidden_global_offset_x - .offset: 80 .size: 8 .value_kind: hidden_global_offset_y - .offset: 88 .size: 8 .value_kind: hidden_global_offset_z - .offset: 96 .size: 2 .value_kind: hidden_grid_dims .group_segment_fixed_size: 0 .kernarg_segment_align: 8 .kernarg_segment_size: 288 .language: OpenCL C .language_version: - 2 - 0 .max_flat_workgroup_size: 1024 .name: _Z32MatrixMultiplication__CudaKernelPiS_S_i .private_segment_fixed_size: 0 .sgpr_count: 18 .sgpr_spill_count: 0 .symbol: _Z32MatrixMultiplication__CudaKernelPiS_S_i.kd .uniform_work_group_size: 1 .uses_dynamic_stack: false .vgpr_count: 10 .vgpr_spill_count: 0 .wavefront_size: 32 .workgroup_processor_mode: 1 amdhsa.target: amdgcn-amd-amdhsa--gfx1100 amdhsa.version: - 1 - 2 ... .end_amdgpu_metadata
You are a helpful coding assistant specialized in converting HIP source code to AMD host assembly.
#include <hip/hip_runtime.h> #include "includes.h" __global__ void MatrixMultiplication__CudaKernel(int* in_tabA, int* in_tabB, int* out_tabC, int outTabWidth) { int row = blockIdx.x * blockDim.x + threadIdx.x; int col = blockIdx.y * blockDim.y + threadIdx.y; //making sure that extra threads will do not any work if (row < outTabWidth && col < outTabWidth) { int tmp_sum = 0; //#pragma unroll for (int i = 0; i < outTabWidth; i++) { tmp_sum += in_tabA[row * outTabWidth + i] * in_tabB[i * outTabWidth + col]; } out_tabC[row * outTabWidth + col] = tmp_sum; } }
.text .file "MatrixMultiplication__CudaKernel.hip" .globl _Z47__device_stub__MatrixMultiplication__CudaKernelPiS_S_i # -- Begin function _Z47__device_stub__MatrixMultiplication__CudaKernelPiS_S_i .p2align 4, 0x90 .type _Z47__device_stub__MatrixMultiplication__CudaKernelPiS_S_i,@function _Z47__device_stub__MatrixMultiplication__CudaKernelPiS_S_i: # @_Z47__device_stub__MatrixMultiplication__CudaKernelPiS_S_i .cfi_startproc # %bb.0: subq $120, %rsp .cfi_def_cfa_offset 128 movq %rdi, 72(%rsp) movq %rsi, 64(%rsp) movq %rdx, 56(%rsp) movl %ecx, 4(%rsp) leaq 72(%rsp), %rax movq %rax, 80(%rsp) leaq 64(%rsp), %rax movq %rax, 88(%rsp) leaq 56(%rsp), %rax movq %rax, 96(%rsp) leaq 4(%rsp), %rax movq %rax, 104(%rsp) leaq 40(%rsp), %rdi leaq 24(%rsp), %rsi leaq 16(%rsp), %rdx leaq 8(%rsp), %rcx callq __hipPopCallConfiguration movq 40(%rsp), %rsi movl 48(%rsp), %edx movq 24(%rsp), %rcx movl 32(%rsp), %r8d leaq 80(%rsp), %r9 movl $_Z32MatrixMultiplication__CudaKernelPiS_S_i, %edi pushq 8(%rsp) .cfi_adjust_cfa_offset 8 pushq 24(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $136, %rsp .cfi_adjust_cfa_offset -136 retq .Lfunc_end0: .size _Z47__device_stub__MatrixMultiplication__CudaKernelPiS_S_i, .Lfunc_end0-_Z47__device_stub__MatrixMultiplication__CudaKernelPiS_S_i .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_ctor .type __hip_module_ctor,@function __hip_module_ctor: # @__hip_module_ctor .cfi_startproc # %bb.0: subq $40, %rsp .cfi_def_cfa_offset 48 cmpq $0, __hip_gpubin_handle(%rip) jne .LBB1_2 # %bb.1: movl $__hip_fatbin_wrapper, %edi callq __hipRegisterFatBinary movq %rax, __hip_gpubin_handle(%rip) .LBB1_2: movq __hip_gpubin_handle(%rip), %rdi xorps %xmm0, %xmm0 movups %xmm0, 16(%rsp) movups %xmm0, (%rsp) movl $_Z32MatrixMultiplication__CudaKernelPiS_S_i, %esi movl $.L__unnamed_1, %edx movl $.L__unnamed_1, %ecx movl $-1, %r8d xorl %r9d, %r9d callq __hipRegisterFunction movl $__hip_module_dtor, %edi addq $40, %rsp .cfi_def_cfa_offset 8 jmp atexit # TAILCALL .Lfunc_end1: .size __hip_module_ctor, .Lfunc_end1-__hip_module_ctor .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_dtor .type __hip_module_dtor,@function __hip_module_dtor: # @__hip_module_dtor .cfi_startproc # %bb.0: movq __hip_gpubin_handle(%rip), %rdi testq %rdi, %rdi je .LBB2_2 # %bb.1: pushq %rax .cfi_def_cfa_offset 16 callq __hipUnregisterFatBinary movq $0, __hip_gpubin_handle(%rip) addq $8, %rsp .cfi_def_cfa_offset 8 .LBB2_2: retq .Lfunc_end2: .size __hip_module_dtor, .Lfunc_end2-__hip_module_dtor .cfi_endproc # -- End function .type _Z32MatrixMultiplication__CudaKernelPiS_S_i,@object # @_Z32MatrixMultiplication__CudaKernelPiS_S_i .section .rodata,"a",@progbits .globl _Z32MatrixMultiplication__CudaKernelPiS_S_i .p2align 3, 0x0 _Z32MatrixMultiplication__CudaKernelPiS_S_i: .quad _Z47__device_stub__MatrixMultiplication__CudaKernelPiS_S_i .size _Z32MatrixMultiplication__CudaKernelPiS_S_i, 8 .type .L__unnamed_1,@object # @0 .section .rodata.str1.1,"aMS",@progbits,1 .L__unnamed_1: .asciz "_Z32MatrixMultiplication__CudaKernelPiS_S_i" .size .L__unnamed_1, 44 .type __hip_fatbin_wrapper,@object # @__hip_fatbin_wrapper .section .hipFatBinSegment,"a",@progbits .p2align 3, 0x0 __hip_fatbin_wrapper: .long 1212764230 # 0x48495046 .long 1 # 0x1 .quad __hip_fatbin .quad 0 .size __hip_fatbin_wrapper, 24 .type __hip_gpubin_handle,@object # @__hip_gpubin_handle .local __hip_gpubin_handle .comm __hip_gpubin_handle,8,8 .section .init_array,"aw",@init_array .p2align 3, 0x0 .quad __hip_module_ctor .type __hip_cuid_,@object # @__hip_cuid_ .bss .globl __hip_cuid_ __hip_cuid_: .byte 0 # 0x0 .size __hip_cuid_, 1 .section ".linker-options","e",@llvm_linker_options .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym _Z47__device_stub__MatrixMultiplication__CudaKernelPiS_S_i .addrsig_sym __hip_module_ctor .addrsig_sym __hip_module_dtor .addrsig_sym _Z32MatrixMultiplication__CudaKernelPiS_S_i .addrsig_sym __hip_fatbin .addrsig_sym __hip_fatbin_wrapper .addrsig_sym __hip_cuid_
You are a helpful coding assistant specialized in converting from CUDA device assembly to AMD device assembly.
code for sm_80 Function : _Z32MatrixMultiplication__CudaKernelPiS_S_i .headerflags @"EF_CUDA_TEXMODE_UNIFIED EF_CUDA_64BIT_ADDRESS EF_CUDA_SM80 EF_CUDA_VIRTUAL_SM(EF_CUDA_SM80)" /*0000*/ MOV R1, c[0x0][0x28] ; /* 0x00000a0000017a02 */ /* 0x000fe40000000f00 */ /*0010*/ S2R R0, SR_CTAID.Y ; /* 0x0000000000007919 */ /* 0x000e280000002600 */ /*0020*/ S2R R5, SR_TID.Y ; /* 0x0000000000057919 */ /* 0x000e280000002200 */ /*0030*/ S2R R3, SR_CTAID.X ; /* 0x0000000000037919 */ /* 0x000e680000002500 */ /*0040*/ S2R R2, SR_TID.X ; /* 0x0000000000027919 */ /* 0x000e620000002100 */ /*0050*/ IMAD R0, R0, c[0x0][0x4], R5 ; /* 0x0000010000007a24 */ /* 0x001fca00078e0205 */ /*0060*/ ISETP.GE.AND P0, PT, R0, c[0x0][0x178], PT ; /* 0x00005e0000007a0c */ /* 0x000fe20003f06270 */ /*0070*/ IMAD R3, R3, c[0x0][0x0], R2 ; /* 0x0000000003037a24 */ /* 0x002fca00078e0202 */ /*0080*/ ISETP.GE.OR P0, PT, R3, c[0x0][0x178], P0 ; /* 0x00005e0003007a0c */ /* 0x000fda0000706670 */ /*0090*/ @P0 EXIT ; /* 0x000000000000094d */ /* 0x000fea0003800000 */ /*00a0*/ MOV R2, c[0x0][0x178] ; /* 0x00005e0000027a02 */ /* 0x000fe20000000f00 */ /*00b0*/ ULDC.64 UR4, c[0x0][0x118] ; /* 0x0000460000047ab9 */ /* 0x000fe20000000a00 */ /*00c0*/ HFMA2.MMA R28, -RZ, RZ, 0, 0 ; /* 0x00000000ff1c7435 */ /* 0x000fe200000001ff */ /*00d0*/ IMAD R3, R3, c[0x0][0x178], RZ ; /* 0x00005e0003037a24 */ /* 0x000fe200078e02ff */ /*00e0*/ ISETP.GE.AND P0, PT, R2, 0x1, PT ; /* 0x000000010200780c */ /* 0x000fda0003f06270 */ /*00f0*/ @!P0 BRA 0xbf0 ; /* 0x00000af000008947 */ /* 0x000fea0003800000 */ /*0100*/ IADD3 R4, R2.reuse, -0x1, RZ ; /* 0xffffffff02047810 */ /* 0x040fe40007ffe0ff */ /*0110*/ LOP3.LUT R5, R2, 0x3, RZ, 0xc0, !PT ; /* 0x0000000302057812 */ /* 0x000fe400078ec0ff */ /*0120*/ ISETP.GE.U32.AND P0, PT, R4, 0x3, PT ; /* 0x000000030400780c */ /* 0x000fe40003f06070 */ /*0130*/ MOV R4, RZ ; /* 0x000000ff00047202 */ /* 0x000fe40000000f00 */ /*0140*/ MOV R28, RZ ; /* 0x000000ff001c7202 */ /* 0x000fd20000000f00 */ /*0150*/ @!P0 BRA 0xaf0 ; /* 0x0000099000008947 */ /* 0x000fea0003800000 */ /*0160*/ IADD3 R6, -R5, c[0x0][0x178], RZ ; /* 0x00005e0005067a10 */ /* 0x000fe20007ffe1ff */ /*0170*/ HFMA2.MMA R25, -RZ, RZ, 0, 2.384185791015625e-07 ; /* 0x00000004ff197435 */ /* 0x000fe200000001ff */ /*0180*/ ULDC.64 UR6, c[0x0][0x160] ; /* 0x0000580000067ab9 */ /* 0x000fe20000000a00 */ /*0190*/ MOV R4, RZ ; /* 0x000000ff00047202 */ /* 0x000fe40000000f00 */ /*01a0*/ ISETP.GT.AND P0, PT, R6, RZ, PT ; /* 0x000000ff0600720c */ /* 0x000fcc0003f04270 */ /*01b0*/ IMAD.WIDE R24, R0, R25, c[0x0][0x168] ; /* 0x00005a0000187625 */ /* 0x000fce00078e0219 */ /*01c0*/ @!P0 BRA 0x960 ; /* 0x0000079000008947 */ /* 0x000fea0003800000 */ /*01d0*/ ISETP.GT.AND P1, PT, R6, 0xc, PT ; /* 0x0000000c0600780c */ /* 0x000fe40003f24270 */ /*01e0*/ PLOP3.LUT P0, PT, PT, PT, PT, 0x80, 0x0 ; /* 0x000000000000781c */ /* 0x000fd60003f0f070 */ /*01f0*/ @!P1 BRA 0x6a0 ; /* 0x000004a000009947 */ /* 0x000fea0003800000 */ /*0200*/ PLOP3.LUT P0, PT, PT, PT, PT, 0x8, 0x0 ; /* 0x000000000000781c */ /* 0x000fe40003f0e170 */ /*0210*/ MOV R12, UR6 ; /* 0x00000006000c7c02 */ /* 0x000fe20008000f00 */ /*0220*/ LDG.E R29, [R24.64] ; /* 0x00000004181d7981 */ /* 0x0000a2000c1e1900 */ /*0230*/ MOV R13, UR7 ; /* 0x00000007000d7c02 */ /* 0x000fca0008000f00 */ /*0240*/ IMAD.WIDE R12, R3, 0x4, R12 ; /* 0x00000004030c7825 */ /* 0x000fca00078e020c */ /*0250*/ LDG.E R27, [R12.64] ; /* 0x000000040c1b7981 */ /* 0x000ea2000c1e1900 */ /*0260*/ IMAD.WIDE R10, R2, 0x4, R24 ; /* 0x00000004020a7825 */ /* 0x000fc600078e0218 */ /*0270*/ LDG.E R17, [R12.64+0x4] ; /* 0x000004040c117981 */ /* 0x000ee6000c1e1900 */ /*0280*/ IMAD.WIDE R18, R2.reuse, 0x4, R10 ; /* 0x0000000402127825 */ /* 0x040fe200078e020a */ /*0290*/ LDG.E R16, [R10.64] ; /* 0x000000040a107981 */ /* 0x0002e8000c1e1900 */ /*02a0*/ LDG.E R7, [R12.64+0xc] ; /* 0x00000c040c077981 */ /* 0x000f22000c1e1900 */ /*02b0*/ IMAD.WIDE R14, R2, 0x4, R18 ; /* 0x00000004020e7825 */ /* 0x000fc600078e0212 */ /*02c0*/ LDG.E R18, [R18.64] ; /* 0x0000000412127981 */ /* 0x000b26000c1e1900 */ /*02d0*/ IMAD.WIDE R20, R2.reuse, 0x4, R14 ; /* 0x0000000402147825 */ /* 0x040fe200078e020e */ /*02e0*/ LDG.E R26, [R14.64] ; /* 0x000000040e1a7981 */ /* 0x000128000c1e1900 */ /*02f0*/ LDG.E R9, [R12.64+0x10] ; /* 0x000010040c097981 */ /* 0x000f28000c1e1900 */ /*0300*/ LDG.E R19, [R12.64+0x8] ; /* 0x000008040c137981 */ /* 0x020f22000c1e1900 */ /*0310*/ IMAD.WIDE R14, R2, 0x4, R20 ; /* 0x00000004020e7825 */ /* 0x001fc600078e0214 */ /*0320*/ LDG.E R20, [R20.64] ; /* 0x0000000414147981 */ /* 0x000166000c1e1900 */ /*0330*/ IMAD.WIDE R22, R2.reuse, 0x4, R14 ; /* 0x0000000402167825 */ /* 0x040fe200078e020e */ /*0340*/ LDG.E R8, [R14.64] ; /* 0x000000040e087981 */ /* 0x000168000c1e1900 */ /*0350*/ LDG.E R11, [R12.64+0x14] ; /* 0x000014040c0b7981 */ /* 0x002f62000c1e1900 */ /*0360*/ IMAD.WIDE R24, R2, 0x4, R22 ; /* 0x0000000402187825 */ /* 0x000fc600078e0216 */ /*0370*/ LDG.E R10, [R22.64] ; /* 0x00000004160a7981 */ /* 0x000368000c1e1900 */ /*0380*/ LDG.E R21, [R12.64+0x18] ; /* 0x000018040c157981 */ /* 0x001f62000c1e1900 */ /*0390*/ IMAD R29, R29, R27, R28 ; /* 0x0000001b1d1d7224 */ /* 0x004fc600078e021c */ /*03a0*/ LDG.E R27, [R12.64+0x1c] ; /* 0x00001c040c1b7981 */ /* 0x000ea8000c1e1900 */ /*03b0*/ LDG.E R28, [R24.64] ; /* 0x00000004181c7981 */ /* 0x0000a2000c1e1900 */ /*03c0*/ IMAD.WIDE R14, R2, 0x4, R24 ; /* 0x00000004020e7825 */ /* 0x000fc800078e0218 */ /*03d0*/ IMAD R29, R16, R17, R29 ; /* 0x00000011101d7224 */ /* 0x008fe400078e021d */ /*03e0*/ IMAD.WIDE R16, R2, 0x4, R14 ; /* 0x0000000402107825 */ /* 0x000fe400078e020e */ /*03f0*/ LDG.E R14, [R14.64] ; /* 0x000000040e0e7981 */ /* 0x0006a4000c1e1900 */ /*0400*/ IMAD R29, R18, R19, R29 ; /* 0x00000013121d7224 */ /* 0x010fe400078e021d */ /*0410*/ IMAD.WIDE R18, R2, 0x4, R16 ; /* 0x0000000402127825 */ /* 0x000fe400078e0210 */ /*0420*/ LDG.E R16, [R16.64] ; /* 0x0000000410107981 */ /* 0x0008a4000c1e1900 */ /*0430*/ IMAD R26, R26, R7, R29 ; /* 0x000000071a1a7224 */ /* 0x000fc400078e021d */ /*0440*/ IMAD.WIDE R22, R2.reuse, 0x4, R18 ; /* 0x0000000402167825 */ /* 0x042fe200078e0212 */ /*0450*/ LDG.E R7, [R12.64+0x20] ; /* 0x000020040c077981 */ /* 0x000ea8000c1e1900 */ /*0460*/ LDG.E R29, [R12.64+0x24] ; /* 0x000024040c1d7981 */ /* 0x000ea2000c1e1900 */ /*0470*/ IMAD.WIDE R24, R2, 0x4, R22 ; /* 0x0000000402187825 */ /* 0x001fc600078e0216 */ /*0480*/ LDG.E R18, [R18.64] ; /* 0x0000000412127981 */ /* 0x0000a2000c1e1900 */ /*0490*/ IMAD R9, R20, R9, R26 ; /* 0x0000000914097224 */ /* 0x020fc600078e021a */ /*04a0*/ LDG.E R26, [R12.64+0x28] ; /* 0x000028040c1a7981 */ /* 0x000f62000c1e1900 */ /*04b0*/ IMAD R11, R8, R11, R9 ; /* 0x0000000b080b7224 */ /* 0x000fe400078e0209 */ /*04c0*/ IMAD.WIDE R8, R2, 0x4, R24 ; /* 0x0000000402087825 */ /* 0x000fe200078e0218 */ /*04d0*/ LDG.E R22, [R22.64] ; /* 0x0000000416167981 */ /* 0x000368000c1e1900 */ /*04e0*/ LDG.E R17, [R12.64+0x2c] ; /* 0x00002c040c117981 */ /* 0x010f22000c1e1900 */ /*04f0*/ IMAD R21, R10, R21, R11 ; /* 0x000000150a157224 */ /* 0x000fc600078e020b */ /*0500*/ LDG.E R15, [R24.64] ; /* 0x00000004180f7981 */ /* 0x008722000c1e1900 */ /*0510*/ IMAD.WIDE R10, R2, 0x4, R8 ; /* 0x00000004020a7825 */ /* 0x000fc600078e0208 */ /*0520*/ LDG.E R19, [R8.64] ; /* 0x0000000408137981 */ /* 0x001128000c1e1900 */ /*0530*/ LDG.E R23, [R10.64] ; /* 0x000000040a177981 */ /* 0x002f28000c1e1900 */ /*0540*/ LDG.E R24, [R12.64+0x30] ; /* 0x000030040c187981 */ /* 0x008ee8000c1e1900 */ /*0550*/ LDG.E R25, [R12.64+0x38] ; /* 0x000038040c197981 */ /* 0x000ee8000c1e1900 */ /*0560*/ LDG.E R8, [R12.64+0x3c] ; /* 0x00003c040c087981 */ /* 0x001ee2000c1e1900 */ /*0570*/ IMAD R9, R28, R27, R21 ; /* 0x0000001b1c097224 */ /* 0x004fc600078e0215 */ /*0580*/ LDG.E R28, [R12.64+0x34] ; /* 0x000034040c1c7981 */ /* 0x000ea2000c1e1900 */ /*0590*/ IMAD.WIDE R20, R2, 0x4, R10 ; /* 0x0000000402147825 */ /* 0x000fca00078e020a */ /*05a0*/ LDG.E R27, [R20.64] ; /* 0x00000004141b7981 */ /* 0x000ea2000c1e1900 */ /*05b0*/ IADD3 R6, R6, -0x10, RZ ; /* 0xfffffff006067810 */ /* 0x000fc80007ffe0ff */ /*05c0*/ ISETP.GT.AND P1, PT, R6, 0xc, PT ; /* 0x0000000c0600780c */ /* 0x000fe20003f24270 */ /*05d0*/ IMAD R7, R14, R7, R9 ; /* 0x000000070e077224 */ /* 0x000fc800078e0209 */ /*05e0*/ IMAD R7, R16, R29, R7 ; /* 0x0000001d10077224 */ /* 0x000fc800078e0207 */ /*05f0*/ IMAD R7, R18, R26, R7 ; /* 0x0000001a12077224 */ /* 0x020fc800078e0207 */ /*0600*/ IMAD R7, R22, R17, R7 ; /* 0x0000001116077224 */ /* 0x010fe200078e0207 */ /*0610*/ UIADD3 UR6, UP0, UR6, 0x40, URZ ; /* 0x0000004006067890 */ /* 0x000fe2000ff1e03f */ /*0620*/ IADD3 R4, R4, 0x10, RZ ; /* 0x0000001004047810 */ /* 0x000fc60007ffe0ff */ /*0630*/ UIADD3.X UR7, URZ, UR7, URZ, UP0, !UPT ; /* 0x000000073f077290 */ /* 0x000fe200087fe43f */ /*0640*/ IMAD R7, R15, R24, R7 ; /* 0x000000180f077224 */ /* 0x008fc800078e0207 */ /*0650*/ IMAD R28, R19, R28, R7 ; /* 0x0000001c131c7224 */ /* 0x004fc800078e0207 */ /*0660*/ IMAD R28, R23, R25, R28 ; /* 0x00000019171c7224 */ /* 0x000fe400078e021c */ /*0670*/ IMAD.WIDE R24, R2, 0x4, R20 ; /* 0x0000000402187825 */ /* 0x000fc800078e0214 */ /*0680*/ IMAD R28, R27, R8, R28 ; /* 0x000000081b1c7224 */ /* 0x000fe200078e021c */ /*0690*/ @P1 BRA 0x210 ; /* 0xfffffb7000001947 */ /* 0x000fea000383ffff */ /*06a0*/ ISETP.GT.AND P1, PT, R6, 0x4, PT ; /* 0x000000040600780c */ /* 0x000fda0003f24270 */ /*06b0*/ @!P1 BRA 0x940 ; /* 0x0000028000009947 */ /* 0x000fea0003800000 */ /*06c0*/ IMAD.WIDE R16, R2, 0x4, R24 ; /* 0x0000000402107825 */ /* 0x000fe200078e0218 */ /*06d0*/ MOV R8, UR6 ; /* 0x0000000600087c02 */ /* 0x000fe20008000f00 */ /*06e0*/ LDG.E R7, [R24.64] ; /* 0x0000000418077981 */ /* 0x0000a2000c1e1900 */ /*06f0*/ MOV R9, UR7 ; /* 0x0000000700097c02 */ /* 0x000fc60008000f00 */ /*0700*/ IMAD.WIDE R12, R2.reuse, 0x4, R16 ; /* 0x00000004020c7825 */ /* 0x040fe200078e0210 */ /*0710*/ LDG.E R21, [R16.64] ; /* 0x0000000410157981 */ /* 0x0002e6000c1e1900 */ /*0720*/ IMAD.WIDE R8, R3, 0x4, R8 ; /* 0x0000000403087825 */ /* 0x000fe200078e0208 */ /*0730*/ LDG.E R23, [R12.64] ; /* 0x000000040c177981 */ /* 0x000966000c1e1900 */ /*0740*/ IMAD.WIDE R14, R2.reuse, 0x4, R12 ; /* 0x00000004020e7825 */ /* 0x040fe200078e020c */ /*0750*/ LDG.E R20, [R8.64] ; /* 0x0000000408147981 */ /* 0x000ea8000c1e1900 */ /*0760*/ LDG.E R22, [R8.64+0x4] ; /* 0x0000040408167981 */ /* 0x000ee2000c1e1900 */ /*0770*/ IMAD.WIDE R10, R2, 0x4, R14 ; /* 0x00000004020a7825 */ /* 0x000fc600078e020e */ /*0780*/ LDG.E R26, [R8.64+0x8] ; /* 0x00000804081a7981 */ /* 0x000f66000c1e1900 */ /*0790*/ IMAD.WIDE R16, R2.reuse, 0x4, R10 ; /* 0x0000000402107825 */ /* 0x042fe200078e020a */ /*07a0*/ LDG.E R14, [R14.64] ; /* 0x000000040e0e7981 */ /* 0x000368000c1e1900 */ /*07b0*/ LDG.E R27, [R8.64+0xc] ; /* 0x00000c04081b7981 */ /* 0x000f62000c1e1900 */ /*07c0*/ IMAD.WIDE R18, R2, 0x4, R16 ; /* 0x0000000402127825 */ /* 0x000fc600078e0210 */ /*07d0*/ LDG.E R10, [R10.64] ; /* 0x000000040a0a7981 */ /* 0x000368000c1e1900 */ /*07e0*/ LDG.E R25, [R8.64+0x10] ; /* 0x0000100408197981 */ /* 0x001f62000c1e1900 */ /*07f0*/ IMAD.WIDE R12, R2, 0x4, R18 ; /* 0x00000004020c7825 */ /* 0x010fc600078e0212 */ /*0800*/ LDG.E R16, [R16.64] ; /* 0x0000000410107981 */ /* 0x000f28000c1e1900 */ /*0810*/ LDG.E R29, [R8.64+0x14] ; /* 0x00001404081d7981 */ /* 0x000f28000c1e1900 */ /*0820*/ LDG.E R24, [R18.64] ; /* 0x0000000412187981 */ /* 0x000128000c1e1900 */ /*0830*/ LDG.E R11, [R8.64+0x18] ; /* 0x00001804080b7981 */ /* 0x002f28000c1e1900 */ /*0840*/ LDG.E R15, [R12.64] ; /* 0x000000040c0f7981 */ /* 0x000f28000c1e1900 */ /*0850*/ LDG.E R18, [R8.64+0x1c] ; /* 0x00001c0408127981 */ /* 0x001f22000c1e1900 */ /*0860*/ UIADD3 UR6, UP0, UR6, 0x20, URZ ; /* 0x0000002006067890 */ /* 0x000fe2000ff1e03f */ /*0870*/ PLOP3.LUT P0, PT, PT, PT, PT, 0x8, 0x0 ; /* 0x000000000000781c */ /* 0x000fc40003f0e170 */ /*0880*/ IADD3 R4, R4, 0x8, RZ ; /* 0x0000000804047810 */ /* 0x000fe40007ffe0ff */ /*0890*/ IADD3 R6, R6, -0x8, RZ ; /* 0xfffffff806067810 */ /* 0x000fe20007ffe0ff */ /*08a0*/ UIADD3.X UR7, URZ, UR7, URZ, UP0, !UPT ; /* 0x000000073f077290 */ /* 0x000fe200087fe43f */ /*08b0*/ IMAD R7, R7, R20, R28 ; /* 0x0000001407077224 */ /* 0x004fc800078e021c */ /*08c0*/ IMAD R7, R21, R22, R7 ; /* 0x0000001615077224 */ /* 0x008fc800078e0207 */ /*08d0*/ IMAD R7, R23, R26, R7 ; /* 0x0000001a17077224 */ /* 0x020fc800078e0207 */ /*08e0*/ IMAD R7, R14, R27, R7 ; /* 0x0000001b0e077224 */ /* 0x000fc800078e0207 */ /*08f0*/ IMAD R7, R10, R25, R7 ; /* 0x000000190a077224 */ /* 0x000fc800078e0207 */ /*0900*/ IMAD R7, R16, R29, R7 ; /* 0x0000001d10077224 */ /* 0x010fc800078e0207 */ /*0910*/ IMAD R7, R24, R11, R7 ; /* 0x0000000b18077224 */ /* 0x000fe400078e0207 */ /*0920*/ IMAD.WIDE R24, R2, 0x4, R12 ; /* 0x0000000402187825 */ /* 0x000fc800078e020c */ /*0930*/ IMAD R28, R15, R18, R7 ; /* 0x000000120f1c7224 */ /* 0x000fe400078e0207 */ /*0940*/ ISETP.NE.OR P0, PT, R6, RZ, P0 ; /* 0x000000ff0600720c */ /* 0x000fda0000705670 */ /*0950*/ @!P0 BRA 0xaf0 ; /* 0x0000019000008947 */ /* 0x000fea0003800000 */ /*0960*/ MOV R8, UR6 ; /* 0x0000000600087c02 */ /* 0x000fe20008000f00 */ /*0970*/ IMAD.WIDE R14, R2, 0x4, R24 ; /* 0x00000004020e7825 */ /* 0x000fe200078e0218 */ /*0980*/ MOV R9, UR7 ; /* 0x0000000700097c02 */ /* 0x000fe20008000f00 */ /*0990*/ LDG.E R25, [R24.64] ; /* 0x0000000418197981 */ /* 0x000ea8000c1e1900 */ /*09a0*/ IMAD.WIDE R8, R3, 0x4, R8 ; /* 0x0000000403087825 */ /* 0x000fc800078e0208 */ /*09b0*/ IMAD.WIDE R12, R2.reuse, 0x4, R14 ; /* 0x00000004020c7825 */ /* 0x040fe200078e020e */ /*09c0*/ LDG.E R7, [R8.64] ; /* 0x0000000408077981 */ /* 0x000ea8000c1e1900 */ /*09d0*/ LDG.E R14, [R14.64] ; /* 0x000000040e0e7981 */ /* 0x000ee2000c1e1900 */ /*09e0*/ IMAD.WIDE R10, R2, 0x4, R12 ; /* 0x00000004020a7825 */ /* 0x000fc600078e020c */ /*09f0*/ LDG.E R16, [R8.64+0x4] ; /* 0x0000040408107981 */ /* 0x000ee8000c1e1900 */ /*0a00*/ LDG.E R18, [R12.64] ; /* 0x000000040c127981 */ /* 0x000f28000c1e1900 */ /*0a10*/ LDG.E R17, [R8.64+0x8] ; /* 0x0000080408117981 */ /* 0x000f28000c1e1900 */ /*0a20*/ LDG.E R19, [R8.64+0xc] ; /* 0x00000c0408137981 */ /* 0x000f68000c1e1900 */ /*0a30*/ LDG.E R20, [R10.64] ; /* 0x000000040a147981 */ /* 0x000f62000c1e1900 */ /*0a40*/ IADD3 R6, R6, -0x4, RZ ; /* 0xfffffffc06067810 */ /* 0x000fc80007ffe0ff */ /*0a50*/ ISETP.NE.AND P0, PT, R6, RZ, PT ; /* 0x000000ff0600720c */ /* 0x000fe20003f05270 */ /*0a60*/ UIADD3 UR6, UP0, UR6, 0x10, URZ ; /* 0x0000001006067890 */ /* 0x000fe2000ff1e03f */ /*0a70*/ IADD3 R4, R4, 0x4, RZ ; /* 0x0000000404047810 */ /* 0x000fc60007ffe0ff */ /*0a80*/ UIADD3.X UR7, URZ, UR7, URZ, UP0, !UPT ; /* 0x000000073f077290 */ /* 0x000fe200087fe43f */ /*0a90*/ IMAD R7, R25, R7, R28 ; /* 0x0000000719077224 */ /* 0x004fc800078e021c */ /*0aa0*/ IMAD R7, R14, R16, R7 ; /* 0x000000100e077224 */ /* 0x008fe400078e0207 */ /*0ab0*/ IMAD.WIDE R24, R2, 0x4, R10 ; /* 0x0000000402187825 */ /* 0x000fc800078e020a */ /*0ac0*/ IMAD R7, R18, R17, R7 ; /* 0x0000001112077224 */ /* 0x010fc800078e0207 */ /*0ad0*/ IMAD R28, R20, R19, R7 ; /* 0x00000013141c7224 */ /* 0x020fe200078e0207 */ /*0ae0*/ @P0 BRA 0x960 ; /* 0xfffffe7000000947 */ /* 0x000fea000383ffff */ /*0af0*/ ISETP.NE.AND P0, PT, R5, RZ, PT ; /* 0x000000ff0500720c */ /* 0x000fda0003f05270 */ /*0b00*/ @!P0 BRA 0xbf0 ; /* 0x000000e000008947 */ /* 0x000fea0003800000 */ /*0b10*/ HFMA2.MMA R9, -RZ, RZ, 0, 2.384185791015625e-07 ; /* 0x00000004ff097435 */ /* 0x000fe200000001ff */ /*0b20*/ IADD3 R6, R3, R4, RZ ; /* 0x0000000403067210 */ /* 0x000fe20007ffe0ff */ /*0b30*/ IMAD R4, R4, c[0x0][0x178], R0 ; /* 0x00005e0004047a24 */ /* 0x000fd000078e0200 */ /*0b40*/ IMAD.WIDE R6, R6, R9, c[0x0][0x160] ; /* 0x0000580006067625 */ /* 0x000fc800078e0209 */ /*0b50*/ IMAD.WIDE R8, R4, R9, c[0x0][0x168] ; /* 0x00005a0004087625 */ /* 0x000fca00078e0209 */ /*0b60*/ LDG.E R11, [R8.64] ; /* 0x00000004080b7981 */ /* 0x0000a8000c1e1900 */ /*0b70*/ LDG.E R4, [R6.64] ; /* 0x0000000406047981 */ /* 0x0002a2000c1e1900 */ /*0b80*/ IADD3 R5, R5, -0x1, RZ ; /* 0xffffffff05057810 */ /* 0x000fc80007ffe0ff */ /*0b90*/ ISETP.NE.AND P0, PT, R5, RZ, PT ; /* 0x000000ff0500720c */ /* 0x000fe20003f05270 */ /*0ba0*/ IMAD.WIDE R8, R2, 0x4, R8 ; /* 0x0000000402087825 */ /* 0x001fe200078e0208 */ /*0bb0*/ IADD3 R6, P1, R6, 0x4, RZ ; /* 0x0000000406067810 */ /* 0x002fc80007f3e0ff */ /*0bc0*/ IADD3.X R7, RZ, R7, RZ, P1, !PT ; /* 0x00000007ff077210 */ /* 0x000fe20000ffe4ff */ /*0bd0*/ IMAD R28, R11, R4, R28 ; /* 0x000000040b1c7224 */ /* 0x004fcc00078e021c */ /*0be0*/ @P0 BRA 0xb60 ; /* 0xffffff7000000947 */ /* 0x000fea000383ffff */ /*0bf0*/ IADD3 R3, R0, R3, RZ ; /* 0x0000000300037210 */ /* 0x000fe40007ffe0ff */ /*0c00*/ MOV R2, 0x4 ; /* 0x0000000400027802 */ /* 0x000fca0000000f00 */ /*0c10*/ IMAD.WIDE R2, R3, R2, c[0x0][0x170] ; /* 0x00005c0003027625 */ /* 0x000fca00078e0202 */ /*0c20*/ STG.E [R2.64], R28 ; /* 0x0000001c02007986 */ /* 0x000fe2000c101904 */ /*0c30*/ EXIT ; /* 0x000000000000794d */ /* 0x000fea0003800000 */ /*0c40*/ BRA 0xc40; /* 0xfffffff000007947 */ /* 0x000fc0000383ffff */ /*0c50*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0c60*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0c70*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0c80*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0c90*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0ca0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0cb0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0cc0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0cd0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0ce0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0cf0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ ..........
