source
stringclasses
13 values
gpu
stringclasses
4 values
model
stringclasses
13 values
kernel_family
stringclasses
7 values
kernel_name
stringclasses
328 values
dtype
stringclasses
2 values
held_out
bool
2 classes
M
float64
1
16.4k
N
float64
8
248k
K
float64
2.05k
29.6k
bs
float64
1
1
seq
float64
1
16.4k
n_heads
float64
8
64
head_dim
float64
64
256
kv_heads
float64
1
64
numel
float64
64
531M
op_type
stringclasses
177 values
gpu_time_duration_ms
float64
0
703
launch_block_size
float64
16
512
launch_grid_size
float64
1
1.04M
dram_bytes_sum
float64
1.27
94.3B
launch_registers_per_thread
float64
16
255
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
void cutlass::Kernel2<cutlass_75_tensorop_bf16_s1688gemm_bf16_64x64_nn_align1>(T1::Params)
bf16
false
1
2,880
2,880
1
128
null
null
null
null
expert_gate_up_tok1
0.01782
128
256
10,550,000
96
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
nvjet_tst_88x64_64x11_1x2_h_bz_bias_TNN
bf16
false
128
3,600
2,880
1
128
null
null
null
null
qkv_fused
0.01981
384
132
25,730,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
nvjet_tst_32x64_64x16_1x4_h_bz_splitK_bias_TNN
bf16
false
128
360
2,880
1
128
null
null
null
null
k_or_v
0.00768
384
8
1,670,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
void cutlass::device_kernel<at::cuda::detail::enable_3x_kernel_for_sm9x<cutlass::gemm::kernel::GemmUniversal<cutlass::gemm::GroupProblemShape<cute::tuple<int, int, int>>, cutlass::gemm::collective::CollectiveMma<cutlass::gemm::MainloopSm90ArrayTmaGmmaWarpSpecialized<(int)4, cute::tuple<cute::C<(int)2>, cute::C<(int)1>,...
bf16
false
128
131,072
2,880
1
128
null
null
null
null
lm_head
0.3793
384
132
922,780,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
void cutlass::device_kernel<at::cuda::detail::enable_3x_kernel_for_sm9x<cutlass::gemm::kernel::GemmUniversal<cutlass::gemm::GroupProblemShape<cute::tuple<int, int, int>>, cutlass::gemm::collective::CollectiveMma<cutlass::gemm::MainloopSm90ArrayTmaGmmaWarpSpecialized<(int)4, cute::tuple<cute::C<(int)2>, cute::C<(int)1>,...
bf16
false
128
131,072
2,880
1
128
null
null
null
null
lm_head
0.20144
384
132
461,680,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
nvjet_tst_128x64_64x8_1x2_h_bz_bias_TNT
bf16
false
128
3,600
2,880
1
128
null
null
null
null
qkv_fused
0.01725
384
128
25,100,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
nvjet_tst_64x24_64x16_2x4_v_bz_bias_TNT
bf16
false
128
360
2,880
1
128
null
null
null
null
k_or_v
0.01098
384
96
4,440,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
nvjet_tst_64x24_64x16_2x4_v_bz_bias_TNT
bf16
false
128
360
2,880
1
128
null
null
null
null
k_or_v
0.01088
384
96
4,440,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
nvjet_tst_128x128_64x6_2x1_v_bz_TNT
bf16
false
128
360
2,880
1
128
null
null
null
null
k_or_v
0.0081
384
132
4,220,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
void cutlass::Kernel2<cutlass_75_tensorop_bf16_s1688gemm_bf16_64x64_nn_align1>(T1::Params)
bf16
false
1
2,880
2,880
1
128
null
null
null
null
expert_gate_up_tok1
0.01802
128
256
10,550,000
96
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
nvjet_tst_88x64_64x11_1x2_h_bz_bias_TNN
bf16
false
128
3,600
2,880
1
128
null
null
null
null
qkv_fused
0.01917
384
132
25,730,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
nvjet_tst_32x64_64x16_1x4_h_bz_splitK_bias_TNN
bf16
false
128
360
2,880
1
128
null
null
null
null
k_or_v
0.00765
384
8
1,680,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
void cutlass::device_kernel<at::cuda::detail::enable_3x_kernel_for_sm9x<cutlass::gemm::kernel::GemmUniversal<cutlass::gemm::GroupProblemShape<cute::tuple<int, int, int>>, cutlass::gemm::collective::CollectiveMma<cutlass::gemm::MainloopSm90ArrayTmaGmmaWarpSpecialized<(int)4, cute::tuple<cute::C<(int)2>, cute::C<(int)1>,...
