Add files using upload-large-folder tool

Prism/LLaDA/LLaDA_Prism/.venv/lib/python3.12/site-packages/nvidia/cuda_cupti/include/Openacc/cupti_openacc.h

@@ -0,0 +1,98 @@
+/*
+ * Copyright 2017 NVIDIA Corporation.  All rights reserved.
+ *
+ * NOTICE TO LICENSEE:
+ *
+ * This source code and/or documentation ("Licensed Deliverables") are
+ * subject to NVIDIA intellectual property rights under U.S. and
+ * international Copyright laws.
+ *
+ * These Licensed Deliverables contained herein is PROPRIETARY and
+ * CONFIDENTIAL to NVIDIA and is being provided under the terms and
+ * conditions of a form of NVIDIA software license agreement by and
+ * between NVIDIA and Licensee ("License Agreement") or electronically
+ * accepted by Licensee.  Notwithstanding any terms or conditions to
+ * the contrary in the License Agreement, reproduction or disclosure
+ * of the Licensed Deliverables to any third party without the express
+ * written consent of NVIDIA is prohibited.
+ *
+ * NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE
+ * LICENSE AGREEMENT, NVIDIA MAKES NO REPRESENTATION ABOUT THE
+ * SUITABILITY OF THESE LICENSED DELIVERABLES FOR ANY PURPOSE.  IT IS
+ * PROVIDED "AS IS" WITHOUT EXPRESS OR IMPLIED WARRANTY OF ANY KIND.
+ * NVIDIA DISCLAIMS ALL WARRANTIES WITH REGARD TO THESE LICENSED
+ * DELIVERABLES, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY,
+ * NONINFRINGEMENT, AND FITNESS FOR A PARTICULAR PURPOSE.
+ * NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE
+ * LICENSE AGREEMENT, IN NO EVENT SHALL NVIDIA BE LIABLE FOR ANY
+ * SPECIAL, INDIRECT, INCIDENTAL, OR CONSEQUENTIAL DAMAGES, OR ANY
+ * DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS,
+ * WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS
+ * ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE
+ * OF THESE LICENSED DELIVERABLES.
+ *
+ * U.S. Government End Users.  These Licensed Deliverables are a
+ * "commercial item" as that term is defined at 48 C.F.R. 2.101 (OCT
+ * 1995), consisting of "commercial computer software" and "commercial
+ * computer software documentation" as such terms are used in 48
+ * C.F.R. 12.212 (SEPT 1995) and is provided to the U.S. Government
+ * only as a commercial end item.  Consistent with 48 C.F.R.12.212 and
+ * 48 C.F.R. 227.7202-1 through 227.7202-4 (JUNE 1995), all
+ * U.S. Government End Users acquire the Licensed Deliverables with
+ * only those rights set forth herein.
+ *
+ * Any use of the Licensed Deliverables in individual and commercial
+ * software must include, in the user documentation and internal
+ * comments to the code, the above Disclaimer and U.S. Government End
+ * Users Notice.
+ */
+
+#include <cuda_stdint.h>
+
+#if !defined(_CUPTI_OPENACC_H_)
+#define _CUPTI_OPENACC_H_
+
+#ifndef CUPTIAPI
+#ifdef _WIN32
+#define CUPTIAPI __stdcall
+#else
+#define CUPTIAPI
+#endif
+#endif
+
+#if defined(__LP64__)
+#define CUPTILP64 1
+#elif defined(_WIN64)
+#define CUPTILP64 1
+#else
+#undef CUPTILP64
+#endif
+
+#if defined(__cplusplus)
+extern "C" {
+#endif
+
+#if defined(__GNUC__) && defined(CUPTI_LIB)
+    #pragma GCC visibility push(default)
+#endif
+
+/**
+ * \brief Initialize OpenACC support
+ *
+ * \param profRegister function of type acc_prof_reg as obtained from acc_register_library
+ * \param profUnregister function of type acc_prof_reg as obtained from acc_register_library
+ * \param profLookup function of type acc_prof_lookup as obtained from acc_register_library
+ */
+CUptiResult CUPTIAPI
+cuptiOpenACCInitialize(void *profRegister, void *profUnregister, void *profLookup);
+
+#if defined(__GNUC__) && defined(CUPTI_LIB)
+    #pragma GCC visibility pop
+#endif
+
+#if defined(__cplusplus)
+}
+#endif
+
+#endif /*_CUPTI_OPENACC_H_*/
+

Prism/LLaDA/LLaDA_Prism/.venv/lib/python3.12/site-packages/nvidia/cuda_cupti/include/Openmp/cupti_openmp.h

@@ -0,0 +1,100 @@
+/*
+ * Copyright 2018 NVIDIA Corporation.  All rights reserved.
+ *
+ * NOTICE TO LICENSEE:
+ *
+ * This source code and/or documentation ("Licensed Deliverables") are
+ * subject to NVIDIA intellectual property rights under U.S. and
+ * international Copyright laws.
+ *
+ * These Licensed Deliverables contained herein is PROPRIETARY and
+ * CONFIDENTIAL to NVIDIA and is being provided under the terms and
+ * conditions of a form of NVIDIA software license agreement by and
+ * between NVIDIA and Licensee ("License Agreement") or electronically
+ * accepted by Licensee.  Notwithstanding any terms or conditions to
+ * the contrary in the License Agreement, reproduction or disclosure
+ * of the Licensed Deliverables to any third party without the express
+ * written consent of NVIDIA is prohibited.
+ *
+ * NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE
+ * LICENSE AGREEMENT, NVIDIA MAKES NO REPRESENTATION ABOUT THE
+ * SUITABILITY OF THESE LICENSED DELIVERABLES FOR ANY PURPOSE.  IT IS
+ * PROVIDED "AS IS" WITHOUT EXPRESS OR IMPLIED WARRANTY OF ANY KIND.
+ * NVIDIA DISCLAIMS ALL WARRANTIES WITH REGARD TO THESE LICENSED
+ * DELIVERABLES, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY,
+ * NONINFRINGEMENT, AND FITNESS FOR A PARTICULAR PURPOSE.
+ * NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE
+ * LICENSE AGREEMENT, IN NO EVENT SHALL NVIDIA BE LIABLE FOR ANY
+ * SPECIAL, INDIRECT, INCIDENTAL, OR CONSEQUENTIAL DAMAGES, OR ANY
+ * DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS,
+ * WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS
+ * ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE
+ * OF THESE LICENSED DELIVERABLES.
+ *
+ * U.S. Government End Users.  These Licensed Deliverables are a
+ * "commercial item" as that term is defined at 48 C.F.R. 2.101 (OCT
+ * 1995), consisting of "commercial computer software" and "commercial
+ * computer software documentation" as such terms are used in 48
+ * C.F.R. 12.212 (SEPT 1995) and is provided to the U.S. Government
+ * only as a commercial end item.  Consistent with 48 C.F.R.12.212 and
+ * 48 C.F.R. 227.7202-1 through 227.7202-4 (JUNE 1995), all
+ * U.S. Government End Users acquire the Licensed Deliverables with
+ * only those rights set forth herein.
+ *
+ * Any use of the Licensed Deliverables in individual and commercial
+ * software must include, in the user documentation and internal
+ * comments to the code, the above Disclaimer and U.S. Government End
+ * Users Notice.
+ */
+
+#include <cuda_stdint.h>
+#include "Openmp/omp-tools.h"
+
+#if !defined(_CUPTI_OPENMP_H_)
+#define _CUPTI_OPENMP_H_
+
+#ifndef CUPTIAPI
+#ifdef _WIN32
+#define CUPTIAPI __stdcall
+#else
+#define CUPTIAPI
+#endif
+#endif
+
+#if defined(__LP64__)
+#define CUPTILP64 1
+#elif defined(_WIN64)
+#define CUPTILP64 1
+#else
+#undef CUPTILP64
+#endif
+
+#if defined(__cplusplus)
+extern "C" {
+#endif
+
+#if defined(__GNUC__) && defined(CUPTI_LIB)
+    #pragma GCC visibility push(default)
+#endif
+
+/**
+ * \brief Initialize OPENMP support (deprecated, used before OpenMP 5.0)
+ *
+ */
+int CUPTIAPI cuptiOpenMpInitialize(ompt_function_lookup_t ompt_fn_lookup, const char *runtime_version, unsigned int ompt_version);
+
+/**
+ * \brief Initialize OPENMP support
+ *
+ */
+int CUPTIAPI cuptiOpenMpInitialize_v2(ompt_function_lookup_t lookup, int initial_device_num, ompt_data_t *tool_data);
+
+#if defined(__GNUC__) && defined(CUPTI_LIB)
+    #pragma GCC visibility pop
+#endif
+
+#if defined(__cplusplus)
+}
+#endif
+
+#endif /*_CUPTI_OPENMP_H_*/

Prism/LLaDA/LLaDA_Prism/.venv/lib/python3.12/site-packages/nvidia/cuda_cupti/include/Openmp/omp-tools.h