.text .amdgcn_target "amdgcn-amd-amdhsa--gfx1100" .protected _Z32MatrixMultiplication__CudaKernelPiS_S_i .globl _Z32MatrixMultiplication__CudaKernelPiS_S_i .p2align 8 .type _Z32MatrixMultiplication__CudaKernelPiS_S_i,@function _Z32MatrixMultiplication__CudaKernelPiS_S_i: s_clause 0x1 s_load_b32 s3, s[0:1], 0x2c s_load_b32 s2, s[0:1], 0x18 v_and_b32_e32 v2, 0x3ff, v0 v_bfe_u32 v3, v0, 10, 10 s_waitcnt lgkmcnt(0) s_and_b32 s4, s3, 0xffff s_lshr_b32 s3, s3, 16 s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1) v_mad_u64_u32 v[0:1], null, s14, s4, v[2:3] v_mad_u64_u32 v[1:2], null, s15, s3, v[3:4] s_mov_b32 s3, exec_lo v_max_i32_e32 v2, v0, v1 s_delay_alu instid0(VALU_DEP_1) v_cmpx_gt_i32_e64 s2, v2 s_cbranch_execz .LBB0_6 s_cmp_lt_i32 s2, 1 s_cbranch_scc1 .LBB0_4 s_load_b128 s[4:7], s[0:1], 0x0 v_mul_lo_u32 v2, v0, s2 s_mov_b32 s3, s2 v_mov_b32_e32 v5, v1 s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1) v_ashrrev_i32_e32 v3, 31, v2 v_lshlrev_b64 v[3:4], 2, v[2:3] v_mov_b32_e32 v2, 0 s_waitcnt lgkmcnt(0) s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3) v_add_co_u32 v3, vcc_lo, s4, v3 v_add_co_ci_u32_e32 v4, vcc_lo, s5, v4, vcc_lo .p2align 6 .LBB0_3: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1) v_ashrrev_i32_e32 v6, 31, v5 s_add_i32 s3, s3, -1 s_cmp_eq_u32 s3, 0 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_lshlrev_b64 v[6:7], 2, v[5:6] v_add_co_u32 v6, vcc_lo, s6, v6 s_delay_alu instid0(VALU_DEP_2) v_add_co_ci_u32_e32 v7, vcc_lo, s7, v7, vcc_lo global_load_b32 v8, v[3:4], off global_load_b32 v9, v[6:7], off s_waitcnt vmcnt(0) v_mad_u64_u32 v[6:7], null, v9, v8, v[2:3] v_add_co_u32 v3, vcc_lo, v3, 4 v_add_co_ci_u32_e32 v4, vcc_lo, 0, v4, vcc_lo s_delay_alu instid0(VALU_DEP_3) v_dual_mov_b32 v2, v6 :: v_dual_add_nc_u32 v5, s2, v5 s_cbranch_scc0 .LBB0_3 s_branch .LBB0_5 .LBB0_4: v_mov_b32_e32 v2, 0 .LBB0_5: s_load_b64 s[0:1], s[0:1], 0x10 s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) v_mad_u64_u32 v[3:4], null, v0, s2, v[1:2] v_ashrrev_i32_e32 v4, 31, v3 s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1) v_lshlrev_b64 v[0:1], 2, v[3:4] s_waitcnt lgkmcnt(0) v_add_co_u32 v0, vcc_lo, s0, v0 s_delay_alu instid0(VALU_DEP_2) v_add_co_ci_u32_e32 v1, vcc_lo, s1, v1, vcc_lo global_store_b32 v[0:1], v2, off .LBB0_6: s_nop 0 s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) s_endpgm .section .rodata,"a",@progbits .p2align 6, 0x0 .amdhsa_kernel _Z32MatrixMultiplication__CudaKernelPiS_S_i .amdhsa_group_segment_fixed_size 0 .amdhsa_private_segment_fixed_size 0 .amdhsa_kernarg_size 288 .amdhsa_user_sgpr_count 14 .amdhsa_user_sgpr_dispatch_ptr 0 .amdhsa_user_sgpr_queue_ptr 0 .amdhsa_user_sgpr_kernarg_segment_ptr 1 .amdhsa_user_sgpr_dispatch_id 0 .amdhsa_user_sgpr_private_segment_size 0 .amdhsa_wavefront_size32 1 .amdhsa_uses_dynamic_stack 0 .amdhsa_enable_private_segment 0 .amdhsa_system_sgpr_workgroup_id_x 1 .amdhsa_system_sgpr_workgroup_id_y 1 .amdhsa_system_sgpr_workgroup_id_z 0 .amdhsa_system_sgpr_workgroup_info 0 .amdhsa_system_vgpr_workitem_id 1 .amdhsa_next_free_vgpr 10 .amdhsa_next_free_sgpr 16 .amdhsa_float_round_mode_32 0 .amdhsa_float_round_mode_16_64 0 .amdhsa_float_denorm_mode_32 3 .amdhsa_float_denorm_mode_16_64 3 .amdhsa_dx10_clamp 1 .amdhsa_ieee_mode 1 .amdhsa_fp16_overflow 0 .amdhsa_workgroup_processor_mode 1 .amdhsa_memory_ordered 1 .amdhsa_forward_progress 0 .amdhsa_shared_vgpr_count 0 .amdhsa_exception_fp_ieee_invalid_op 0 .amdhsa_exception_fp_denorm_src 0 .amdhsa_exception_fp_ieee_div_zero 0 .amdhsa_exception_fp_ieee_overflow 0 .amdhsa_exception_fp_ieee_underflow 0 .amdhsa_exception_fp_ieee_inexact 0 .amdhsa_exception_int_div_zero 0 .end_amdhsa_kernel .text .Lfunc_end0: .size _Z32MatrixMultiplication__CudaKernelPiS_S_i, .Lfunc_end0-_Z32MatrixMultiplication__CudaKernelPiS_S_i .section .AMDGPU.csdata,"",@progbits .text .p2alignl 7, 3214868480 .fill 96, 4, 3214868480 .type __hip_cuid_,@object .section .bss,"aw",@nobits .globl __hip_cuid_ __hip_cuid_: .byte 0 .size __hip_cuid_, 1 .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym __hip_cuid_ .amdgpu_metadata --- amdhsa.kernels: - .args: - .address_space: global .offset: 0 .size: 8 .value_kind: global_buffer - .address_space: global .offset: 8 .size: 8 .value_kind: global_buffer - .address_space: global .offset: 16 .size: 8 .value_kind: global_buffer - .offset: 24 .size: 4 .value_kind: by_value - .offset: 32 .size: 4 .value_kind: hidden_block_count_x - .offset: 36 .size: 4 .value_kind: hidden_block_count_y - .offset: 40 .size: 4 .value_kind: hidden_block_count_z - .offset: 44 .size: 2 .value_kind: hidden_group_size_x - .offset: 46 .size: 2 .value_kind: hidden_group_size_y - .offset: 48 .size: 2 .value_kind: hidden_group_size_z - .offset: 50 .size: 2 .value_kind: hidden_remainder_x - .offset: 52 .size: 2 .value_kind: hidden_remainder_y - .offset: 54 .size: 2 .value_kind: hidden_remainder_z - .offset: 72 .size: 8 .value_kind: hidden_global_offset_x - .offset: 80 .size: 8 .value_kind: hidden_global_offset_y - .offset: 88 .size: 8 .value_kind: hidden_global_offset_z - .offset: 96 .size: 2 .value_kind: hidden_grid_dims .group_segment_fixed_size: 0 .kernarg_segment_align: 8 .kernarg_segment_size: 288 .language: OpenCL C .language_version: - 2 - 0 .max_flat_workgroup_size: 1024 .name: _Z32MatrixMultiplication__CudaKernelPiS_S_i .private_segment_fixed_size: 0 .sgpr_count: 18 .sgpr_spill_count: 0 .symbol: _Z32MatrixMultiplication__CudaKernelPiS_S_i.kd .uniform_work_group_size: 1 .uses_dynamic_stack: false .vgpr_count: 10 .vgpr_spill_count: 0 .wavefront_size: 32 .workgroup_processor_mode: 1 amdhsa.target: amdgcn-amd-amdhsa--gfx1100 amdhsa.version: - 1 - 2 ... .end_amdgpu_metadata
You are a helpful coding assistant specialized in converting from CUDA host assembly to AMD host assembly.
.file "tmpxft_0011c356_00000000-6_MatrixMultiplication__CudaKernel.cudafe1.cpp" .text #APP #NO_APP .type _ZL26__cudaUnregisterBinaryUtilv, @function _ZL26__cudaUnregisterBinaryUtilv: .LFB2029: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaUnregisterFatBinary@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2029: .size _ZL26__cudaUnregisterBinaryUtilv, .-_ZL26__cudaUnregisterBinaryUtilv .globl _Z57__device_stub__Z32MatrixMultiplication__CudaKernelPiS_S_iPiS_S_i .type _Z57__device_stub__Z32MatrixMultiplication__CudaKernelPiS_S_iPiS_S_i, @function _Z57__device_stub__Z32MatrixMultiplication__CudaKernelPiS_S_iPiS_S_i: .LFB2051: .cfi_startproc endbr64 subq $152, %rsp .cfi_def_cfa_offset 160 movq %rdi, 24(%rsp) movq %rsi, 16(%rsp) movq %rdx, 8(%rsp) movl %ecx, 4(%rsp) movq %fs:40, %rax movq %rax, 136(%rsp) xorl %eax, %eax leaq 24(%rsp), %rax movq %rax, 96(%rsp) leaq 16(%rsp), %rax movq %rax, 104(%rsp) leaq 8(%rsp), %rax movq %rax, 112(%rsp) leaq 4(%rsp), %rax movq %rax, 120(%rsp) movl $1, 48(%rsp) movl $1, 52(%rsp) movl $1, 56(%rsp) movl $1, 60(%rsp) movl $1, 64(%rsp) movl $1, 68(%rsp) leaq 40(%rsp), %rcx leaq 32(%rsp), %rdx leaq 60(%rsp), %rsi leaq 48(%rsp), %rdi call __cudaPopCallConfiguration@PLT testl %eax, %eax je .L7 .L3: movq 136(%rsp), %rax subq %fs:40, %rax jne .L8 addq $152, %rsp .cfi_remember_state .cfi_def_cfa_offset 8 ret .L7: .cfi_restore_state pushq 40(%rsp) .cfi_def_cfa_offset 168 pushq 40(%rsp) .cfi_def_cfa_offset 176 leaq 112(%rsp), %r9 movq 76(%rsp), %rcx movl 84(%rsp), %r8d movq 64(%rsp), %rsi movl 72(%rsp), %edx leaq _Z32MatrixMultiplication__CudaKernelPiS_S_i(%rip), %rdi call cudaLaunchKernel@PLT addq $16, %rsp .cfi_def_cfa_offset 160 jmp .L3 .L8: call __stack_chk_fail@PLT .cfi_endproc .LFE2051: .size _Z57__device_stub__Z32MatrixMultiplication__CudaKernelPiS_S_iPiS_S_i, .-_Z57__device_stub__Z32MatrixMultiplication__CudaKernelPiS_S_iPiS_S_i .globl _Z32MatrixMultiplication__CudaKernelPiS_S_i .type _Z32MatrixMultiplication__CudaKernelPiS_S_i, @function _Z32MatrixMultiplication__CudaKernelPiS_S_i: .LFB2052: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 call _Z57__device_stub__Z32MatrixMultiplication__CudaKernelPiS_S_iPiS_S_i addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2052: .size _Z32MatrixMultiplication__CudaKernelPiS_S_i, .