bf16
false
128
131,072
2,880
1
128
null
null
null
null
lm_head
0.41302
384
132
999,090,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
void cutlass::device_kernel<at::cuda::detail::enable_3x_kernel_for_sm9x<cutlass::gemm::kernel::GemmUniversal<cutlass::gemm::GroupProblemShape<cute::tuple<int, int, int>>, cutlass::gemm::collective::CollectiveMma<cutlass::gemm::MainloopSm90ArrayTmaGmmaWarpSpecialized<(int)4, cute::tuple<cute::C<(int)2>, cute::C<(int)1>,...
bf16
false
128
131,072
2,880
1
128
null
null
null
null
lm_head
0.224
384
132
494,550,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
nvjet_tst_256x128_64x4_1x2_h_bz_coopA_TNT
bf16
false
128
131,072
2,880
1
128
null
null
null
null
lm_head
0.50211
384
132
1,270,000,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
void cutlass::Kernel2<cutlass_80_simt_sgemm_128x32_8x5_nn_align1>(T1::Params)
bf16
false
128
32
2,880
1
128
null
null
null
null
moe_router
0.00547
128
4
32,000
80
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
nvjet_tst_128x128_64x6_2x1_v_bz_bias_TNT
bf16
false
128
3,600
2,880
1
128
null
null
null
null
qkv_fused
0.02496
384
128
26,640,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
nvjet_tst_64x48_64x15_2x4_v_bz_bias_TNT
bf16
false
128
360
2,880
1
128
null
null
null
null
k_or_v
0.01184
384
96
5,920,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
nvjet_tst_64x48_64x15_2x4_v_bz_bias_TNT
bf16
false
128
360
2,880
1
128
null
null
null
null
k_or_v
0.01203
384
96
5,920,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
nvjet_tst_256x128_64x4_1x2_h_bz_coopA_TNT
bf16
false
17
2,880
2,880
1
128
null
null
null
null
expert_gate_up_tok17
0.01539
384
132
16,780,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
void cutlass::Kernel2<cutlass_75_tensorop_bf16_s1688gemm_bf16_64x64_nn_align1>(T1::Params)
bf16
false
128
3,600
2,880
1
128
null
null
null
null
qkv_fused
0.04531
128
512
39,680,000
96
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
nvjet_tst_88x128_64x7_1x2_h_bz_bias_TNN
bf16
false
128
3,600
2,880
1
128
null
null
null
null
qkv_fused
0.02496
384
132
27,890,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
nvjet_tst_8x64_64x16_4x2_v_bz_splitK_bias_TNN
bf16
false
128
360
2,880
1
128
null
null
null
null
k_or_v
0.00749
384
64
3,150,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
void cutlass::device_kernel<at::cuda::detail::enable_3x_kernel_for_sm9x<cutlass::gemm::kernel::GemmUniversal<cutlass::gemm::GroupProblemShape<cute::tuple<int, int, int>>, cutlass::gemm::collective::CollectiveMma<cutlass::gemm::MainloopSm90ArrayTmaGmmaWarpSpecialized<(int)4, cute::tuple<cute::C<(int)2>, cute::C<(int)1>,...
bf16
false
128
131,072
2,880
1
128
null
null
null
null
lm_head
0.46749
384
132
1,110,000,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
void cutlass::device_kernel<at::cuda::detail::enable_3x_kernel_for_sm9x<cutlass::gemm::kernel::GemmUniversal<cutlass::gemm::GroupProblemShape<cute::tuple<int, int, int>>, cutlass::gemm::collective::CollectiveMma<cutlass::gemm::MainloopSm90ArrayTmaGmmaWarpSpecialized<(int)4, cute::tuple<cute::C<(int)2>, cute::C<(int)1>,...