@@ -0,0 +1,1083 @@
+/*
+ * include/50/omp-tools.h.var
+ */
+
+//===----------------------------------------------------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is dual licensed under the MIT and the University of Illinois Open
+// Source Licenses. See LICENSE.txt for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef __OMPT__
+#define __OMPT__
+
+/*****************************************************************************
+ * system include files
+ *****************************************************************************/
+
+#include <stdint.h>
+#include <stddef.h>
+
+/*****************************************************************************
+ * iteration macros
+ *****************************************************************************/
+
+#define FOREACH_OMPT_INQUIRY_FN(macro)      \
+    macro (ompt_enumerate_states)           \
+    macro (ompt_enumerate_mutex_impls)      \
+                                            \
+    macro (ompt_set_callback)               \
+    macro (ompt_get_callback)               \
+                                            \
+    macro (ompt_get_state)                  \
+                                            \
+    macro (ompt_get_parallel_info)          \
+    macro (ompt_get_task_info)              \
+    macro (ompt_get_task_memory)            \
+    macro (ompt_get_thread_data)            \
+    macro (ompt_get_unique_id)              \
+    macro (ompt_finalize_tool)              \
+                                            \
+    macro(ompt_get_num_procs)               \
+    macro(ompt_get_num_places)              \
+    macro(ompt_get_place_proc_ids)          \
+    macro(ompt_get_place_num)               \
+    macro(ompt_get_partition_place_nums)    \
+    macro(ompt_get_proc_id)                 \
+                                            \
+    macro(ompt_get_target_info)             \
+    macro(ompt_get_num_devices)
+
+#define FOREACH_OMPT_STATE(macro)                                                                \
+                                                                                                \
+    /* first available state */                                                                 \
+    macro (ompt_state_undefined, 0x102)      /* undefined thread state */                        \
+                                                                                                \
+    /* work states (0..15) */                                                                   \
+    macro (ompt_state_work_serial, 0x000)    /* working outside parallel */                      \
+    macro (ompt_state_work_parallel, 0x001)  /* working within parallel */                       \
+    macro (ompt_state_work_reduction, 0x002) /* performing a reduction */                        \
+                                                                                                \
+    /* barrier wait states (16..31) */                                                          \
+    macro (ompt_state_wait_barrier, 0x010)   /* waiting at a barrier */                          \
+    macro (ompt_state_wait_barrier_implicit_parallel, 0x011)                                     \
+                                            /* implicit barrier at the end of parallel region */\
+    macro (ompt_state_wait_barrier_implicit_workshare, 0x012)                                    \
+                                            /* implicit barrier at the end of worksharing */    \
+    macro (ompt_state_wait_barrier_implicit, 0x013)  /* implicit barrier */                      \
+    macro (ompt_state_wait_barrier_explicit, 0x014)  /* explicit barrier */                      \
+                                                                                                \
+    /* task wait states (32..63) */                                                             \
+    macro (ompt_state_wait_taskwait, 0x020)  /* waiting at a taskwait */                         \
+    macro (ompt_state_wait_taskgroup, 0x021) /* waiting at a taskgroup */                        \
+                                                                                                \
+    /* mutex wait states (64..127) */                                                           \
+    macro (ompt_state_wait_mutex, 0x040)                                                         \
+    macro (ompt_state_wait_lock, 0x041)      /* waiting for lock */                              \
+    macro (ompt_state_wait_critical, 0x042)  /* waiting for critical */                          \
+    macro (ompt_state_wait_atomic, 0x043)    /* waiting for atomic */                            \
+    macro (ompt_state_wait_ordered, 0x044)   /* waiting for ordered */                           \
+                                                                                                \
+    /* target wait states (128..255) */                                                         \
+    macro (ompt_state_wait_target, 0x080)        /* waiting for target region */                 \
+    macro (ompt_state_wait_target_map, 0x081)    /* waiting for target data mapping operation */ \
+    macro (ompt_state_wait_target_update, 0x082) /* waiting for target update operation */       \
+                                                                                                \
+    /* misc (256..511) */                                                                       \
+    macro (ompt_state_idle, 0x100)           /* waiting for work */                              \
+    macro (ompt_state_overhead, 0x101)       /* overhead excluding wait states */                \
+                                                                                                \
+    /* implementation-specific states (512..) */
+
+
+#define FOREACH_KMP_MUTEX_IMPL(macro)                                                \
+    macro (kmp_mutex_impl_none, 0)         /* unknown implementation */              \
+    macro (kmp_mutex_impl_spin, 1)         /* based on spin */                       \
+    macro (kmp_mutex_impl_queuing, 2)      /* based on some fair policy */           \
+    macro (kmp_mutex_impl_speculative, 3)  /* based on HW-supported speculation */
+
+#define FOREACH_OMPT_EVENT(macro)                                                                                        \
+                                                                                                                         \
+    /*--- Mandatory Events ---*/                                                                                         \
+    macro (ompt_callback_thread_begin,      ompt_callback_thread_begin_t,       1) /* thread begin                    */ \
+    macro (ompt_callback_thread_end,        ompt_callback_thread_end_t,         2) /* thread end                      */ \
+                                                                                                                         \
+    macro (ompt_callback_parallel_begin,    ompt_callback_parallel_begin_t,     3) /* parallel begin                  */ \
+    macro (ompt_callback_parallel_end,      ompt_callback_parallel_end_t,       4) /* parallel end                    */ \
+                                                                                                                         \
+    macro (ompt_callback_task_create,       ompt_callback_task_create_t,        5) /* task begin                      */ \
+    macro (ompt_callback_task_schedule,     ompt_callback_task_schedule_t,      6) /* task schedule                   */ \
+    macro (ompt_callback_implicit_task,     ompt_callback_implicit_task_t,      7) /* implicit task                   */ \
+                                                                                                                         \
+    macro (ompt_callback_target,            ompt_callback_target_t,             8) /* target                          */ \
+    macro (ompt_callback_target_data_op,    ompt_callback_target_data_op_t,     9) /* target data op                  */ \
+    macro (ompt_callback_target_submit,     ompt_callback_target_submit_t,     10) /* target  submit                  */ \
+                                                                                                                         \
+    macro (ompt_callback_control_tool,      ompt_callback_control_tool_t,      11) /* control tool                    */ \
+                                                                                                                         \
+    macro (ompt_callback_device_initialize, ompt_callback_device_initialize_t, 12) /* device initialize               */ \
+    macro (ompt_callback_device_finalize,   ompt_callback_device_finalize_t,   13) /* device finalize                 */ \
+                                                                                                                         \
+    macro (ompt_callback_device_load,       ompt_callback_device_load_t,       14) /* device load                     */ \
+    macro (ompt_callback_device_unload,     ompt_callback_device_unload_t,     15) /* device unload                   */ \
+                                                                                                                         \
+    /* Optional Events */                                                                                                \
+    macro (ompt_callback_sync_region_wait,  ompt_callback_sync_region_t,       16) /* sync region wait begin or end   */ \
+                                                                                                                         \
+    macro (ompt_callback_mutex_released,    ompt_callback_mutex_t,             17) /* mutex released                  */ \
+                                                                                                                         \
+    macro (ompt_callback_dependences,       ompt_callback_dependences_t,       18) /* report task dependences         */ \
+    macro (ompt_callback_task_dependence,   ompt_callback_task_dependence_t,   19) /* report task dependence          */ \
+                                                                                                                         \
+    macro (ompt_callback_work,              ompt_callback_work_t,              20) /* task at work begin or end       */ \
+                                                                                                                         \
+    macro (ompt_callback_master,            ompt_callback_master_t,            21) /* task at master begin or end     */ \
+                                                                                                                         \
+    macro (ompt_callback_target_map,        ompt_callback_target_map_t,        22) /* target map                      */ \
+                                                                                                                         \
+    macro (ompt_callback_sync_region,       ompt_callback_sync_region_t,       23) /* sync region begin or end        */ \
+                                                                                                                         \
+    macro (ompt_callback_lock_init,         ompt_callback_mutex_acquire_t,     24) /* lock init                       */ \
+    macro (ompt_callback_lock_destroy,      ompt_callback_mutex_t,             25) /* lock destroy                    */ \
+                                                                                                                         \
+    macro (ompt_callback_mutex_acquire,     ompt_callback_mutex_acquire_t,     26) /* mutex acquire                   */ \
+    macro (ompt_callback_mutex_acquired,    ompt_callback_mutex_t,             27) /* mutex acquired                  */ \
+                                                                                                                         \
+    macro (ompt_callback_nest_lock,         ompt_callback_nest_lock_t,         28) /* nest lock                       */ \
+                                                                                                                         \
+    macro (ompt_callback_flush,             ompt_callback_flush_t,             29) /* after executing flush           */ \
+                                                                                                                         \
+    macro (ompt_callback_cancel,            ompt_callback_cancel_t,            30) /* cancel innermost binding region */ \
+                                                                                                                         \
+    macro (ompt_callback_reduction,         ompt_callback_sync_region_t,       31) /* reduction                       */ \
+                                                                                                                         \
+    macro (ompt_callback_dispatch,          ompt_callback_dispatch_t,          32) /* dispatch of work                */
+
+/*****************************************************************************
+ * implementation specific types
+ *****************************************************************************/
+
+typedef enum kmp_mutex_impl_t {
+#define kmp_mutex_impl_macro(impl, code) impl = code,
+    FOREACH_KMP_MUTEX_IMPL(kmp_mutex_impl_macro)
+#undef kmp_mutex_impl_macro
+} kmp_mutex_impl_t;
+
+/*****************************************************************************
+ * definitions generated from spec
+ *****************************************************************************/
+
+typedef enum ompt_callbacks_t {
+  ompt_callback_thread_begin             = 1,
+  ompt_callback_thread_end               = 2,
+  ompt_callback_parallel_begin           = 3,
+  ompt_callback_parallel_end             = 4,
+  ompt_callback_task_create              = 5,
+  ompt_callback_task_schedule            = 6,
+  ompt_callback_implicit_task            = 7,
+  ompt_callback_target                   = 8,
+  ompt_callback_target_data_op           = 9,
+  ompt_callback_target_submit            = 10,
+  ompt_callback_control_tool             = 11,
+  ompt_callback_device_initialize        = 12,
+  ompt_callback_device_finalize          = 13,
+  ompt_callback_device_load              = 14,
+  ompt_callback_device_unload            = 15,
+  ompt_callback_sync_region_wait         = 16,
+  ompt_callback_mutex_released           = 17,
+  ompt_callback_dependences              = 18,
+  ompt_callback_task_dependence          = 19,
+  ompt_callback_work                     = 20,
+  ompt_callback_master                   = 21,
+  ompt_callback_target_map               = 22,
+  ompt_callback_sync_region              = 23,
+  ompt_callback_lock_init                = 24,
+  ompt_callback_lock_destroy             = 25,
+  ompt_callback_mutex_acquire            = 26,
+  ompt_callback_mutex_acquired           = 27,
+  ompt_callback_nest_lock                = 28,
+  ompt_callback_flush                    = 29,
+  ompt_callback_cancel                   = 30,
+  ompt_callback_reduction                = 31,
+  ompt_callback_dispatch                 = 32
+} ompt_callbacks_t;
+
+typedef enum ompt_record_t {
+  ompt_record_ompt               = 1,
+  ompt_record_native             = 2,
+  ompt_record_invalid            = 3
+} ompt_record_t;
+
+typedef enum ompt_record_native_t {
+  ompt_record_native_info  = 1,
+  ompt_record_native_event = 2
+} ompt_record_native_t;
+
+typedef enum ompt_set_result_t {
+  ompt_set_error            = 0,
+  ompt_set_never            = 1,
+  ompt_set_impossible       = 2,
+  ompt_set_sometimes        = 3,
+  ompt_set_sometimes_paired = 4,
+  ompt_set_always           = 5
+} ompt_set_result_t;
+
+typedef uint64_t ompt_id_t;
+
+typedef uint64_t ompt_device_time_t;
+
+typedef uint64_t ompt_buffer_cursor_t;
+
+typedef enum ompt_thread_t {
+  ompt_thread_initial                 = 1,
+  ompt_thread_worker                  = 2,
+  ompt_thread_other                   = 3,
+  ompt_thread_unknown                 = 4
+} ompt_thread_t;
+
+typedef enum ompt_scope_endpoint_t {
+  ompt_scope_begin                    = 1,
+  ompt_scope_end                      = 2
+} ompt_scope_endpoint_t;
+
+typedef enum ompt_dispatch_t {
+  ompt_dispatch_iteration             = 1,
+  ompt_dispatch_section               = 2
+} ompt_dispatch_t;
+
+typedef enum ompt_sync_region_t {
+  ompt_sync_region_barrier                = 1,
+  ompt_sync_region_barrier_implicit       = 2,
+  ompt_sync_region_barrier_explicit       = 3,
+  ompt_sync_region_barrier_implementation = 4,
+  ompt_sync_region_taskwait               = 5,
+  ompt_sync_region_taskgroup              = 6,
+  ompt_sync_region_reduction              = 7
+} ompt_sync_region_t;
+
+typedef enum ompt_target_data_op_t {
+  ompt_target_data_alloc                = 1,
+  ompt_target_data_transfer_to_device   = 2,
+  ompt_target_data_transfer_from_device = 3,
+  ompt_target_data_delete               = 4,
+  ompt_target_data_associate            = 5,
+  ompt_target_data_disassociate         = 6
+} ompt_target_data_op_t;
+
+typedef enum ompt_work_t {
+  ompt_work_loop               = 1,
+  ompt_work_sections           = 2,
+  ompt_work_single_executor    = 3,
+  ompt_work_single_other       = 4,
+  ompt_work_workshare          = 5,
+  ompt_work_distribute         = 6,
+  ompt_work_taskloop           = 7
+} ompt_work_t;
+
+typedef enum ompt_mutex_t {
+  ompt_mutex_lock                     = 1,
+  ompt_mutex_test_lock                = 2,
+  ompt_mutex_nest_lock                = 3,
+  ompt_mutex_test_nest_lock           = 4,
+  ompt_mutex_critical                 = 5,
+  ompt_mutex_atomic                   = 6,
+  ompt_mutex_ordered                  = 7
+} ompt_mutex_t;
+
+typedef enum ompt_native_mon_flag_t {
+  ompt_native_data_motion_explicit    = 0x01,
+  ompt_native_data_motion_implicit    = 0x02,
+  ompt_native_kernel_invocation       = 0x04,
+  ompt_native_kernel_execution        = 0x08,
+  ompt_native_driver                  = 0x10,
+  ompt_native_runtime                 = 0x20,
+  ompt_native_overhead                = 0x40,
+  ompt_native_idleness                = 0x80
+} ompt_native_mon_flag_t;
+
+typedef enum ompt_task_flag_t {
+  ompt_task_initial                   = 0x00000001,
+  ompt_task_implicit                  = 0x00000002,
+  ompt_task_explicit                  = 0x00000004,
+  ompt_task_target                    = 0x00000008,
+  ompt_task_undeferred                = 0x08000000,
+  ompt_task_untied                    = 0x10000000,
+  ompt_task_final                     = 0x20000000,
+  ompt_task_mergeable                 = 0x40000000,
+  ompt_task_merged                    = 0x80000000
+} ompt_task_flag_t;
+
+typedef enum ompt_task_status_t {
+  ompt_task_complete      = 1,
+  ompt_task_yield         = 2,
+  ompt_task_cancel        = 3,
+  ompt_task_detach        = 4,
+  ompt_task_early_fulfill = 5,
+  ompt_task_late_fulfill  = 6,
+  ompt_task_switch        = 7
+} ompt_task_status_t;
+
+typedef enum ompt_target_t {
+  ompt_target                         = 1,
+  ompt_target_enter_data              = 2,
+  ompt_target_exit_data               = 3,
+  ompt_target_update                  = 4
+} ompt_target_t;
+
+typedef enum ompt_parallel_flag_t {
+  ompt_parallel_invoker_program = 0x00000001,
+  ompt_parallel_invoker_runtime = 0x00000002,
+  ompt_parallel_league          = 0x40000000,
+  ompt_parallel_team            = 0x80000000
+} ompt_parallel_flag_t;
+
+typedef enum ompt_target_map_flag_t {
+  ompt_target_map_flag_to             = 0x01,
+  ompt_target_map_flag_from           = 0x02,
+  ompt_target_map_flag_alloc          = 0x04,
+  ompt_target_map_flag_release        = 0x08,
+  ompt_target_map_flag_delete         = 0x10,
+  ompt_target_map_flag_implicit       = 0x20
+} ompt_target_map_flag_t;
+
+typedef enum ompt_dependence_type_t {
+  ompt_dependence_type_in              = 1,
+  ompt_dependence_type_out             = 2,
+  ompt_dependence_type_inout           = 3,
+  ompt_dependence_type_mutexinoutset   = 4,
+  ompt_dependence_type_source          = 5,
+  ompt_dependence_type_sink            = 6
+} ompt_dependence_type_t;
+
+typedef enum ompt_cancel_flag_t {
+  ompt_cancel_parallel       = 0x01,
+  ompt_cancel_sections       = 0x02,
+  ompt_cancel_loop           = 0x04,
+  ompt_cancel_taskgroup      = 0x08,
+  ompt_cancel_activated      = 0x10,
+  ompt_cancel_detected       = 0x20,
+  ompt_cancel_discarded_task = 0x40
+} ompt_cancel_flag_t;
+
+typedef uint64_t ompt_hwid_t;
+
+typedef uint64_t ompt_wait_id_t;
+
+typedef enum ompt_frame_flag_t {
+  ompt_frame_runtime        = 0x00,
+  ompt_frame_application    = 0x01,
+  ompt_frame_cfa            = 0x10,
+  ompt_frame_framepointer   = 0x20,
+  ompt_frame_stackaddress   = 0x30
+} ompt_frame_flag_t; 
+
+typedef enum ompt_state_t {
+  ompt_state_work_serial                      = 0x000,
+  ompt_state_work_parallel                    = 0x001,
+  ompt_state_work_reduction                   = 0x002,
+
+  ompt_state_wait_barrier                     = 0x010,
+  ompt_state_wait_barrier_implicit_parallel   = 0x011,
+  ompt_state_wait_barrier_implicit_workshare  = 0x012,
+  ompt_state_wait_barrier_implicit            = 0x013,
+  ompt_state_wait_barrier_explicit            = 0x014,
+
+  ompt_state_wait_taskwait                    = 0x020,
+  ompt_state_wait_taskgroup                   = 0x021,
+
+  ompt_state_wait_mutex                       = 0x040,
+  ompt_state_wait_lock                        = 0x041,
+  ompt_state_wait_critical                    = 0x042,
+  ompt_state_wait_atomic                      = 0x043,
+  ompt_state_wait_ordered                     = 0x044,
+
+  ompt_state_wait_target                      = 0x080,
+  ompt_state_wait_target_map                  = 0x081,
+  ompt_state_wait_target_update               = 0x082,
+
+  ompt_state_idle                             = 0x100,
+  ompt_state_overhead                         = 0x101,
+  ompt_state_undefined                        = 0x102
+} ompt_state_t;
+
+typedef uint64_t (*ompt_get_unique_id_t) (void);
+
+typedef uint64_t ompd_size_t;
+
+typedef uint64_t ompd_wait_id_t;
+
+typedef uint64_t ompd_addr_t;
+typedef int64_t  ompd_word_t;
+typedef uint64_t ompd_seg_t;
+
+typedef uint64_t ompd_device_t;
+
+typedef uint64_t ompd_thread_id_t;
+
+typedef enum ompd_scope_t {
+  ompd_scope_global = 1,
+  ompd_scope_address_space = 2,
+  ompd_scope_thread = 3,
+  ompd_scope_parallel = 4,
+  ompd_scope_implicit_task = 5,
+  ompd_scope_task = 6
+} ompd_scope_t;
+
+typedef uint64_t ompd_icv_id_t;
+
+typedef enum ompd_rc_t {
+  ompd_rc_ok = 0,
+  ompd_rc_unavailable = 1,
+  ompd_rc_stale_handle = 2,
+  ompd_rc_bad_input = 3,
+  ompd_rc_error = 4,
+  ompd_rc_unsupported = 5,
+  ompd_rc_needs_state_tracking = 6,
+  ompd_rc_incompatible = 7,
+  ompd_rc_device_read_error = 8,
+  ompd_rc_device_write_error = 9,
+  ompd_rc_nomem = 10,
+} ompd_rc_t;
+
+typedef void (*ompt_interface_fn_t) (void);
+
+typedef ompt_interface_fn_t (*ompt_function_lookup_t) (
+  const char *interface_function_name
+);
+
+typedef union ompt_data_t {
+  uint64_t value;
+  void *ptr;
+} ompt_data_t;
+
+typedef struct ompt_frame_t {
+  ompt_data_t exit_frame;
+  ompt_data_t enter_frame;
+  int exit_frame_flags;
+  int enter_frame_flags;
+} ompt_frame_t;
+
+typedef void (*ompt_callback_t) (void);
+
+typedef void ompt_device_t;
+
+typedef void ompt_buffer_t;
+
+typedef void (*ompt_callback_buffer_request_t) (
+  int device_num,
+  ompt_buffer_t **buffer,
+  size_t *bytes
+);
+
+typedef void (*ompt_callback_buffer_complete_t) (
+  int device_num,
+  ompt_buffer_t *buffer,
+  size_t bytes,
+  ompt_buffer_cursor_t begin,
+  int buffer_owned
+);
+
+typedef void (*ompt_finalize_t) (
+  ompt_data_t *tool_data
+);
+
+typedef int (*ompt_initialize_t) (
+  ompt_function_lookup_t lookup,
+  int initial_device_num,
+  ompt_data_t *tool_data
+);
+
+typedef struct ompt_start_tool_result_t {
+  ompt_initialize_t initialize;
+  ompt_finalize_t finalize;
+  ompt_data_t tool_data;
+} ompt_start_tool_result_t;
+
+typedef struct ompt_record_abstract_t {
+  ompt_record_native_t rclass;
+  const char *type;
+  ompt_device_time_t start_time;
+  ompt_device_time_t end_time;
+  ompt_hwid_t hwid;
+} ompt_record_abstract_t;
+
+typedef struct ompt_dependence_t {
+  ompt_data_t variable;
+  ompt_dependence_type_t dependence_type;
+} ompt_dependence_t;
+
+typedef int (*ompt_enumerate_states_t) (
+  int current_state,
+  int *next_state,
+  const char **next_state_name
+);
+
+typedef int (*ompt_enumerate_mutex_impls_t) (
+  int current_impl,
+  int *next_impl,
+  const char **next_impl_name
+);
+
+typedef ompt_set_result_t (*ompt_set_callback_t) (
+  ompt_callbacks_t event,
+  ompt_callback_t callback
+);
+
+typedef int (*ompt_get_callback_t) (
+  ompt_callbacks_t event,
+  ompt_callback_t *callback
+);
+
+typedef ompt_data_t *(*ompt_get_thread_data_t) (void);
+
+typedef int (*ompt_get_num_procs_t) (void);
+
+typedef int (*ompt_get_num_places_t) (void);
+
+typedef int (*ompt_get_place_proc_ids_t) (
+  int place_num,
+  int ids_size,
+  int *ids
+);
+
+typedef int (*ompt_get_place_num_t) (void);
+
+typedef int (*ompt_get_partition_place_nums_t) (
+  int place_nums_size,
+  int *place_nums
+);
+
+typedef int (*ompt_get_proc_id_t) (void);
+
+typedef int (*ompt_get_state_t) (
+  ompt_wait_id_t *wait_id
+);
+
+typedef int (*ompt_get_parallel_info_t) (
+  int ancestor_level,
+  ompt_data_t **parallel_data,
+  int *team_size
+);
+
+typedef int (*ompt_get_task_info_t) (
+  int ancestor_level,
+  int *flags,
+  ompt_data_t **task_data,
+  ompt_frame_t **task_frame,
+  ompt_data_t **parallel_data,
+  int *thread_num
+);
+
+typedef int (*ompt_get_task_memory_t)(
+  void **addr,
+  size_t *size,
+  int block
+);
+
+typedef int (*ompt_get_target_info_t) (
+  uint64_t *device_num,
+  ompt_id_t *target_id,
+  ompt_id_t *host_op_id
+);
+
+typedef int (*ompt_get_num_devices_t) (void);
+
+typedef void (*ompt_finalize_tool_t) (void);
+
+typedef int (*ompt_get_device_num_procs_t) (
+  ompt_device_t *device
+);
+
+typedef ompt_device_time_t (*ompt_get_device_time_t) (
+  ompt_device_t *device
+);
+
+typedef double (*ompt_translate_time_t) (
+  ompt_device_t *device,
+  ompt_device_time_t time
+);
+
+typedef ompt_set_result_t (*ompt_set_trace_ompt_t) (
+  ompt_device_t *device,
+  unsigned int enable,
+  unsigned int etype
+);
+
+typedef ompt_set_result_t (*ompt_set_trace_native_t) (
+  ompt_device_t *device,
+  int enable,
+  int flags
+);
+
+typedef int (*ompt_start_trace_t) (
+  ompt_device_t *device,
+  ompt_callback_buffer_request_t request,
+  ompt_callback_buffer_complete_t complete
+);
+
+typedef int (*ompt_pause_trace_t) (
+  ompt_device_t *device,
+  int begin_pause
+);
+
+typedef int (*ompt_flush_trace_t) (
+  ompt_device_t *device
+);
+
+typedef int (*ompt_stop_trace_t) (
+  ompt_device_t *device
+);
+
+typedef int (*ompt_advance_buffer_cursor_t) (
+  ompt_device_t *device,
+  ompt_buffer_t *buffer,
+  size_t size,
+  ompt_buffer_cursor_t current,
+  ompt_buffer_cursor_t *next
+);
+
+typedef ompt_record_t (*ompt_get_record_type_t) (
+  ompt_buffer_t *buffer,
+  ompt_buffer_cursor_t current
+);
+
+typedef void *(*ompt_get_record_native_t) (
+  ompt_buffer_t *buffer,
+  ompt_buffer_cursor_t current,
+  ompt_id_t *host_op_id
+);
+
+typedef ompt_record_abstract_t *
+(*ompt_get_record_abstract_t) (
+  void *native_record
+);
+
+typedef void (*ompt_callback_thread_begin_t) (
+  ompt_thread_t thread_type,
+  ompt_data_t *thread_data
+);
+
+typedef struct ompt_record_thread_begin_t {
+  ompt_thread_t thread_type;
+} ompt_record_thread_begin_t;
+
+typedef void (*ompt_callback_thread_end_t) (
+  ompt_data_t *thread_data
+);
+
+typedef void (*ompt_callback_parallel_begin_t) (
+  ompt_data_t *encountering_task_data,
+  const ompt_frame_t *encountering_task_frame,
+  ompt_data_t *parallel_data,
+  unsigned int requested_parallelism,
+  int flags,
+  const void *codeptr_ra
+);
+
+typedef struct ompt_record_parallel_begin_t {
+  ompt_id_t encountering_task_id;
+  ompt_id_t parallel_id;
+  unsigned int requested_parallelism;
+  int flags;
+  const void *codeptr_ra;
+} ompt_record_parallel_begin_t;
+
+typedef void (*ompt_callback_parallel_end_t) (
+  ompt_data_t *parallel_data,
+  ompt_data_t *encountering_task_data,
+  int flags,
+  const void *codeptr_ra
+);
+
+typedef struct ompt_record_parallel_end_t {
+  ompt_id_t parallel_id;
+  ompt_id_t encountering_task_id;
+  int flags;
+  const void *codeptr_ra;
+} ompt_record_parallel_end_t;
+
+typedef void (*ompt_callback_work_t) (
+  ompt_work_t wstype,
+  ompt_scope_endpoint_t endpoint,
+  ompt_data_t *parallel_data,
+  ompt_data_t *task_data,
+  uint64_t count,
+  const void *codeptr_ra
+);
+
+typedef struct ompt_record_work_t {
+  ompt_work_t wstype;
+  ompt_scope_endpoint_t endpoint;
+  ompt_id_t parallel_id;
+  ompt_id_t task_id;
+  uint64_t count;
+  const void *codeptr_ra;
+} ompt_record_work_t;
+
+typedef void (*ompt_callback_dispatch_t) (
+  ompt_data_t *parallel_data,
+  ompt_data_t *task_data,
+  ompt_dispatch_t kind,
+  ompt_data_t instance 
+);
+
+typedef struct ompt_record_dispatch_t {
+  ompt_id_t parallel_id;
+  ompt_id_t task_id;
+  ompt_dispatch_t kind;
+  ompt_data_t instance; 
+} ompt_record_dispatch_t;
+
+typedef void (*ompt_callback_task_create_t) (
+  ompt_data_t *encountering_task_data,
+  const ompt_frame_t *encountering_task_frame,
+  ompt_data_t *new_task_data,
+  int flags,
+  int has_dependences,
+  const void *codeptr_ra
+);
+
+typedef struct ompt_record_task_create_t {
+  ompt_id_t encountering_task_id;
+  ompt_id_t new_task_id;
+  int flags;
+  int has_dependences;
+  const void *codeptr_ra;
+} ompt_record_task_create_t;
+
+typedef void (*ompt_callback_dependences_t) (
+  ompt_data_t *task_data,
+  const ompt_dependence_t *deps,
+  int ndeps
+);
+
+typedef struct ompt_record_dependences_t {
+  ompt_id_t task_id;
+  ompt_dependence_t dep;
+  int ndeps;
+} ompt_record_dependences_t;
+
+typedef void (*ompt_callback_task_dependence_t) (
+  ompt_data_t *src_task_data,
+  ompt_data_t *sink_task_data
+);
+
+typedef struct ompt_record_task_dependence_t {
+  ompt_id_t src_task_id;
+  ompt_id_t sink_task_id;
+} ompt_record_task_dependence_t;
+
+typedef void (*ompt_callback_task_schedule_t) (
+  ompt_data_t *prior_task_data,
+  ompt_task_status_t prior_task_status,
+  ompt_data_t *next_task_data
+);
+
+typedef struct ompt_record_task_schedule_t {
+  ompt_id_t prior_task_id;
+  ompt_task_status_t prior_task_status;
+  ompt_id_t next_task_id;
+} ompt_record_task_schedule_t;
+
+typedef void (*ompt_callback_implicit_task_t) (
+  ompt_scope_endpoint_t endpoint,
+  ompt_data_t *parallel_data,
+  ompt_data_t *task_data,
+  unsigned int actual_parallelism,
+  unsigned int index,
+  int flags
+);
+
+typedef struct ompt_record_implicit_task_t {
+  ompt_scope_endpoint_t endpoint;
+  ompt_id_t parallel_id;
+  ompt_id_t task_id;
+  unsigned int actual_parallelism;
+  unsigned int index;
+  int flags;
+} ompt_record_implicit_task_t;
+
+typedef void (*ompt_callback_master_t) (
+  ompt_scope_endpoint_t endpoint,
+  ompt_data_t *parallel_data,
+  ompt_data_t *task_data,
+  const void *codeptr_ra
+);
+
+typedef struct ompt_record_master_t {
+  ompt_scope_endpoint_t endpoint;
+  ompt_id_t parallel_id;
+  ompt_id_t task_id;
+  const void *codeptr_ra;
+} ompt_record_master_t;
+
+typedef void (*ompt_callback_sync_region_t) (
+  ompt_sync_region_t kind,
+  ompt_scope_endpoint_t endpoint,
+  ompt_data_t *parallel_data,
+  ompt_data_t *task_data,
+  const void *codeptr_ra
+);
+
+typedef struct ompt_record_sync_region_t {
+  ompt_sync_region_t kind;
+  ompt_scope_endpoint_t endpoint;
+  ompt_id_t parallel_id;
+  ompt_id_t task_id;
+  const void *codeptr_ra;
+} ompt_record_sync_region_t;
+
+typedef void (*ompt_callback_mutex_acquire_t) (
+  ompt_mutex_t kind,
+  unsigned int hint,
+  unsigned int impl,
+  ompt_wait_id_t wait_id,
+  const void *codeptr_ra
+);
+
+typedef struct ompt_record_mutex_acquire_t {
+  ompt_mutex_t kind;
+  unsigned int hint;
+  unsigned int impl;
+  ompt_wait_id_t wait_id;
+  const void *codeptr_ra;
+} ompt_record_mutex_acquire_t;
+
+typedef void (*ompt_callback_mutex_t) (
+  ompt_mutex_t kind,
+  ompt_wait_id_t wait_id,
+  const void *codeptr_ra
+);
+
+typedef struct ompt_record_mutex_t {
+  ompt_mutex_t kind;
+  ompt_wait_id_t wait_id;
+  const void *codeptr_ra;
+} ompt_record_mutex_t;
+
+typedef void (*ompt_callback_nest_lock_t) (
+  ompt_scope_endpoint_t endpoint,
+  ompt_wait_id_t wait_id,
+  const void *codeptr_ra
+);
+
+typedef struct ompt_record_nest_lock_t {
+  ompt_scope_endpoint_t endpoint;
+  ompt_wait_id_t wait_id;
+  const void *codeptr_ra;
+} ompt_record_nest_lock_t;
+
+typedef void (*ompt_callback_flush_t) (
+  ompt_data_t *thread_data,
+  const void *codeptr_ra
+);
+
+typedef struct ompt_record_flush_t {
+  const void *codeptr_ra;
+} ompt_record_flush_t;
+
+typedef void (*ompt_callback_cancel_t) (
+  ompt_data_t *task_data,
+  int flags,
+  const void *codeptr_ra
+);
+
+typedef struct ompt_record_cancel_t {
+  ompt_id_t task_id;
+  int flags;
+  const void *codeptr_ra;
+} ompt_record_cancel_t;
+
+typedef void (*ompt_callback_device_initialize_t) (
+  int device_num,
+  const char *type,
+  ompt_device_t *device,
+  ompt_function_lookup_t lookup,
+  const char *documentation
+);
+
+typedef void (*ompt_callback_device_finalize_t) (
+  int device_num
+);
+
+typedef void (*ompt_callback_device_load_t) (
+  int device_num,
+  const char *filename,
+  int64_t offset_in_file,
+  void *vma_in_file,
+  size_t bytes,
+  void *host_addr,
+  void *device_addr,
+  uint64_t module_id
+);
+
+typedef void (*ompt_callback_device_unload_t) (
+  int device_num,
+  uint64_t module_id
+);
+
+typedef void (*ompt_callback_target_data_op_t) (
+  ompt_id_t target_id,
+  ompt_id_t host_op_id,
+  ompt_target_data_op_t optype,
+  void *src_addr,
+  int src_device_num,
+  void *dest_addr,
+  int dest_device_num,
+  size_t bytes,
+  const void *codeptr_ra
+);
+
+typedef struct ompt_record_target_data_op_t {
+  ompt_id_t host_op_id;
+  ompt_target_data_op_t optype;
+  void *src_addr;
+  int src_device_num;
+  void *dest_addr;
+  int dest_device_num;
+  size_t bytes;
+  ompt_device_time_t end_time;
+  const void *codeptr_ra;
+} ompt_record_target_data_op_t;
+
+typedef void (*ompt_callback_target_t) (
+  ompt_target_t kind,
+  ompt_scope_endpoint_t endpoint,
+  int device_num,
+  ompt_data_t *task_data,
+  ompt_id_t target_id,
+  const void *codeptr_ra
+);
+
+typedef struct ompt_record_target_t {
+  ompt_target_t kind;
+  ompt_scope_endpoint_t endpoint;
+  int device_num;
+  ompt_id_t task_id;
+  ompt_id_t target_id;
+  const void *codeptr_ra;
+} ompt_record_target_t;
+
+typedef void (*ompt_callback_target_map_t) (
+  ompt_id_t target_id,
+  unsigned int nitems,
+  void **host_addr,
+  void **device_addr,
+  size_t *bytes,
+  unsigned int *mapping_flags,
+  const void *codeptr_ra
+);
+
+typedef struct ompt_record_target_map_t {
+  ompt_id_t target_id;
+  unsigned int nitems;
+  void **host_addr;
+  void **device_addr;
+  size_t *bytes;
+  unsigned int *mapping_flags;
+  const void *codeptr_ra;
+} ompt_record_target_map_t;
+
+typedef void (*ompt_callback_target_submit_t) (
+  ompt_id_t target_id,
+  ompt_id_t host_op_id,
+  unsigned int requested_num_teams
+);
+
+typedef struct ompt_record_target_kernel_t {
+  ompt_id_t host_op_id;
+  unsigned int requested_num_teams;
+  unsigned int granted_num_teams;
+  ompt_device_time_t end_time;
+} ompt_record_target_kernel_t;
+
+typedef int (*ompt_callback_control_tool_t) (
+  uint64_t command,
+  uint64_t modifier,
+  void *arg,
+  const void *codeptr_ra
+);
+
+typedef struct ompt_record_control_tool_t {
+  uint64_t command;
+  uint64_t modifier;
+  const void *codeptr_ra;
+} ompt_record_control_tool_t;
+
+typedef struct ompd_address_t {
+  ompd_seg_t segment;
+  ompd_addr_t address;
+} ompd_address_t;
+
+typedef struct ompd_frame_info_t {
+  ompd_address_t frame_address;
+  ompd_word_t frame_flag;
+} ompd_frame_info_t;
+
+typedef struct _ompd_aspace_handle ompd_address_space_handle_t;
+typedef struct _ompd_thread_handle ompd_thread_handle_t;
+typedef struct _ompd_parallel_handle ompd_parallel_handle_t;
+typedef struct _ompd_task_handle ompd_task_handle_t;
+
+typedef struct _ompd_aspace_cont ompd_address_space_context_t;
+typedef struct _ompd_thread_cont ompd_thread_context_t;
+
+typedef struct ompd_device_type_sizes_t {
+  uint8_t sizeof_char;
+  uint8_t sizeof_short;
+  uint8_t sizeof_int;
+  uint8_t sizeof_long;
+  uint8_t sizeof_long_long;
+  uint8_t sizeof_pointer;
+} ompd_device_type_sizes_t;
+
+typedef struct ompt_record_ompt_t {
+  ompt_callbacks_t type;
+  ompt_device_time_t time;
+  ompt_id_t thread_id;
+  ompt_id_t target_id;
+  union {
+    ompt_record_thread_begin_t thread_begin;
+    ompt_record_parallel_begin_t parallel_begin;
+    ompt_record_parallel_end_t parallel_end;
+    ompt_record_work_t work;
+    ompt_record_dispatch_t dispatch;
+    ompt_record_task_create_t task_create;
+    ompt_record_dependences_t dependences;
+    ompt_record_task_dependence_t task_dependence;
+    ompt_record_task_schedule_t task_schedule;
+    ompt_record_implicit_task_t implicit_task;
+    ompt_record_master_t master;
+    ompt_record_sync_region_t sync_region;
+    ompt_record_mutex_acquire_t mutex_acquire;
+    ompt_record_mutex_t mutex;
+    ompt_record_nest_lock_t nest_lock;
+    ompt_record_flush_t flush;
+    ompt_record_cancel_t cancel;
+    ompt_record_target_t target;
+    ompt_record_target_data_op_t target_data_op;
+    ompt_record_target_map_t target_map;
+    ompt_record_target_kernel_t target_kernel;
+    ompt_record_control_tool_t control_tool;
+  } record;
+} ompt_record_ompt_t;
+
+typedef ompt_record_ompt_t *(*ompt_get_record_ompt_t) (
+  ompt_buffer_t *buffer,
+  ompt_buffer_cursor_t current
+);
+
+#define ompt_id_none 0
+#define ompt_data_none {0}
+#define ompt_time_none 0
+#define ompt_hwid_none 0
+#define ompt_addr_none ~0
+#define ompt_mutex_impl_none 0
+#define ompt_wait_id_none 0
+
+#define ompd_segment_none 0
+
+#endif /* __OMPT__ */

Prism/LLaDA/LLaDA_Prism/.venv/lib/python3.12/site-packages/nvidia/cuda_cupti/include/cuda_stdint.h

@@ -0,0 +1,112 @@
+/*
+ * Copyright 2009-2017 NVIDIA Corporation.  All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *  * Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *  * Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *  * Neither the name of NVIDIA CORPORATION nor the names of its
+ *    contributors may be used to endorse or promote products derived
+ *    from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
+ * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+ * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT OWNER OR
+ * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
+ * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
+ * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
+ * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
+ * OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef __cuda_stdint_h__
+#define __cuda_stdint_h__
+
+// Compiler-specific treatment for C99's stdint.h
+//
+// By default, this header will use the standard headers (so it
+// is your responsibility to make sure they are available), except
+// on MSVC before Visual Studio 2010, when they were not provided.
+// To support old MSVC, a few of the commonly-used definitions are
+// provided here.  If more definitions are needed, add them here,
+// or replace these definitions with a complete implementation,
+// such as the ones available from Google, Boost, or MSVC10.  You
+// can prevent the definition of any of these types (in order to
+// use your own) by #defining CU_STDINT_TYPES_ALREADY_DEFINED.
+
+#if !defined(CU_STDINT_TYPES_ALREADY_DEFINED)
+
+// In VS including stdint.h forces the C++ runtime dep - provide an opt-out
+// (CU_STDINT_VS_FORCE_NO_STDINT_H) for users that care (notably static
+// cudart).
+#if defined(_MSC_VER) && ((_MSC_VER < 1600) || defined(CU_STDINT_VS_FORCE_NO_STDINT_H))
+
+// These definitions can be used with MSVC 8 and 9,
+// which don't ship with stdint.h:
+
+typedef unsigned   char   uint8_t;
+
+typedef            short  int16_t;
+typedef unsigned   short uint16_t;
+
+// To keep it consistent with all MSVC build. define those types
+// in the exact same way they are defined with the MSVC headers
+#if defined(_MSC_VER)
+typedef signed     char    int8_t;
+
+typedef            int     int32_t;
+typedef unsigned   int     uint32_t;
+
+typedef long long          int64_t;
+typedef unsigned long long uint64_t;
+#else
+typedef            char    int8_t;
+
+typedef            long   int32_t;
+typedef unsigned   long  uint32_t;
+
+typedef          __int64  int64_t;
+typedef unsigned __int64 uint64_t;
+#endif
+
+#elif defined(__DJGPP__)
+
+// These definitions can be used when compiling
+// C code with DJGPP, which only provides stdint.h
+// when compiling C++ code with TR1 enabled.
+
+typedef               char    int8_t;
+typedef unsigned      char   uint8_t;
+
+typedef               short  int16_t;
+typedef unsigned      short uint16_t;
+
+typedef               long   int32_t;
+typedef unsigned      long  uint32_t;
+
+typedef          long long   int64_t;
+typedef unsigned long long  uint64_t;
+
+#else
+
+// Use standard headers, as specified by C99 and C++ TR1.
+// Known to be provided by:
+// - gcc/glibc, supported by all versions of glibc
+// - djgpp, supported since 2001
+// - MSVC, supported by Visual Studio 2010 and later
+
+#include <stdint.h>
+
+#endif
+
+#endif // !defined(CU_STDINT_TYPES_ALREADY_DEFINED)
+
+
+#endif // file guard

Prism/LLaDA/LLaDA_Prism/.venv/lib/python3.12/site-packages/nvidia/cuda_cupti/include/cupti.h

@@ -0,0 +1,123 @@
+/*
+ * Copyright 2010-2017 NVIDIA Corporation.  All rights reserved.
+ *
+ * NOTICE TO LICENSEE:
+ *
+ * This source code and/or documentation ("Licensed Deliverables") are
+ * subject to NVIDIA intellectual property rights under U.S. and
+ * international Copyright laws.
+ *
+ * These Licensed Deliverables contained herein is PROPRIETARY and
+ * CONFIDENTIAL to NVIDIA and is being provided under the terms and
+ * conditions of a form of NVIDIA software license agreement by and
+ * between NVIDIA and Licensee ("License Agreement") or electronically
+ * accepted by Licensee.  Notwithstanding any terms or conditions to
+ * the contrary in the License Agreement, reproduction or disclosure
+ * of the Licensed Deliverables to any third party without the express
+ * written consent of NVIDIA is prohibited.
+ *
+ * NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE
+ * LICENSE AGREEMENT, NVIDIA MAKES NO REPRESENTATION ABOUT THE
+ * SUITABILITY OF THESE LICENSED DELIVERABLES FOR ANY PURPOSE.  IT IS
+ * PROVIDED "AS IS" WITHOUT EXPRESS OR IMPLIED WARRANTY OF ANY KIND.
+ * NVIDIA DISCLAIMS ALL WARRANTIES WITH REGARD TO THESE LICENSED
+ * DELIVERABLES, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY,
+ * NONINFRINGEMENT, AND FITNESS FOR A PARTICULAR PURPOSE.
+ * NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE
+ * LICENSE AGREEMENT, IN NO EVENT SHALL NVIDIA BE LIABLE FOR ANY
+ * SPECIAL, INDIRECT, INCIDENTAL, OR CONSEQUENTIAL DAMAGES, OR ANY
+ * DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS,
+ * WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS
+ * ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE
+ * OF THESE LICENSED DELIVERABLES.
+ *
+ * U.S. Government End Users.  These Licensed Deliverables are a
+ * "commercial item" as that term is defined at 48 C.F.R. 2.101 (OCT
+ * 1995), consisting of "commercial computer software" and "commercial
+ * computer software documentation" as such terms are used in 48
+ * C.F.R. 12.212 (SEPT 1995) and is provided to the U.S. Government
+ * only as a commercial end item.  Consistent with 48 C.F.R.12.212 and
+ * 48 C.F.R. 227.7202-1 through 227.7202-4 (JUNE 1995), all
+ * U.S. Government End Users acquire the Licensed Deliverables with
+ * only those rights set forth herein.
+ *
+ * Any use of the Licensed Deliverables in individual and commercial
+ * software must include, in the user documentation and internal
+ * comments to the code, the above Disclaimer and U.S. Government End
+ * Users Notice.
+ */
+
+#if !defined(_CUPTI_H_)
+#define _CUPTI_H_
+
+#ifdef _WIN32
+#ifndef WIN32_LEAN_AND_MEAN
+#define WIN32_LEAN_AND_MEAN
+#endif
+#ifdef NOMINMAX
+#include <windows.h>
+#else
+#define NOMINMAX
+#include <windows.h>
+#undef NOMINMAX
+#endif
+#endif
+
+#include <cuda.h>
+#include <cupti_result.h>
+#include <cupti_version.h>
+
+/* Activity, callback, event and metric APIs */
+#include <cupti_activity.h>
+#include <cupti_callbacks.h>
+#include <cupti_events.h>
+#include <cupti_metrics.h>
+
+/* Runtime, driver, and nvtx function identifiers */
+#include <cupti_driver_cbid.h>
+#include <cupti_runtime_cbid.h>
+#include <cupti_nvtx_cbid.h>
+
+/* To support function parameter structures for obsoleted API. See
+   cuda.h for the actual definition of these structures. */
+typedef unsigned int CUdeviceptr_v1;
+typedef struct CUDA_MEMCPY2D_v1_st { int dummy; } CUDA_MEMCPY2D_v1;
+typedef struct CUDA_MEMCPY3D_v1_st { int dummy; } CUDA_MEMCPY3D_v1;
+typedef struct CUDA_ARRAY_DESCRIPTOR_v1_st { int dummy; } CUDA_ARRAY_DESCRIPTOR_v1;
+typedef struct CUDA_ARRAY3D_DESCRIPTOR_v1_st { int dummy; } CUDA_ARRAY3D_DESCRIPTOR_v1;
+
+/* Function parameter structures */
+#include <generated_cuda_runtime_api_meta.h>
+#include <generated_cuda_meta.h>
+
+/* The following parameter structures cannot be included unless a
+   header that defines GL_VERSION is included before including them.
+   If these are needed then make sure such a header is included
+   already. */
+#ifdef GL_VERSION
+#include <generated_cuda_gl_interop_meta.h>
+#include <generated_cudaGL_meta.h>
+#endif
+
+//#include <generated_nvtx_meta.h>
+
+/* The following parameter structures cannot be included by default as
+   they are not guaranteed to be available on all systems. Uncomment
+   the includes that are available, or use the include explicitly. */
+#if defined(__linux__)
+//#include <generated_cuda_vdpau_interop_meta.h>
+//#include <generated_cudaVDPAU_meta.h>
+#endif
+
+#ifdef _WIN32
+//#include <generated_cuda_d3d9_interop_meta.h>
+//#include <generated_cuda_d3d10_interop_meta.h>
+//#include <generated_cuda_d3d11_interop_meta.h>
+//#include <generated_cudaD3D9_meta.h>
+//#include <generated_cudaD3D10_meta.h>
+//#include <generated_cudaD3D11_meta.h>
+#endif
+
+#endif /*_CUPTI_H_*/
+
+