-_Z32MatrixMultiplication__CudaKernelPiS_S_i .section .rodata.str1.8,"aMS",@progbits,1 .align 8 .LC0: .string "_Z32MatrixMultiplication__CudaKernelPiS_S_i" .text .type _ZL24__sti____cudaRegisterAllv, @function _ZL24__sti____cudaRegisterAllv: .LFB2054: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 leaq _ZL15__fatDeviceText(%rip), %rdi call __cudaRegisterFatBinary@PLT movq %rax, %rdi movq %rax, _ZL20__cudaFatCubinHandle(%rip) pushq $0 .cfi_def_cfa_offset 24 pushq $0 .cfi_def_cfa_offset 32 pushq $0 .cfi_def_cfa_offset 40 pushq $0 .cfi_def_cfa_offset 48 movl $0, %r9d movl $-1, %r8d leaq .LC0(%rip), %rdx movq %rdx, %rcx leaq _Z32MatrixMultiplication__CudaKernelPiS_S_i(%rip), %rsi call __cudaRegisterFunction@PLT addq $32, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaRegisterFatBinaryEnd@PLT leaq _ZL26__cudaUnregisterBinaryUtilv(%rip), %rdi call atexit@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2054: .size _ZL24__sti____cudaRegisterAllv, .-_ZL24__sti____cudaRegisterAllv .section .init_array,"aw" .align 8 .quad _ZL24__sti____cudaRegisterAllv .section .nvFatBinSegment,"aw" .align 8 .type _ZL15__fatDeviceText, @object .size _ZL15__fatDeviceText, 24 _ZL15__fatDeviceText: .long 1180844977 .long 1 .quad fatbinData .quad 0 .local _ZL20__cudaFatCubinHandle .comm _ZL20__cudaFatCubinHandle,8,8 .ident "GCC: (Ubuntu 13.3.0-6ubuntu2~24.04) 13.3.0" .section .note.GNU-stack,"",@progbits .section .note.gnu.property,"a" .align 8 .long 1f - 0f .long 4f - 1f .long 5 0: .string "GNU" 1: .align 8 .long 0xc0000002 .long 3f - 2f 2: .long 0x3 3: .align 8 4:
.text .file "MatrixMultiplication__CudaKernel.hip" .globl _Z47__device_stub__MatrixMultiplication__CudaKernelPiS_S_i # -- Begin function _Z47__device_stub__MatrixMultiplication__CudaKernelPiS_S_i .p2align 4, 0x90 .type _Z47__device_stub__MatrixMultiplication__CudaKernelPiS_S_i,@function _Z47__device_stub__MatrixMultiplication__CudaKernelPiS_S_i: # @_Z47__device_stub__MatrixMultiplication__CudaKernelPiS_S_i .cfi_startproc # %bb.0: subq $120, %rsp .cfi_def_cfa_offset 128 movq %rdi, 72(%rsp) movq %rsi, 64(%rsp) movq %rdx, 56(%rsp) movl %ecx, 4(%rsp) leaq 72(%rsp), %rax movq %rax, 80(%rsp) leaq 64(%rsp), %rax movq %rax, 88(%rsp) leaq 56(%rsp), %rax movq %rax, 96(%rsp) leaq 4(%rsp), %rax movq %rax, 104(%rsp) leaq 40(%rsp), %rdi leaq 24(%rsp), %rsi leaq 16(%rsp), %rdx leaq 8(%rsp), %rcx callq __hipPopCallConfiguration movq 40(%rsp), %rsi movl 48(%rsp), %edx movq 24(%rsp), %rcx movl 32(%rsp), %r8d leaq 80(%rsp), %r9 movl $_Z32MatrixMultiplication__CudaKernelPiS_S_i, %edi pushq 8(%rsp) .cfi_adjust_cfa_offset 8 pushq 24(%rsp) .cfi_adjust_cfa_offset 8 callq hipLaunchKernel addq $136, %rsp .cfi_adjust_cfa_offset -136 retq .Lfunc_end0: .size _Z47__device_stub__MatrixMultiplication__CudaKernelPiS_S_i, .Lfunc_end0-_Z47__device_stub__MatrixMultiplication__CudaKernelPiS_S_i .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_ctor .type __hip_module_ctor,@function __hip_module_ctor: # @__hip_module_ctor .cfi_startproc # %bb.0: subq $40, %rsp .cfi_def_cfa_offset 48 cmpq $0, __hip_gpubin_handle(%rip) jne .LBB1_2 # %bb.1: movl $__hip_fatbin_wrapper, %edi callq __hipRegisterFatBinary movq %rax, __hip_gpubin_handle(%rip) .LBB1_2: movq __hip_gpubin_handle(%rip), %rdi xorps %xmm0, %xmm0 movups %xmm0, 16(%rsp) movups %xmm0, (%rsp) movl $_Z32MatrixMultiplication__CudaKernelPiS_S_i, %esi movl $.L__unnamed_1, %edx movl $.L__unnamed_1, %ecx movl $-1, %r8d xorl %r9d, %r9d callq __hipRegisterFunction movl $__hip_module_dtor, %edi addq $40, %rsp .cfi_def_cfa_offset 8 jmp atexit # TAILCALL .Lfunc_end1: .size __hip_module_ctor, .Lfunc_end1-__hip_module_ctor .cfi_endproc # -- End function .p2align 4, 0x90 # -- Begin function __hip_module_dtor .type __hip_module_dtor,@function __hip_module_dtor: # @__hip_module_dtor .cfi_startproc # %bb.0: movq __hip_gpubin_handle(%rip), %rdi testq %rdi, %rdi je .LBB2_2 # %bb.1: pushq %rax .cfi_def_cfa_offset 16 callq __hipUnregisterFatBinary movq $0, __hip_gpubin_handle(%rip) addq $8, %rsp .cfi_def_cfa_offset 8 .LBB2_2: retq .Lfunc_end2: .size __hip_module_dtor, .Lfunc_end2-__hip_module_dtor .cfi_endproc # -- End function .type _Z32MatrixMultiplication__CudaKernelPiS_S_i,@object # @_Z32MatrixMultiplication__CudaKernelPiS_S_i .section .rodata,"a",@progbits .globl _Z32MatrixMultiplication__CudaKernelPiS_S_i .p2align 3, 0x0 _Z32MatrixMultiplication__CudaKernelPiS_S_i: .quad _Z47__device_stub__MatrixMultiplication__CudaKernelPiS_S_i .size _Z32MatrixMultiplication__CudaKernelPiS_S_i, 8 .type .L__unnamed_1,@object # @0 .section .rodata.str1.1,"aMS",@progbits,1 .L__unnamed_1: .asciz "_Z32MatrixMultiplication__CudaKernelPiS_S_i" .size .L__unnamed_1, 44 .type __hip_fatbin_wrapper,@object # @__hip_fatbin_wrapper .section .hipFatBinSegment,"a",@progbits .p2align 3, 0x0 __hip_fatbin_wrapper: .long 1212764230 # 0x48495046 .long 1 # 0x1 .quad __hip_fatbin .quad 0 .size __hip_fatbin_wrapper, 24 .type __hip_gpubin_handle,@object # @__hip_gpubin_handle .local __hip_gpubin_handle .comm __hip_gpubin_handle,8,8 .section .init_array,"aw",@init_array .p2align 3, 0x0 .quad __hip_module_ctor .type __hip_cuid_,@object # @__hip_cuid_ .bss .globl __hip_cuid_ __hip_cuid_: .byte 0 # 0x0 .size __hip_cuid_, 1 .section ".linker-options","e",@llvm_linker_options .ident "AMD clang version 18.0.0git (https://github.com/RadeonOpenCompute/llvm-project roc-6.3.2 25012 e5bf7e55c91490b07c49d8960fa7983d864936c4)" .section ".note.GNU-stack","",@progbits .addrsig .addrsig_sym _Z47__device_stub__MatrixMultiplication__CudaKernelPiS_S_i .addrsig_sym __hip_module_ctor .addrsig_sym __hip_module_dtor .addrsig_sym _Z32MatrixMultiplication__CudaKernelPiS_S_i .addrsig_sym __hip_fatbin .addrsig_sym __hip_fatbin_wrapper .addrsig_sym __hip_cuid_
You are a helpful coding assistant specialized in converting CUDA source code to CUDA device assembly.
#include <cuda.h> #include <stdio.h> #include <stdlib.h> #define N 3 // dim of matrix //Fattened matrix multiplication . Kernel does not support x,y addressing __global__ void mat_multiply(int* d_mat1, int* d_mat2, int* d_mat3, int width) { int k,sum=0; int col = blockDim.x * blockIdx.x + threadIdx.x; int row = blockDim.y * blockIdx.y + threadIdx.y; if(row<width && col<width) { for(k=0;k<width;k++) { sum += d_mat1[row*width+k] * d_mat2[k*width+col]; } d_mat3[row*width+col] = sum; } } int main() { int i,j; int SIZE = N*N; //int BYTES = SIZE*sizeof(int); int *d_mat1, *d_mat2, *d_mat3; // allocate memory on the device cudaMallocManaged(&d_mat1,N*N*sizeof(int)); cudaMallocManaged(&d_mat2,N*N*sizeof(int)); cudaMallocManaged(&d_mat3,N*N*sizeof(int)); // generate matrix on host for(i=0;i<N*N;i++) //linearize array { d_mat1[i] = 1; d_mat2[i] = 1; d_mat3[i] = 0; } dim3 dimGrid(1,1); dim3 dimBlock(N,N); // lauch kernel mat_multiply<<<dimGrid,dimBlock>>>(d_mat1,d_mat2,d_mat3,N); cudaDeviceSynchronize(); for(i=0;i<N*N;i++) { printf("%d ",d_mat3[i]); if(i%N==0 && i>N) printf("\n"); } printf("\n"); }
code for sm_80 Function : _Z12mat_multiplyPiS_S_i .headerflags @"EF_CUDA_TEXMODE_UNIFIED EF_CUDA_64BIT_ADDRESS EF_CUDA_SM80 EF_CUDA_VIRTUAL_SM(EF_CUDA_SM80)" /*0000*/ MOV R1, c[0x0][0x28] ; /* 0x00000a0000017a02 */ /* 0x000fe40000000f00 */ /*0010*/ S2R R0, SR_CTAID.X ; /* 0x0000000000007919 */ /* 0x000e280000002500 */ /*0020*/ S2R R3, SR_TID.X ; /* 0x0000000000037919 */ /* 0x000e280000002100 */ /*0030*/ S2R R2, SR_CTAID.Y ; /* 0x0000000000027919 */ /* 0x000e680000002600 */ /*0040*/ S2R R5, SR_TID.Y ; /* 0x0000000000057919 */ /* 0x000e620000002200 */ /*0050*/ IMAD R0, R0, c[0x0][0x0], R3 ; /* 0x0000000000007a24 */ /* 0x001fca00078e0203 */ /*0060*/ ISETP.GE.AND P0, PT, R0, c[0x0][0x178], PT ; /* 0x00005e0000007a0c */ /* 0x000fe20003f06270 */ /*0070*/ IMAD R3, R2, c[0x0][0x4], R5 ; /* 0x0000010002037a24 */ /* 0x002fca00078e0205 */ /*0080*/ ISETP.GE.OR P0, PT, R3, c[0x0][0x178], P0 ; /* 0x00005e0003007a0c */ /* 0x000fda0000706670 */ /*0090*/ @P0 EXIT ; /* 0x000000000000094d */ /* 0x000fea0003800000 */ /*00a0*/ MOV R2, c[0x0][0x178] ; /* 0x00005e0000027a02 */ /* 0x000fe20000000f00 */ /*00b0*/ ULDC.64 UR4, c[0x0][0x118] ; /* 0x0000460000047ab9 */ /* 0x000fe20000000a00 */ /*00c0*/ HFMA2.MMA R28, -RZ, RZ, 0, 0 ; /* 0x00000000ff1c7435 */ /* 0x000fe200000001ff */ /*00d0*/ IMAD R3, R3, c[0x0][0x178], RZ ; /* 0x00005e0003037a24 */ /* 