bf16
false
128
131,072
2,880
1
128
null
null
null
null
lm_head
0.24925
384
132
555,420,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
nvjet_tst_128x128_64x6_2x1_v_bz_bias_TNT
bf16
false
128
3,600
2,880
1
128
null
null
null
null
qkv_fused
0.02506
384
128
26,610,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
nvjet_tst_64x48_64x15_2x4_v_bz_bias_TNT
bf16
false
128
360
2,880
1
128
null
null
null
null
k_or_v
0.01184
384
96
5,920,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
nvjet_tst_64x48_64x15_2x4_v_bz_bias_TNT
bf16
false
128
360
2,880
1
128
null
null
null
null
k_or_v
0.01184
384
96
5,920,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
nvjet_tst_256x128_64x4_1x2_h_bz_coopA_TNT
bf16
false
32
2,880
2,880
1
128
null
null
null
null
expert_gate_up_tok32
0.01606
384
132
16,960,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
void cutlass::Kernel2<cutlass_75_tensorop_bf16_s1688gemm_bf16_64x64_nn_align1>(T1::Params)
bf16
false
128
3,600
2,880
1
128
null
null
null
null
qkv_fused
0.04544
128
512
39,560,000
96
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
nvjet_tst_88x128_64x7_1x2_h_bz_bias_TNN
bf16
false
128
3,600
2,880
1
128
null
null
null
null
qkv_fused
0.02493
384
132
27,860,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
nvjet_tst_8x64_64x16_4x2_v_bz_splitK_bias_TNN
bf16
false
128
360
2,880
1
128
null
null
null
null
k_or_v
0.00752
384
64
3,150,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
void cutlass::device_kernel<at::cuda::detail::enable_3x_kernel_for_sm9x<cutlass::gemm::kernel::GemmUniversal<cutlass::gemm::GroupProblemShape<cute::tuple<int, int, int>>, cutlass::gemm::collective::CollectiveMma<cutlass::gemm::MainloopSm90ArrayTmaGmmaWarpSpecialized<(int)4, cute::tuple<cute::C<(int)2>, cute::C<(int)1>,...
bf16
false
128
131,072
2,880
1
128
null
null
null
null
lm_head
0.51082
384
132
1,120,000,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
void cutlass::device_kernel<at::cuda::detail::enable_3x_kernel_for_sm9x<cutlass::gemm::kernel::GemmUniversal<cutlass::gemm::GroupProblemShape<cute::tuple<int, int, int>>, cutlass::gemm::collective::CollectiveMma<cutlass::gemm::MainloopSm90ArrayTmaGmmaWarpSpecialized<(int)4, cute::tuple<cute::C<(int)2>, cute::C<(int)1>,...
bf16
false
128
131,072
2,880
1
128
null
null
null
null
lm_head
0.27357
384
132
559,860,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
nvjet_tst_128x128_64x6_2x1_v_bz_bias_TNT
bf16
false
128
3,600
2,880
1
128
null
null
null
null
qkv_fused
0.02512
384
128
26,590,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
nvjet_tst_64x48_64x15_2x4_v_bz_bias_TNT
bf16
false
128
360
2,880
1
128
null
null
null
null
k_or_v
0.012
384
96
5,920,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
nvjet_tst_64x48_64x15_2x4_v_bz_bias_TNT
bf16
false
128
360
2,880
1
128
null
null
null
null
k_or_v
0.0119
384
96
5,920,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
nvjet_tst_256x128_64x4_1x2_h_bz_coopA_TNT
bf16
false
34
2,880
2,880
1
128
null
null
null
null
expert_gate_up_tok34
0.01587
384
132
16,980,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
void cutlass::Kernel2<cutlass_75_tensorop_bf16_s1688gemm_bf16_64x64_nn_align1>(T1::Params)
bf16
false
128
3,600
2,880
1
128
null
null
null
null
qkv_fused
0.04518
128
512
39,700,000
96
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
nvjet_tst_88x128_64x7_1x2_h_bz_bias_TNN
bf16
false
128
3,600
2,880
1
128
null
null
null
null
qkv_fused
0.02512
384
132
27,870,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
nvjet_tst_8x64_64x16_4x2_v_bz_splitK_bias_TNN
bf16
false
128
360
2,880
1
128
null
null
null
null
k_or_v
0.00755
384
64
3,150,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
void cutlass::device_kernel<at::cuda::detail::enable_3x_kernel_for_sm9x<cutlass::gemm::kernel::GemmUniversal<cutlass::gemm::GroupProblemShape<cute::tuple<int, int, int>>, cutlass::gemm::collective::CollectiveMma<cutlass::gemm::MainloopSm90ArrayTmaGmmaWarpSpecialized<(int)4, cute::tuple<cute::C<(int)2>, cute::C<(int)1>,...