Prism/LLaDA/LLaDA_Prism/.venv/lib/python3.12/site-packages/nvidia/cuda_cupti/include/cupti_callbacks.h

@@ -0,0 +1,863 @@
+/*
+ * Copyright 2010-2023 NVIDIA Corporation.  All rights reserved.
+ *
+ * NOTICE TO LICENSEE:
+ *
+ * This source code and/or documentation ("Licensed Deliverables") are
+ * subject to NVIDIA intellectual property rights under U.S. and
+ * international Copyright laws.
+ *
+ * These Licensed Deliverables contained herein is PROPRIETARY and
+ * CONFIDENTIAL to NVIDIA and is being provided under the terms and
+ * conditions of a form of NVIDIA software license agreement by and
+ * between NVIDIA and Licensee ("License Agreement") or electronically
+ * accepted by Licensee.  Notwithstanding any terms or conditions to
+ * the contrary in the License Agreement, reproduction or disclosure
+ * of the Licensed Deliverables to any third party without the express
+ * written consent of NVIDIA is prohibited.
+ *
+ * NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE
+ * LICENSE AGREEMENT, NVIDIA MAKES NO REPRESENTATION ABOUT THE
+ * SUITABILITY OF THESE LICENSED DELIVERABLES FOR ANY PURPOSE.  IT IS
+ * PROVIDED "AS IS" WITHOUT EXPRESS OR IMPLIED WARRANTY OF ANY KIND.
+ * NVIDIA DISCLAIMS ALL WARRANTIES WITH REGARD TO THESE LICENSED
+ * DELIVERABLES, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY,
+ * NONINFRINGEMENT, AND FITNESS FOR A PARTICULAR PURPOSE.
+ * NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE
+ * LICENSE AGREEMENT, IN NO EVENT SHALL NVIDIA BE LIABLE FOR ANY
+ * SPECIAL, INDIRECT, INCIDENTAL, OR CONSEQUENTIAL DAMAGES, OR ANY
+ * DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS,
+ * WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS
+ * ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE
+ * OF THESE LICENSED DELIVERABLES.
+ *
+ * U.S. Government End Users.  These Licensed Deliverables are a
+ * "commercial item" as that term is defined at 48 C.F.R. 2.101 (OCT
+ * 1995), consisting of "commercial computer software" and "commercial
+ * computer software documentation" as such terms are used in 48
+ * C.F.R. 12.212 (SEPT 1995) and is provided to the U.S. Government
+ * only as a commercial end item.  Consistent with 48 C.F.R.12.212 and
+ * 48 C.F.R. 227.7202-1 through 227.7202-4 (JUNE 1995), all
+ * U.S. Government End Users acquire the Licensed Deliverables with
+ * only those rights set forth herein.
+ *
+ * Any use of the Licensed Deliverables in individual and commercial
+ * software must include, in the user documentation and internal
+ * comments to the code, the above Disclaimer and U.S. Government End
+ * Users Notice.
+ */
+
+#if !defined(__CUPTI_CALLBACKS_H__)
+#define __CUPTI_CALLBACKS_H__
+
+#include <cuda.h>
+#include <builtin_types.h>
+#include <string.h>
+#include <cuda_stdint.h>
+#include <cupti_result.h>
+
+#ifndef CUPTIAPI
+#ifdef _WIN32
+#define CUPTIAPI __stdcall
+#else
+#define CUPTIAPI
+#endif
+#endif
+
+#if defined(__cplusplus)
+extern "C" {
+#endif
+
+#if defined(__GNUC__) && defined(CUPTI_LIB)
+    #pragma GCC visibility push(default)
+#endif
+
+/**
+ * \defgroup CUPTI_CALLBACK_API CUPTI Callback API
+ * Functions, types, and enums that implement the CUPTI Callback API.
+ * @{
+ */
+
+/**
+ * \brief Specifies the point in an API call that a callback is issued.
+ *
+ * Specifies the point in an API call that a callback is issued. This
+ * value is communicated to the callback function via \ref
+ * CUpti_CallbackData::callbackSite.
+ */
+typedef enum {
+  /**
+   * The callback is at the entry of the API call.
+   */
+  CUPTI_API_ENTER                 = 0,
+  /**
+   * The callback is at the exit of the API call.
+   */
+  CUPTI_API_EXIT                  = 1,
+  CUPTI_API_CBSITE_FORCE_INT     = 0x7fffffff
+} CUpti_ApiCallbackSite;
+
+/**
+ * \brief Callback domains.
+ *
+ * Callback domains. Each domain represents callback points for a
+ * group of related API functions or CUDA driver activity.
+ */
+typedef enum {
+  /**
+   * Invalid domain.
+   */
+  CUPTI_CB_DOMAIN_INVALID           = 0,
+  /**
+   * Domain containing callback points for all driver API functions.
+   */
+  CUPTI_CB_DOMAIN_DRIVER_API        = 1,
+  /**
+   * Domain containing callback points for all runtime API
+   * functions.
+   */
+  CUPTI_CB_DOMAIN_RUNTIME_API       = 2,
+  /**
+   * Domain containing callback points for CUDA resource tracking.
+   */
+  CUPTI_CB_DOMAIN_RESOURCE          = 3,
+  /**
+   * Domain containing callback points for CUDA synchronization.
+   */
+  CUPTI_CB_DOMAIN_SYNCHRONIZE       = 4,
+  /**
+   * Domain containing callback points for NVTX API functions.
+   */
+  CUPTI_CB_DOMAIN_NVTX              = 5,
+  /**
+   * Domain containing callback points for various states.
+   */
+  CUPTI_CB_DOMAIN_STATE             = 6,
+
+  CUPTI_CB_DOMAIN_SIZE,
+
+  CUPTI_CB_DOMAIN_FORCE_INT         = 0x7fffffff
+} CUpti_CallbackDomain;
+
+/**
+ * \brief Callback IDs for resource domain.
+ *
+ * Callback IDs for resource domain, CUPTI_CB_DOMAIN_RESOURCE.  This
+ * value is communicated to the callback function via the \p cbid
+ * parameter.
+ */
+typedef enum {
+  /**
+   * Invalid resource callback ID.
+   */
+  CUPTI_CBID_RESOURCE_INVALID                               = 0,
+  /**
+   * A new context has been created.
+   */
+  CUPTI_CBID_RESOURCE_CONTEXT_CREATED                       = 1,
+  /**
+   * A context is about to be destroyed.
+   */
+  CUPTI_CBID_RESOURCE_CONTEXT_DESTROY_STARTING              = 2,
+  /**
+   * A new stream has been created.
+   */
+  CUPTI_CBID_RESOURCE_STREAM_CREATED                        = 3,
+  /**
+   * A stream is about to be destroyed.
+   */
+  CUPTI_CBID_RESOURCE_STREAM_DESTROY_STARTING               = 4,
+  /**
+   * The driver has finished initializing.
+   */
+  CUPTI_CBID_RESOURCE_CU_INIT_FINISHED                      = 5,
+  /**
+   * A module has been loaded.
+   */
+  CUPTI_CBID_RESOURCE_MODULE_LOADED                         = 6,
+  /**
+   * A module is about to be unloaded.
+   */
+  CUPTI_CBID_RESOURCE_MODULE_UNLOAD_STARTING                = 7,
+  /**
+   * The current module which is being profiled.
+   */
+  CUPTI_CBID_RESOURCE_MODULE_PROFILED                       = 8,
+  /**
+   * CUDA graph has been created.
+   */
+  CUPTI_CBID_RESOURCE_GRAPH_CREATED                         = 9,
+  /**
+   * CUDA graph is about to be destroyed.
+   */
+  CUPTI_CBID_RESOURCE_GRAPH_DESTROY_STARTING                = 10,
+  /**
+   * CUDA graph is cloned.
+   */
+  CUPTI_CBID_RESOURCE_GRAPH_CLONED                          = 11,
+  /**
+   * CUDA graph node is about to be created
+   */
+  CUPTI_CBID_RESOURCE_GRAPHNODE_CREATE_STARTING             = 12,
+  /**
+   * CUDA graph node is created.
+   */
+  CUPTI_CBID_RESOURCE_GRAPHNODE_CREATED                     = 13,
+  /**
+   * CUDA graph node is about to be destroyed.
+   */
+  CUPTI_CBID_RESOURCE_GRAPHNODE_DESTROY_STARTING            = 14,
+  /**
+   * Dependency on a CUDA graph node is created.
+   */
+  CUPTI_CBID_RESOURCE_GRAPHNODE_DEPENDENCY_CREATED          = 15,
+  /**
+   * Dependency on a CUDA graph node is destroyed.
+   */
+  CUPTI_CBID_RESOURCE_GRAPHNODE_DEPENDENCY_DESTROY_STARTING = 16,
+  /**
+   * An executable CUDA graph is about to be created.
+   */
+  CUPTI_CBID_RESOURCE_GRAPHEXEC_CREATE_STARTING             = 17,
+  /**
+   * An executable CUDA graph is created.
+   */
+  CUPTI_CBID_RESOURCE_GRAPHEXEC_CREATED                     = 18,
+  /**
+   * An executable CUDA graph is about to be destroyed.
+   */
+  CUPTI_CBID_RESOURCE_GRAPHEXEC_DESTROY_STARTING            = 19,
+  /**
+   * CUDA graph node is cloned.
+   */
+  CUPTI_CBID_RESOURCE_GRAPHNODE_CLONED                      = 20,
+  /**
+   * CUDA stream attribute is changed.
+   */
+  CUPTI_CBID_RESOURCE_STREAM_ATTRIBUTE_CHANGED              = 21,
+
+  CUPTI_CBID_RESOURCE_SIZE,
+  CUPTI_CBID_RESOURCE_FORCE_INT                   = 0x7fffffff
+} CUpti_CallbackIdResource;
+
+/**
+ * \brief Callback IDs for synchronization domain.
+ *
+ * Callback IDs for synchronization domain,
+ * CUPTI_CB_DOMAIN_SYNCHRONIZE.  This value is communicated to the
+ * callback function via the \p cbid parameter.
+ */
+typedef enum {
+  /**
+   * Invalid synchronize callback ID.
+   */
+  CUPTI_CBID_SYNCHRONIZE_INVALID                  = 0,
+  /**
+   * Stream synchronization has completed for the stream.
+   */
+  CUPTI_CBID_SYNCHRONIZE_STREAM_SYNCHRONIZED      = 1,
+  /**
+   * Context synchronization has completed for the context.
+   */
+  CUPTI_CBID_SYNCHRONIZE_CONTEXT_SYNCHRONIZED     = 2,
+  CUPTI_CBID_SYNCHRONIZE_SIZE,
+  CUPTI_CBID_SYNCHRONIZE_FORCE_INT                = 0x7fffffff
+} CUpti_CallbackIdSync;
+
+
+/**
+ * \brief Callback IDs for state domain.
+ *
+ * Callback IDs for state domain,
+ * CUPTI_CB_DOMAIN_STATE. This value is communicated to the
+ * callback function via the \p cbid parameter.
+ */
+typedef enum {
+  /**
+   * Invalid state callback ID.
+   */
+  CUPTI_CBID_STATE_INVALID                        = 0,
+  /**
+   * Notification of fatal errors - high impact, non-recoverable
+   * When encountered, CUPTI automatically invokes cuptiFinalize()
+   * User can control behavior of the application in future from 
+   * receiving this callback - such as continuing without profiling, or
+   * terminating the whole application.
+   */
+  CUPTI_CBID_STATE_FATAL_ERROR                    = 1,
+  /**
+   * Notification of non fatal errors - high impact, but recoverable
+   * This notification is not issued in the current release.
+   */
+  CUPTI_CBID_STATE_ERROR                          = 2,
+  /**
+   * Notification of warnings - low impact, recoverable.
+   */
+  CUPTI_CBID_STATE_WARNING                        = 3,
+
+  CUPTI_CBID_STATE_SIZE,
+  CUPTI_CBID_STATE_FORCE_INT         = 0x7fffffff
+} CUpti_CallbackIdState;
+
+
+/**
+ * \brief Data passed into a runtime or driver API callback function.
+ *
+ * Data passed into a runtime or driver API callback function as the
+ * \p cbdata argument to \ref CUpti_CallbackFunc. The \p cbdata will
+ * be this type for \p domain equal to CUPTI_CB_DOMAIN_DRIVER_API or
+ * CUPTI_CB_DOMAIN_RUNTIME_API. The callback data is valid only within
+ * the invocation of the callback function that is passed the data. If
+ * you need to retain some data for use outside of the callback, you
+ * must make a copy of that data. For example, if you make a shallow
+ * copy of CUpti_CallbackData within a callback, you cannot
+ * dereference \p functionParams outside of that callback to access
+ * the function parameters. \p functionName is an exception: the
+ * string pointed to by \p functionName is a global constant and so
+ * may be accessed outside of the callback.
+ */
+typedef struct {
+  /**
+   * Point in the runtime or driver function from where the callback
+   * was issued.
+   */
+  CUpti_ApiCallbackSite callbackSite;
+
+  /**
+   * Name of the runtime or driver API function which issued the
+   * callback. This string is a global constant and so may be
+   * accessed outside of the callback.
+   */
+  const char *functionName;
+
+  /**
+   * Pointer to the arguments passed to the runtime or driver API
+   * call. See generated_cuda_runtime_api_meta.h and
+   * generated_cuda_meta.h for structure definitions for the
+   * parameters for each runtime and driver API function.
+   */
+  const void *functionParams;
+
+  /**
+   * Pointer to the return value of the runtime or driver API
+   * call. This field is only valid within the exit::CUPTI_API_EXIT
+   * callback. For a runtime API \p functionReturnValue points to a
+   * \p cudaError_t. For a driver API \p functionReturnValue points
+   * to a \p CUresult.
+   */
+  void *functionReturnValue;
+
+  /**
+   * Name of the symbol operated on by the runtime or driver API
+   * function which issued the callback. This entry is valid only for
+   * driver and runtime launch callbacks, where it returns the name of
+   * the kernel.
+   */
+  const char *symbolName;
+
+  /**
+   * Driver context current to the thread, or null if no context is
+   * current. This value can change from the entry to exit callback
+   * of a runtime API function if the runtime initializes a context.
+   */
+  CUcontext context;
+
+  /**
+   * Unique ID for the CUDA context associated with the thread. The
+   * UIDs are assigned sequentially as contexts are created and are
+   * unique within a process.
+   */
+  uint32_t contextUid;
+
+  /**
+   * Pointer to data shared between the entry and exit callbacks of
+   * a given runtime or drive API function invocation. This field
+   * can be used to pass 64-bit values from the entry callback to
+   * the corresponding exit callback.
+   */
+  uint64_t *correlationData;
+
+  /**
+   * The activity record correlation ID for this callback. For a
+   * driver domain callback (i.e. \p domain
+   * CUPTI_CB_DOMAIN_DRIVER_API) this ID will equal the correlation ID
+   * in the CUpti_ActivityAPI record corresponding to the CUDA driver
+   * function call. For a runtime domain callback (i.e. \p domain
+   * CUPTI_CB_DOMAIN_RUNTIME_API) this ID will equal the correlation
+   * ID in the CUpti_ActivityAPI record corresponding to the CUDA
+   * runtime function call. Within the callback, this ID can be
+   * recorded to correlate user data with the activity record. This
+   * field is new in 4.1.
+   */
+  uint32_t correlationId;
+
+} CUpti_CallbackData;
+
+/**
+ * \brief Data passed into a resource callback function.
+ *
+ * Data passed into a resource callback function as the \p cbdata
+ * argument to \ref CUpti_CallbackFunc. The \p cbdata will be this
+ * type for \p domain equal to CUPTI_CB_DOMAIN_RESOURCE. The callback
+ * data is valid only within the invocation of the callback function
+ * that is passed the data. If you need to retain some data for use
+ * outside of the callback, you must make a copy of that data.
+ */
+typedef struct {
+  /**
+   * For CUPTI_CBID_RESOURCE_CONTEXT_CREATED and
+   * CUPTI_CBID_RESOURCE_CONTEXT_DESTROY_STARTING, the context being
+   * created or destroyed. For CUPTI_CBID_RESOURCE_STREAM_CREATED and
+   * CUPTI_CBID_RESOURCE_STREAM_DESTROY_STARTING, the context
+   * containing the stream being created or destroyed.
+   */
+  CUcontext context;
+
+  union {
+    /**
+     * For CUPTI_CBID_RESOURCE_STREAM_CREATED and
+     * CUPTI_CBID_RESOURCE_STREAM_DESTROY_STARTING, the stream being
+     * created or destroyed.
+     */
+    CUstream stream;
+  } resourceHandle;
+
+  /**
+   * Reserved for future use.
+   */
+  void *resourceDescriptor;
+} CUpti_ResourceData;
+
+
+/**
+ * \brief Module data passed into a resource callback function.
+ *
+ * CUDA module data passed into a resource callback function as the \p cbdata
+ * argument to \ref CUpti_CallbackFunc. The \p cbdata will be this
+ * type for \p domain equal to CUPTI_CB_DOMAIN_RESOURCE. The module
+ * data is valid only within the invocation of the callback function
+ * that is passed the data. If you need to retain some data for use
+ * outside of the callback, you must make a copy of that data.
+ */
+
+typedef struct {
+  /**
+   * Identifier to associate with the CUDA module.
+   */
+    uint32_t moduleId;
+
+  /**
+   * The size of the cubin.
+   */
+    size_t cubinSize;
+
+  /**
+   * Pointer to the associated cubin.
+   */
+    const char *pCubin;
+} CUpti_ModuleResourceData;
+
+/**
+ * \brief CUDA graphs data passed into a resource callback function.
+ *
+ * CUDA graphs data passed into a resource callback function as the \p cbdata
+ * argument to \ref CUpti_CallbackFunc. The \p cbdata will be this
+ * type for \p domain equal to CUPTI_CB_DOMAIN_RESOURCE. The graph
+ * data is valid only within the invocation of the callback function
+ * that is passed the data. If you need to retain some data for use
+ * outside of the callback, you must make a copy of that data.
+ */
+
+typedef struct {
+  /**
+   * CUDA graph
+   */
+    CUgraph graph;
+  /**
+   * The original CUDA graph from which \param graph is cloned
+   */
+    CUgraph originalGraph;
+  /**
+   * CUDA graph node
+   */
+    CUgraphNode node;
+  /**
+   * The original CUDA graph node from which \param node is cloned
+   */
+    CUgraphNode originalNode;
+  /**
+   * Type of the \param node
+   */
+    CUgraphNodeType nodeType;
+  /**
+   * The dependent graph node
+   * The size of the array is \param numDependencies.
+   */
+    CUgraphNode dependency;
+  /**
+   * CUDA executable graph
+   */
+    CUgraphExec graphExec;
+} CUpti_GraphData;
+
+/**
+ * \brief Data passed into a synchronize callback function.
+ *
+ * Data passed into a synchronize callback function as the \p cbdata
+ * argument to \ref CUpti_CallbackFunc. The \p cbdata will be this
+ * type for \p domain equal to CUPTI_CB_DOMAIN_SYNCHRONIZE. The
+ * callback data is valid only within the invocation of the callback
+ * function that is passed the data. If you need to retain some data
+ * for use outside of the callback, you must make a copy of that data.
+ */
+typedef struct {
+  /**
+   * The context of the stream being synchronized.
+   */
+  CUcontext context;
+  /**
+   * The stream being synchronized.
+   */
+  CUstream  stream;
+} CUpti_SynchronizeData;
+
+/**
+ * \brief Data passed into a NVTX callback function.
+ *
+ * Data passed into a NVTX callback function as the \p cbdata argument
+ * to \ref CUpti_CallbackFunc. The \p cbdata will be this type for \p
+ * domain equal to CUPTI_CB_DOMAIN_NVTX. Unless otherwise notes, the
+ * callback data is valid only within the invocation of the callback
+ * function that is passed the data. If you need to retain some data
+ * for use outside of the callback, you must make a copy of that data.
+ */
+typedef struct {
+  /**
+   * Name of the NVTX API function which issued the callback. This
+   * string is a global constant and so may be accessed outside of the
+   * callback.
+   */
+  const char *functionName;
+
+  /**
+   * Pointer to the arguments passed to the NVTX API call. See
+   * generated_nvtx_meta.h for structure definitions for the
+   * parameters for each NVTX API function.
+   */
+  const void *functionParams;
+
+  /**
+   * Pointer to the return value of the NVTX API call. See
+   * nvToolsExt.h for each NVTX API function's return value.
+   */
+  const void *functionReturnValue;
+} CUpti_NvtxData;
+
+/**
+ * \brief Stream attribute data passed into a resource callback function
+ * for CUPTI_CBID_RESOURCE_STREAM_ATTRIBUTE_CHANGED callback
+
+ * Data passed into a resource callback function as the \p cbdata
+ * argument to \ref CUpti_CallbackFunc. The \p cbdata will be this
+ * type for \p domain equal to CUPTI_CB_DOMAIN_RESOURCE. The
+ * stream attribute data is valid only within the invocation of the callback
+ * function that is passed the data. If you need to retain some data
+ * for use outside of the callback, you must make a copy of that data.
+ */
+typedef struct {
+  /**
+   * The CUDA stream handle for the attribute
+   */
+  CUstream stream;
+
+  /**
+   * The type of the CUDA stream attribute
+   */
+  CUstreamAttrID attr;
+
+  /**
+   * The value of the CUDA stream attribute
+   */
+  const CUstreamAttrValue *value;
+} CUpti_StreamAttrData;
+
+/**
+ * \brief Data passed into a State callback function.
+ *
+ * Data passed into a State callback function as the \p cbdata argument
+ * to \ref CUpti_CallbackFunc. The \p cbdata will be this type for \p
+ * domain equal to CUPTI_CB_DOMAIN_STATE and callback Ids belonging to CUpti_CallbackIdState. 
+ * Unless otherwise noted, the callback data is valid only within the invocation of the callback
+ * function that is passed the data. If you need to retain some data
+ * for use outside of the callback, you must make a copy of that data.
+ */
+typedef struct {
+  union {
+    /**
+     * Data passed along with the callback Ids 
+     * Enum CUpti_CallbackIdState used to denote callback ids
+     */
+    struct {
+      /**
+       * Error code
+       */
+      CUptiResult result;
+      /**
+       * String containing more details. It can be NULL.
+       */
+      const char *message;
+    } notification;
+  };
+} CUpti_StateData;
+
+/**
+ * \brief An ID for a driver API, runtime API, resource or
+ * synchronization callback.
+ *
+ * An ID for a driver API, runtime API, resource or synchronization
+ * callback. Within a driver API callback this should be interpreted
+ * as a CUpti_driver_api_trace_cbid value (these values are defined in
+ * cupti_driver_cbid.h). Within a runtime API callback this should be
+ * interpreted as a CUpti_runtime_api_trace_cbid value (these values
+ * are defined in cupti_runtime_cbid.h). Within a resource API
+ * callback this should be interpreted as a \ref
+ * CUpti_CallbackIdResource value. Within a synchronize API callback
+ * this should be interpreted as a \ref CUpti_CallbackIdSync value.
+ */
+typedef uint32_t CUpti_CallbackId;
+
+/**
+ * \brief Function type for a callback.
+ *
+ * Function type for a callback. The type of the data passed to the
+ * callback in \p cbdata depends on the \p domain. If \p domain is
+ * CUPTI_CB_DOMAIN_DRIVER_API or CUPTI_CB_DOMAIN_RUNTIME_API the type
+ * of \p cbdata will be CUpti_CallbackData. If \p domain is
+ * CUPTI_CB_DOMAIN_RESOURCE the type of \p cbdata will be
+ * CUpti_ResourceData. If \p domain is CUPTI_CB_DOMAIN_SYNCHRONIZE the
+ * type of \p cbdata will be CUpti_SynchronizeData. If \p domain is
+ * CUPTI_CB_DOMAIN_NVTX the type of \p cbdata will be CUpti_NvtxData.
+ *
+ * \param userdata User data supplied at subscription of the callback
+ * \param domain The domain of the callback
+ * \param cbid The ID of the callback
+ * \param cbdata Data passed to the callback.
+ */
+typedef void (CUPTIAPI *CUpti_CallbackFunc)(
+    void *userdata,
+    CUpti_CallbackDomain domain,
+    CUpti_CallbackId cbid,
+    const void *cbdata);
+
+/**
+ * \brief A callback subscriber.
+ */
+typedef struct CUpti_Subscriber_st *CUpti_SubscriberHandle;
+
+/**
+ * \brief Pointer to an array of callback domains.
+ */
+typedef CUpti_CallbackDomain *CUpti_DomainTable;
+
+/**
+ * \brief Get the available callback domains.
+ *
+ * Returns in \p *domainTable an array of size \p *domainCount of all
+ * the available callback domains.
+ * \note \b Thread-safety: this function is thread safe.
+ *
+ * \param domainCount Returns number of callback domains
+ * \param domainTable Returns pointer to array of available callback domains
+ *
+ * \retval CUPTI_SUCCESS on success
+ * \retval CUPTI_ERROR_NOT_INITIALIZED if unable to initialize CUPTI
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p domainCount or \p domainTable are NULL
+ */
+CUptiResult CUPTIAPI cuptiSupportedDomains(size_t *domainCount,
+                                           CUpti_DomainTable *domainTable);
+
+/**
+ * \brief Initialize a callback subscriber with a callback function
+ * and user data.
+ *
+ * Initializes a callback subscriber with a callback function and
+ * (optionally) a pointer to user data. The returned subscriber handle
+ * can be used to enable and disable the callback for specific domains
+ * and callback IDs.
+ * \note Only a single subscriber can be registered at a time. To ensure
+ * that no other CUPTI client interrupts the profiling session, it's the
+ * responsibility of all the CUPTI clients to call this function before
+ * starting the profling session. In case profiling session is already
+ * started by another CUPTI client, this function returns the error code
+ * CUPTI_ERROR_MULTIPLE_SUBSCRIBERS_NOT_SUPPORTED.
+ * Note that this function returns the same error when application is
+ * launched using NVIDIA tools like nvprof, Visual Profiler, Nsight Systems,
+ * Nsight Compute, cuda-gdb and cuda-memcheck.
+ * \note This function does not enable any callbacks.
+ * \note \b Thread-safety: this function is thread safe.
+ *
+ * \param subscriber Returns handle to initialize subscriber
+ * \param callback The callback function
+ * \param userdata A pointer to user data. This data will be passed to
+ * the callback function via the \p userdata parameter.
+ *
+ * \retval CUPTI_SUCCESS on success
+ * \retval CUPTI_ERROR_NOT_INITIALIZED if unable to initialize CUPTI
+ * \retval CUPTI_ERROR_MULTIPLE_SUBSCRIBERS_NOT_SUPPORTED if there is already a CUPTI subscriber
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p subscriber is NULL
+ */
+CUptiResult CUPTIAPI cuptiSubscribe(CUpti_SubscriberHandle *subscriber,
+                                    CUpti_CallbackFunc callback,
+                                    void *userdata);
+
+/**
+ * \brief Unregister a callback subscriber.
+ *
+ * Removes a callback subscriber so that no future callbacks will be
+ * issued to that subscriber.
+ * \note \b Thread-safety: this function is thread safe.
+ *
+ * \param subscriber Handle to the initialize subscriber
+ *
+ * \retval CUPTI_SUCCESS on success
+ * \retval CUPTI_ERROR_NOT_INITIALIZED if unable to initialized CUPTI
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p subscriber is NULL or not initialized
+ */
+CUptiResult CUPTIAPI cuptiUnsubscribe(CUpti_SubscriberHandle subscriber);
+
+/**
+ * \brief Get the current enabled/disabled state of a callback for a specific
+ * domain and function ID.
+ *
+ * Returns non-zero in \p *enable if the callback for a domain and
+ * callback ID is enabled, and zero if not enabled.
+ *
+ * \note \b Thread-safety: a subscriber must serialize access to
+ * cuptiGetCallbackState, cuptiEnableCallback, cuptiEnableDomain, and
+ * cuptiEnableAllDomains. For example, if cuptiGetCallbackState(sub,
+ * d, c) and cuptiEnableCallback(sub, d, c) are called concurrently,
+ * the results are undefined.
+ *
+ * \param enable Returns non-zero if callback enabled, zero if not enabled
+ * \param subscriber Handle to the initialize subscriber
+ * \param domain The domain of the callback
+ * \param cbid The ID of the callback
+ *
+ * \retval CUPTI_SUCCESS on success
+ * \retval CUPTI_ERROR_NOT_INITIALIZED if unable to initialized CUPTI
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p enabled is NULL, or if \p
+ * subscriber, \p domain or \p cbid is invalid.
+ */
+CUptiResult CUPTIAPI cuptiGetCallbackState(uint32_t *enable,
+                                           CUpti_SubscriberHandle subscriber,
+                                           CUpti_CallbackDomain domain,
+                                           CUpti_CallbackId cbid);
+
+/**
+ * \brief Enable or disabled callbacks for a specific domain and
+ * callback ID.
+ *
+ * Enable or disabled callbacks for a subscriber for a specific domain
+ * and callback ID.
+ *
+ * \note \b Thread-safety: a subscriber must serialize access to
+ * cuptiGetCallbackState, cuptiEnableCallback, cuptiEnableDomain, and
+ * cuptiEnableAllDomains. For example, if cuptiGetCallbackState(sub,
+ * d, c) and cuptiEnableCallback(sub, d, c) are called concurrently,
+ * the results are undefined.
+ *
+ * \param enable New enable state for the callback. Zero disables the
+ * callback, non-zero enables the callback.
+ * \param subscriber - Handle to callback subscription
+ * \param domain The domain of the callback
+ * \param cbid The ID of the callback
+ *
+ * \retval CUPTI_SUCCESS on success
+ * \retval CUPTI_ERROR_NOT_INITIALIZED if unable to initialized CUPTI
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p subscriber, \p domain or \p
+ * cbid is invalid.
+ */
+CUptiResult CUPTIAPI cuptiEnableCallback(uint32_t enable,
+                                         CUpti_SubscriberHandle subscriber,
+                                         CUpti_CallbackDomain domain,
+                                         CUpti_CallbackId cbid);
+
+/**
+ * \brief Enable or disabled all callbacks for a specific domain.
+ *
+ * Enable or disabled all callbacks for a specific domain.
+ *
+ * \note \b Thread-safety: a subscriber must serialize access to
+ * cuptiGetCallbackState, cuptiEnableCallback, cuptiEnableDomain, and
+ * cuptiEnableAllDomains. For example, if cuptiGetCallbackEnabled(sub,
+ * d, *) and cuptiEnableDomain(sub, d) are called concurrently, the
+ * results are undefined.
+ *
+ * \param enable New enable state for all callbacks in the
+ * domain. Zero disables all callbacks, non-zero enables all
+ * callbacks.
+ * \param subscriber - Handle to callback subscription
+ * \param domain The domain of the callback
+ *
+ * \retval CUPTI_SUCCESS on success
+ * \retval CUPTI_ERROR_NOT_INITIALIZED if unable to initialized CUPTI
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p subscriber or \p domain is invalid
+ */
+CUptiResult CUPTIAPI cuptiEnableDomain(uint32_t enable,
+                                       CUpti_SubscriberHandle subscriber,
+                                       CUpti_CallbackDomain domain);
+
+/**
+ * \brief Enable or disable all callbacks in all domains.
+ *
+ * Enable or disable all callbacks in all domains.
+ *
+ * \note \b Thread-safety: a subscriber must serialize access to
+ * cuptiGetCallbackState, cuptiEnableCallback, cuptiEnableDomain, and
+ * cuptiEnableAllDomains. For example, if cuptiGetCallbackState(sub,
+ * d, *) and cuptiEnableAllDomains(sub) are called concurrently, the
+ * results are undefined.
+ *
+ * \param enable New enable state for all callbacks in all
+ * domain. Zero disables all callbacks, non-zero enables all
+ * callbacks.
+ * \param subscriber - Handle to callback subscription
+ *
+ * \retval CUPTI_SUCCESS on success
+ * \retval CUPTI_ERROR_NOT_INITIALIZED if unable to initialized CUPTI
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p subscriber is invalid
+ */
+CUptiResult CUPTIAPI cuptiEnableAllDomains(uint32_t enable,
+                                           CUpti_SubscriberHandle subscriber);
+
+/**
+ * \brief Get the name of a callback for a specific domain and callback ID.
+ *
+ * Returns a pointer to the name c_string in \p **name.
+ *
+ * \note \b Names are available only for the DRIVER and RUNTIME domains.
+ *
+ * \param domain The domain of the callback
+ * \param cbid The ID of the callback
+ * \param name Returns pointer to the name string on success, NULL otherwise
+ *
+ * \retval CUPTI_SUCCESS on success
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p name is NULL, or if
+ * \p domain or \p cbid is invalid.
+ */
+CUptiResult CUPTIAPI cuptiGetCallbackName(CUpti_CallbackDomain domain,
+                                          uint32_t cbid,
+                                          const char **name);
+
+/** @} */ /* END CUPTI_CALLBACK_API */
+
+#if defined(__GNUC__) && defined(CUPTI_LIB)
+    #pragma GCC visibility pop
+#endif
+
+#if defined(__cplusplus)
+}
+#endif
+
+#endif  // file guard
+