0x000fe200078e02ff */ /*00e0*/ ISETP.GE.AND P0, PT, R2, 0x1, PT ; /* 0x000000010200780c */ /* 0x000fda0003f06270 */ /*00f0*/ @!P0 BRA 0xbf0 ; /* 0x00000af000008947 */ /* 0x000fea0003800000 */ /*0100*/ IADD3 R4, R2.reuse, -0x1, RZ ; /* 0xffffffff02047810 */ /* 0x040fe40007ffe0ff */ /*0110*/ LOP3.LUT R5, R2, 0x3, RZ, 0xc0, !PT ; /* 0x0000000302057812 */ /* 0x000fe400078ec0ff */ /*0120*/ ISETP.GE.U32.AND P0, PT, R4, 0x3, PT ; /* 0x000000030400780c */ /* 0x000fe40003f06070 */ /*0130*/ MOV R28, RZ ; /* 0x000000ff001c7202 */ /* 0x000fe40000000f00 */ /*0140*/ MOV R4, RZ ; /* 0x000000ff00047202 */ /* 0x000fd20000000f00 */ /*0150*/ @!P0 BRA 0xaf0 ; /* 0x0000099000008947 */ /* 0x000fea0003800000 */ /*0160*/ IADD3 R6, -R5, c[0x0][0x178], RZ ; /* 0x00005e0005067a10 */ /* 0x000fe20007ffe1ff */ /*0170*/ HFMA2.MMA R25, -RZ, RZ, 0, 2.384185791015625e-07 ; /* 0x00000004ff197435 */ /* 0x000fe200000001ff */ /*0180*/ ULDC.64 UR6, c[0x0][0x160] ; /* 0x0000580000067ab9 */ /* 0x000fe20000000a00 */ /*0190*/ MOV R28, RZ ; /* 0x000000ff001c7202 */ /* 0x000fe40000000f00 */ /*01a0*/ ISETP.GT.AND P0, PT, R6, RZ, PT ; /* 0x000000ff0600720c */ /* 0x000fcc0003f04270 */ /*01b0*/ IMAD.WIDE R24, R0, R25, c[0x0][0x168] ; /* 0x00005a0000187625 */ /* 0x000fce00078e0219 */ /*01c0*/ @!P0 BRA 0x960 ; /* 0x0000079000008947 */ /* 0x000fea0003800000 */ /*01d0*/ ISETP.GT.AND P1, PT, R6, 0xc, PT ; /* 0x0000000c0600780c */ /* 0x000fe40003f24270 */ /*01e0*/ PLOP3.LUT P0, PT, PT, PT, PT, 0x80, 0x0 ; /* 0x000000000000781c */ /* 0x000fd60003f0f070 */ /*01f0*/ @!P1 BRA 0x6a0 ; /* 0x000004a000009947 */ /* 0x000fea0003800000 */ /*0200*/ PLOP3.LUT P0, PT, PT, PT, PT, 0x8, 0x0 ; /* 0x000000000000781c */ /* 0x000fe40003f0e170 */ /*0210*/ MOV R12, UR6 ; /* 0x00000006000c7c02 */ /* 0x000fe20008000f00 */ /*0220*/ LDG.E R29, [R24.64] ; /* 0x00000004181d7981 */ /* 0x0000a2000c1e1900 */ /*0230*/ MOV R13, UR7 ; /* 0x00000007000d7c02 */ /* 0x000fca0008000f00 */ /*0240*/ IMAD.WIDE R12, R3, 0x4, R12 ; /* 0x00000004030c7825 */ /* 0x000fca00078e020c */ /*0250*/ LDG.E R27, [R12.64] ; /* 0x000000040c1b7981 */ /* 0x000ea2000c1e1900 */ /*0260*/ IMAD.WIDE R10, R2, 0x4, R24 ; /* 0x00000004020a7825 */ /* 0x000fc600078e0218 */ /*0270*/ LDG.E R17, [R12.64+0x4] ; /* 0x000004040c117981 */ /* 0x000ee6000c1e1900 */ /*0280*/ IMAD.WIDE R18, R2.reuse, 0x4, R10 ; /* 0x0000000402127825 */ /* 0x040fe200078e020a */ /*0290*/ LDG.E R16, [R10.64] ; /* 0x000000040a107981 */ /* 0x0002e8000c1e1900 */ /*02a0*/ LDG.E R7, [R12.64+0xc] ; /* 0x00000c040c077981 */ /* 0x000f22000c1e1900 */ /*02b0*/ IMAD.WIDE R14, R2, 0x4, R18 ; /* 0x00000004020e7825 */ /* 0x000fc600078e0212 */ /*02c0*/ LDG.E R18, [R18.64] ; /* 0x0000000412127981 */ /* 0x000b26000c1e1900 */ /*02d0*/ IMAD.WIDE R20, R2.reuse, 0x4, R14 ; /* 0x0000000402147825 */ /* 0x040fe200078e020e */ /*02e0*/ LDG.E R26, [R14.64] ; /* 0x000000040e1a7981 */ /* 0x000128000c1e1900 */ /*02f0*/ LDG.E R9, [R12.64+0x10] ; /* 0x000010040c097981 */ /* 0x000f28000c1e1900 */ /*0300*/ LDG.E R19, [R12.64+0x8] ; /* 0x000008040c137981 */ /* 0x020f22000c1e1900 */ /*0310*/ IMAD.WIDE R14, R2, 0x4, R20 ; /* 0x00000004020e7825 */ /* 0x001fc600078e0214 */ /*0320*/ LDG.E R20, [R20.64] ; /* 0x0000000414147981 */ /* 0x000166000c1e1900 */ /*0330*/ IMAD.WIDE R22, R2.reuse, 0x4, R14 ; /* 0x0000000402167825 */ /* 0x040fe200078e020e */ /*0340*/ LDG.E R8, [R14.64] ; /* 0x000000040e087981 */ /* 0x000168000c1e1900 */ /*0350*/ LDG.E R11, [R12.64+0x14] ; /* 0x000014040c0b7981 */ /* 0x002f62000c1e1900 */ /*0360*/ IMAD.WIDE R24, R2, 0x4, R22 ; /* 0x0000000402187825 */ /* 0x000fc600078e0216 */ /*0370*/ LDG.E R10, [R22.64] ; /* 0x00000004160a7981 */ /* 0x000368000c1e1900 */ /*0380*/ LDG.E R21, [R12.64+0x18] ; /* 0x000018040c157981 */ /* 0x001f62000c1e1900 */ /*0390*/ IMAD R29, R29, R27, R28 ; /* 0x0000001b1d1d7224 */ /* 0x004fc600078e021c */ /*03a0*/ LDG.E R27, [R12.64+0x1c] ; /* 0x00001c040c1b7981 */ /* 0x000ea8000c1e1900 */ /*03b0*/ LDG.E R28, [R24.64] ; /* 0x00000004181c7981 */ /* 0x0000a2000c1e1900 */ /*03c0*/ IMAD.WIDE R14, R2, 0x4, R24 ; /* 0x00000004020e7825 */ /* 0x000fc800078e0218 */ /*03d0*/ IMAD R29, R16, R17, R29 ; /* 0x00000011101d7224 */ /* 0x008fe400078e021d */ /*03e0*/ IMAD.WIDE R16, R2, 0x4, R14 ; /* 0x0000000402107825 */ /* 0x000fe400078e020e */ /*03f0*/ LDG.E R14, [R14.64] ; /* 0x000000040e0e7981 */ /* 0x0006a4000c1e1900 */ /*0400*/ IMAD R29, R18, R19, R29 ; /* 0x00000013121d7224 */ /* 0x010fe400078e021d */ /*0410*/ IMAD.WIDE R18, R2, 0x4, R16 ; /* 0x0000000402127825 */ /* 0x000fe400078e0210 */ /*0420*/ LDG.E R16, [R16.64] ; /* 0x0000000410107981 */ /* 0x0008a4000c1e1900 */ /*0430*/ IMAD R26, R26, R7, R29 ; /* 0x000000071a1a7224 */ /* 0x000fc400078e021d */ /*0440*/ IMAD.WIDE R22, R2.reuse, 0x4, R18 ; /* 0x0000000402167825 */ /* 0x042fe200078e0212 */ /*0450*/ LDG.E R7, [R12.64+0x20] ; /* 0x000020040c077981 */ /* 0x000ea8000c1e1900 */ /*0460*/ LDG.E R29, [R12.64+0x24] ; /* 0x000024040c1d7981 */ /* 0x000ea2000c1e1900 */ /*0470*/ IMAD.WIDE R24, R2, 0x4, R22 ; /* 0x0000000402187825 */ /* 0x001fc600078e0216 */ /*0480*/ LDG.E R18, [R18.64] ; /* 0x0000000412127981 */ /* 0x0000a2000c1e1900 */ /*0490*/ IMAD R9, R20, R9, R26 ; /* 0x0000000914097224 */ /* 0x020fc600078e021a */ /*04a0*/ LDG.E R26, [R12.64+0x28] ; /* 0x000028040c1a7981 */ /* 0x000f62000c1e1900 */ /*04b0*/ IMAD R11, R8, R11, R9 ; /* 0x0000000b080b7224 */ /* 0x000fe400078e0209 */ /*04c0*/ IMAD.WIDE R8, R2, 0x4, R24 ; /* 0x0000000402087825 */ /* 0x000fe200078e0218 */ /*04d0*/ LDG.E R22, [R22.64] ; /* 0x0000000416167981 */ /* 0x000368000c1e1900 */ /*04e0*/ LDG.E R17, [R12.64+0x2c] ; /* 0x00002c040c117981 */ /* 0x010f22000c1e1900 */ /*04f0*/ IMAD R21, R10, R21, R11 ; /* 0x000000150a157224 */ /* 0x000fc600078e020b */ /*0500*/ LDG.E R15, [R24.64] ; /* 0x00000004180f7981 */ /* 0x008722000c1e1900 */ /*0510*/ IMAD.WIDE R10, R2, 0x4, R8 ; /* 0x00000004020a7825 */ /* 0x000fc600078e0208 */ /*0520*/ LDG.E R19, [R8.64] ; /* 0x0000000408137981 */ /* 0x001128000c1e1900 */ /*0530*/ LDG.E R23, [R10.64] ; /* 0x000000040a177981 */ /* 0x002f28000c1e1900 */ /*0540*/ LDG.E R24, [R12.64+0x30] ; /* 0x000030040c187981 */ /* 0x008ee8000c1e1900 */ /*0550*/ LDG.E R25, [R12.64+0x38] ; /* 0x000038040c197981 */ /* 0x000ee8000c1e1900 */ /*0560*/ LDG.E R8, [R12.64+0x3c] ; /* 0x00003c040c087981 */ /* 0x001ee2000c1e1900 */ /*0570*/ IMAD R9, R28, R27, R21 ; /* 0x0000001b1c097224 */ /* 0x004fc600078e0215 */ /*0580*/ LDG.E R28, [R12.64+0x34] ; /* 0x000034040c1c7981 */ /* 0x000ea2000c1e1900 */ /*0590*/ IMAD.WIDE R20, R2, 0x4, R10 ; /* 0x0000000402147825 */ /* 0x000fca00078e020a */ /*05a0*/ LDG.E R27, [R20.64] ; /* 0x00000004141b7981 */ /* 0x000ea2000c1e1900 */ /*05b0*/ IADD3 R6, R6, -0x10, RZ ; /* 0xfffffff006067810 */ /* 0x000fc80007ffe0ff */ /*05c0*/ ISETP.GT.AND P1, PT, R6, 0xc, PT ; /* 0x0000000c0600780c */ /* 0x000fe20003f24270 */ /*05d0*/ IMAD R7, R14, R7, R9 ; /* 0x000000070e077224 */ /* 0x000fc800078e0209 */ /*05e0*/ IMAD R7, R16, R29, R7 ; /* 0x0000001d10077224 */ /* 0x000fc800078e0207 */ /*05f0*/ IMAD R7, R18, R26, R7 ; /* 0x0000001a12077224 */ /* 0x020fc800078e0207 */ /*0600*/ IMAD R7, R22, R17, R7 ; /* 0x0000001116077224 */ /* 0x010fe200078e0207 */ /*0610*/ UIADD3 UR6, UP0, UR6, 0x40, URZ ; /* 0x0000004006067890 */ /* 0x000fe2000ff1e03f */ /*0620*/ IADD3 R4, R4, 0x10, RZ ; /* 0x0000001004047810 */ /* 0x000fc60007ffe0ff */ /*0630*/ UIADD3.X UR7, URZ, UR7, URZ, UP0, !UPT ; /* 0x000000073f077290 */ /* 0x000fe200087fe43f */ /*0640*/ IMAD R7, R15, R24, R7 ; /* 0x000000180f077224 */ /* 0x008fc800078e0207 */ /*0650*/ IMAD R28, R19, R28, R7 ; /* 0x0000001c131c7224 */ /* 0x004fc800078e0207 */ /*0660*/ IMAD R28, R23, R25, R28 ; /* 0x00000019171c7224 */ /* 0x000fe400078e021c */ /*0670*/ IMAD.WIDE R24, R2, 0x4, R20 ; /* 0x0000000402187825 */ /* 0x000fc800078e0214 */ /*0680*/ IMAD R28, R27, R8, R28 ; /* 0x000000081b1c7224 */ /* 0x000fe200078e021c */ /*0690*/ @P1 BRA 0x210 ; /* 0xfffffb7000001947 */ /* 0x000fea000383ffff */ /*06a0*/ ISETP.GT.AND P1, PT, R6, 0x4, PT ; /* 0x000000040600780c */ /* 0x000fda0003f24270 */ /*06b0*/ @!P1 BRA 0x940 ; /* 0x0000028000009947 */ /* 0x000fea0003800000 */ /*06c0*/ IMAD.WIDE R16, R2, 0x4, R24 ; /* 0x0000000402107825 */ /* 0x000fe200078e0218 */ /*06d0*/ MOV R8, UR6 ; /* 