bf16
false
128
131,072
2,880
1
128
null
null
null
null
lm_head
0.47171
384
132
989,390,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
void cutlass::device_kernel<at::cuda::detail::enable_3x_kernel_for_sm9x<cutlass::gemm::kernel::GemmUniversal<cutlass::gemm::GroupProblemShape<cute::tuple<int, int, int>>, cutlass::gemm::collective::CollectiveMma<cutlass::gemm::MainloopSm90ArrayTmaGmmaWarpSpecialized<(int)4, cute::tuple<cute::C<(int)2>, cute::C<(int)1>,...
bf16
false
128
131,072
2,880
1
128
null
null
null
null
lm_head
0.24643
384
132
487,780,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
nvjet_tst_128x128_64x6_2x1_v_bz_bias_TNT
bf16
false
128
3,600
2,880
1
128
null
null
null
null
qkv_fused
0.0248
384
128
26,600,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
nvjet_tst_64x48_64x15_2x4_v_bz_bias_TNT
bf16
false
128
360
2,880
1
128
null
null
null
null
k_or_v
0.01206
384
96
5,920,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
nvjet_tst_64x48_64x15_2x4_v_bz_bias_TNT
bf16
false
128
360
2,880
1
128
null
null
null
null
k_or_v
0.0121
384
96
5,920,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
nvjet_tst_256x128_64x4_1x2_h_bz_coopA_TNT
bf16
false
36
2,880
2,880
1
128
null
null
null
null
expert_gate_up_tok36
0.01562
384
132
17,000,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
void cutlass::Kernel2<cutlass_75_tensorop_bf16_s1688gemm_bf16_64x64_nn_align1>(T1::Params)
bf16
false
128
3,600
2,880
1
128
null
null
null
null
qkv_fused
0.04531
128
512
39,600,000
96
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
nvjet_tst_88x128_64x7_1x2_h_bz_bias_TNN
bf16
false
128
3,600
2,880
1
128
null
null
null
null
qkv_fused
0.02512
384
132
27,900,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
nvjet_tst_8x64_64x16_4x2_v_bz_splitK_bias_TNN
bf16
false
128
360
2,880
1
128
null
null
null
null
k_or_v
0.00755
384
64
3,150,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
void cutlass::device_kernel<at::cuda::detail::enable_3x_kernel_for_sm9x<cutlass::gemm::kernel::GemmUniversal<cutlass::gemm::GroupProblemShape<cute::tuple<int, int, int>>, cutlass::gemm::collective::CollectiveMma<cutlass::gemm::MainloopSm90ArrayTmaGmmaWarpSpecialized<(int)4, cute::tuple<cute::C<(int)2>, cute::C<(int)1>,...
bf16
false
128
131,072
2,880
1
128
null
null
null
null
lm_head
0.50525
384
132
1,100,000,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
void cutlass::device_kernel<at::cuda::detail::enable_3x_kernel_for_sm9x<cutlass::gemm::kernel::GemmUniversal<cutlass::gemm::GroupProblemShape<cute::tuple<int, int, int>>, cutlass::gemm::collective::CollectiveMma<cutlass::gemm::MainloopSm90ArrayTmaGmmaWarpSpecialized<(int)4, cute::tuple<cute::C<(int)2>, cute::C<(int)1>,...