Prism/LLaDA/LLaDA_Prism/.venv/lib/python3.12/site-packages/nvidia/cuda_cupti/include/cupti_checkpoint.h

@@ -0,0 +1,127 @@
+#pragma once
+
+#include <cuda.h>
+#include <cupti_result.h>
+
+#include <stddef.h>
+#include <stdint.h>
+
+namespace NV { namespace Cupti { namespace Checkpoint {
+
+#ifdef __cplusplus
+extern "C"
+{
+#endif
+
+/**
+ * \defgroup CUPTI_CHECKPOINT_API CUPTI Checkpoint API
+ * Functions, types, and enums that implement the CUPTI Checkpoint API.
+ * @{
+ */
+
+/**
+ * \brief Specifies optimization options for a checkpoint, may be OR'd together to specify multiple options.
+ */
+typedef enum
+{
+    CUPTI_CHECKPOINT_OPT_NONE     = 0, //!< Default behavior
+    CUPTI_CHECKPOINT_OPT_TRANSFER = 1, //!< Determine which mem blocks have changed, and only restore those. This optimization is cached, which means cuptiCheckpointRestore must always be called at the same point in the application when this option is enabled, or the result may be incorrect.
+} CUpti_CheckpointOptimizations;
+
+/**
+ * \brief Configuration and handle for a CUPTI Checkpoint
+ *
+ * A CUptiCheckpoint object should be initialized with desired options prior to passing into any
+ * CUPTI Checkpoint API function.  The first call into a Checkpoint API function will initialize internal
+ * state based on these options.  Subsequent changes to these options will not have any effect.
+ *
+ * Checkpoint data is saved in device, host, and filesystem space.  There are options to reserve memory
+ * at each level (device, host, filesystem) which are intended to allow a guarantee that a certain amount
+ * of memory will remain free for use after the checkpoint is saved.
+ * Note, however, that falling back to slower levels of memory (host, and then filesystem) to save the checkpoint
+ * will result in performance degradation.
+ * Currently, the filesystem limitation is not implemented.  Note that falling back to filesystem storage may
+ * significantly impact the performance for saving and restoring a checkpoint.
+ */
+typedef struct
+{
+   size_t structSize;      //!< [in] Must be set to CUpti_Checkpoint_STRUCT_SIZE
+
+   CUcontext ctx;          //!< [in] Set to context to save from, or will use current context if NULL
+
+   size_t reserveDeviceMB; //!< [in] Restrict checkpoint from using last N MB of device memory (-1 = use no device memory)
+   size_t reserveHostMB;   //!< [in] Restrict checkpoint from using last N MB of host memory (-1 = use no host memory)
+   uint8_t allowOverwrite; //!< [in] Boolean, Allow checkpoint to save over existing checkpoint
+   uint8_t optimizations;  //!< [in] Mask of CUpti_CheckpointOptimizations flags for this checkpoint
+
+   void * pPriv;           //!< [in] Assign to NULL
+} CUpti_Checkpoint;
+
+#define CUpti_Checkpoint_STRUCT_SIZE  \
+(offsetof(CUpti_Checkpoint, pPriv) +  \
+sizeof(((CUpti_Checkpoint*)(nullptr))->pPriv))
+
+#if defined(__GNUC__) && defined(CUPTI_LIB)
+    #pragma GCC visibility push(default)
+#endif
+
+/**
+ * \brief Initialize and save a checkpoint of the device state associated with the handle context
+ *
+ * Uses the handle options to configure and save a checkpoint of the device state associated with the specified context.
+ *
+ * \param handle A pointer to a CUpti_Checkpoint object
+ *
+ * \retval CUPTI_SUCCESS if a checkpoint was successfully initialized and saved
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p handle does not appear to refer to a valid CUpti_Checkpoint
+ * \retval CUPTI_ERROR_INVALID_CONTEXT
+ * \retval CUPTI_ERROR_INVALID_DEVICE if device associated with context is not compatible with checkpoint API
+ * \retval CUPTI_ERROR_INVALID_OPERATION if Save is requested over an existing checkpoint, but \p allowOverwrite was not originally specified
+ * \retval CUPTI_ERROR_OUT_OF_MEMORY if as configured, not enough backing storage space to save the checkpoint
+ */
+CUptiResult cuptiCheckpointSave(CUpti_Checkpoint * const handle);
+
+/**
+ * \brief Restore a checkpoint to the device associated with its context
+ *
+ * Restores device, pinned, and allocated memory to the state when the checkpoint was saved
+ *
+ * \param handle A pointer to a previously saved CUpti_Checkpoint object
+ *
+ * \retval CUTPI_SUCCESS if the checkpoint was successfully restored
+ * \retval CUPTI_ERROR_NOT_INITIALIZED if the checkpoint was not previously initialized
+ * \retval CUPTI_ERROR_INVALID_CONTEXT
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if the handle appears invalid
+ * \retval CUPTI_ERROR_UNKNOWN if the restore or optimization operation fails
+ */
+CUptiResult cuptiCheckpointRestore(CUpti_Checkpoint * const handle);
+
+/**
+ * \brief Free the backing data for a checkpoint
+ *
+ * Frees all associated device, host memory and filesystem storage used for this context.
+ * After freeing a handle, it may be re-used as if it was new - options may be re-configured and will
+ * take effect on the next call to \p cuptiCheckpointSave.
+ *
+ * \param handle A pointer to a previously saved CUpti_Checkpoint object
+ *
+ * \retval CUPTI_SUCCESS if the handle was successfully freed
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if the handle was already freed or appears invalid
+ * \retval CUPTI_ERROR_INVALID_CONTEXT if the context is no longer valid
+ */
+CUptiResult cuptiCheckpointFree(CUpti_Checkpoint * const handle);
+
+#if defined(__GNUC__) && defined(CUPTI_LIB)
+    #pragma GCC visibility pop
+#endif
+
+/**
+ * @}
+ */
+
+#ifdef __cplusplus
+}
+#endif
+
+// Exit namespace NV::Cupti::Checkpoint
+}}}

Prism/LLaDA/LLaDA_Prism/.venv/lib/python3.12/site-packages/nvidia/cuda_cupti/include/cupti_common.h

@@ -0,0 +1,93 @@
+/*
+ * Copyright 2023 NVIDIA Corporation.  All rights reserved.
+ *
+ * NOTICE TO LICENSEE:
+ *
+ * This source code and/or documentation ("Licensed Deliverables") are
+ * subject to NVIDIA intellectual property rights under U.S. and
+ * international Copyright laws.
+ *
+ * These Licensed Deliverables contained herein is PROPRIETARY and
+ * CONFIDENTIAL to NVIDIA and is being provided under the terms and
+ * conditions of a form of NVIDIA software license agreement by and
+ * between NVIDIA and Licensee ("License Agreement") or electronically
+ * accepted by Licensee.  Notwithstanding any terms or conditions to
+ * the contrary in the License Agreement, reproduction or disclosure
+ * of the Licensed Deliverables to any third party without the express
+ * written consent of NVIDIA is prohibited.
+ *
+ * NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE
+ * LICENSE AGREEMENT, NVIDIA MAKES NO REPRESENTATION ABOUT THE
+ * SUITABILITY OF THESE LICENSED DELIVERABLES FOR ANY PURPOSE.  IT IS
+ * PROVIDED "AS IS" WITHOUT EXPRESS OR IMPLIED WARRANTY OF ANY KIND.
+ * NVIDIA DISCLAIMS ALL WARRANTIES WITH REGARD TO THESE LICENSED
+ * DELIVERABLES, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY,
+ * NONINFRINGEMENT, AND FITNESS FOR A PARTICULAR PURPOSE.
+ * NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE
+ * LICENSE AGREEMENT, IN NO EVENT SHALL NVIDIA BE LIABLE FOR ANY
+ * SPECIAL, INDIRECT, INCIDENTAL, OR CONSEQUENTIAL DAMAGES, OR ANY
+ * DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS,
+ * WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS
+ * ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE
+ * OF THESE LICENSED DELIVERABLES.
+ *
+ * U.S. Government End Users.  These Licensed Deliverables are a
+ * "commercial item" as that term is defined at 48 C.F.R. 2.101 (OCT
+ * 1995), consisting of "commercial computer software" and "commercial
+ * computer software documentation" as such terms are used in 48
+ * C.F.R. 12.212 (SEPT 1995) and is provided to the U.S. Government
+ * only as a commercial end item.  Consistent with 48 C.F.R.12.212 and
+ * 48 C.F.R. 227.7202-1 through 227.7202-4 (JUNE 1995), all
+ * U.S. Government End Users acquire the Licensed Deliverables with
+ * only those rights set forth herein.
+ *
+ * Any use of the Licensed Deliverables in individual and commercial
+ * software must include, in the user documentation and internal
+ * comments to the code, the above Disclaimer and U.S. Government End
+ * Users Notice.
+ */
+ 
+#if !defined(__CUPTI_COMMON_H__)
+#define __CUPTI_COMMON_H__
+
+#ifndef CUPTIAPI
+#ifdef _WIN32
+#define CUPTIAPI __stdcall
+#else
+#define CUPTIAPI
+#endif
+#endif
+
+#ifndef CUPTIUTILAPI
+#ifdef _WIN32
+#define CUPTIUTILAPI __stdcall
+#else
+#define CUPTIUTILAPI
+#endif
+#endif
+
+#if defined(__LP64__)
+#define CUPTILP64 1
+#elif defined(_WIN64)
+#define CUPTILP64 1
+#else
+#undef CUPTILP64
+#endif
+
+#define ACTIVITY_RECORD_ALIGNMENT 8
+#if defined(_WIN32) // Windows 32- and 64-bit
+#define START_PACKED_ALIGNMENT __pragma(pack(push,1)) // exact fit - no padding
+#define PACKED_ALIGNMENT __declspec(align(ACTIVITY_RECORD_ALIGNMENT))
+#define END_PACKED_ALIGNMENT __pragma(pack(pop))
+#elif defined(__GNUC__) // GCC
+#define START_PACKED_ALIGNMENT
+#define PACKED_ALIGNMENT __attribute__ ((__packed__)) __attribute__ ((aligned (ACTIVITY_RECORD_ALIGNMENT)))
+#define END_PACKED_ALIGNMENT
+#else // all other compilers
+#define START_PACKED_ALIGNMENT
+#define PACKED_ALIGNMENT
+#define END_PACKED_ALIGNMENT
+#endif
+
+#endif /*__CUPTI_COMMON_H__*/
+

Prism/LLaDA/LLaDA_Prism/.venv/lib/python3.12/site-packages/nvidia/cuda_cupti/include/cupti_driver_cbid.h