0x0000000600087c02 */ /* 0x000fe20008000f00 */ /*06e0*/ LDG.E R7, [R24.64] ; /* 0x0000000418077981 */ /* 0x0000a2000c1e1900 */ /*06f0*/ MOV R9, UR7 ; /* 0x0000000700097c02 */ /* 0x000fc60008000f00 */ /*0700*/ IMAD.WIDE R12, R2.reuse, 0x4, R16 ; /* 0x00000004020c7825 */ /* 0x040fe200078e0210 */ /*0710*/ LDG.E R21, [R16.64] ; /* 0x0000000410157981 */ /* 0x0002e6000c1e1900 */ /*0720*/ IMAD.WIDE R8, R3, 0x4, R8 ; /* 0x0000000403087825 */ /* 0x000fe200078e0208 */ /*0730*/ LDG.E R23, [R12.64] ; /* 0x000000040c177981 */ /* 0x000966000c1e1900 */ /*0740*/ IMAD.WIDE R14, R2.reuse, 0x4, R12 ; /* 0x00000004020e7825 */ /* 0x040fe200078e020c */ /*0750*/ LDG.E R20, [R8.64] ; /* 0x0000000408147981 */ /* 0x000ea8000c1e1900 */ /*0760*/ LDG.E R22, [R8.64+0x4] ; /* 0x0000040408167981 */ /* 0x000ee2000c1e1900 */ /*0770*/ IMAD.WIDE R10, R2, 0x4, R14 ; /* 0x00000004020a7825 */ /* 0x000fc600078e020e */ /*0780*/ LDG.E R26, [R8.64+0x8] ; /* 0x00000804081a7981 */ /* 0x000f66000c1e1900 */ /*0790*/ IMAD.WIDE R16, R2.reuse, 0x4, R10 ; /* 0x0000000402107825 */ /* 0x042fe200078e020a */ /*07a0*/ LDG.E R14, [R14.64] ; /* 0x000000040e0e7981 */ /* 0x000368000c1e1900 */ /*07b0*/ LDG.E R27, [R8.64+0xc] ; /* 0x00000c04081b7981 */ /* 0x000f62000c1e1900 */ /*07c0*/ IMAD.WIDE R18, R2, 0x4, R16 ; /* 0x0000000402127825 */ /* 0x000fc600078e0210 */ /*07d0*/ LDG.E R10, [R10.64] ; /* 0x000000040a0a7981 */ /* 0x000368000c1e1900 */ /*07e0*/ LDG.E R25, [R8.64+0x10] ; /* 0x0000100408197981 */ /* 0x001f62000c1e1900 */ /*07f0*/ IMAD.WIDE R12, R2, 0x4, R18 ; /* 0x00000004020c7825 */ /* 0x010fc600078e0212 */ /*0800*/ LDG.E R16, [R16.64] ; /* 0x0000000410107981 */ /* 0x000f28000c1e1900 */ /*0810*/ LDG.E R29, [R8.64+0x14] ; /* 0x00001404081d7981 */ /* 0x000f28000c1e1900 */ /*0820*/ LDG.E R24, [R18.64] ; /* 0x0000000412187981 */ /* 0x000128000c1e1900 */ /*0830*/ LDG.E R11, [R8.64+0x18] ; /* 0x00001804080b7981 */ /* 0x002f28000c1e1900 */ /*0840*/ LDG.E R15, [R12.64] ; /* 0x000000040c0f7981 */ /* 0x000f28000c1e1900 */ /*0850*/ LDG.E R18, [R8.64+0x1c] ; /* 0x00001c0408127981 */ /* 0x001f22000c1e1900 */ /*0860*/ UIADD3 UR6, UP0, UR6, 0x20, URZ ; /* 0x0000002006067890 */ /* 0x000fe2000ff1e03f */ /*0870*/ PLOP3.LUT P0, PT, PT, PT, PT, 0x8, 0x0 ; /* 0x000000000000781c */ /* 0x000fc40003f0e170 */ /*0880*/ IADD3 R4, R4, 0x8, RZ ; /* 0x0000000804047810 */ /* 0x000fe40007ffe0ff */ /*0890*/ IADD3 R6, R6, -0x8, RZ ; /* 0xfffffff806067810 */ /* 0x000fe20007ffe0ff */ /*08a0*/ UIADD3.X UR7, URZ, UR7, URZ, UP0, !UPT ; /* 0x000000073f077290 */ /* 0x000fe200087fe43f */ /*08b0*/ IMAD R7, R7, R20, R28 ; /* 0x0000001407077224 */ /* 0x004fc800078e021c */ /*08c0*/ IMAD R7, R21, R22, R7 ; /* 0x0000001615077224 */ /* 0x008fc800078e0207 */ /*08d0*/ IMAD R7, R23, R26, R7 ; /* 0x0000001a17077224 */ /* 0x020fc800078e0207 */ /*08e0*/ IMAD R7, R14, R27, R7 ; /* 0x0000001b0e077224 */ /* 0x000fc800078e0207 */ /*08f0*/ IMAD R7, R10, R25, R7 ; /* 0x000000190a077224 */ /* 0x000fc800078e0207 */ /*0900*/ IMAD R7, R16, R29, R7 ; /* 0x0000001d10077224 */ /* 0x010fc800078e0207 */ /*0910*/ IMAD R7, R24, R11, R7 ; /* 0x0000000b18077224 */ /* 0x000fe400078e0207 */ /*0920*/ IMAD.WIDE R24, R2, 0x4, R12 ; /* 0x0000000402187825 */ /* 0x000fc800078e020c */ /*0930*/ IMAD R28, R15, R18, R7 ; /* 0x000000120f1c7224 */ /* 0x000fe400078e0207 */ /*0940*/ ISETP.NE.OR P0, PT, R6, RZ, P0 ; /* 0x000000ff0600720c */ /* 0x000fda0000705670 */ /*0950*/ @!P0 BRA 0xaf0 ; /* 0x0000019000008947 */ /* 0x000fea0003800000 */ /*0960*/ MOV R8, UR6 ; /* 0x0000000600087c02 */ /* 0x000fe20008000f00 */ /*0970*/ IMAD.WIDE R14, R2, 0x4, R24 ; /* 0x00000004020e7825 */ /* 0x000fe200078e0218 */ /*0980*/ MOV R9, UR7 ; /* 0x0000000700097c02 */ /* 0x000fe20008000f00 */ /*0990*/ LDG.E R25, [R24.64] ; /* 0x0000000418197981 */ /* 0x000ea8000c1e1900 */ /*09a0*/ IMAD.WIDE R8, R3, 0x4, R8 ; /* 0x0000000403087825 */ /* 0x000fc800078e0208 */ /*09b0*/ IMAD.WIDE R12, R2.reuse, 0x4, R14 ; /* 0x00000004020c7825 */ /* 0x040fe200078e020e */ /*09c0*/ LDG.E R7, [R8.64] ; /* 0x0000000408077981 */ /* 0x000ea8000c1e1900 */ /*09d0*/ LDG.E R14, [R14.64] ; /* 0x000000040e0e7981 */ /* 0x000ee2000c1e1900 */ /*09e0*/ IMAD.WIDE R10, R2, 0x4, R12 ; /* 0x00000004020a7825 */ /* 0x000fc600078e020c */ /*09f0*/ LDG.E R16, [R8.64+0x4] ; /* 0x0000040408107981 */ /* 0x000ee8000c1e1900 */ /*0a00*/ LDG.E R18, [R12.64] ; /* 0x000000040c127981 */ /* 0x000f28000c1e1900 */ /*0a10*/ LDG.E R17, [R8.64+0x8] ; /* 0x0000080408117981 */ /* 0x000f28000c1e1900 */ /*0a20*/ LDG.E R19, [R8.64+0xc] ; /* 0x00000c0408137981 */ /* 0x000f68000c1e1900 */ /*0a30*/ LDG.E R20, [R10.64] ; /* 0x000000040a147981 */ /* 0x000f62000c1e1900 */ /*0a40*/ IADD3 R6, R6, -0x4, RZ ; /* 0xfffffffc06067810 */ /* 0x000fc80007ffe0ff */ /*0a50*/ ISETP.NE.AND P0, PT, R6, RZ, PT ; /* 0x000000ff0600720c */ /* 0x000fe20003f05270 */ /*0a60*/ UIADD3 UR6, UP0, UR6, 0x10, URZ ; /* 0x0000001006067890 */ /* 0x000fe2000ff1e03f */ /*0a70*/ IADD3 R4, R4, 0x4, RZ ; /* 0x0000000404047810 */ /* 0x000fc60007ffe0ff */ /*0a80*/ UIADD3.X UR7, URZ, UR7, URZ, UP0, !UPT ; /* 0x000000073f077290 */ /* 0x000fe200087fe43f */ /*0a90*/ IMAD R7, R25, R7, R28 ; /* 0x0000000719077224 */ /* 0x004fc800078e021c */ /*0aa0*/ IMAD R7, R14, R16, R7 ; /* 0x000000100e077224 */ /* 0x008fe400078e0207 */ /*0ab0*/ IMAD.WIDE R24, R2, 0x4, R10 ; /* 0x0000000402187825 */ /* 0x000fc800078e020a */ /*0ac0*/ IMAD R7, R18, R17, R7 ; /* 0x0000001112077224 */ /* 0x010fc800078e0207 */ /*0ad0*/ IMAD R28, R20, R19, R7 ; /* 0x00000013141c7224 */ /* 0x020fe200078e0207 */ /*0ae0*/ @P0 BRA 0x960 ; /* 0xfffffe7000000947 */ /* 0x000fea000383ffff */ /*0af0*/ ISETP.NE.AND P0, PT, R5, RZ, PT ; /* 0x000000ff0500720c */ /* 0x000fda0003f05270 */ /*0b00*/ @!P0 BRA 0xbf0 ; /* 0x000000e000008947 */ /* 0x000fea0003800000 */ /*0b10*/ HFMA2.MMA R9, -RZ, RZ, 0, 2.384185791015625e-07 ; /* 0x00000004ff097435 */ /* 0x000fe200000001ff */ /*0b20*/ IADD3 R6, R3, R4, RZ ; /* 0x0000000403067210 */ /* 0x000fe20007ffe0ff */ /*0b30*/ IMAD R4, R4, c[0x0][0x178], R0 ; /* 0x00005e0004047a24 */ /* 0x000fd000078e0200 */ /*0b40*/ IMAD.WIDE R6, R6, R9, c[0x0][0x160] ; /* 0x0000580006067625 */ /* 0x000fc800078e0209 */ /*0b50*/ IMAD.WIDE R8, R4, R9, c[0x0][0x168] ; /* 0x00005a0004087625 */ /* 0x000fca00078e0209 */ /*0b60*/ LDG.E R11, [R8.64] ; /* 0x00000004080b7981 */ /* 0x0000a8000c1e1900 */ /*0b70*/ LDG.E R4, [R6.64] ; /* 0x0000000406047981 */ /* 0x0002a2000c1e1900 */ /*0b80*/ IADD3 R5, R5, -0x1, RZ ; /* 0xffffffff05057810 */ /* 0x000fc80007ffe0ff */ /*0b90*/ ISETP.NE.AND P0, PT, R5, RZ, PT ; /* 0x000000ff0500720c */ /* 0x000fe20003f05270 */ /*0ba0*/ IMAD.WIDE R8, R2, 0x4, R8 ; /* 0x0000000402087825 */ /* 0x001fe200078e0208 */ /*0bb0*/ IADD3 R6, P1, R6, 0x4, RZ ; /* 0x0000000406067810 */ /* 0x002fc80007f3e0ff */ /*0bc0*/ IADD3.X R7, RZ, R7, RZ, P1, !PT ; /* 0x00000007ff077210 */ /* 0x000fe20000ffe4ff */ /*0bd0*/ IMAD R28, R11, R4, R28 ; /* 0x000000040b1c7224 */ /* 0x004fcc00078e021c */ /*0be0*/ @P0 BRA 0xb60 ; /* 0xffffff7000000947 */ /* 0x000fea000383ffff */ /*0bf0*/ IADD3 R3, R0, R3, RZ ; /* 0x0000000300037210 */ /* 0x000fe40007ffe0ff */ /*0c00*/ MOV R2, 0x4 ; /* 0x0000000400027802 */ /* 0x000fca0000000f00 */ /*0c10*/ IMAD.WIDE R2, R3, R2, c[0x0][0x170] ; /* 0x00005c0003027625 */ /* 0x000fca00078e0202 */ /*0c20*/ STG.E [R2.64], R28 ; /* 0x0000001c02007986 */ /* 0x000fe2000c101904 */ /*0c30*/ EXIT ; /* 0x000000000000794d */ /* 0x000fea0003800000 */ /*0c40*/ BRA 0xc40; /* 0xfffffff000007947 */ /* 0x000fc0000383ffff */ /*0c50*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0c60*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0c70*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0c80*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0c90*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0ca0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0cb0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0cc0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0cd0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0ce0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ /*0cf0*/ NOP; /* 0x0000000000007918 */ /* 0x000fc00000000000 */ ..........
You are a helpful coding assistant specialized in converting CUDA source code to CUDA host assembly.