bf16
false
128
131,072
2,880
1
128
null
null
null
null
lm_head
0.26726
384
132
537,900,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
gemm
nvjet_tst_256x128_64x4_1x2_h_bz_coopA_TNT
bf16
false
128
131,072
2,880
1
128
null
null
null
null
lm_head
0.87453
384
132
1,430,000,000
168
gpt-oss-20b_prefill
H100
gpt-oss-20b
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)1, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
360
2,880
null
null
null
null
null
null
null
0.0048
512
64
274,430
32
gpt-oss-20b_prefill
H100
gpt-oss-20b
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)1, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
360
2,880
null
null
null
null
null
null
null
0.0048
512
64
273,660
32
gpt-oss-20b_prefill
H100
gpt-oss-20b
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)1, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
32
2,880
null
null
null
null
null
null
null
0.00451
512
4
27,390
32
gpt-oss-20b_prefill
H100
gpt-oss-20b
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)1, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
360
2,880
null
null
null
null
null
null
null
0.00499
512
64
273,660
32
gpt-oss-20b_prefill
H100
gpt-oss-20b
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)1, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
360
2,880
null
null
null
null
null
null
null
0.0048
512
64
273,920
32
gpt-oss-20b_prefill
H100
gpt-oss-20b
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)1, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
32
2,880
null
null
null
null
null
null
null
0.00467
512
4
29,180
32
gpt-oss-20b_prefill
H100
gpt-oss-20b
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)1, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
360
2,880
null
null
null
null
null
null
null
0.00483
512
64
273,660
32
gpt-oss-20b_prefill
H100
gpt-oss-20b
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)1, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
360
2,880
null
null
null
null
null
null
null
0.00483
512
64
273,920
32
gpt-oss-20b_prefill
H100
gpt-oss-20b
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)1, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
32
2,880
null
null
null
null
null
null
null
0.00451
512
4
27,140
32
gpt-oss-20b_prefill
H100
gpt-oss-20b
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)1, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
360
2,880
null
null
null
null
null
null
null
0.00477
512
64
273,660
32
gpt-oss-20b_prefill
H100
gpt-oss-20b
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)1, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
360
2,880
null
null
null
null
null
null
null
0.00493
512
64
273,410
32
gpt-oss-20b_prefill
H100
gpt-oss-20b
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)1, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
32
2,880
null
null
null
null
null
null
null
0.00454
512
4
27,140
32
gpt-oss-20b_prefill
H100
gpt-oss-20b
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)1, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
360
2,880
null
null
null
null
null
null
null
0.00474
512
128
535,550
32
gpt-oss-20b_prefill
H100
gpt-oss-20b
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)1, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
360
2,880
null
null
null
null
null
null
null
0.00499
512
128
536,060
32
gpt-oss-20b_prefill
H100
gpt-oss-20b
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)1, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
32
2,880
null
null
null
null
null
null
null
0.00461
512
8
43,260
32
gpt-oss-20b_prefill
H100
gpt-oss-20b
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)1, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
360
2,880
null
null
null
null
null
null
null
0.00477
512
128
535,810
32
gpt-oss-20b_prefill
H100
gpt-oss-20b
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)1, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
360
2,880
null
null
null
null
null
null
null
0.00486
512
128
536,320
32
gpt-oss-20b_prefill
H100
gpt-oss-20b
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)1, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
32
2,880
null
null
null
null
null
null
null
0.00458
512
8
43,520
32
gpt-oss-20b_prefill
H100
gpt-oss-20b
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)1, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
360
2,880
null
null
null
null
null
null
null
0.00474
512
128
535,810
32
gpt-oss-20b_prefill
H100
gpt-oss-20b
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)1, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
360
2,880
null
null
null
null
null
null
null
0.00486
512
128
535,550
32
gpt-oss-20b_prefill
H100
gpt-oss-20b
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)1, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
32
2,880
null
null
null
null
null
null
null
0.00458
512
8
43,260
32