@@ -0,0 +1,799 @@
+
+// *************************************************************************
+//      Definitions of indices for API functions, unique across entire API
+// *************************************************************************
+
+// This file is generated.  Any changes you make will be lost during the next clean build.
+// CUDA public interface, for type definitions and cu* function prototypes
+
+#if !defined(_CUPTI_DRIVER_CBID_H_)
+#define _CUPTI_DRIVER_CBID_H_
+
+typedef enum CUpti_driver_api_trace_cbid_enum {
+    CUPTI_DRIVER_TRACE_CBID_INVALID                                                        = 0,
+    CUPTI_DRIVER_TRACE_CBID_cuInit                                                         = 1,
+    CUPTI_DRIVER_TRACE_CBID_cuDriverGetVersion                                             = 2,
+    CUPTI_DRIVER_TRACE_CBID_cuDeviceGet                                                    = 3,
+    CUPTI_DRIVER_TRACE_CBID_cuDeviceGetCount                                               = 4,
+    CUPTI_DRIVER_TRACE_CBID_cuDeviceGetName                                                = 5,
+    CUPTI_DRIVER_TRACE_CBID_cuDeviceComputeCapability                                      = 6,
+    CUPTI_DRIVER_TRACE_CBID_cuDeviceTotalMem                                               = 7,
+    CUPTI_DRIVER_TRACE_CBID_cuDeviceGetProperties                                          = 8,
+    CUPTI_DRIVER_TRACE_CBID_cuDeviceGetAttribute                                           = 9,
+    CUPTI_DRIVER_TRACE_CBID_cuCtxCreate                                                    = 10,
+    CUPTI_DRIVER_TRACE_CBID_cuCtxDestroy                                                   = 11,
+    CUPTI_DRIVER_TRACE_CBID_cuCtxAttach                                                    = 12,
+    CUPTI_DRIVER_TRACE_CBID_cuCtxDetach                                                    = 13,
+    CUPTI_DRIVER_TRACE_CBID_cuCtxPushCurrent                                               = 14,
+    CUPTI_DRIVER_TRACE_CBID_cuCtxPopCurrent                                                = 15,
+    CUPTI_DRIVER_TRACE_CBID_cuCtxGetDevice                                                 = 16,
+    CUPTI_DRIVER_TRACE_CBID_cuCtxSynchronize                                               = 17,
+    CUPTI_DRIVER_TRACE_CBID_cuModuleLoad                                                   = 18,
+    CUPTI_DRIVER_TRACE_CBID_cuModuleLoadData                                               = 19,
+    CUPTI_DRIVER_TRACE_CBID_cuModuleLoadDataEx                                             = 20,
+    CUPTI_DRIVER_TRACE_CBID_cuModuleLoadFatBinary                                          = 21,
+    CUPTI_DRIVER_TRACE_CBID_cuModuleUnload                                                 = 22,
+    CUPTI_DRIVER_TRACE_CBID_cuModuleGetFunction                                            = 23,
+    CUPTI_DRIVER_TRACE_CBID_cuModuleGetGlobal                                              = 24,
+    CUPTI_DRIVER_TRACE_CBID_cu64ModuleGetGlobal                                            = 25,
+    CUPTI_DRIVER_TRACE_CBID_cuModuleGetTexRef                                              = 26,
+    CUPTI_DRIVER_TRACE_CBID_cuMemGetInfo                                                   = 27,
+    CUPTI_DRIVER_TRACE_CBID_cu64MemGetInfo                                                 = 28,
+    CUPTI_DRIVER_TRACE_CBID_cuMemAlloc                                                     = 29,
+    CUPTI_DRIVER_TRACE_CBID_cu64MemAlloc                                                   = 30,
+    CUPTI_DRIVER_TRACE_CBID_cuMemAllocPitch                                                = 31,
+    CUPTI_DRIVER_TRACE_CBID_cu64MemAllocPitch                                              = 32,
+    CUPTI_DRIVER_TRACE_CBID_cuMemFree                                                      = 33,
+    CUPTI_DRIVER_TRACE_CBID_cu64MemFree                                                    = 34,
+    CUPTI_DRIVER_TRACE_CBID_cuMemGetAddressRange                                           = 35,
+    CUPTI_DRIVER_TRACE_CBID_cu64MemGetAddressRange                                         = 36,
+    CUPTI_DRIVER_TRACE_CBID_cuMemAllocHost                                                 = 37,
+    CUPTI_DRIVER_TRACE_CBID_cuMemFreeHost                                                  = 38,
+    CUPTI_DRIVER_TRACE_CBID_cuMemHostAlloc                                                 = 39,
+    CUPTI_DRIVER_TRACE_CBID_cuMemHostGetDevicePointer                                      = 40,
+    CUPTI_DRIVER_TRACE_CBID_cu64MemHostGetDevicePointer                                    = 41,
+    CUPTI_DRIVER_TRACE_CBID_cuMemHostGetFlags                                              = 42,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyHtoD                                                   = 43,
+    CUPTI_DRIVER_TRACE_CBID_cu64MemcpyHtoD                                                 = 44,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyDtoH                                                   = 45,
+    CUPTI_DRIVER_TRACE_CBID_cu64MemcpyDtoH                                                 = 46,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyDtoD                                                   = 47,
+    CUPTI_DRIVER_TRACE_CBID_cu64MemcpyDtoD                                                 = 48,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyDtoA                                                   = 49,
+    CUPTI_DRIVER_TRACE_CBID_cu64MemcpyDtoA                                                 = 50,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyAtoD                                                   = 51,
+    CUPTI_DRIVER_TRACE_CBID_cu64MemcpyAtoD                                                 = 52,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyHtoA                                                   = 53,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyAtoH                                                   = 54,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyAtoA                                                   = 55,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpy2D                                                     = 56,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpy2DUnaligned                                            = 57,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpy3D                                                     = 58,
+    CUPTI_DRIVER_TRACE_CBID_cu64Memcpy3D                                                   = 59,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyHtoDAsync                                              = 60,
+    CUPTI_DRIVER_TRACE_CBID_cu64MemcpyHtoDAsync                                            = 61,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyDtoHAsync                                              = 62,
+    CUPTI_DRIVER_TRACE_CBID_cu64MemcpyDtoHAsync                                            = 63,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyDtoDAsync                                              = 64,
+    CUPTI_DRIVER_TRACE_CBID_cu64MemcpyDtoDAsync                                            = 65,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyHtoAAsync                                              = 66,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyAtoHAsync                                              = 67,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpy2DAsync                                                = 68,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpy3DAsync                                                = 69,
+    CUPTI_DRIVER_TRACE_CBID_cu64Memcpy3DAsync                                              = 70,
+    CUPTI_DRIVER_TRACE_CBID_cuMemsetD8                                                     = 71,
+    CUPTI_DRIVER_TRACE_CBID_cu64MemsetD8                                                   = 72,
+    CUPTI_DRIVER_TRACE_CBID_cuMemsetD16                                                    = 73,
+    CUPTI_DRIVER_TRACE_CBID_cu64MemsetD16                                                  = 74,
+    CUPTI_DRIVER_TRACE_CBID_cuMemsetD32                                                    = 75,
+    CUPTI_DRIVER_TRACE_CBID_cu64MemsetD32                                                  = 76,
+    CUPTI_DRIVER_TRACE_CBID_cuMemsetD2D8                                                   = 77,
+    CUPTI_DRIVER_TRACE_CBID_cu64MemsetD2D8                                                 = 78,
+    CUPTI_DRIVER_TRACE_CBID_cuMemsetD2D16                                                  = 79,
+    CUPTI_DRIVER_TRACE_CBID_cu64MemsetD2D16                                                = 80,
+    CUPTI_DRIVER_TRACE_CBID_cuMemsetD2D32                                                  = 81,
+    CUPTI_DRIVER_TRACE_CBID_cu64MemsetD2D32                                                = 82,
+    CUPTI_DRIVER_TRACE_CBID_cuFuncSetBlockShape                                            = 83,
+    CUPTI_DRIVER_TRACE_CBID_cuFuncSetSharedSize                                            = 84,
+    CUPTI_DRIVER_TRACE_CBID_cuFuncGetAttribute                                             = 85,
+    CUPTI_DRIVER_TRACE_CBID_cuFuncSetCacheConfig                                           = 86,
+    CUPTI_DRIVER_TRACE_CBID_cuArrayCreate                                                  = 87,
+    CUPTI_DRIVER_TRACE_CBID_cuArrayGetDescriptor                                           = 88,
+    CUPTI_DRIVER_TRACE_CBID_cuArrayDestroy                                                 = 89,
+    CUPTI_DRIVER_TRACE_CBID_cuArray3DCreate                                                = 90,
+    CUPTI_DRIVER_TRACE_CBID_cuArray3DGetDescriptor                                         = 91,
+    CUPTI_DRIVER_TRACE_CBID_cuTexRefCreate                                                 = 92,
+    CUPTI_DRIVER_TRACE_CBID_cuTexRefDestroy                                                = 93,
+    CUPTI_DRIVER_TRACE_CBID_cuTexRefSetArray                                               = 94,
+    CUPTI_DRIVER_TRACE_CBID_cuTexRefSetAddress                                             = 95,
+    CUPTI_DRIVER_TRACE_CBID_cu64TexRefSetAddress                                           = 96,
+    CUPTI_DRIVER_TRACE_CBID_cuTexRefSetAddress2D                                           = 97,
+    CUPTI_DRIVER_TRACE_CBID_cu64TexRefSetAddress2D                                         = 98,
+    CUPTI_DRIVER_TRACE_CBID_cuTexRefSetFormat                                              = 99,
+    CUPTI_DRIVER_TRACE_CBID_cuTexRefSetAddressMode                                         = 100,
+    CUPTI_DRIVER_TRACE_CBID_cuTexRefSetFilterMode                                          = 101,
+    CUPTI_DRIVER_TRACE_CBID_cuTexRefSetFlags                                               = 102,
+    CUPTI_DRIVER_TRACE_CBID_cuTexRefGetAddress                                             = 103,
+    CUPTI_DRIVER_TRACE_CBID_cu64TexRefGetAddress                                           = 104,
+    CUPTI_DRIVER_TRACE_CBID_cuTexRefGetArray                                               = 105,
+    CUPTI_DRIVER_TRACE_CBID_cuTexRefGetAddressMode                                         = 106,
+    CUPTI_DRIVER_TRACE_CBID_cuTexRefGetFilterMode                                          = 107,
+    CUPTI_DRIVER_TRACE_CBID_cuTexRefGetFormat                                              = 108,
+    CUPTI_DRIVER_TRACE_CBID_cuTexRefGetFlags                                               = 109,
+    CUPTI_DRIVER_TRACE_CBID_cuParamSetSize                                                 = 110,
+    CUPTI_DRIVER_TRACE_CBID_cuParamSeti                                                    = 111,
+    CUPTI_DRIVER_TRACE_CBID_cuParamSetf                                                    = 112,
+    CUPTI_DRIVER_TRACE_CBID_cuParamSetv                                                    = 113,
+    CUPTI_DRIVER_TRACE_CBID_cuParamSetTexRef                                               = 114,
+    CUPTI_DRIVER_TRACE_CBID_cuLaunch                                                       = 115,
+    CUPTI_DRIVER_TRACE_CBID_cuLaunchGrid                                                   = 116,
+    CUPTI_DRIVER_TRACE_CBID_cuLaunchGridAsync                                              = 117,
+    CUPTI_DRIVER_TRACE_CBID_cuEventCreate                                                  = 118,
+    CUPTI_DRIVER_TRACE_CBID_cuEventRecord                                                  = 119,
+    CUPTI_DRIVER_TRACE_CBID_cuEventQuery                                                   = 120,
+    CUPTI_DRIVER_TRACE_CBID_cuEventSynchronize                                             = 121,
+    CUPTI_DRIVER_TRACE_CBID_cuEventDestroy                                                 = 122,
+    CUPTI_DRIVER_TRACE_CBID_cuEventElapsedTime                                             = 123,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamCreate                                                 = 124,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamQuery                                                  = 125,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamSynchronize                                            = 126,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamDestroy                                                = 127,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphicsUnregisterResource                                   = 128,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphicsSubResourceGetMappedArray                            = 129,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphicsResourceGetMappedPointer                             = 130,
+    CUPTI_DRIVER_TRACE_CBID_cu64GraphicsResourceGetMappedPointer                           = 131,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphicsResourceSetMapFlags                                  = 132,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphicsMapResources                                         = 133,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphicsUnmapResources                                       = 134,
+    CUPTI_DRIVER_TRACE_CBID_cuGetExportTable                                               = 135,
+    CUPTI_DRIVER_TRACE_CBID_cuCtxSetLimit                                                  = 136,
+    CUPTI_DRIVER_TRACE_CBID_cuCtxGetLimit                                                  = 137,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D10GetDevice                                               = 138,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D10CtxCreate                                               = 139,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphicsD3D10RegisterResource                                = 140,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D10RegisterResource                                        = 141,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D10UnregisterResource                                      = 142,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D10MapResources                                            = 143,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D10UnmapResources                                          = 144,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D10ResourceSetMapFlags                                     = 145,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D10ResourceGetMappedArray                                  = 146,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D10ResourceGetMappedPointer                                = 147,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D10ResourceGetMappedSize                                   = 148,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D10ResourceGetMappedPitch                                  = 149,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D10ResourceGetSurfaceDimensions                            = 150,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D11GetDevice                                               = 151,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D11CtxCreate                                               = 152,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphicsD3D11RegisterResource                                = 153,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D9GetDevice                                                = 154,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D9CtxCreate                                                = 155,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphicsD3D9RegisterResource                                 = 156,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D9GetDirect3DDevice                                        = 157,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D9RegisterResource                                         = 158,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D9UnregisterResource                                       = 159,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D9MapResources                                             = 160,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D9UnmapResources                                           = 161,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D9ResourceSetMapFlags                                      = 162,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D9ResourceGetSurfaceDimensions                             = 163,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D9ResourceGetMappedArray                                   = 164,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D9ResourceGetMappedPointer                                 = 165,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D9ResourceGetMappedSize                                    = 166,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D9ResourceGetMappedPitch                                   = 167,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D9Begin                                                    = 168,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D9End                                                      = 169,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D9RegisterVertexBuffer                                     = 170,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D9MapVertexBuffer                                          = 171,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D9UnmapVertexBuffer                                        = 172,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D9UnregisterVertexBuffer                                   = 173,
+    CUPTI_DRIVER_TRACE_CBID_cuGLCtxCreate                                                  = 174,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphicsGLRegisterBuffer                                     = 175,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphicsGLRegisterImage                                      = 176,
+    CUPTI_DRIVER_TRACE_CBID_cuWGLGetDevice                                                 = 177,
+    CUPTI_DRIVER_TRACE_CBID_cuGLInit                                                       = 178,
+    CUPTI_DRIVER_TRACE_CBID_cuGLRegisterBufferObject                                       = 179,
+    CUPTI_DRIVER_TRACE_CBID_cuGLMapBufferObject                                            = 180,
+    CUPTI_DRIVER_TRACE_CBID_cuGLUnmapBufferObject                                          = 181,
+    CUPTI_DRIVER_TRACE_CBID_cuGLUnregisterBufferObject                                     = 182,
+    CUPTI_DRIVER_TRACE_CBID_cuGLSetBufferObjectMapFlags                                    = 183,
+    CUPTI_DRIVER_TRACE_CBID_cuGLMapBufferObjectAsync                                       = 184,
+    CUPTI_DRIVER_TRACE_CBID_cuGLUnmapBufferObjectAsync                                     = 185,
+    CUPTI_DRIVER_TRACE_CBID_cuVDPAUGetDevice                                               = 186,
+    CUPTI_DRIVER_TRACE_CBID_cuVDPAUCtxCreate                                               = 187,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphicsVDPAURegisterVideoSurface                            = 188,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphicsVDPAURegisterOutputSurface                           = 189,
+    CUPTI_DRIVER_TRACE_CBID_cuModuleGetSurfRef                                             = 190,
+    CUPTI_DRIVER_TRACE_CBID_cuSurfRefCreate                                                = 191,
+    CUPTI_DRIVER_TRACE_CBID_cuSurfRefDestroy                                               = 192,
+    CUPTI_DRIVER_TRACE_CBID_cuSurfRefSetFormat                                             = 193,
+    CUPTI_DRIVER_TRACE_CBID_cuSurfRefSetArray                                              = 194,
+    CUPTI_DRIVER_TRACE_CBID_cuSurfRefGetFormat                                             = 195,
+    CUPTI_DRIVER_TRACE_CBID_cuSurfRefGetArray                                              = 196,
+    CUPTI_DRIVER_TRACE_CBID_cu64DeviceTotalMem                                             = 197,
+    CUPTI_DRIVER_TRACE_CBID_cu64D3D10ResourceGetMappedPointer                              = 198,
+    CUPTI_DRIVER_TRACE_CBID_cu64D3D10ResourceGetMappedSize                                 = 199,
+    CUPTI_DRIVER_TRACE_CBID_cu64D3D10ResourceGetMappedPitch                                = 200,
+    CUPTI_DRIVER_TRACE_CBID_cu64D3D10ResourceGetSurfaceDimensions                          = 201,
+    CUPTI_DRIVER_TRACE_CBID_cu64D3D9ResourceGetSurfaceDimensions                           = 202,
+    CUPTI_DRIVER_TRACE_CBID_cu64D3D9ResourceGetMappedPointer                               = 203,
+    CUPTI_DRIVER_TRACE_CBID_cu64D3D9ResourceGetMappedSize                                  = 204,
+    CUPTI_DRIVER_TRACE_CBID_cu64D3D9ResourceGetMappedPitch                                 = 205,
+    CUPTI_DRIVER_TRACE_CBID_cu64D3D9MapVertexBuffer                                        = 206,
+    CUPTI_DRIVER_TRACE_CBID_cu64GLMapBufferObject                                          = 207,
+    CUPTI_DRIVER_TRACE_CBID_cu64GLMapBufferObjectAsync                                     = 208,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D11GetDevices                                              = 209,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D11CtxCreateOnDevice                                       = 210,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D10GetDevices                                              = 211,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D10CtxCreateOnDevice                                       = 212,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D9GetDevices                                               = 213,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D9CtxCreateOnDevice                                        = 214,
+    CUPTI_DRIVER_TRACE_CBID_cu64MemHostAlloc                                               = 215,
+    CUPTI_DRIVER_TRACE_CBID_cuMemsetD8Async                                                = 216,
+    CUPTI_DRIVER_TRACE_CBID_cu64MemsetD8Async                                              = 217,
+    CUPTI_DRIVER_TRACE_CBID_cuMemsetD16Async                                               = 218,
+    CUPTI_DRIVER_TRACE_CBID_cu64MemsetD16Async                                             = 219,
+    CUPTI_DRIVER_TRACE_CBID_cuMemsetD32Async                                               = 220,
+    CUPTI_DRIVER_TRACE_CBID_cu64MemsetD32Async                                             = 221,
+    CUPTI_DRIVER_TRACE_CBID_cuMemsetD2D8Async                                              = 222,
+    CUPTI_DRIVER_TRACE_CBID_cu64MemsetD2D8Async                                            = 223,
+    CUPTI_DRIVER_TRACE_CBID_cuMemsetD2D16Async                                             = 224,
+    CUPTI_DRIVER_TRACE_CBID_cu64MemsetD2D16Async                                           = 225,
+    CUPTI_DRIVER_TRACE_CBID_cuMemsetD2D32Async                                             = 226,
+    CUPTI_DRIVER_TRACE_CBID_cu64MemsetD2D32Async                                           = 227,
+    CUPTI_DRIVER_TRACE_CBID_cu64ArrayCreate                                                = 228,
+    CUPTI_DRIVER_TRACE_CBID_cu64ArrayGetDescriptor                                         = 229,
+    CUPTI_DRIVER_TRACE_CBID_cu64Array3DCreate                                              = 230,
+    CUPTI_DRIVER_TRACE_CBID_cu64Array3DGetDescriptor                                       = 231,
+    CUPTI_DRIVER_TRACE_CBID_cu64Memcpy2D                                                   = 232,
+    CUPTI_DRIVER_TRACE_CBID_cu64Memcpy2DUnaligned                                          = 233,
+    CUPTI_DRIVER_TRACE_CBID_cu64Memcpy2DAsync                                              = 234,
+    CUPTI_DRIVER_TRACE_CBID_cuCtxCreate_v2                                                 = 235,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D10CtxCreate_v2                                            = 236,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D11CtxCreate_v2                                            = 237,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D9CtxCreate_v2                                             = 238,
+    CUPTI_DRIVER_TRACE_CBID_cuGLCtxCreate_v2                                               = 239,
+    CUPTI_DRIVER_TRACE_CBID_cuVDPAUCtxCreate_v2                                            = 240,
+    CUPTI_DRIVER_TRACE_CBID_cuModuleGetGlobal_v2                                           = 241,
+    CUPTI_DRIVER_TRACE_CBID_cuMemGetInfo_v2                                                = 242,
+    CUPTI_DRIVER_TRACE_CBID_cuMemAlloc_v2                                                  = 243,
+    CUPTI_DRIVER_TRACE_CBID_cuMemAllocPitch_v2                                             = 244,
+    CUPTI_DRIVER_TRACE_CBID_cuMemFree_v2                                                   = 245,
+    CUPTI_DRIVER_TRACE_CBID_cuMemGetAddressRange_v2                                        = 246,
+    CUPTI_DRIVER_TRACE_CBID_cuMemHostGetDevicePointer_v2                                   = 247,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpy_v2                                                    = 248,
+    CUPTI_DRIVER_TRACE_CBID_cuMemsetD8_v2                                                  = 249,
+    CUPTI_DRIVER_TRACE_CBID_cuMemsetD16_v2                                                 = 250,
+    CUPTI_DRIVER_TRACE_CBID_cuMemsetD32_v2                                                 = 251,
+    CUPTI_DRIVER_TRACE_CBID_cuMemsetD2D8_v2                                                = 252,
+    CUPTI_DRIVER_TRACE_CBID_cuMemsetD2D16_v2                                               = 253,
+    CUPTI_DRIVER_TRACE_CBID_cuMemsetD2D32_v2                                               = 254,
+    CUPTI_DRIVER_TRACE_CBID_cuTexRefSetAddress_v2                                          = 255,
+    CUPTI_DRIVER_TRACE_CBID_cuTexRefSetAddress2D_v2                                        = 256,
+    CUPTI_DRIVER_TRACE_CBID_cuTexRefGetAddress_v2                                          = 257,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphicsResourceGetMappedPointer_v2                          = 258,
+    CUPTI_DRIVER_TRACE_CBID_cuDeviceTotalMem_v2                                            = 259,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D10ResourceGetMappedPointer_v2                             = 260,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D10ResourceGetMappedSize_v2                                = 261,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D10ResourceGetMappedPitch_v2                               = 262,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D10ResourceGetSurfaceDimensions_v2                         = 263,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D9ResourceGetSurfaceDimensions_v2                          = 264,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D9ResourceGetMappedPointer_v2                              = 265,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D9ResourceGetMappedSize_v2                                 = 266,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D9ResourceGetMappedPitch_v2                                = 267,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D9MapVertexBuffer_v2                                       = 268,
+    CUPTI_DRIVER_TRACE_CBID_cuGLMapBufferObject_v2                                         = 269,
+    CUPTI_DRIVER_TRACE_CBID_cuGLMapBufferObjectAsync_v2                                    = 270,
+    CUPTI_DRIVER_TRACE_CBID_cuMemHostAlloc_v2                                              = 271,
+    CUPTI_DRIVER_TRACE_CBID_cuArrayCreate_v2                                               = 272,
+    CUPTI_DRIVER_TRACE_CBID_cuArrayGetDescriptor_v2                                        = 273,
+    CUPTI_DRIVER_TRACE_CBID_cuArray3DCreate_v2                                             = 274,
+    CUPTI_DRIVER_TRACE_CBID_cuArray3DGetDescriptor_v2                                      = 275,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyHtoD_v2                                                = 276,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyHtoDAsync_v2                                           = 277,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyDtoH_v2                                                = 278,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyDtoHAsync_v2                                           = 279,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyDtoD_v2                                                = 280,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyDtoDAsync_v2                                           = 281,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyAtoH_v2                                                = 282,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyAtoHAsync_v2                                           = 283,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyAtoD_v2                                                = 284,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyDtoA_v2                                                = 285,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyAtoA_v2                                                = 286,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpy2D_v2                                                  = 287,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpy2DUnaligned_v2                                         = 288,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpy2DAsync_v2                                             = 289,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpy3D_v2                                                  = 290,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpy3DAsync_v2                                             = 291,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyHtoA_v2                                                = 292,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyHtoAAsync_v2                                           = 293,
+    CUPTI_DRIVER_TRACE_CBID_cuMemAllocHost_v2                                              = 294,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamWaitEvent                                              = 295,
+    CUPTI_DRIVER_TRACE_CBID_cuCtxGetApiVersion                                             = 296,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D10GetDirect3DDevice                                       = 297,
+    CUPTI_DRIVER_TRACE_CBID_cuD3D11GetDirect3DDevice                                       = 298,
+    CUPTI_DRIVER_TRACE_CBID_cuCtxGetCacheConfig                                            = 299,
+    CUPTI_DRIVER_TRACE_CBID_cuCtxSetCacheConfig                                            = 300,
+    CUPTI_DRIVER_TRACE_CBID_cuMemHostRegister                                              = 301,
+    CUPTI_DRIVER_TRACE_CBID_cuMemHostUnregister                                            = 302,
+    CUPTI_DRIVER_TRACE_CBID_cuCtxSetCurrent                                                = 303,
+    CUPTI_DRIVER_TRACE_CBID_cuCtxGetCurrent                                                = 304,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpy                                                       = 305,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyAsync                                                  = 306,
+    CUPTI_DRIVER_TRACE_CBID_cuLaunchKernel                                                 = 307,
+    CUPTI_DRIVER_TRACE_CBID_cuProfilerStart                                                = 308,
+    CUPTI_DRIVER_TRACE_CBID_cuProfilerStop                                                 = 309,
+    CUPTI_DRIVER_TRACE_CBID_cuPointerGetAttribute                                          = 310,
+    CUPTI_DRIVER_TRACE_CBID_cuProfilerInitialize                                           = 311,
+    CUPTI_DRIVER_TRACE_CBID_cuDeviceCanAccessPeer                                          = 312,
+    CUPTI_DRIVER_TRACE_CBID_cuCtxEnablePeerAccess                                          = 313,
+    CUPTI_DRIVER_TRACE_CBID_cuCtxDisablePeerAccess                                         = 314,
+    CUPTI_DRIVER_TRACE_CBID_cuMemPeerRegister                                              = 315,
+    CUPTI_DRIVER_TRACE_CBID_cuMemPeerUnregister                                            = 316,
+    CUPTI_DRIVER_TRACE_CBID_cuMemPeerGetDevicePointer                                      = 317,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyPeer                                                   = 318,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyPeerAsync                                              = 319,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpy3DPeer                                                 = 320,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpy3DPeerAsync                                            = 321,
+    CUPTI_DRIVER_TRACE_CBID_cuCtxDestroy_v2                                                = 322,
+    CUPTI_DRIVER_TRACE_CBID_cuCtxPushCurrent_v2                                            = 323,
+    CUPTI_DRIVER_TRACE_CBID_cuCtxPopCurrent_v2                                             = 324,
+    CUPTI_DRIVER_TRACE_CBID_cuEventDestroy_v2                                              = 325,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamDestroy_v2                                             = 326,
+    CUPTI_DRIVER_TRACE_CBID_cuTexRefSetAddress2D_v3                                        = 327,
+    CUPTI_DRIVER_TRACE_CBID_cuIpcGetMemHandle                                              = 328,
+    CUPTI_DRIVER_TRACE_CBID_cuIpcOpenMemHandle                                             = 329,
+    CUPTI_DRIVER_TRACE_CBID_cuIpcCloseMemHandle                                            = 330,
+    CUPTI_DRIVER_TRACE_CBID_cuDeviceGetByPCIBusId                                          = 331,
+    CUPTI_DRIVER_TRACE_CBID_cuDeviceGetPCIBusId                                            = 332,
+    CUPTI_DRIVER_TRACE_CBID_cuGLGetDevices                                                 = 333,
+    CUPTI_DRIVER_TRACE_CBID_cuIpcGetEventHandle                                            = 334,
+    CUPTI_DRIVER_TRACE_CBID_cuIpcOpenEventHandle                                           = 335,
+    CUPTI_DRIVER_TRACE_CBID_cuCtxSetSharedMemConfig                                        = 336,
+    CUPTI_DRIVER_TRACE_CBID_cuCtxGetSharedMemConfig                                        = 337,
+    CUPTI_DRIVER_TRACE_CBID_cuFuncSetSharedMemConfig                                       = 338,
+    CUPTI_DRIVER_TRACE_CBID_cuTexObjectCreate                                              = 339,
+    CUPTI_DRIVER_TRACE_CBID_cuTexObjectDestroy                                             = 340,
+    CUPTI_DRIVER_TRACE_CBID_cuTexObjectGetResourceDesc                                     = 341,
+    CUPTI_DRIVER_TRACE_CBID_cuTexObjectGetTextureDesc                                      = 342,
+    CUPTI_DRIVER_TRACE_CBID_cuSurfObjectCreate                                             = 343,
+    CUPTI_DRIVER_TRACE_CBID_cuSurfObjectDestroy                                            = 344,
+    CUPTI_DRIVER_TRACE_CBID_cuSurfObjectGetResourceDesc                                    = 345,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamAddCallback                                            = 346,
+    CUPTI_DRIVER_TRACE_CBID_cuMipmappedArrayCreate                                         = 347,
+    CUPTI_DRIVER_TRACE_CBID_cuMipmappedArrayGetLevel                                       = 348,
+    CUPTI_DRIVER_TRACE_CBID_cuMipmappedArrayDestroy                                        = 349,
+    CUPTI_DRIVER_TRACE_CBID_cuTexRefSetMipmappedArray                                      = 350,
+    CUPTI_DRIVER_TRACE_CBID_cuTexRefSetMipmapFilterMode                                    = 351,
+    CUPTI_DRIVER_TRACE_CBID_cuTexRefSetMipmapLevelBias                                     = 352,
+    CUPTI_DRIVER_TRACE_CBID_cuTexRefSetMipmapLevelClamp                                    = 353,
+    CUPTI_DRIVER_TRACE_CBID_cuTexRefSetMaxAnisotropy                                       = 354,
+    CUPTI_DRIVER_TRACE_CBID_cuTexRefGetMipmappedArray                                      = 355,
+    CUPTI_DRIVER_TRACE_CBID_cuTexRefGetMipmapFilterMode                                    = 356,
+    CUPTI_DRIVER_TRACE_CBID_cuTexRefGetMipmapLevelBias                                     = 357,
+    CUPTI_DRIVER_TRACE_CBID_cuTexRefGetMipmapLevelClamp                                    = 358,
+    CUPTI_DRIVER_TRACE_CBID_cuTexRefGetMaxAnisotropy                                       = 359,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphicsResourceGetMappedMipmappedArray                      = 360,
+    CUPTI_DRIVER_TRACE_CBID_cuTexObjectGetResourceViewDesc                                 = 361,
+    CUPTI_DRIVER_TRACE_CBID_cuLinkCreate                                                   = 362,
+    CUPTI_DRIVER_TRACE_CBID_cuLinkAddData                                                  = 363,
+    CUPTI_DRIVER_TRACE_CBID_cuLinkAddFile                                                  = 364,
+    CUPTI_DRIVER_TRACE_CBID_cuLinkComplete                                                 = 365,
+    CUPTI_DRIVER_TRACE_CBID_cuLinkDestroy                                                  = 366,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamCreateWithPriority                                     = 367,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamGetPriority                                            = 368,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamGetFlags                                               = 369,
+    CUPTI_DRIVER_TRACE_CBID_cuCtxGetStreamPriorityRange                                    = 370,
+    CUPTI_DRIVER_TRACE_CBID_cuMemAllocManaged                                              = 371,
+    CUPTI_DRIVER_TRACE_CBID_cuGetErrorString                                               = 372,
+    CUPTI_DRIVER_TRACE_CBID_cuGetErrorName                                                 = 373,
+    CUPTI_DRIVER_TRACE_CBID_cuOccupancyMaxActiveBlocksPerMultiprocessor                    = 374,
+    CUPTI_DRIVER_TRACE_CBID_cuCompilePtx                                                   = 375,
+    CUPTI_DRIVER_TRACE_CBID_cuBinaryFree                                                   = 376,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamAttachMemAsync                                         = 377,
+    CUPTI_DRIVER_TRACE_CBID_cuPointerSetAttribute                                          = 378,
+    CUPTI_DRIVER_TRACE_CBID_cuMemHostRegister_v2                                           = 379,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphicsResourceSetMapFlags_v2                               = 380,
+    CUPTI_DRIVER_TRACE_CBID_cuLinkCreate_v2                                                = 381,
+    CUPTI_DRIVER_TRACE_CBID_cuLinkAddData_v2                                               = 382,
+    CUPTI_DRIVER_TRACE_CBID_cuLinkAddFile_v2                                               = 383,
+    CUPTI_DRIVER_TRACE_CBID_cuOccupancyMaxPotentialBlockSize                               = 384,
+    CUPTI_DRIVER_TRACE_CBID_cuGLGetDevices_v2                                              = 385,
+    CUPTI_DRIVER_TRACE_CBID_cuDevicePrimaryCtxRetain                                       = 386,
+    CUPTI_DRIVER_TRACE_CBID_cuDevicePrimaryCtxRelease                                      = 387,
+    CUPTI_DRIVER_TRACE_CBID_cuDevicePrimaryCtxSetFlags                                     = 388,
+    CUPTI_DRIVER_TRACE_CBID_cuDevicePrimaryCtxReset                                        = 389,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphicsEGLRegisterImage                                     = 390,
+    CUPTI_DRIVER_TRACE_CBID_cuCtxGetFlags                                                  = 391,
+    CUPTI_DRIVER_TRACE_CBID_cuDevicePrimaryCtxGetState                                     = 392,
+    CUPTI_DRIVER_TRACE_CBID_cuEGLStreamConsumerConnect                                     = 393,
+    CUPTI_DRIVER_TRACE_CBID_cuEGLStreamConsumerDisconnect                                  = 394,