#include <cuda.h> #include <stdio.h> #include <stdlib.h> #define N 3 // dim of matrix //Fattened matrix multiplication . Kernel does not support x,y addressing __global__ void mat_multiply(int* d_mat1, int* d_mat2, int* d_mat3, int width) { int k,sum=0; int col = blockDim.x * blockIdx.x + threadIdx.x; int row = blockDim.y * blockIdx.y + threadIdx.y; if(row<width && col<width) { for(k=0;k<width;k++) { sum += d_mat1[row*width+k] * d_mat2[k*width+col]; } d_mat3[row*width+col] = sum; } } int main() { int i,j; int SIZE = N*N; //int BYTES = SIZE*sizeof(int); int *d_mat1, *d_mat2, *d_mat3; // allocate memory on the device cudaMallocManaged(&d_mat1,N*N*sizeof(int)); cudaMallocManaged(&d_mat2,N*N*sizeof(int)); cudaMallocManaged(&d_mat3,N*N*sizeof(int)); // generate matrix on host for(i=0;i<N*N;i++) //linearize array { d_mat1[i] = 1; d_mat2[i] = 1; d_mat3[i] = 0; } dim3 dimGrid(1,1); dim3 dimBlock(N,N); // lauch kernel mat_multiply<<<dimGrid,dimBlock>>>(d_mat1,d_mat2,d_mat3,N); cudaDeviceSynchronize(); for(i=0;i<N*N;i++) { printf("%d ",d_mat3[i]); if(i%N==0 && i>N) printf("\n"); } printf("\n"); }
.file "tmpxft_00125150_00000000-6_q3.cudafe1.cpp" .text #APP #NO_APP .type _ZL26__cudaUnregisterBinaryUtilv, @function _ZL26__cudaUnregisterBinaryUtilv: .LFB2060: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaUnregisterFatBinary@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2060: .size _ZL26__cudaUnregisterBinaryUtilv, .-_ZL26__cudaUnregisterBinaryUtilv .globl _Z37__device_stub__Z12mat_multiplyPiS_S_iPiS_S_i .type _Z37__device_stub__Z12mat_multiplyPiS_S_iPiS_S_i, @function _Z37__device_stub__Z12mat_multiplyPiS_S_iPiS_S_i: .LFB2082: .cfi_startproc endbr64 subq $152, %rsp .cfi_def_cfa_offset 160 movq %rdi, 24(%rsp) movq %rsi, 16(%rsp) movq %rdx, 8(%rsp) movl %ecx, 4(%rsp) movq %fs:40, %rax movq %rax, 136(%rsp) xorl %eax, %eax leaq 24(%rsp), %rax movq %rax, 96(%rsp) leaq 16(%rsp), %rax movq %rax, 104(%rsp) leaq 8(%rsp), %rax movq %rax, 112(%rsp) leaq 4(%rsp), %rax movq %rax, 120(%rsp) movl $1, 48(%rsp) movl $1, 52(%rsp) movl $1, 56(%rsp) movl $1, 60(%rsp) movl $1, 64(%rsp) movl $1, 68(%rsp) leaq 40(%rsp), %rcx leaq 32(%rsp), %rdx leaq 60(%rsp), %rsi leaq 48(%rsp), %rdi call __cudaPopCallConfiguration@PLT testl %eax, %eax je .L7 .L3: movq 136(%rsp), %rax subq %fs:40, %rax jne .L8 addq $152, %rsp .cfi_remember_state .cfi_def_cfa_offset 8 ret .L7: .cfi_restore_state pushq 40(%rsp) .cfi_def_cfa_offset 168 pushq 40(%rsp) .cfi_def_cfa_offset 176 leaq 112(%rsp), %r9 movq 76(%rsp), %rcx movl 84(%rsp), %r8d movq 64(%rsp), %rsi movl 72(%rsp), %edx leaq _Z12mat_multiplyPiS_S_i(%rip), %rdi call cudaLaunchKernel@PLT addq $16, %rsp .cfi_def_cfa_offset 160 jmp .L3 .L8: call __stack_chk_fail@PLT .cfi_endproc .LFE2082: .size _Z37__device_stub__Z12mat_multiplyPiS_S_iPiS_S_i, .-_Z37__device_stub__Z12mat_multiplyPiS_S_iPiS_S_i .globl _Z12mat_multiplyPiS_S_i .type _Z12mat_multiplyPiS_S_i, @function _Z12mat_multiplyPiS_S_i: .LFB2083: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 call _Z37__device_stub__Z12mat_multiplyPiS_S_iPiS_S_i addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2083: .size _Z12mat_multiplyPiS_S_i, .-_Z12mat_multiplyPiS_S_i .section .rodata.str1.1,"aMS",@progbits,1 .LC0: .string "%d " .LC1: .string "\n" .text .globl main .type main, @function main: .LFB2057: .cfi_startproc endbr64 pushq %r12 .cfi_def_cfa_offset 16 .cfi_offset 12, -16 pushq %rbp .cfi_def_cfa_offset 24 .cfi_offset 6, -24 pushq %rbx .cfi_def_cfa_offset 32 .cfi_offset 3, -32 subq $64, %rsp .cfi_def_cfa_offset 96 movq %fs:40, %rax movq %rax, 56(%rsp) xorl %eax, %eax leaq 8(%rsp), %rdi movl $1, %edx movl $36, %esi call cudaMallocManaged@PLT leaq 16(%rsp), %rdi movl $1, %edx movl $36, %esi call cudaMallocManaged@PLT leaq 24(%rsp), %rdi movl $1, %edx movl $36, %esi call cudaMallocManaged@PLT movl $0, %eax .L12: movq 8(%rsp), %rdx movl $1, (%rdx,%rax) movq 16(%rsp), %rdx movl $1, (%rdx,%rax) movq 24(%rsp), %rdx movl $0, (%rdx,%rax) addq $4, %rax cmpq $36, %rax jne .L12 movl $1, 32(%rsp) movl $1, 36(%rsp) movl $1, 40(%rsp) movl $3, 44(%rsp) movl $3, 48(%rsp) movl $1, 52(%rsp) movl $0, %r9d movl $0, %r8d movq 44(%rsp), %rdx movl $1, %ecx movq 32(%rsp), %rdi movl $1, %esi call __cudaPushCallConfiguration@PLT testl %eax, %eax je .L20 .L13: call cudaDeviceSynchronize@PLT movl $0, %ebx leaq .LC0(%rip), %rbp leaq .LC1(%rip), %r12 jmp .L15 .L20: movl $3, %ecx movq 24(%rsp), %rdx movq 16(%rsp), %rsi movq 8(%rsp), %rdi call _Z37__device_stub__Z12mat_multiplyPiS_S_iPiS_S_i jmp .L13 .L14: addq $1, %rbx cmpq $9, %rbx je .L21 .L15: movq 24(%rsp), %rax movl (%rax,%rbx,4), %edx movq %rbp, %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT movslq %ebx, %rax imulq $1431655766, %rax, %rax shrq $32, %rax movl %ebx, %edx sarl $31, %edx subl %edx, %eax leal (%rax,%rax,2), %eax cmpl %ebx, %eax jne .L14 cmpl $3, %ebx jle .L14 movq %r12, %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT jmp .L14 .L21: leaq .LC1(%rip), %rsi movl $2, %edi movl $0, %eax call __printf_chk@PLT movq 56(%rsp), %rax subq %fs:40, %rax jne .L22 movl $0, %eax addq $64, %rsp .cfi_remember_state .cfi_def_cfa_offset 32 popq %rbx .cfi_def_cfa_offset 24 popq %rbp .cfi_def_cfa_offset 16 popq %r12 .cfi_def_cfa_offset 8 ret .L22: .cfi_restore_state call __stack_chk_fail@PLT .cfi_endproc .LFE2057: .size main, .-main .section .rodata.str1.1 .LC2: .string "_Z12mat_multiplyPiS_S_i" .text .type _ZL24__sti____cudaRegisterAllv, @function _ZL24__sti____cudaRegisterAllv: .LFB2085: .cfi_startproc endbr64 subq $8, %rsp .cfi_def_cfa_offset 16 leaq _ZL15__fatDeviceText(%rip), %rdi call __cudaRegisterFatBinary@PLT movq %rax, %rdi movq %rax, _ZL20__cudaFatCubinHandle(%rip) pushq $0 .cfi_def_cfa_offset 24 pushq $0 .cfi_def_cfa_offset 32 pushq $0 .cfi_def_cfa_offset 40 pushq $0 .cfi_def_cfa_offset 48 movl $0, %r9d movl $-1, %r8d leaq .LC2(%rip), %rdx movq %rdx, %rcx leaq _Z12mat_multiplyPiS_S_i(%rip), %rsi call __cudaRegisterFunction@PLT addq $32, %rsp .cfi_def_cfa_offset 16 movq _ZL20__cudaFatCubinHandle(%rip), %rdi call __cudaRegisterFatBinaryEnd@PLT leaq _ZL26__cudaUnregisterBinaryUtilv(%rip), %rdi call atexit@PLT addq $8, %rsp .cfi_def_cfa_offset 8 ret .cfi_endproc .LFE2085: .size _ZL24__sti____cudaRegisterAllv, .-_ZL24__sti____cudaRegisterAllv .section .init_array,"aw" .align 8 .quad _ZL24__sti____cudaRegisterAllv .section .nvFatBinSegment,"aw" .align 8 .type _ZL15__fatDeviceText, @object .size _ZL15__fatDeviceText, 24 _ZL15__fatDeviceText: .long 1180844977 .long 1 .quad fatbinData .quad 0 .local _ZL20__cudaFatCubinHandle .comm _ZL20__cudaFatCubinHandle,8,8 .ident "GCC: (Ubuntu 13.3.0-6ubuntu2~24.04) 13.3.0" .section .note.GNU-stack,"",@progbits .section .note.gnu.property,"a" .align 8 .long 1f - 0f .long 4f - 1f .long 5 0: .string "GNU" 1: .align 8 .long 0xc0000002 .long 3f - 2f 2: .long 0x3 3: .align 8 4:
You are a helpful coding assistant specialized in converting CUDA source code to HIP/AMD source code.
#include <cuda.h> #include <stdio.h> #include <stdlib.h> #define N 3 // dim of matrix //Fattened matrix multiplication . Kernel does not support x,y addressing __global__ void mat_multiply(int* d_mat1, int* d_mat2, int* d_mat3, int width) { int k,sum=0; int col = blockDim.x * blockIdx.x + threadIdx.x; int row = blockDim.y * blockIdx.y + threadIdx.y; if(row<width && col<width) { for(k=0;k<width;k++) { sum += d_mat1[row*width+k] * d_mat2[k*width+col]; } d_mat3[row*width+col] = sum; } } int main() { int i,j; int SIZE = N*N; //int BYTES = SIZE*sizeof(int); int *d_mat1, *d_mat2, *d_mat3; // allocate memory on the device cudaMallocManaged(&d_mat1,N*N*sizeof(int)); cudaMallocManaged(&d_mat2,N*N*sizeof(int)); cudaMallocManaged(&d_mat3,N*N*sizeof(int)); // generate matrix on host for(i=0;i<N*N;i++) //linearize array { d_mat1[i] = 1; d_mat2[i] = 1; d_mat3[i] = 0; } dim3 dimGrid(1,1); dim3 dimBlock(N,N); // lauch kernel mat_multiply<<<dimGrid,dimBlock>>>(d_mat1,d_mat2,d_mat3,N); cudaDeviceSynchronize(); for(i=0;i<N*N;i++) { printf("%d ",d_mat3[i]); if(i%N==0 && i>N) printf("\n"); } printf("\n"); }
#include <hip/hip_runtime.h> #include <stdio.h> #include <stdlib.h> #define N 3 // dim of matrix //Fattened matrix multiplication . Kernel does not support x,y addressing __global__ void mat_multiply(int* d_mat1, int* d_mat2, int* d_mat3, int width) { int k,sum=0; int col = blockDim.x * blockIdx.x + threadIdx.x; int row = blockDim.y * blockIdx.y + threadIdx.y; if(row<width && col<width) { for(k=0;k<width;k++) { sum += d_mat1[row*width+k] * d_mat2[k*width+col]; } d_mat3[row*width+col] = sum; } } int main() { int i,j; int SIZE = N*N; //int BYTES = SIZE*sizeof(int); int *d_mat1, *d_mat2, *d_mat3; // allocate memory on the device hipMallocManaged(&d_mat1,N*N*sizeof(int)); hipMallocManaged(&d_mat2,N*N*sizeof(int)); hipMallocManaged(&d_mat3,N*N*sizeof(int)); // generate matrix on host for(i=0;i<N*N;i++) //linearize array { d_mat1[i] = 1; d_mat2[i] = 1; d_mat3[i] = 0; } dim3 dimGrid(1,1); dim3 dimBlock(N,N); // lauch kernel mat_multiply<<<dimGrid,dimBlock>>>(d_mat1,d_mat2,d_mat3,N); hipDeviceSynchronize(); for(i=0;i<N*N;i++) { printf("%d ",d_mat3[i]); if(i%N==0 && i>N) printf("\n"); } printf("\n"); }