gpt-oss-20b_prefill
H100
gpt-oss-20b
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)1, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
360
2,880
null
null
null
null
null
null
null
0.00474
512
128
536,580
32
gpt-oss-20b_prefill
H100
gpt-oss-20b
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)1, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
360
2,880
null
null
null
null
null
null
null
0.00499
512
128
536,060
32
gpt-oss-20b_prefill
H100
gpt-oss-20b
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)1, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
32
2,880
null
null
null
null
null
null
null
0.00448
512
8
43,780
32
gpt-oss-20b_prefill
H100
gpt-oss-20b
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)1, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
360
2,880
null
null
null
null
null
null
null
0.00467
512
16
76,290
32
gpt-oss-20b_prefill
H100
gpt-oss-20b
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)1, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
360
2,880
null
null
null
null
null
null
null
0.00467
512
16
76,540
32
gpt-oss-20b_prefill
H100
gpt-oss-20b
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)1, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
360
2,880
null
null
null
null
null
null
null
0.0048
512
16
76,540
32
gpt-oss-20b_prefill
H100
gpt-oss-20b
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)1, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
360
2,880
null
null
null
null
null
null
null
0.00458
512
16
76,290
32
gpt-oss-20b_prefill
H100
gpt-oss-20b
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)1, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
360
2,880
null
null
null
null
null
null
null
0.00467
512
32
141,820
32
gpt-oss-20b_prefill
H100
gpt-oss-20b
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)1, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
360
2,880
null
null
null
null
null
null
null
0.0048
512
32
142,080
32
gpt-oss-20b_prefill
H100
gpt-oss-20b
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)1, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
360
2,880
null
null
null
null
null
null
null
0.00474
512
32
141,820
32
gpt-oss-20b_prefill
H100
gpt-oss-20b
splitk_reduce
void cublasLt::splitKreduce_kernel<(int)32, (int)16, int, float, __nv_bfloat16, float, __nv_bfloat16, (bool)0, float, __nv_bfloat16, __nv_bfloat16, (bool)1, (bool)1, (bool)0>(cublasLt::cublasSplitKParams<T6>, const T4 *, const T10 *, T9 *, T5 *, const T6 *, const T6 *, const T11 *, const T4 *, T11 *, void *, long, T6 *...
bf16
false
128
360
2,880
null
null
null
null
null
null
null
0.00467
512
32
142,080
32
gpt-oss-20b_prefill
H100
gpt-oss-20b
cast
void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 3)]::operator ()() const::[lambda() (instance 3)]::operator ()() const::[lambda(int) (instance 1)], std::array<char *, (unsigned long)2>, (int)4, TrivialOffsetCalculator<(int)1, unsigned int>,...
bf16
false
null
null
null
1
128
null
null
null
8,294,400
cast
0.01955
128
16,200
18,550,000
32
gpt-oss-20b_prefill
H100
gpt-oss-20b
elementwise
void at::native::vectorized_elementwise_kernel<(int)4, at::native::CUDAFunctorOnSelf_add<int>, std::array<char *, (unsigned long)2>>(int, T2, T3)
bf16
false
null
null
null
1
128
null
null
null
4,147,200
residual
0.02029
128
8,100
43,580,000
32
gpt-oss-20b_prefill
H100
gpt-oss-20b
elementwise
void at::native::vectorized_elementwise_kernel<(int)4, at::native::AUnaryFunctor<unsigned char, unsigned char, unsigned char, at::native::BitwiseAndFunctor<unsigned char>>, std::array<char *, (unsigned long)2>>(int, T2, T3)
bf16
false
null
null
null
1
128
null
null
null
66,355,200
other
0.09366
128
129,600
245,040,000
30
gpt-oss-20b_prefill
H100
gpt-oss-20b
cast
void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 3)]::operator ()() const::[lambda() (instance 3)]::operator ()() const::[lambda(int) (instance 1)], std::array<char *, (unsigned long)2>, (int)4, TrivialOffsetCalculator<(int)1, unsigned int>,...
bf16
false
null
null
null
1
128
null
null
null
132,710,400
cast
0.26867
128
259,200
638,710,000
32
gpt-oss-20b_prefill
H100
gpt-oss-20b
cast
void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 3)]::operator ()() const::[lambda() (instance 4)]::operator ()() const::[lambda(long) (instance 1)], std::array<char *, (unsigned long)2>, (int)4, TrivialOffsetCalculator<(int)1, unsigned int>...
bf16
false
null
null
null
1
128
null
null
null
132,710,400
cast
0.57654
128
259,200
1,570,000,000
32
gpt-oss-20b_prefill
H100
gpt-oss-20b
copy
void at::native::index_elementwise_kernel<(int)128, (int)4, void at::native::gpu_index_kernel<void at::native::index_kernel_impl<at::native::OpaqueType<(int)2>>(at::TensorIteratorBase &, c10::ArrayRef<long>, c10::ArrayRef<long>)::[lambda(char *, const char *, long) (instance 1)]>(at::TensorIteratorBase &, c10::ArrayRef...