+    CUPTI_DRIVER_TRACE_CBID_cuEGLStreamConsumerAcquireFrame                                = 395,
+    CUPTI_DRIVER_TRACE_CBID_cuEGLStreamConsumerReleaseFrame                                = 396,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyHtoD_v2_ptds                                           = 397,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyDtoH_v2_ptds                                           = 398,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyDtoD_v2_ptds                                           = 399,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyDtoA_v2_ptds                                           = 400,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyAtoD_v2_ptds                                           = 401,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyHtoA_v2_ptds                                           = 402,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyAtoH_v2_ptds                                           = 403,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyAtoA_v2_ptds                                           = 404,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpy2D_v2_ptds                                             = 405,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpy2DUnaligned_v2_ptds                                    = 406,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpy3D_v2_ptds                                             = 407,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpy_ptds                                                  = 408,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyPeer_ptds                                              = 409,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpy3DPeer_ptds                                            = 410,
+    CUPTI_DRIVER_TRACE_CBID_cuMemsetD8_v2_ptds                                             = 411,
+    CUPTI_DRIVER_TRACE_CBID_cuMemsetD16_v2_ptds                                            = 412,
+    CUPTI_DRIVER_TRACE_CBID_cuMemsetD32_v2_ptds                                            = 413,
+    CUPTI_DRIVER_TRACE_CBID_cuMemsetD2D8_v2_ptds                                           = 414,
+    CUPTI_DRIVER_TRACE_CBID_cuMemsetD2D16_v2_ptds                                          = 415,
+    CUPTI_DRIVER_TRACE_CBID_cuMemsetD2D32_v2_ptds                                          = 416,
+    CUPTI_DRIVER_TRACE_CBID_cuGLMapBufferObject_v2_ptds                                    = 417,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyAsync_ptsz                                             = 418,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyHtoAAsync_v2_ptsz                                      = 419,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyAtoHAsync_v2_ptsz                                      = 420,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyHtoDAsync_v2_ptsz                                      = 421,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyDtoHAsync_v2_ptsz                                      = 422,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyDtoDAsync_v2_ptsz                                      = 423,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpy2DAsync_v2_ptsz                                        = 424,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpy3DAsync_v2_ptsz                                        = 425,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyPeerAsync_ptsz                                         = 426,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpy3DPeerAsync_ptsz                                       = 427,
+    CUPTI_DRIVER_TRACE_CBID_cuMemsetD8Async_ptsz                                           = 428,
+    CUPTI_DRIVER_TRACE_CBID_cuMemsetD16Async_ptsz                                          = 429,
+    CUPTI_DRIVER_TRACE_CBID_cuMemsetD32Async_ptsz                                          = 430,
+    CUPTI_DRIVER_TRACE_CBID_cuMemsetD2D8Async_ptsz                                         = 431,
+    CUPTI_DRIVER_TRACE_CBID_cuMemsetD2D16Async_ptsz                                        = 432,
+    CUPTI_DRIVER_TRACE_CBID_cuMemsetD2D32Async_ptsz                                        = 433,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamGetPriority_ptsz                                       = 434,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamGetFlags_ptsz                                          = 435,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamWaitEvent_ptsz                                         = 436,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamAddCallback_ptsz                                       = 437,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamAttachMemAsync_ptsz                                    = 438,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamQuery_ptsz                                             = 439,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamSynchronize_ptsz                                       = 440,
+    CUPTI_DRIVER_TRACE_CBID_cuEventRecord_ptsz                                             = 441,
+    CUPTI_DRIVER_TRACE_CBID_cuLaunchKernel_ptsz                                            = 442,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphicsMapResources_ptsz                                    = 443,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphicsUnmapResources_ptsz                                  = 444,
+    CUPTI_DRIVER_TRACE_CBID_cuGLMapBufferObjectAsync_v2_ptsz                               = 445,
+    CUPTI_DRIVER_TRACE_CBID_cuEGLStreamProducerConnect                                     = 446,
+    CUPTI_DRIVER_TRACE_CBID_cuEGLStreamProducerDisconnect                                  = 447,
+    CUPTI_DRIVER_TRACE_CBID_cuEGLStreamProducerPresentFrame                                = 448,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphicsResourceGetMappedEglFrame                            = 449,
+    CUPTI_DRIVER_TRACE_CBID_cuPointerGetAttributes                                         = 450,
+    CUPTI_DRIVER_TRACE_CBID_cuOccupancyMaxActiveBlocksPerMultiprocessorWithFlags           = 451,
+    CUPTI_DRIVER_TRACE_CBID_cuOccupancyMaxPotentialBlockSizeWithFlags                      = 452,
+    CUPTI_DRIVER_TRACE_CBID_cuEGLStreamProducerReturnFrame                                 = 453,
+    CUPTI_DRIVER_TRACE_CBID_cuDeviceGetP2PAttribute                                        = 454,
+    CUPTI_DRIVER_TRACE_CBID_cuTexRefSetBorderColor                                         = 455,
+    CUPTI_DRIVER_TRACE_CBID_cuTexRefGetBorderColor                                         = 456,
+    CUPTI_DRIVER_TRACE_CBID_cuMemAdvise                                                    = 457,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamWaitValue32                                            = 458,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamWaitValue32_ptsz                                       = 459,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamWriteValue32                                           = 460,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamWriteValue32_ptsz                                      = 461,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamBatchMemOp                                             = 462,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamBatchMemOp_ptsz                                        = 463,
+    CUPTI_DRIVER_TRACE_CBID_cuNVNbufferGetPointer                                          = 464,
+    CUPTI_DRIVER_TRACE_CBID_cuNVNtextureGetArray                                           = 465,
+    CUPTI_DRIVER_TRACE_CBID_cuNNSetAllocator                                               = 466,
+    CUPTI_DRIVER_TRACE_CBID_cuMemPrefetchAsync                                             = 467,
+    CUPTI_DRIVER_TRACE_CBID_cuMemPrefetchAsync_ptsz                                        = 468,
+    CUPTI_DRIVER_TRACE_CBID_cuEventCreateFromNVNSync                                       = 469,
+    CUPTI_DRIVER_TRACE_CBID_cuEGLStreamConsumerConnectWithFlags                            = 470,
+    CUPTI_DRIVER_TRACE_CBID_cuMemRangeGetAttribute                                         = 471,
+    CUPTI_DRIVER_TRACE_CBID_cuMemRangeGetAttributes                                        = 472,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamWaitValue64                                            = 473,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamWaitValue64_ptsz                                       = 474,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamWriteValue64                                           = 475,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamWriteValue64_ptsz                                      = 476,
+    CUPTI_DRIVER_TRACE_CBID_cuLaunchCooperativeKernel                                      = 477,
+    CUPTI_DRIVER_TRACE_CBID_cuLaunchCooperativeKernel_ptsz                                 = 478,
+    CUPTI_DRIVER_TRACE_CBID_cuEventCreateFromEGLSync                                       = 479,
+    CUPTI_DRIVER_TRACE_CBID_cuLaunchCooperativeKernelMultiDevice                           = 480,
+    CUPTI_DRIVER_TRACE_CBID_cuFuncSetAttribute                                             = 481,
+    CUPTI_DRIVER_TRACE_CBID_cuDeviceGetUuid                                                = 482,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamGetCtx                                                 = 483,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamGetCtx_ptsz                                            = 484,
+    CUPTI_DRIVER_TRACE_CBID_cuImportExternalMemory                                         = 485,
+    CUPTI_DRIVER_TRACE_CBID_cuExternalMemoryGetMappedBuffer                                = 486,
+    CUPTI_DRIVER_TRACE_CBID_cuExternalMemoryGetMappedMipmappedArray                        = 487,
+    CUPTI_DRIVER_TRACE_CBID_cuDestroyExternalMemory                                        = 488,
+    CUPTI_DRIVER_TRACE_CBID_cuImportExternalSemaphore                                      = 489,
+    CUPTI_DRIVER_TRACE_CBID_cuSignalExternalSemaphoresAsync                                = 490,
+    CUPTI_DRIVER_TRACE_CBID_cuSignalExternalSemaphoresAsync_ptsz                           = 491,
+    CUPTI_DRIVER_TRACE_CBID_cuWaitExternalSemaphoresAsync                                  = 492,
+    CUPTI_DRIVER_TRACE_CBID_cuWaitExternalSemaphoresAsync_ptsz                             = 493,
+    CUPTI_DRIVER_TRACE_CBID_cuDestroyExternalSemaphore                                     = 494,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamBeginCapture                                           = 495,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamBeginCapture_ptsz                                      = 496,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamEndCapture                                             = 497,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamEndCapture_ptsz                                        = 498,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamIsCapturing                                            = 499,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamIsCapturing_ptsz                                       = 500,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphCreate                                                  = 501,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphAddKernelNode                                           = 502,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphKernelNodeGetParams                                     = 503,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphAddMemcpyNode                                           = 504,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphMemcpyNodeGetParams                                     = 505,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphAddMemsetNode                                           = 506,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphMemsetNodeGetParams                                     = 507,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphMemsetNodeSetParams                                     = 508,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphNodeGetType                                             = 509,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphGetRootNodes                                            = 510,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphNodeGetDependencies                                     = 511,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphNodeGetDependentNodes                                   = 512,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphInstantiate                                             = 513,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphLaunch                                                  = 514,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphLaunch_ptsz                                             = 515,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphExecDestroy                                             = 516,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphDestroy                                                 = 517,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphAddDependencies                                         = 518,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphRemoveDependencies                                      = 519,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphMemcpyNodeSetParams                                     = 520,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphKernelNodeSetParams                                     = 521,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphDestroyNode                                             = 522,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphClone                                                   = 523,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphNodeFindInClone                                         = 524,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphAddChildGraphNode                                       = 525,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphAddEmptyNode                                            = 526,
+    CUPTI_DRIVER_TRACE_CBID_cuLaunchHostFunc                                               = 527,
+    CUPTI_DRIVER_TRACE_CBID_cuLaunchHostFunc_ptsz                                          = 528,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphChildGraphNodeGetGraph                                  = 529,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphAddHostNode                                             = 530,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphHostNodeGetParams                                       = 531,
+    CUPTI_DRIVER_TRACE_CBID_cuDeviceGetLuid                                                = 532,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphHostNodeSetParams                                       = 533,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphGetNodes                                                = 534,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphGetEdges                                                = 535,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamGetCaptureInfo                                         = 536,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamGetCaptureInfo_ptsz                                    = 537,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphExecKernelNodeSetParams                                 = 538,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamBeginCapture_v2                                        = 539,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamBeginCapture_v2_ptsz                                   = 540,
+    CUPTI_DRIVER_TRACE_CBID_cuThreadExchangeStreamCaptureMode                              = 541,
+    CUPTI_DRIVER_TRACE_CBID_cuDeviceGetNvSciSyncAttributes                                 = 542,
+    CUPTI_DRIVER_TRACE_CBID_cuOccupancyAvailableDynamicSMemPerBlock                        = 543,
+    CUPTI_DRIVER_TRACE_CBID_cuDevicePrimaryCtxRelease_v2                                   = 544,
+    CUPTI_DRIVER_TRACE_CBID_cuDevicePrimaryCtxReset_v2                                     = 545,
+    CUPTI_DRIVER_TRACE_CBID_cuDevicePrimaryCtxSetFlags_v2                                  = 546,
+    CUPTI_DRIVER_TRACE_CBID_cuMemAddressReserve                                            = 547,
+    CUPTI_DRIVER_TRACE_CBID_cuMemAddressFree                                               = 548,
+    CUPTI_DRIVER_TRACE_CBID_cuMemCreate                                                    = 549,
+    CUPTI_DRIVER_TRACE_CBID_cuMemRelease                                                   = 550,
+    CUPTI_DRIVER_TRACE_CBID_cuMemMap                                                       = 551,
+    CUPTI_DRIVER_TRACE_CBID_cuMemUnmap                                                     = 552,
+    CUPTI_DRIVER_TRACE_CBID_cuMemSetAccess                                                 = 553,
+    CUPTI_DRIVER_TRACE_CBID_cuMemExportToShareableHandle                                   = 554,
+    CUPTI_DRIVER_TRACE_CBID_cuMemImportFromShareableHandle                                 = 555,
+    CUPTI_DRIVER_TRACE_CBID_cuMemGetAllocationGranularity                                  = 556,
+    CUPTI_DRIVER_TRACE_CBID_cuMemGetAllocationPropertiesFromHandle                         = 557,
+    CUPTI_DRIVER_TRACE_CBID_cuMemGetAccess                                                 = 558,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamSetFlags                                               = 559,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamSetFlags_ptsz                                          = 560,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphExecUpdate                                              = 561,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphExecMemcpyNodeSetParams                                 = 562,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphExecMemsetNodeSetParams                                 = 563,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphExecHostNodeSetParams                                   = 564,
+    CUPTI_DRIVER_TRACE_CBID_cuMemRetainAllocationHandle                                    = 565,
+    CUPTI_DRIVER_TRACE_CBID_cuFuncGetModule                                                = 566,
+    CUPTI_DRIVER_TRACE_CBID_cuIpcOpenMemHandle_v2                                          = 567,
+    CUPTI_DRIVER_TRACE_CBID_cuCtxResetPersistingL2Cache                                    = 568,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphKernelNodeCopyAttributes                                = 569,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphKernelNodeGetAttribute                                  = 570,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphKernelNodeSetAttribute                                  = 571,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamCopyAttributes                                         = 572,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamCopyAttributes_ptsz                                    = 573,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamGetAttribute                                           = 574,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamGetAttribute_ptsz                                      = 575,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamSetAttribute                                           = 576,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamSetAttribute_ptsz                                      = 577,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphInstantiate_v2                                          = 578,
+    CUPTI_DRIVER_TRACE_CBID_cuDeviceGetTexture1DLinearMaxWidth                             = 579,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphUpload                                                  = 580,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphUpload_ptsz                                             = 581,
+    CUPTI_DRIVER_TRACE_CBID_cuArrayGetSparseProperties                                     = 582,
+    CUPTI_DRIVER_TRACE_CBID_cuMipmappedArrayGetSparseProperties                            = 583,
+    CUPTI_DRIVER_TRACE_CBID_cuMemMapArrayAsync                                             = 584,
+    CUPTI_DRIVER_TRACE_CBID_cuMemMapArrayAsync_ptsz                                        = 585,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphExecChildGraphNodeSetParams                             = 586,
+    CUPTI_DRIVER_TRACE_CBID_cuEventRecordWithFlags                                         = 587,
+    CUPTI_DRIVER_TRACE_CBID_cuEventRecordWithFlags_ptsz                                    = 588,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphAddEventRecordNode                                      = 589,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphAddEventWaitNode                                        = 590,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphEventRecordNodeGetEvent                                 = 591,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphEventWaitNodeGetEvent                                   = 592,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphEventRecordNodeSetEvent                                 = 593,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphEventWaitNodeSetEvent                                   = 594,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphExecEventRecordNodeSetEvent                             = 595,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphExecEventWaitNodeSetEvent                               = 596,
+    CUPTI_DRIVER_TRACE_CBID_cuArrayGetPlane                                                = 597,
+    CUPTI_DRIVER_TRACE_CBID_cuMemAllocAsync                                                = 598,
+    CUPTI_DRIVER_TRACE_CBID_cuMemAllocAsync_ptsz                                           = 599,
+    CUPTI_DRIVER_TRACE_CBID_cuMemFreeAsync                                                 = 600,
+    CUPTI_DRIVER_TRACE_CBID_cuMemFreeAsync_ptsz                                            = 601,
+    CUPTI_DRIVER_TRACE_CBID_cuMemPoolTrimTo                                                = 602,
+    CUPTI_DRIVER_TRACE_CBID_cuMemPoolSetAttribute                                          = 603,
+    CUPTI_DRIVER_TRACE_CBID_cuMemPoolGetAttribute                                          = 604,
+    CUPTI_DRIVER_TRACE_CBID_cuMemPoolSetAccess                                             = 605,
+    CUPTI_DRIVER_TRACE_CBID_cuDeviceGetDefaultMemPool                                      = 606,
+    CUPTI_DRIVER_TRACE_CBID_cuMemPoolCreate                                                = 607,
+    CUPTI_DRIVER_TRACE_CBID_cuMemPoolDestroy                                               = 608,
+    CUPTI_DRIVER_TRACE_CBID_cuDeviceSetMemPool                                             = 609,
+    CUPTI_DRIVER_TRACE_CBID_cuDeviceGetMemPool                                             = 610,
+    CUPTI_DRIVER_TRACE_CBID_cuMemAllocFromPoolAsync                                        = 611,
+    CUPTI_DRIVER_TRACE_CBID_cuMemAllocFromPoolAsync_ptsz                                   = 612,
+    CUPTI_DRIVER_TRACE_CBID_cuMemPoolExportToShareableHandle                               = 613,
+    CUPTI_DRIVER_TRACE_CBID_cuMemPoolImportFromShareableHandle                             = 614,
+    CUPTI_DRIVER_TRACE_CBID_cuMemPoolExportPointer                                         = 615,
+    CUPTI_DRIVER_TRACE_CBID_cuMemPoolImportPointer                                         = 616,
+    CUPTI_DRIVER_TRACE_CBID_cuMemPoolGetAccess                                             = 617,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphAddExternalSemaphoresSignalNode                         = 618,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphExternalSemaphoresSignalNodeGetParams                   = 619,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphExternalSemaphoresSignalNodeSetParams                   = 620,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphAddExternalSemaphoresWaitNode                           = 621,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphExternalSemaphoresWaitNodeGetParams                     = 622,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphExternalSemaphoresWaitNodeSetParams                     = 623,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphExecExternalSemaphoresSignalNodeSetParams               = 624,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphExecExternalSemaphoresWaitNodeSetParams                 = 625,
+    CUPTI_DRIVER_TRACE_CBID_cuGetProcAddress                                               = 626,
+    CUPTI_DRIVER_TRACE_CBID_cuFlushGPUDirectRDMAWrites                                     = 627,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphDebugDotPrint                                           = 628,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamGetCaptureInfo_v2                                      = 629,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamGetCaptureInfo_v2_ptsz                                 = 630,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamUpdateCaptureDependencies                              = 631,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamUpdateCaptureDependencies_ptsz                         = 632,
+    CUPTI_DRIVER_TRACE_CBID_cuUserObjectCreate                                             = 633,
+    CUPTI_DRIVER_TRACE_CBID_cuUserObjectRetain                                             = 634,
+    CUPTI_DRIVER_TRACE_CBID_cuUserObjectRelease                                            = 635,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphRetainUserObject                                        = 636,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphReleaseUserObject                                       = 637,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphAddMemAllocNode                                         = 638,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphAddMemFreeNode                                          = 639,
+    CUPTI_DRIVER_TRACE_CBID_cuDeviceGraphMemTrim                                           = 640,
+    CUPTI_DRIVER_TRACE_CBID_cuDeviceGetGraphMemAttribute                                   = 641,
+    CUPTI_DRIVER_TRACE_CBID_cuDeviceSetGraphMemAttribute                                   = 642,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphInstantiateWithFlags                                    = 643,
+    CUPTI_DRIVER_TRACE_CBID_cuDeviceGetExecAffinitySupport                                 = 644,
+    CUPTI_DRIVER_TRACE_CBID_cuCtxCreate_v3                                                 = 645,
+    CUPTI_DRIVER_TRACE_CBID_cuCtxGetExecAffinity                                           = 646,
+    CUPTI_DRIVER_TRACE_CBID_cuDeviceGetUuid_v2                                             = 647,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphMemAllocNodeGetParams                                   = 648,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphMemFreeNodeGetParams                                    = 649,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphNodeSetEnabled                                          = 650,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphNodeGetEnabled                                          = 651,
+    CUPTI_DRIVER_TRACE_CBID_cuLaunchKernelEx                                               = 652,
+    CUPTI_DRIVER_TRACE_CBID_cuLaunchKernelEx_ptsz                                          = 653,
+    CUPTI_DRIVER_TRACE_CBID_cuArrayGetMemoryRequirements                                   = 654,
+    CUPTI_DRIVER_TRACE_CBID_cuMipmappedArrayGetMemoryRequirements                          = 655,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphInstantiateWithParams                                   = 656,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphInstantiateWithParams_ptsz                              = 657,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphExecGetFlags                                            = 658,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamWaitValue32_v2                                         = 659,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamWaitValue32_v2_ptsz                                    = 660,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamWaitValue64_v2                                         = 661,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamWaitValue64_v2_ptsz                                    = 662,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamWriteValue32_v2                                        = 663,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamWriteValue32_v2_ptsz                                   = 664,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamWriteValue64_v2                                        = 665,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamWriteValue64_v2_ptsz                                   = 666,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamBatchMemOp_v2                                          = 667,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamBatchMemOp_v2_ptsz                                     = 668,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphAddBatchMemOpNode                                       = 669,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphBatchMemOpNodeGetParams                                 = 670,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphBatchMemOpNodeSetParams                                 = 671,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphExecBatchMemOpNodeSetParams                             = 672,
+    CUPTI_DRIVER_TRACE_CBID_cuModuleGetLoadingMode                                         = 673,
+    CUPTI_DRIVER_TRACE_CBID_cuMemGetHandleForAddressRange                                  = 674,
+    CUPTI_DRIVER_TRACE_CBID_cuOccupancyMaxPotentialClusterSize                             = 675,
+    CUPTI_DRIVER_TRACE_CBID_cuOccupancyMaxActiveClusters                                   = 676,
+    CUPTI_DRIVER_TRACE_CBID_cuGetProcAddress_v2                                            = 677,
+    CUPTI_DRIVER_TRACE_CBID_cuLibraryLoadData                                              = 678,
+    CUPTI_DRIVER_TRACE_CBID_cuLibraryLoadFromFile                                          = 679,
+    CUPTI_DRIVER_TRACE_CBID_cuLibraryUnload                                                = 680,
+    CUPTI_DRIVER_TRACE_CBID_cuLibraryGetKernel                                             = 681,
+    CUPTI_DRIVER_TRACE_CBID_cuLibraryGetModule                                             = 682,
+    CUPTI_DRIVER_TRACE_CBID_cuKernelGetFunction                                            = 683,
+    CUPTI_DRIVER_TRACE_CBID_cuLibraryGetGlobal                                             = 684,
+    CUPTI_DRIVER_TRACE_CBID_cuLibraryGetManaged                                            = 685,
+    CUPTI_DRIVER_TRACE_CBID_cuKernelGetAttribute                                           = 686,
+    CUPTI_DRIVER_TRACE_CBID_cuKernelSetAttribute                                           = 687,
+    CUPTI_DRIVER_TRACE_CBID_cuKernelSetCacheConfig                                         = 688,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphAddKernelNode_v2                                        = 689,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphKernelNodeGetParams_v2                                  = 690,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphKernelNodeSetParams_v2                                  = 691,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphExecKernelNodeSetParams_v2                              = 692,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamGetId                                                  = 693,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamGetId_ptsz                                             = 694,
+    CUPTI_DRIVER_TRACE_CBID_cuCtxGetId                                                     = 695,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphExecUpdate_v2                                           = 696,
+    CUPTI_DRIVER_TRACE_CBID_cuTensorMapEncodeTiled                                         = 697,
+    CUPTI_DRIVER_TRACE_CBID_cuTensorMapEncodeIm2col                                        = 698,
+    CUPTI_DRIVER_TRACE_CBID_cuTensorMapReplaceAddress                                      = 699,
+    CUPTI_DRIVER_TRACE_CBID_cuLibraryGetUnifiedFunction                                    = 700,
+    CUPTI_DRIVER_TRACE_CBID_cuCoredumpGetAttribute                                         = 701,
+    CUPTI_DRIVER_TRACE_CBID_cuCoredumpGetAttributeGlobal                                   = 702,
+    CUPTI_DRIVER_TRACE_CBID_cuCoredumpSetAttribute                                         = 703,
+    CUPTI_DRIVER_TRACE_CBID_cuCoredumpSetAttributeGlobal                                   = 704,
+    CUPTI_DRIVER_TRACE_CBID_cuCtxSetFlags                                                  = 705,
+    CUPTI_DRIVER_TRACE_CBID_cuMulticastCreate                                              = 706,
+    CUPTI_DRIVER_TRACE_CBID_cuMulticastAddDevice                                           = 707,
+    CUPTI_DRIVER_TRACE_CBID_cuMulticastBindMem                                             = 708,
+    CUPTI_DRIVER_TRACE_CBID_cuMulticastBindAddr                                            = 709,
+    CUPTI_DRIVER_TRACE_CBID_cuMulticastUnbind                                              = 710,
+    CUPTI_DRIVER_TRACE_CBID_cuMulticastGetGranularity                                      = 711,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphAddNode                                                 = 712,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphNodeSetParams                                           = 713,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphExecNodeSetParams                                       = 714,
+    CUPTI_DRIVER_TRACE_CBID_cuMemAdvise_v2                                                 = 715,
+    CUPTI_DRIVER_TRACE_CBID_cuMemPrefetchAsync_v2                                          = 716,
+    CUPTI_DRIVER_TRACE_CBID_cuMemPrefetchAsync_v2_ptsz                                     = 717,
+    CUPTI_DRIVER_TRACE_CBID_cuFuncGetName                                                  = 718,
+    CUPTI_DRIVER_TRACE_CBID_cuKernelGetName                                                = 719,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamBeginCaptureToGraph                                    = 720,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamBeginCaptureToGraph_ptsz                               = 721,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphConditionalHandleCreate                                 = 722,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphAddNode_v2                                              = 723,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphGetEdges_v2                                             = 724,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphNodeGetDependencies_v2                                  = 725,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphNodeGetDependentNodes_v2                                = 726,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphAddDependencies_v2                                      = 727,
+    CUPTI_DRIVER_TRACE_CBID_cuGraphRemoveDependencies_v2                                   = 728,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamGetCaptureInfo_v3                                      = 729,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamGetCaptureInfo_v3_ptsz                                 = 730,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamUpdateCaptureDependencies_v2                           = 731,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamUpdateCaptureDependencies_v2_ptsz                      = 732,
+    CUPTI_DRIVER_TRACE_CBID_cuFuncGetParamInfo                                             = 733,
+    CUPTI_DRIVER_TRACE_CBID_cuKernelGetParamInfo                                           = 734,
+    CUPTI_DRIVER_TRACE_CBID_cuDeviceRegisterAsyncNotification                              = 735,
+    CUPTI_DRIVER_TRACE_CBID_cuDeviceUnregisterAsyncNotification                            = 736,
+    CUPTI_DRIVER_TRACE_CBID_cuModuleGetFunctionCount                                       = 737,
+    CUPTI_DRIVER_TRACE_CBID_cuModuleEnumerateFunctions                                     = 738,
+    CUPTI_DRIVER_TRACE_CBID_cuLibraryGetKernelCount                                        = 739,
+    CUPTI_DRIVER_TRACE_CBID_cuLibraryEnumerateKernels                                      = 740,
+    CUPTI_DRIVER_TRACE_CBID_cuFuncIsLoaded                                                 = 741,
+    CUPTI_DRIVER_TRACE_CBID_cuFuncLoad                                                     = 742,
+    CUPTI_DRIVER_TRACE_CBID_cuGreenCtxCreate                                               = 743,
+    CUPTI_DRIVER_TRACE_CBID_cuGreenCtxDestroy                                              = 744,
+    CUPTI_DRIVER_TRACE_CBID_cuDeviceGetDevResource                                         = 745,
+    CUPTI_DRIVER_TRACE_CBID_cuCtxGetDevResource                                            = 746,
+    CUPTI_DRIVER_TRACE_CBID_cuGreenCtxGetDevResource                                       = 747,
+    CUPTI_DRIVER_TRACE_CBID_cuDevResourceGenerateDesc                                      = 748,
+    CUPTI_DRIVER_TRACE_CBID_cuGreenCtxRecordEvent                                          = 749,
+    CUPTI_DRIVER_TRACE_CBID_cuGreenCtxWaitEvent                                            = 750,
+    CUPTI_DRIVER_TRACE_CBID_cuDevSmResourceSplitByCount                                    = 751,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamGetGreenCtx                                            = 752,
+    CUPTI_DRIVER_TRACE_CBID_cuCtxFromGreenCtx                                              = 753,
+    CUPTI_DRIVER_TRACE_CBID_cuKernelGetLibrary                                             = 754,
+    CUPTI_DRIVER_TRACE_CBID_cuCtxRecordEvent                                               = 755,
+    CUPTI_DRIVER_TRACE_CBID_cuCtxWaitEvent                                                 = 756,
+    CUPTI_DRIVER_TRACE_CBID_cuCtxCreate_v4                                                 = 757,
+    CUPTI_DRIVER_TRACE_CBID_cuGreenCtxStreamCreate                                         = 758,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamGetCtx_v2                                              = 759,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamGetCtx_v2_ptsz                                         = 760,
+    CUPTI_DRIVER_TRACE_CBID_cuMemBatchDecompressAsync                                      = 761,
+    CUPTI_DRIVER_TRACE_CBID_cuMemBatchDecompressAsync_ptsz                                 = 762,
+    CUPTI_DRIVER_TRACE_CBID_cuLogsRegisterCallback                                         = 763,
+    CUPTI_DRIVER_TRACE_CBID_cuLogsUnregisterCallback                                       = 764,
+    CUPTI_DRIVER_TRACE_CBID_cuLogsCurrent                                                  = 765,
+    CUPTI_DRIVER_TRACE_CBID_cuLogsDumpToFile                                               = 766,
+    CUPTI_DRIVER_TRACE_CBID_cuLogsDumpToMemory                                             = 767,
+    CUPTI_DRIVER_TRACE_CBID_cuCheckpointProcessGetRestoreThreadId                          = 768,
+    CUPTI_DRIVER_TRACE_CBID_cuCheckpointProcessGetState                                    = 769,
+    CUPTI_DRIVER_TRACE_CBID_cuCheckpointProcessLock                                        = 770,
+    CUPTI_DRIVER_TRACE_CBID_cuCheckpointProcessCheckpoint                                  = 771,
+    CUPTI_DRIVER_TRACE_CBID_cuCheckpointProcessRestore                                     = 772,
+    CUPTI_DRIVER_TRACE_CBID_cuCheckpointProcessUnlock                                      = 773,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamGetDevice                                              = 774,
+    CUPTI_DRIVER_TRACE_CBID_cuStreamGetDevice_ptsz                                         = 775,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyBatchAsync                                             = 776,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpyBatchAsync_ptsz                                        = 777,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpy3DBatchAsync                                           = 778,
+    CUPTI_DRIVER_TRACE_CBID_cuMemcpy3DBatchAsync_ptsz                                      = 779,
+    CUPTI_DRIVER_TRACE_CBID_cuEventElapsedTime_v2                                          = 780,
+    CUPTI_DRIVER_TRACE_CBID_cuTensorMapEncodeIm2colWide                                    = 781,
+    CUPTI_DRIVER_TRACE_CBID_SIZE                                                           = 782,
+    CUPTI_DRIVER_TRACE_CBID_FORCE_INT                                                      = 0x7fffffff
+} CUpti_driver_api_trace_cbid;
+
+#endif