bf16
false
null
null
null
1
128
null
null
null
132,710,400
other
0.67344
128
259,200
1,320,000,000
36
gpt-oss-20b_prefill
H100
gpt-oss-20b
cast
void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl_nocast<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 3)]::operator ()() const::[lambda() (instance 12)]::operator ()() const::[lambda(c10::BFloat16) (instance 1)]>(at::TensorIteratorBase &, const ...
bf16
false
null
null
null
1
128
null
null
null
132,710,400
cast
0.46998
128
259,200
1,310,000,000
18
gpt-oss-20b_prefill
H100
gpt-oss-20b
elementwise
void at::native::vectorized_elementwise_kernel<(int)4, at::native::BUnaryFunctor<unsigned char, unsigned char, unsigned char, at::native::rshift_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 1)]::operator ()() const::[lambda() (instance 1)]::operator ()() const::[lambda(unsigned char, unsigned char) (insta...
bf16
false
null
null
null
1
128
null
null
null
66,355,200
other
0.09398
128
129,600
244,860,000
32
gpt-oss-20b_prefill
H100
gpt-oss-20b
cast
void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 3)]::operator ()() const::[lambda() (instance 3)]::operator ()() const::[lambda(int) (instance 1)], std::array<char *, (unsigned long)2>, (int)4, TrivialOffsetCalculator<(int)1, unsigned int>,...
bf16
false
null
null
null
1
128
null
null
null
132,710,400
cast
0.2656
128
259,200
638,410,000
32
gpt-oss-20b_prefill
H100
gpt-oss-20b
cast
void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 3)]::operator ()() const::[lambda() (instance 4)]::operator ()() const::[lambda(long) (instance 1)], std::array<char *, (unsigned long)2>, (int)4, TrivialOffsetCalculator<(int)1, unsigned int>...
bf16
false
null
null
null
1
128
null
null
null
132,710,400
cast
0.57674
128
259,200
1,570,000,000
32
gpt-oss-20b_prefill
H100
gpt-oss-20b
copy
void at::native::index_elementwise_kernel<(int)128, (int)4, void at::native::gpu_index_kernel<void at::native::index_kernel_impl<at::native::OpaqueType<(int)2>>(at::TensorIteratorBase &, c10::ArrayRef<long>, c10::ArrayRef<long>)::[lambda(char *, const char *, long) (instance 1)]>(at::TensorIteratorBase &, c10::ArrayRef...
bf16
false
null
null
null
1
128
null
null
null
132,710,400
other
0.67261
128
259,200
1,320,000,000
36
gpt-oss-20b_prefill
H100
gpt-oss-20b
cast
void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl_nocast<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 3)]::operator ()() const::[lambda() (instance 12)]::operator ()() const::[lambda(c10::BFloat16) (instance 1)]>(at::TensorIteratorBase &, const ...
bf16
false
null
null
null
1
128
null
null
null
132,710,400
cast
0.46749
128
259,200
1,310,000,000
18
gpt-oss-20b_prefill
H100
gpt-oss-20b
elementwise
void at::native::vectorized_elementwise_kernel<(int)8, at::native::FillFunctor<c10::BFloat16>, std::array<char *, (unsigned long)1>>(int, T2, T3)
bf16
false
null
null
null
1
128
null
null
null
512
fill
0.00298
128
1
4,100
16
gpt-oss-20b_prefill
H100
gpt-oss-20b
elementwise
void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl<at::native::<unnamed>::pow_tensor_tensor_kernel(at::TensorIteratorBase &)::[lambda() (instance 1)]::operator ()() const::[lambda() (instance 11)]::operator ()() const::[lambda(c10::BFloat16, c10::BFloat16) (instance 1)]>(at::TensorIt...
bf16
false
null
null
null
1
128
null
null
null
8,294,400
other
0.06883
128
16,200
38,260,000
32
gpt-oss-20b_prefill
H100
gpt-oss-20b
elementwise
void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl_nocast<at::native::BinaryFunctor<c10::BFloat16, c10::BFloat16, c10::BFloat16, at::native::binary_internal::MulFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3)
bf16
false
null
null
null
1
128
null
null
null
265,420,800
mul
0.82858
128
518,400
1,060,000,000
18