Prism/LLaDA/LLaDA_Prism/.venv/lib/python3.12/site-packages/nvidia/cuda_cupti/include/cupti_events.h

@@ -0,0 +1,1349 @@
+/*
+ * Copyright 2010-2024 NVIDIA Corporation.  All rights reserved.
+ *
+ * NOTICE TO LICENSEE:
+ *
+ * This source code and/or documentation ("Licensed Deliverables") are
+ * subject to NVIDIA intellectual property rights under U.S. and
+ * international Copyright laws.
+ *
+ * These Licensed Deliverables contained herein is PROPRIETARY and
+ * CONFIDENTIAL to NVIDIA and is being provided under the terms and
+ * conditions of a form of NVIDIA software license agreement by and
+ * between NVIDIA and Licensee ("License Agreement") or electronically
+ * accepted by Licensee.  Notwithstanding any terms or conditions to
+ * the contrary in the License Agreement, reproduction or disclosure
+ * of the Licensed Deliverables to any third party without the express
+ * written consent of NVIDIA is prohibited.
+ *
+ * NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE
+ * LICENSE AGREEMENT, NVIDIA MAKES NO REPRESENTATION ABOUT THE
+ * SUITABILITY OF THESE LICENSED DELIVERABLES FOR ANY PURPOSE.  IT IS
+ * PROVIDED "AS IS" WITHOUT EXPRESS OR IMPLIED WARRANTY OF ANY KIND.
+ * NVIDIA DISCLAIMS ALL WARRANTIES WITH REGARD TO THESE LICENSED
+ * DELIVERABLES, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY,
+ * NONINFRINGEMENT, AND FITNESS FOR A PARTICULAR PURPOSE.
+ * NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE
+ * LICENSE AGREEMENT, IN NO EVENT SHALL NVIDIA BE LIABLE FOR ANY
+ * SPECIAL, INDIRECT, INCIDENTAL, OR CONSEQUENTIAL DAMAGES, OR ANY
+ * DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS,
+ * WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS
+ * ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE
+ * OF THESE LICENSED DELIVERABLES.
+ *
+ * U.S. Government End Users.  These Licensed Deliverables are a
+ * "commercial item" as that term is defined at 48 C.F.R. 2.101 (OCT
+ * 1995), consisting of "commercial computer software" and "commercial
+ * computer software documentation" as such terms are used in 48
+ * C.F.R. 12.212 (SEPT 1995) and is provided to the U.S. Government
+ * only as a commercial end item.  Consistent with 48 C.F.R.12.212 and
+ * 48 C.F.R. 227.7202-1 through 227.7202-4 (JUNE 1995), all
+ * U.S. Government End Users acquire the Licensed Deliverables with
+ * only those rights set forth herein.
+ *
+ * Any use of the Licensed Deliverables in individual and commercial
+ * software must include, in the user documentation and internal
+ * comments to the code, the above Disclaimer and U.S. Government End
+ * Users Notice.
+ */
+
+#if !defined(_CUPTI_EVENTS_H_)
+#define _CUPTI_EVENTS_H_
+
+#include <cuda.h>
+#include <string.h>
+#include <cuda_stdint.h>
+#include <cupti_result.h>
+
+#ifndef CUPTIAPI
+#ifdef _WIN32
+#define CUPTIAPI __stdcall
+#else
+#define CUPTIAPI
+#endif
+#endif
+
+#if defined(__cplusplus)
+extern "C" {
+#endif
+
+#if defined(__GNUC__) && defined(CUPTI_LIB)
+    #pragma GCC visibility push(default)
+#endif
+
+/**
+ * \defgroup CUPTI_EVENT_API CUPTI Event API
+ * Functions, types, and enums that implement the CUPTI Event API.
+ *
+ * \note The CUPTI event API from the header cupti_events.h is not supported on devices
+ * with compute capability 7.5 and higher (i.e. Turing and later GPU architectures).
+ * This API is deprecated in CUDA 12.8 release and will be removed in a future CUDA release.
+ * This is replaced by the host profiling API in the header cupti_profiler_host.h and
+ * target profiling API in the header cupti_range_profiler.h which are supported on
+ * devices with compute capability 7.0 and higher (i.e. Volta and later GPU architectures).
+ *
+ * @{
+ */
+
+/**
+ * \brief ID for an event.
+ *
+ * An event represents a countable activity, action, or occurrence on
+ * the device.
+ */
+typedef uint32_t CUpti_EventID;
+
+/**
+ * \brief ID for an event domain.
+ *
+ * ID for an event domain. An event domain represents a group of
+ * related events. A device may have multiple instances of a domain,
+ * indicating that the device can simultaneously record multiple
+ * instances of each event within that domain.
+ */
+typedef uint32_t CUpti_EventDomainID;
+
+/**
+ * \brief A group of events.
+ *
+ * An event group is a collection of events that are managed
+ * together. All events in an event group must belong to the same
+ * domain.
+ */
+typedef void *CUpti_EventGroup;
+
+/**
+ * \brief Device class.
+ *
+ * Enumeration of device classes for device attribute
+ * CUPTI_DEVICE_ATTR_DEVICE_CLASS.
+ */
+typedef enum {
+  CUPTI_DEVICE_ATTR_DEVICE_CLASS_TESLA              = 0,
+  CUPTI_DEVICE_ATTR_DEVICE_CLASS_QUADRO             = 1,
+  CUPTI_DEVICE_ATTR_DEVICE_CLASS_GEFORCE            = 2,
+  CUPTI_DEVICE_ATTR_DEVICE_CLASS_TEGRA              = 3,
+} CUpti_DeviceAttributeDeviceClass;
+
+/**
+ * \brief Device attributes.
+ *
+ * CUPTI device attributes. These attributes can be read using \ref
+ * cuptiDeviceGetAttribute.
+ */
+typedef enum {
+  /**
+   * Number of event IDs for a device. Value is a uint32_t.
+   */
+  CUPTI_DEVICE_ATTR_MAX_EVENT_ID                            = 1,
+  /**
+   * Number of event domain IDs for a device. Value is a uint32_t.
+   */
+  CUPTI_DEVICE_ATTR_MAX_EVENT_DOMAIN_ID                     = 2,
+  /**
+   * Get global memory bandwidth in Kbytes/sec. Value is a uint64_t.
+   */
+  CUPTI_DEVICE_ATTR_GLOBAL_MEMORY_BANDWIDTH                 = 3,
+  /**
+   * Get theoretical maximum number of instructions per cycle. Value
+   * is a uint32_t.
+   */
+  CUPTI_DEVICE_ATTR_INSTRUCTION_PER_CYCLE                   = 4,
+  /**
+   * Get theoretical maximum number of single precision instructions
+   * that can be executed per second. Value is a uint64_t.
+   */
+  CUPTI_DEVICE_ATTR_INSTRUCTION_THROUGHPUT_SINGLE_PRECISION = 5,
+  /**
+   * Get number of frame buffers for device.  Value is a uint64_t.
+   */
+  CUPTI_DEVICE_ATTR_MAX_FRAME_BUFFERS                       = 6,
+  /**
+   * Get PCIE link rate in Mega bits/sec for device. Return 0 if bus-type
+   * is non-PCIE. Value is a uint64_t.
+   */
+  CUPTI_DEVICE_ATTR_PCIE_LINK_RATE                          = 7,
+  /**
+   * Get PCIE link width for device. Return 0 if bus-type
+   * is non-PCIE. Value is a uint64_t.
+   */
+  CUPTI_DEVICE_ATTR_PCIE_LINK_WIDTH                         = 8,
+  /**
+   * Get PCIE generation for device. Return 0 if bus-type
+   * is non-PCIE. Value is a uint64_t.
+   */
+  CUPTI_DEVICE_ATTR_PCIE_GEN                                = 9,
+  /**
+   * Get the class for the device. Value is a
+   * CUpti_DeviceAttributeDeviceClass.
+   */
+  CUPTI_DEVICE_ATTR_DEVICE_CLASS                            = 10,
+  /**
+   * Get the peak single precision flop per cycle. Value is a uint64_t.
+   */
+  CUPTI_DEVICE_ATTR_FLOP_SP_PER_CYCLE                       = 11,
+  /**
+   * Get the peak double precision flop per cycle. Value is a uint64_t.
+   */
+  CUPTI_DEVICE_ATTR_FLOP_DP_PER_CYCLE                       = 12,
+  /**
+   * Get number of L2 units. Value is a uint64_t.
+   */
+  CUPTI_DEVICE_ATTR_MAX_L2_UNITS                           = 13,
+  /**
+   * Get the maximum shared memory for the CU_FUNC_CACHE_PREFER_SHARED
+   * preference. Value is a uint64_t.
+   */
+  CUPTI_DEVICE_ATTR_MAX_SHARED_MEMORY_CACHE_CONFIG_PREFER_SHARED = 14,
+  /**
+   * Get the maximum shared memory for the CU_FUNC_CACHE_PREFER_L1
+   * preference. Value is a uint64_t.
+   */
+  CUPTI_DEVICE_ATTR_MAX_SHARED_MEMORY_CACHE_CONFIG_PREFER_L1 = 15,
+  /**
+   * Get the maximum shared memory for the CU_FUNC_CACHE_PREFER_EQUAL
+   * preference. Value is a uint64_t.
+   */
+  CUPTI_DEVICE_ATTR_MAX_SHARED_MEMORY_CACHE_CONFIG_PREFER_EQUAL = 16,
+  /**
+   * Get the peak half precision flop per cycle. Value is a uint64_t.
+   */
+  CUPTI_DEVICE_ATTR_FLOP_HP_PER_CYCLE                       = 17,
+  /**
+   * Check if Nvlink is connected to device. Returns 1, if at least one
+   * Nvlink is connected to the device, returns 0 otherwise.
+   * Value is a uint32_t.
+   */
+  CUPTI_DEVICE_ATTR_NVLINK_PRESENT                          = 18,
+    /**
+   * Check if Nvlink is present between GPU and CPU. Returns Bandwidth,
+   * in Bytes/sec, if Nvlink is present, returns 0 otherwise.
+   * Value is a uint64_t.
+   */
+  CUPTI_DEVICE_ATTR_GPU_CPU_NVLINK_BW                       = 19,
+  /**
+   * Check if NVSwitch is present in the underlying topology.
+   * Returns 1, if present, returns 0 otherwise.
+   * Value is a uint32_t.
+   */
+  CUPTI_DEVICE_ATTR_NVSWITCH_PRESENT                        = 20,
+  CUPTI_DEVICE_ATTR_FORCE_INT                               = 0x7fffffff,
+} CUpti_DeviceAttribute;
+
+/**
+ * \brief Event domain attributes.
+ *
+ * Event domain attributes. Except where noted, all the attributes can
+ * be read using either \ref cuptiDeviceGetEventDomainAttribute or
+ * \ref cuptiEventDomainGetAttribute.
+ */
+typedef enum {
+  /**
+   * Event domain name. Value is a null terminated const c-string.
+   */
+  CUPTI_EVENT_DOMAIN_ATTR_NAME                 = 0,
+  /**
+   * Number of instances of the domain for which event counts will be
+   * collected.  The domain may have additional instances that cannot
+   * be profiled (see CUPTI_EVENT_DOMAIN_ATTR_TOTAL_INSTANCE_COUNT).
+   * Can be read only with \ref
+   * cuptiDeviceGetEventDomainAttribute. Value is a uint32_t.
+   */
+  CUPTI_EVENT_DOMAIN_ATTR_INSTANCE_COUNT       = 1,
+  /**
+   * Total number of instances of the domain, including instances that
+   * cannot be profiled.  Use CUPTI_EVENT_DOMAIN_ATTR_INSTANCE_COUNT
+   * to get the number of instances that can be profiled. Can be read
+   * only with \ref cuptiDeviceGetEventDomainAttribute. Value is a
+   * uint32_t.
+   */
+  CUPTI_EVENT_DOMAIN_ATTR_TOTAL_INSTANCE_COUNT = 3,
+  /**
+   * Collection method used for events contained in the event domain.
+   * Value is a \ref CUpti_EventCollectionMethod.
+   */
+  CUPTI_EVENT_DOMAIN_ATTR_COLLECTION_METHOD    = 4,
+
+  CUPTI_EVENT_DOMAIN_ATTR_FORCE_INT      = 0x7fffffff,
+} CUpti_EventDomainAttribute;
+
+/**
+ * \brief The collection method used for an event.
+ *
+ * The collection method indicates how an event is collected.
+ */
+typedef enum {
+  /**
+   * Event is collected using a hardware global performance monitor.
+   */
+  CUPTI_EVENT_COLLECTION_METHOD_PM                  = 0,
+  /**
+   * Event is collected using a hardware SM performance monitor.
+   */
+  CUPTI_EVENT_COLLECTION_METHOD_SM                  = 1,
+  /**
+   * Event is collected using software instrumentation.
+   */
+  CUPTI_EVENT_COLLECTION_METHOD_INSTRUMENTED        = 2,
+  /**
+   * Event is collected using NvLink throughput counter method.
+   */
+  CUPTI_EVENT_COLLECTION_METHOD_NVLINK_TC           = 3,
+  CUPTI_EVENT_COLLECTION_METHOD_FORCE_INT           = 0x7fffffff
+} CUpti_EventCollectionMethod;
+
+/**
+ * \brief Event group attributes.
+ *
+ * Event group attributes. These attributes can be read using \ref
+ * cuptiEventGroupGetAttribute. Attributes marked [rw] can also be
+ * written using \ref cuptiEventGroupSetAttribute.
+ */
+typedef enum {
+  /**
+   * The domain to which the event group is bound. This attribute is
+   * set when the first event is added to the group.  Value is a
+   * CUpti_EventDomainID.
+   */
+  CUPTI_EVENT_GROUP_ATTR_EVENT_DOMAIN_ID              = 0,
+  /**
+   * [rw] Profile all the instances of the domain for this
+   * eventgroup. This feature can be used to get load balancing
+   * across all instances of a domain. Value is an integer.
+   */
+  CUPTI_EVENT_GROUP_ATTR_PROFILE_ALL_DOMAIN_INSTANCES = 1,
+  /**
+   * [rw] Reserved for user data.
+   */
+  CUPTI_EVENT_GROUP_ATTR_USER_DATA                    = 2,
+  /**
+   * Number of events in the group. Value is a uint32_t.
+   */
+  CUPTI_EVENT_GROUP_ATTR_NUM_EVENTS                   = 3,
+  /**
+   * Enumerates events in the group. Value is a pointer to buffer of
+   * size sizeof(CUpti_EventID) * num_of_events in the eventgroup.
+   * num_of_events can be queried using
+   * CUPTI_EVENT_GROUP_ATTR_NUM_EVENTS.
+   */
+  CUPTI_EVENT_GROUP_ATTR_EVENTS                       = 4,
+  /**
+   * Number of instances of the domain bound to this event group that
+   * will be counted.  Value is a uint32_t.
+   */
+  CUPTI_EVENT_GROUP_ATTR_INSTANCE_COUNT               = 5,
+  /**
+   * Event group scope can be set to CUPTI_EVENT_PROFILING_SCOPE_DEVICE or
+   * CUPTI_EVENT_PROFILING_SCOPE_CONTEXT for an eventGroup, before
+   * adding any event.
+   * Sets the scope of eventgroup as CUPTI_EVENT_PROFILING_SCOPE_DEVICE or
+   * CUPTI_EVENT_PROFILING_SCOPE_CONTEXT when the scope of the events
+   * that will be added is CUPTI_EVENT_PROFILING_SCOPE_BOTH.
+   * If profiling scope of event is either
+   * CUPTI_EVENT_PROFILING_SCOPE_DEVICE or CUPTI_EVENT_PROFILING_SCOPE_CONTEXT
+   * then setting this attribute will not affect the default scope.
+   * It is not allowed to add events of different scope to same eventgroup.
+   * Value is a uint32_t.
+   */
+  CUPTI_EVENT_GROUP_ATTR_PROFILING_SCOPE               = 6,
+  CUPTI_EVENT_GROUP_ATTR_FORCE_INT                     = 0x7fffffff,
+} CUpti_EventGroupAttribute;
+
+/**
+* \brief Profiling scope for event.
+*
+* Profiling scope of event indicates if the event can be collected at context
+* scope or device scope or both i.e. it can be collected at any of context or
+* device scope.
+*/
+typedef enum {
+  /**
+   * Event is collected at context scope.
+   */
+  CUPTI_EVENT_PROFILING_SCOPE_CONTEXT                 = 0,
+  /**
+   * Event is collected at device scope.
+   */
+  CUPTI_EVENT_PROFILING_SCOPE_DEVICE                  = 1,
+  /**
+   * Event can be collected at device or context scope.
+   * The scope can be set using \ref cuptiEventGroupSetAttribute API.
+   */
+  CUPTI_EVENT_PROFILING_SCOPE_BOTH                    = 2,
+  CUPTI_EVENT_PROFILING_SCOPE_FORCE_INT               = 0x7fffffff
+} CUpti_EventProfilingScope;
+
+/**
+ * \brief Event attributes.
+ *
+ * Event attributes. These attributes can be read using \ref
+ * cuptiEventGetAttribute.
+ */
+typedef enum {
+  /**
+   * Event name. Value is a null terminated const c-string.
+   */
+  CUPTI_EVENT_ATTR_NAME              = 0,
+  /**
+   * Short description of event. Value is a null terminated const
+   * c-string.
+   */
+  CUPTI_EVENT_ATTR_SHORT_DESCRIPTION = 1,
+  /**
+   * Long description of event. Value is a null terminated const
+   * c-string.
+   */
+  CUPTI_EVENT_ATTR_LONG_DESCRIPTION  = 2,
+  /**
+   * Category of event. Value is CUpti_EventCategory.
+   */
+  CUPTI_EVENT_ATTR_CATEGORY          = 3,
+  /**
+   * Profiling scope of the events. It can be either device or context or both.
+   * Value is a \ref CUpti_EventProfilingScope.
+   */
+  CUPTI_EVENT_ATTR_PROFILING_SCOPE   = 5,
+
+  CUPTI_EVENT_ATTR_FORCE_INT         = 0x7fffffff,
+} CUpti_EventAttribute;
+
+/**
+ * \brief Event collection modes.
+ *
+ * The event collection mode determines the period over which the
+ * events within the enabled event groups will be collected.
+ */
+typedef enum {
+  /**
+   * Events are collected for the entire duration between the
+   * cuptiEventGroupEnable and cuptiEventGroupDisable calls.
+   * Event values are reset when the events are read.
+   * For CUDA toolkit v6.0 and older this was the default mode.
+   */
+  CUPTI_EVENT_COLLECTION_MODE_CONTINUOUS          = 0,
+  /**
+   * Events are collected only for the durations of kernel executions
+   * that occur between the cuptiEventGroupEnable and
+   * cuptiEventGroupDisable calls. Event collection begins when a
+   * kernel execution begins, and stops when kernel execution
+   * completes. Event values are reset to zero when each kernel
+   * execution begins. If multiple kernel executions occur between the
+   * cuptiEventGroupEnable and cuptiEventGroupDisable calls then the
+   * event values must be read after each kernel launch if those
+   * events need to be associated with the specific kernel launch.
+   * Note that collection in this mode may significantly change the
+   * overall performance characteristics of the application because
+   * kernel executions that occur between the cuptiEventGroupEnable and
+   * cuptiEventGroupDisable calls are serialized on the GPU.
+   * This is the default mode from CUDA toolkit v6.5
+   */
+  CUPTI_EVENT_COLLECTION_MODE_KERNEL              = 1,
+  CUPTI_EVENT_COLLECTION_MODE_FORCE_INT           = 0x7fffffff
+} CUpti_EventCollectionMode;
+
+/**
+ * \brief An event category.
+ *
+ * Each event is assigned to a category that represents the general
+ * type of the event. A event's category is accessed using \ref
+ * cuptiEventGetAttribute and the CUPTI_EVENT_ATTR_CATEGORY attribute.
+ */
+typedef enum {
+  /**
+   * An instruction related event.
+   */
+  CUPTI_EVENT_CATEGORY_INSTRUCTION     = 0,
+  /**
+   * A memory related event.
+   */
+  CUPTI_EVENT_CATEGORY_MEMORY          = 1,
+  /**
+   * A cache related event.
+   */
+  CUPTI_EVENT_CATEGORY_CACHE           = 2,
+  /**
+   * A profile-trigger event.
+   */
+  CUPTI_EVENT_CATEGORY_PROFILE_TRIGGER = 3,
+  /**
+   * A system event.
+   */
+  CUPTI_EVENT_CATEGORY_SYSTEM  = 4,
+  CUPTI_EVENT_CATEGORY_FORCE_INT       = 0x7fffffff
+} CUpti_EventCategory;
+
+/**
+ * \brief The overflow value for a CUPTI event.
+ *
+ * The CUPTI event value that indicates an overflow.
+ */
+#define CUPTI_EVENT_OVERFLOW ((uint64_t)0xFFFFFFFFFFFFFFFFULL)
+
+/**
+ * \brief The value that indicates the event value is invalid
+ */
+#define CUPTI_EVENT_INVALID ((uint64_t)0xFFFFFFFFFFFFFFFEULL)
+
+/**
+ * \brief Flags for cuptiEventGroupReadEvent an
+ * cuptiEventGroupReadAllEvents.
+ *
+ * Flags for \ref cuptiEventGroupReadEvent an \ref
+ * cuptiEventGroupReadAllEvents.
+ */
+typedef enum {
+  /**
+   * No flags.
+   */
+  CUPTI_EVENT_READ_FLAG_NONE          = 0,
+  CUPTI_EVENT_READ_FLAG_FORCE_INT     = 0x7fffffff,
+} CUpti_ReadEventFlags;
+
+
+/**
+ * \brief A set of event groups.
+ *
+ * A set of event groups. When returned by \ref
+ * cuptiEventGroupSetsCreate and \ref cuptiMetricCreateEventGroupSets
+ * a set indicates that event groups that can be enabled at the same
+ * time (i.e. all the events in the set can be collected
+ * simultaneously).
+ */
+typedef struct {
+  /**
+   * The number of event groups in the set.
+   */
+  uint32_t numEventGroups;
+  /**
+   * An array of \p numEventGroups event groups.
+   */
+  CUpti_EventGroup *eventGroups;
+} CUpti_EventGroupSet;
+
+/**
+ * \brief A set of event group sets.
+ *
+ * A set of event group sets. When returned by \ref
+ * cuptiEventGroupSetsCreate and \ref cuptiMetricCreateEventGroupSets
+ * a CUpti_EventGroupSets indicates the number of passes required to
+ * collect all the events, and the event groups that should be
+ * collected during each pass.
+ */
+typedef struct {
+  /**
+   * Number of event group sets.
+   */
+  uint32_t numSets;
+  /**
+   * An array of \p numSets event group sets.
+   */
+  CUpti_EventGroupSet *sets;
+} CUpti_EventGroupSets;
+
+/**
+ * \brief Set the event collection mode.
+ *
+ * Set the event collection mode for a \p context.  The \p mode
+ * controls the event collection behavior of all events in event
+ * groups created in the \p context. This API is invalid in kernel
+ * replay mode.
+ * \note \b Thread-safety: this function is thread safe.
+ *
+ * \param context The context
+ * \param mode The event collection mode
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_NOT_INITIALIZED
+ * \retval CUPTI_ERROR_INVALID_CONTEXT
+ * \retval CUPTI_ERROR_INVALID_OPERATION if called when replay mode is enabled
+ * \retval CUPTI_ERROR_NOT_SUPPORTED if mode is not supported on the device
+ */
+
+CUptiResult CUPTIAPI cuptiSetEventCollectionMode(CUcontext context,
+                                                 CUpti_EventCollectionMode mode);
+
+/**
+ * \brief Read a device attribute.
+ *
+ * Read a device attribute and return it in \p *value.
+ * \note \b Thread-safety: this function is thread safe.
+ *
+ * \param device The CUDA device
+ * \param attrib The attribute to read
+ * \param valueSize Size of buffer pointed by the value, and
+ * returns the number of bytes written to \p value
+ * \param value Returns the value of the attribute
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_NOT_INITIALIZED
+ * \retval CUPTI_ERROR_INVALID_DEVICE
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p valueSize or \p value
+ * is NULL, or if \p attrib is not a device attribute
+ * \retval CUPTI_ERROR_PARAMETER_SIZE_NOT_SUFFICIENT For non-c-string
+ * attribute values, indicates that the \p value buffer is too small
+ * to hold the attribute value.
+ */
+CUptiResult CUPTIAPI cuptiDeviceGetAttribute(CUdevice device,
+                                             CUpti_DeviceAttribute attrib,
+                                             size_t *valueSize,
+                                             void *value);
+
+/**
+ * \brief Get the number of domains for a device.
+ *
+ * Returns the number of domains in \p numDomains for a device.
+ * \note \b Thread-safety: this function is thread safe.
+ *
+ * \param device The CUDA device
+ * \param numDomains Returns the number of domains
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_NOT_INITIALIZED
+ * \retval CUPTI_ERROR_INVALID_DEVICE
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p numDomains is NULL
+ */
+CUptiResult CUPTIAPI cuptiDeviceGetNumEventDomains(CUdevice device,
+                                                   uint32_t *numDomains);
+
+/**
+ * \brief Get the event domains for a device.
+ *
+ * Returns the event domains IDs in \p domainArray for a device.  The
+ * size of the \p domainArray buffer is given by \p
+ * *arraySizeBytes. The size of the \p domainArray buffer must be at
+ * least \p numdomains * sizeof(CUpti_EventDomainID) or else all
+ * domains will not be returned. The value returned in \p
+ * *arraySizeBytes contains the number of bytes returned in \p
+ * domainArray.
+ * \note \b Thread-safety: this function is thread safe.
+ *
+ * \param device The CUDA device
+ * \param arraySizeBytes The size of \p domainArray in bytes, and
+ * returns the number of bytes written to \p domainArray
+ * \param domainArray Returns the IDs of the event domains for the device
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_NOT_INITIALIZED
+ * \retval CUPTI_ERROR_INVALID_DEVICE
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p arraySizeBytes or
+ * \p domainArray are NULL
+ */
+CUptiResult CUPTIAPI cuptiDeviceEnumEventDomains(CUdevice device,
+                                                 size_t *arraySizeBytes,
+                                                 CUpti_EventDomainID *domainArray);
+
+/**
+ * \brief Read an event domain attribute.
+ *
+ * Returns an event domain attribute in \p *value. The size of the \p
+ * value buffer is given by \p *valueSize. The value returned in \p
+ * *valueSize contains the number of bytes returned in \p value.
+ *
+ * If the attribute value is a c-string that is longer than \p
+ * *valueSize, then only the first \p *valueSize characters will be
+ * returned and there will be no terminating null byte.
+ * \note \b Thread-safety: this function is thread safe.
+ *
+ * \param device The CUDA device
+ * \param eventDomain ID of the event domain
+ * \param attrib The event domain attribute to read
+ * \param valueSize The size of the \p value buffer in bytes, and
+ * returns the number of bytes written to \p value
+ * \param value Returns the attribute's value
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_NOT_INITIALIZED
+ * \retval CUPTI_ERROR_INVALID_DEVICE
+ * \retval CUPTI_ERROR_INVALID_EVENT_DOMAIN_ID
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p valueSize or \p value
+ * is NULL, or if \p attrib is not an event domain attribute
+ * \retval CUPTI_ERROR_PARAMETER_SIZE_NOT_SUFFICIENT For non-c-string
+ * attribute values, indicates that the \p value buffer is too small
+ * to hold the attribute value.
+ */
+CUptiResult CUPTIAPI cuptiDeviceGetEventDomainAttribute(CUdevice device,
+                                                        CUpti_EventDomainID eventDomain,
+                                                        CUpti_EventDomainAttribute attrib,
+                                                        size_t *valueSize,
+                                                        void *value);
+
+/**
+ * \brief Get the number of event domains available on any device.
+ *
+ * Returns the total number of event domains available on any
+ * CUDA-capable device.
+ * \note \b Thread-safety: this function is thread safe.
+ *
+ * \param numDomains Returns the number of domains
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p numDomains is NULL
+ */
+CUptiResult CUPTIAPI cuptiGetNumEventDomains(uint32_t *numDomains);
+
+/**
+ * \brief Get the event domains available on any device.
+ *
+ * Returns all the event domains available on any CUDA-capable device.
+ * Event domain IDs are returned in \p domainArray. The size of the \p
+ * domainArray buffer is given by \p *arraySizeBytes. The size of the
+ * \p domainArray buffer must be at least \p numDomains *
+ * sizeof(CUpti_EventDomainID) or all domains will not be
+ * returned. The value returned in \p *arraySizeBytes contains the
+ * number of bytes returned in \p domainArray.
+ * \note \b Thread-safety: this function is thread safe.
+ *
+ * \param arraySizeBytes The size of \p domainArray in bytes, and
+ * returns the number of bytes written to \p domainArray
+ * \param domainArray Returns all the event domains
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p arraySizeBytes or
+ * \p domainArray are NULL
+ */
+CUptiResult CUPTIAPI cuptiEnumEventDomains(size_t *arraySizeBytes,
+                                           CUpti_EventDomainID *domainArray);
+
+/**
+ * \brief Read an event domain attribute.
+ *
+ * Returns an event domain attribute in \p *value. The size of the \p
+ * value buffer is given by \p *valueSize. The value returned in \p
+ * *valueSize contains the number of bytes returned in \p value.
+ *
+ * If the attribute value is a c-string that is longer than \p
+ * *valueSize, then only the first \p *valueSize characters will be
+ * returned and there will be no terminating null byte.
+ * \note \b Thread-safety: this function is thread safe.
+ *
+ * \param eventDomain ID of the event domain
+ * \param attrib The event domain attribute to read
+ * \param valueSize The size of the \p value buffer in bytes, and
+ * returns the number of bytes written to \p value
+ * \param value Returns the attribute's value
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_NOT_INITIALIZED
+ * \retval CUPTI_ERROR_INVALID_EVENT_DOMAIN_ID
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p valueSize or \p value
+ * is NULL, or if \p attrib is not an event domain attribute
+ * \retval CUPTI_ERROR_PARAMETER_SIZE_NOT_SUFFICIENT For non-c-string
+ * attribute values, indicates that the \p value buffer is too small
+ * to hold the attribute value.
+ */
+CUptiResult CUPTIAPI cuptiEventDomainGetAttribute(CUpti_EventDomainID eventDomain,
+                                                  CUpti_EventDomainAttribute attrib,
+                                                  size_t *valueSize,
+                                                  void *value);
+
+/**
+ * \brief Get number of events in a domain.
+ *
+ * Returns the number of events in \p numEvents for a domain.
+ * \note \b Thread-safety: this function is thread safe.
+ *
+ * \param eventDomain ID of the event domain
+ * \param numEvents Returns the number of events in the domain
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_NOT_INITIALIZED
+ * \retval CUPTI_ERROR_INVALID_EVENT_DOMAIN_ID
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p numEvents is NULL
+ */
+CUptiResult CUPTIAPI cuptiEventDomainGetNumEvents(CUpti_EventDomainID eventDomain,
+                                                  uint32_t *numEvents);
+
+/**
+ * \brief Get the events in a domain.
+ *
+ * Returns the event IDs in \p eventArray for a domain.  The size of
+ * the \p eventArray buffer is given by \p *arraySizeBytes. The size
+ * of the \p eventArray buffer must be at least \p numdomainevents *
+ * sizeof(CUpti_EventID) or else all events will not be returned. The
+ * value returned in \p *arraySizeBytes contains the number of bytes
+ * returned in \p eventArray.
+ * \note \b Thread-safety: this function is thread safe.
+ *
+ * \param eventDomain ID of the event domain
+ * \param arraySizeBytes The size of \p eventArray in bytes, and
+ * returns the number of bytes written to \p eventArray
+ * \param eventArray Returns the IDs of the events in the domain
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_NOT_INITIALIZED
+ * \retval CUPTI_ERROR_INVALID_EVENT_DOMAIN_ID
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p arraySizeBytes or \p
+ * eventArray are NULL
+ */
+CUptiResult CUPTIAPI cuptiEventDomainEnumEvents(CUpti_EventDomainID eventDomain,
+                                                size_t *arraySizeBytes,
+                                                CUpti_EventID *eventArray);
+
+/**
+ * \brief Get an event attribute.
+ *
+ * Returns an event attribute in \p *value. The size of the \p
+ * value buffer is given by \p *valueSize. The value returned in \p
+ * *valueSize contains the number of bytes returned in \p value.
+ *
+ * If the attribute value is a c-string that is longer than \p
+ * *valueSize, then only the first \p *valueSize characters will be
+ * returned and there will be no terminating null byte.
+ * \note \b Thread-safety: this function is thread safe.
+ *
+ * \param event ID of the event
+ * \param attrib The event attribute to read
+ * \param valueSize The size of the \p value buffer in bytes, and
+ * returns the number of bytes written to \p value
+ * \param value Returns the attribute's value
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_NOT_INITIALIZED
+ * \retval CUPTI_ERROR_INVALID_EVENT_ID
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p valueSize or \p value
+ * is NULL, or if \p attrib is not an event attribute
+ * \retval CUPTI_ERROR_PARAMETER_SIZE_NOT_SUFFICIENT For non-c-string
+ * attribute values, indicates that the \p value buffer is too small
+ * to hold the attribute value.
+ */
+CUptiResult CUPTIAPI cuptiEventGetAttribute(CUpti_EventID event,
+                                            CUpti_EventAttribute attrib,
+                                            size_t *valueSize,
+                                            void *value);
+
+/**
+ * \brief Find an event by name.
+ *
+ * Find an event by name and return the event ID in \p *event.
+ * \note \b Thread-safety: this function is thread safe.
+ *
+ * \param device The CUDA device
+ * \param eventName The name of the event to find
+ * \param event Returns the ID of the found event or undefined if
+ * unable to find the event
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_NOT_INITIALIZED
+ * \retval CUPTI_ERROR_INVALID_DEVICE
+ * \retval CUPTI_ERROR_INVALID_EVENT_NAME if unable to find an event
+ * with name \p eventName. In this case \p *event is undefined
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p eventName or \p event are NULL
+ */
+CUptiResult CUPTIAPI cuptiEventGetIdFromName(CUdevice device,
+                                             const char *eventName,
+                                             CUpti_EventID *event);
+
+/**
+ * \brief Create a new event group for a context.
+ *
+ * Creates a new event group for \p context and returns the new group
+ * in \p *eventGroup.
+ * \note \p flags are reserved for future use and should be set to zero.
+ * \note \b Thread-safety: this function is thread safe.
+ *
+ * \param context The context for the event group
+ * \param eventGroup Returns the new event group
+ * \param flags Reserved - must be zero
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_NOT_INITIALIZED
+ * \retval CUPTI_ERROR_INVALID_CONTEXT
+ * \retval CUPTI_ERROR_OUT_OF_MEMORY
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p eventGroup is NULL
+ */
+CUptiResult CUPTIAPI cuptiEventGroupCreate(CUcontext context,
+                                           CUpti_EventGroup *eventGroup,
+                                           uint32_t flags);
+
+/**
+ * \brief Destroy an event group.
+ *
+ * Destroy an \p eventGroup and free its resources. An event group
+ * cannot be destroyed if it is enabled.
+ * \note \b Thread-safety: this function is thread safe.
+ *
+ * \param eventGroup The event group to destroy
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_NOT_INITIALIZED
+ * \retval CUPTI_ERROR_INVALID_OPERATION if the event group is enabled
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if eventGroup is NULL
+ */
+CUptiResult CUPTIAPI cuptiEventGroupDestroy(CUpti_EventGroup eventGroup);
+
+/**
+ * \brief Read an event group attribute.
+ *
+ * Read an event group attribute and return it in \p *value.
+ * \note \b Thread-safety: this function is thread safe but client
+ * must guard against simultaneous destruction or modification of \p
+ * eventGroup (for example, client must guard against simultaneous
+ * calls to \ref cuptiEventGroupDestroy, \ref cuptiEventGroupAddEvent,
+ * etc.), and must guard against simultaneous destruction of the
+ * context in which \p eventGroup was created (for example, client
+ * must guard against simultaneous calls to cudaDeviceReset,
+ * cuCtxDestroy, etc.).
+ *
+ * \param eventGroup The event group
+ * \param attrib The attribute to read
+ * \param valueSize Size of buffer pointed by the value, and
+ * returns the number of bytes written to \p value
+ * \param value Returns the value of the attribute
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_NOT_INITIALIZED
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p valueSize or \p value
+ * is NULL, or if \p attrib is not an eventgroup attribute
+ * \retval CUPTI_ERROR_PARAMETER_SIZE_NOT_SUFFICIENT For non-c-string
+ * attribute values, indicates that the \p value buffer is too small
+ * to hold the attribute value.
+ */
+CUptiResult CUPTIAPI cuptiEventGroupGetAttribute(CUpti_EventGroup eventGroup,
+                                                 CUpti_EventGroupAttribute attrib,
+                                                 size_t *valueSize,
+                                                 void *value);
+
+/**
+ * \brief Write an event group attribute.
+ *
+ * Write an event group attribute.
+ * \note \b Thread-safety: this function is thread safe.
+ *
+ * \param eventGroup The event group
+ * \param attrib The attribute to write
+ * \param valueSize The size, in bytes, of the value
+ * \param value The attribute value to write
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_NOT_INITIALIZED
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p valueSize or \p value
+ * is NULL, or if \p attrib is not an event group attribute, or if
+ * \p attrib is not a writable attribute
+ * \retval CUPTI_ERROR_PARAMETER_SIZE_NOT_SUFFICIENT Indicates that
+ * the \p value buffer is too small to hold the attribute value.
+ */
+CUptiResult CUPTIAPI cuptiEventGroupSetAttribute(CUpti_EventGroup eventGroup,
+                                                 CUpti_EventGroupAttribute attrib,
+                                                 size_t valueSize,
+                                                 void *value);
+
+/**
+ * \brief Add an event to an event group.
+ *
+ * Add an event to an event group. The event add can fail for a number of reasons:
+ * \li The event group is enabled
+ * \li The event does not belong to the same event domain as the
+ * events that are already in the event group
+ * \li Device limitations on the events that can belong to the same group
+ * \li The event group is full
+ *
+ * \note \b Thread-safety: this function is thread safe.
+ *
+ * \param eventGroup The event group
+ * \param event The event to add to the group
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_NOT_INITIALIZED
+ * \retval CUPTI_ERROR_INVALID_EVENT_ID
+ * \retval CUPTI_ERROR_OUT_OF_MEMORY
+ * \retval CUPTI_ERROR_INVALID_OPERATION if \p eventGroup is enabled
+ * \retval CUPTI_ERROR_NOT_COMPATIBLE if \p event belongs to a
+ * different event domain than the events already in \p eventGroup, or
+ * if a device limitation prevents \p event from being collected at
+ * the same time as the events already in \p eventGroup
+ * \retval CUPTI_ERROR_MAX_LIMIT_REACHED if \p eventGroup is full
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p eventGroup is NULL
+ */
+CUptiResult CUPTIAPI cuptiEventGroupAddEvent(CUpti_EventGroup eventGroup,
+                                             CUpti_EventID event);
+
+/**
+ * \brief Remove an event from an event group.
+ *
+ * Remove \p event from the an event group. The event cannot be
+ * removed if the event group is enabled.
+ * \note \b Thread-safety: this function is thread safe.
+ *
+ * \param eventGroup The event group
+ * \param event The event to remove from the group
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_NOT_INITIALIZED
+ * \retval CUPTI_ERROR_INVALID_EVENT_ID
+ * \retval CUPTI_ERROR_INVALID_OPERATION if \p eventGroup is enabled
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p eventGroup is NULL
+ */
+CUptiResult CUPTIAPI cuptiEventGroupRemoveEvent(CUpti_EventGroup eventGroup,
+                                                CUpti_EventID event);
+
+/**
+ * \brief Remove all events from an event group.
+ *
+ * Remove all events from an event group. Events cannot be removed if
+ * the event group is enabled.
+ * \note \b Thread-safety: this function is thread safe.
+ *
+ * \param eventGroup The event group
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_NOT_INITIALIZED
+ * \retval CUPTI_ERROR_INVALID_OPERATION if \p eventGroup is enabled
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p eventGroup is NULL
+ */
+CUptiResult CUPTIAPI cuptiEventGroupRemoveAllEvents(CUpti_EventGroup eventGroup);
+
+/**
+ * \brief Zero all the event counts in an event group.
+ *
+ * Zero all the event counts in an event group.
+ * \note \b Thread-safety: this function is thread safe but client
+ * must guard against simultaneous destruction or modification of \p
+ * eventGroup (for example, client must guard against simultaneous
+ * calls to \ref cuptiEventGroupDestroy, \ref cuptiEventGroupAddEvent,
+ * etc.), and must guard against simultaneous destruction of the
+ * context in which \p eventGroup was created (for example, client
+ * must guard against simultaneous calls to cudaDeviceReset,
+ * cuCtxDestroy, etc.).
+ *
+ * \param eventGroup The event group
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_NOT_INITIALIZED
+ * \retval CUPTI_ERROR_HARDWARE
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p eventGroup is NULL
+ */
+CUptiResult CUPTIAPI cuptiEventGroupResetAllEvents(CUpti_EventGroup eventGroup);
+
+/**
+ * \brief Enable an event group.
+ *
+ * Enable an event group. Enabling an event group zeros the value of
+ * all the events in the group and then starts collection of those
+ * events.
+ * \note \b Thread-safety: this function is thread safe.
+ *
+ * \param eventGroup The event group
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_NOT_INITIALIZED
+ * \retval CUPTI_ERROR_HARDWARE
+ * \retval CUPTI_ERROR_NOT_READY if \p eventGroup does not contain any events
+ * \retval CUPTI_ERROR_NOT_COMPATIBLE if \p eventGroup cannot be
+ * enabled due to other already enabled event groups
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p eventGroup is NULL
+ * \retval CUPTI_ERROR_HARDWARE_BUSY if another client is profiling
+ * and hardware is busy
+ */
+CUptiResult CUPTIAPI cuptiEventGroupEnable(CUpti_EventGroup eventGroup);
+
+/**
+ * \brief Disable an event group.
+ *
+ * Disable an event group. Disabling an event group stops collection
+ * of events contained in the group.
+ * \note \b Thread-safety: this function is thread safe.
+ *
+ * \param eventGroup The event group
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_NOT_INITIALIZED
+ * \retval CUPTI_ERROR_HARDWARE
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p eventGroup is NULL
+ */
+CUptiResult CUPTIAPI cuptiEventGroupDisable(CUpti_EventGroup eventGroup);
+
+/**
+ * \brief Read the value for an event in an event group.
+ *
+ * Read the value for an event in an event group. The event value is
+ * returned in the \p eventValueBuffer buffer. \p
+ * eventValueBufferSizeBytes indicates the size of the \p
+ * eventValueBuffer buffer. The buffer must be at least sizeof(uint64)
+ * if ::CUPTI_EVENT_GROUP_ATTR_PROFILE_ALL_DOMAIN_INSTANCES is not set
+ * on the group containing the event.  The buffer must be at least
+ * (sizeof(uint64) * number of domain instances) if
+ * ::CUPTI_EVENT_GROUP_ATTR_PROFILE_ALL_DOMAIN_INSTANCES is set on the
+ * group.
+ *
+ * If any instance of an event counter overflows, the value returned
+ * for that event instance will be ::CUPTI_EVENT_OVERFLOW.
+ *
+ * The only allowed value for \p flags is ::CUPTI_EVENT_READ_FLAG_NONE.
+ *
+ * Reading an event from a disabled event group is not allowed. After
+ * being read, an event's value is reset to zero.
+ * \note \b Thread-safety: this function is thread safe but client
+ * must guard against simultaneous destruction or modification of \p
+ * eventGroup (for example, client must guard against simultaneous
+ * calls to \ref cuptiEventGroupDestroy, \ref cuptiEventGroupAddEvent,
+ * etc.), and must guard against simultaneous destruction of the
+ * context in which \p eventGroup was created (for example, client
+ * must guard against simultaneous calls to cudaDeviceReset,
+ * cuCtxDestroy, etc.). If \ref cuptiEventGroupResetAllEvents is
+ * called simultaneously with this function, then returned event
+ * values are undefined.
+ *
+ * \param eventGroup The event group
+ * \param flags Flags controlling the reading mode
+ * \param event The event to read
+ * \param eventValueBufferSizeBytes The size of \p eventValueBuffer
+ * in bytes, and returns the number of bytes written to \p
+ * eventValueBuffer
+ * \param eventValueBuffer Returns the event value(s)
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_NOT_INITIALIZED
+ * \retval CUPTI_ERROR_INVALID_EVENT_ID
+ * \retval CUPTI_ERROR_HARDWARE
+ * \retval CUPTI_ERROR_INVALID_OPERATION if \p eventGroup is disabled
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p eventGroup, \p
+ * eventValueBufferSizeBytes or \p eventValueBuffer is NULL
+ * \retval CUPTI_ERROR_PARAMETER_SIZE_NOT_SUFFICIENT if size of \p eventValueBuffer
+ * is not sufficient
+ */
+CUptiResult CUPTIAPI cuptiEventGroupReadEvent(CUpti_EventGroup eventGroup,
+                                              CUpti_ReadEventFlags flags,
+                                              CUpti_EventID event,
+                                              size_t *eventValueBufferSizeBytes,
+                                              uint64_t *eventValueBuffer);
+
+/**
+ * \brief Read the values for all the events in an event group.
+ *
+ * Read the values for all the events in an event group. The event
+ * values are returned in the \p eventValueBuffer buffer. \p
+ * eventValueBufferSizeBytes indicates the size of \p
+ * eventValueBuffer.  The buffer must be at least (sizeof(uint64) *
+ * number of events in group) if
+ * ::CUPTI_EVENT_GROUP_ATTR_PROFILE_ALL_DOMAIN_INSTANCES is not set on
+ * the group containing the events.  The buffer must be at least
+ * (sizeof(uint64) * number of domain instances * number of events in
+ * group) if ::CUPTI_EVENT_GROUP_ATTR_PROFILE_ALL_DOMAIN_INSTANCES is
+ * set on the group.
+ *
+ * The data format returned in \p eventValueBuffer is:
+ *    - domain instance 0: event0 event1 ... eventN
+ *    - domain instance 1: event0 event1 ... eventN
+ *    - ...
+ *    - domain instance M: event0 event1 ... eventN
+ *
+ * The event order in \p eventValueBuffer is returned in \p
+ * eventIdArray. The size of \p eventIdArray is specified in \p
+ * eventIdArraySizeBytes. The size should be at least
+ * (sizeof(CUpti_EventID) * number of events in group).
+ *
+ * If any instance of any event counter overflows, the value returned
+ * for that event instance will be ::CUPTI_EVENT_OVERFLOW.
+ *
+ * The only allowed value for \p flags is ::CUPTI_EVENT_READ_FLAG_NONE.
+ *
+ * Reading events from a disabled event group is not allowed. After
+ * being read, an event's value is reset to zero.
+ * \note \b Thread-safety: this function is thread safe but client
+ * must guard against simultaneous destruction or modification of \p
+ * eventGroup (for example, client must guard against simultaneous
+ * calls to \ref cuptiEventGroupDestroy, \ref cuptiEventGroupAddEvent,
+ * etc.), and must guard against simultaneous destruction of the
+ * context in which \p eventGroup was created (for example, client
+ * must guard against simultaneous calls to cudaDeviceReset,
+ * cuCtxDestroy, etc.). If \ref cuptiEventGroupResetAllEvents is
+ * called simultaneously with this function, then returned event
+ * values are undefined.
+ *
+ * \param eventGroup The event group
+ * \param flags Flags controlling the reading mode
+ * \param eventValueBufferSizeBytes The size of \p eventValueBuffer in
+ * bytes, and returns the number of bytes written to \p
+ * eventValueBuffer
+ * \param eventValueBuffer Returns the event values
+ * \param eventIdArraySizeBytes The size of \p eventIdArray in bytes,
+ * and returns the number of bytes written to \p eventIdArray
+ * \param eventIdArray Returns the IDs of the events in the same order
+ * as the values return in eventValueBuffer.
+ * \param numEventIdsRead Returns the number of event IDs returned
+ * in \p eventIdArray
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_NOT_INITIALIZED
+ * \retval CUPTI_ERROR_HARDWARE
+ * \retval CUPTI_ERROR_INVALID_OPERATION if \p eventGroup is disabled
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p eventGroup, \p
+ * eventValueBufferSizeBytes, \p eventValueBuffer, \p
+ * eventIdArraySizeBytes, \p eventIdArray or \p numEventIdsRead is
+ * NULL
+ * \retval CUPTI_ERROR_PARAMETER_SIZE_NOT_SUFFICIENT if size of \p eventValueBuffer
+ * or \p eventIdArray is not sufficient
+ */
+CUptiResult CUPTIAPI cuptiEventGroupReadAllEvents(CUpti_EventGroup       eventGroup,
+                                                  CUpti_ReadEventFlags   flags,
+                                                  size_t                 *eventValueBufferSizeBytes,
+                                                  uint64_t               *eventValueBuffer,
+                                                  size_t                 *eventIdArraySizeBytes,
+                                                  CUpti_EventID          *eventIdArray,
+                                                  size_t                 *numEventIdsRead);
+
+/**
+ * \brief For a set of events, get the grouping that indicates the
+ * number of passes and the event groups necessary to collect the
+ * events.
+ *
+ * The number of events that can be collected simultaneously varies by
+ * device and by the type of the events. When events can be collected
+ * simultaneously, they may need to be grouped into multiple event
+ * groups because they are from different event domains. This function
+ * takes a set of events and determines how many passes are required
+ * to collect all those events, and which events can be collected
+ * simultaneously in each pass.
+ *
+ * The CUpti_EventGroupSets returned in \p eventGroupPasses indicates
+ * how many passes are required to collect the events with the \p
+ * numSets field. Within each event group set, the \p sets array
+ * indicates the event groups that should be collected on each pass.
+ * \note \b Thread-safety: this function is thread safe, but client
+ * must guard against another thread simultaneously destroying \p
+ * context.
+ *
+ * \param context The context for event collection
+ * \param eventIdArraySizeBytes Size of \p eventIdArray in bytes
+ * \param eventIdArray Array of event IDs that need to be grouped
+ * \param eventGroupPasses Returns a CUpti_EventGroupSets object that
+ * indicates the number of passes required to collect the events and
+ * the events to collect on each pass
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_NOT_INITIALIZED
+ * \retval CUPTI_ERROR_INVALID_CONTEXT
+ * \retval CUPTI_ERROR_INVALID_EVENT_ID
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p eventIdArray or
+ * \p eventGroupPasses is NULL
+ */
+CUptiResult CUPTIAPI cuptiEventGroupSetsCreate(CUcontext context,
+                                               size_t eventIdArraySizeBytes,
+                                               CUpti_EventID *eventIdArray,
+                                               CUpti_EventGroupSets **eventGroupPasses);
+
+/**
+ * \brief Destroy a event group sets object.
+ *
+ * Destroy a CUpti_EventGroupSets object.
+ * \note \b Thread-safety: this function is thread safe.
+ *
+ * \param eventGroupSets The object to destroy
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_NOT_INITIALIZED
+ * \retval CUPTI_ERROR_INVALID_OPERATION if any of the event groups
+ * contained in the sets is enabled
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p eventGroupSets is NULL
+ */
+CUptiResult CUPTIAPI cuptiEventGroupSetsDestroy(CUpti_EventGroupSets *eventGroupSets);
+
+
+/**
+ * \brief Enable an event group set.
+ *
+ * Enable a set of event groups. Enabling a set of event groups zeros the value of
+ * all the events in all the groups and then starts collection of those events.
+ * \note \b Thread-safety: this function is thread safe.
+ *
+ * \param eventGroupSet The pointer to the event group set
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_NOT_INITIALIZED
+ * \retval CUPTI_ERROR_HARDWARE
+ * \retval CUPTI_ERROR_NOT_READY if \p eventGroup does not contain any events
+ * \retval CUPTI_ERROR_NOT_COMPATIBLE if \p eventGroup cannot be
+ * enabled due to other already enabled event groups
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p eventGroupSet is NULL
+ * \retval CUPTI_ERROR_HARDWARE_BUSY if other client is profiling and hardware is
+ * busy
+ */
+CUptiResult CUPTIAPI cuptiEventGroupSetEnable(CUpti_EventGroupSet *eventGroupSet);
+
+/**
+ * \brief Disable an event group set.
+ *
+ * Disable a set of event groups. Disabling a set of event groups
+ * stops collection of events contained in the groups.
+ * \note \b Thread-safety: this function is thread safe.
+ * \note \b If this call fails, some of the event groups in the set may be disabled
+ * and other event groups may remain enabled.
+ *
+ * \param eventGroupSet The pointer to the event group set
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_NOT_INITIALIZED
+ * \retval CUPTI_ERROR_HARDWARE
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p eventGroupSet is NULL
+ */
+CUptiResult CUPTIAPI cuptiEventGroupSetDisable(CUpti_EventGroupSet *eventGroupSet);
+
+/**
+ * \brief Enable kernel replay mode.
+ *
+ * Set profiling mode for the context to replay mode. In this mode,
+ * any number of events can be collected in one run of the kernel. The
+ * event collection mode will automatically switch to
+ * CUPTI_EVENT_COLLECTION_MODE_KERNEL.  In this mode, \ref
+ * cuptiSetEventCollectionMode will return
+ * CUPTI_ERROR_INVALID_OPERATION.
+ * \note \b Kernels might take longer to run if many events are enabled.
+ * \note \b Thread-safety: this function is thread safe.
+ *
+ * \param context The context
+ * \retval CUPTI_SUCCESS
+ */
+CUptiResult CUPTIAPI cuptiEnableKernelReplayMode(CUcontext context);
+
+/**
+ * \brief Disable kernel replay mode.
+ *
+ * Set profiling mode for the context to non-replay (default)
+ * mode. Event collection mode will be set to
+ * CUPTI_EVENT_COLLECTION_MODE_KERNEL.  All previously enabled
+ * event groups and event group sets will be disabled.
+ * \note \b Thread-safety: this function is thread safe.
+ *
+ * \param context The context
+ * \retval CUPTI_SUCCESS
+ */
+CUptiResult CUPTIAPI cuptiDisableKernelReplayMode(CUcontext context);
+
+/**
+ * \brief Function type for getting updates on kernel replay.
+ *
+ * \param kernelName The mangled kernel name
+ * \param numReplaysDone Number of replays done so far
+ * \param customData Pointer of any custom data passed in when subscribing
+ */
+typedef void (CUPTIAPI *CUpti_KernelReplayUpdateFunc)(
+    const char *kernelName,
+    int numReplaysDone,
+    void *customData);
+
+/**
+ * \brief Subscribe to kernel replay updates.
+ *
+ * When subscribed, the function pointer passed in will be called each time a
+ * kernel run is finished during kernel replay. Previously subscribed function
+ * pointer will be replaced. Pass in NULL as the function pointer unsubscribes
+ * the update.
+ *
+ * \param updateFunc The update function pointer
+ * \param customData Pointer to any custom data
+ * \retval CUPTI_SUCCESS
+ */
+CUptiResult CUPTIAPI cuptiKernelReplaySubscribeUpdate(CUpti_KernelReplayUpdateFunc updateFunc, void *customData);
+
+/** @} */ /* END CUPTI_EVENT_API */
+
+#if defined(__GNUC__) && defined(CUPTI_LIB)
+    #pragma GCC visibility pop
+#endif
+
+#if defined(__cplusplus)
+}
+#endif
+
+#endif /*_CUPTI_EVENTS_H_*/
+
+

Prism/LLaDA/LLaDA_Prism/.venv/lib/python3.12/site-packages/nvidia/cuda_cupti/include/cupti_metrics.h

@@ -0,0 +1,824 @@
+/*
+ * Copyright 2011-2024   NVIDIA Corporation.  All rights reserved.
+ *
+ * NOTICE TO LICENSEE:
+ *
+ * This source code and/or documentation ("Licensed Deliverables") are
+ * subject to NVIDIA intellectual property rights under U.S. and
+ * international Copyright laws.
+ *
+ * These Licensed Deliverables contained herein is PROPRIETARY and
+ * CONFIDENTIAL to NVIDIA and is being provided under the terms and
+ * conditions of a form of NVIDIA software license agreement by and
+ * between NVIDIA and Licensee ("License Agreement") or electronically
+ * accepted by Licensee.  Notwithstanding any terms or conditions to
+ * the contrary in the License Agreement, reproduction or disclosure
+ * of the Licensed Deliverables to any third party without the express
+ * written consent of NVIDIA is prohibited.
+ *
+ * NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE
+ * LICENSE AGREEMENT, NVIDIA MAKES NO REPRESENTATION ABOUT THE
+ * SUITABILITY OF THESE LICENSED DELIVERABLES FOR ANY PURPOSE.  IT IS
+ * PROVIDED "AS IS" WITHOUT EXPRESS OR IMPLIED WARRANTY OF ANY KIND.
+ * NVIDIA DISCLAIMS ALL WARRANTIES WITH REGARD TO THESE LICENSED
+ * DELIVERABLES, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY,
+ * NONINFRINGEMENT, AND FITNESS FOR A PARTICULAR PURPOSE.
+ * NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE
+ * LICENSE AGREEMENT, IN NO EVENT SHALL NVIDIA BE LIABLE FOR ANY
+ * SPECIAL, INDIRECT, INCIDENTAL, OR CONSEQUENTIAL DAMAGES, OR ANY
+ * DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS,
+ * WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS
+ * ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE
+ * OF THESE LICENSED DELIVERABLES.
+ *
+ * U.S. Government End Users.  These Licensed Deliverables are a
+ * "commercial item" as that term is defined at 48 C.F.R. 2.101 (OCT
+ * 1995), consisting of "commercial computer software" and "commercial
+ * computer software documentation" as such terms are used in 48
+ * C.F.R. 12.212 (SEPT 1995) and is provided to the U.S. Government
+ * only as a commercial end item.  Consistent with 48 C.F.R.12.212 and
+ * 48 C.F.R. 227.7202-1 through 227.7202-4 (JUNE 1995), all
+ * U.S. Government End Users acquire the Licensed Deliverables with
+ * only those rights set forth herein.
+ *
+ * Any use of the Licensed Deliverables in individual and commercial
+ * software must include, in the user documentation and internal
+ * comments to the code, the above Disclaimer and U.S. Government End
+ * Users Notice.
+ */
+
+#if !defined(_CUPTI_METRIC_H_)
+#define _CUPTI_METRIC_H_
+
+#include <cuda.h>
+#include <string.h>
+#include <cuda_stdint.h>
+#include <cupti_result.h>
+
+#ifndef CUPTIAPI
+#ifdef _WIN32
+#define CUPTIAPI __stdcall
+#else
+#define CUPTIAPI
+#endif
+#endif
+
+#if defined(__cplusplus)
+extern "C" {
+#endif
+
+#if defined(__GNUC__) && defined(CUPTI_LIB)
+    #pragma GCC visibility push(default)
+#endif
+
+/**
+ * \defgroup CUPTI_METRIC_API CUPTI Metric API
+ * Functions, types, and enums that implement the CUPTI Metric API.
+ *
+ * \note The CUPTI metric API from the header cupti_metrics.h is not supported on devices
+ * with compute capability 7.5 and higher (i.e. Turing and later GPU architectures).
+ * This API is deprecated in CUDA 12.8 release and will be removed in a future CUDA release.
+ * This is replaced by the host profiling API in the header cupti_profiler_host.h and
+ * target profiling API in the header cupti_range_profiler.h which are supported on
+ * devices with compute capability 7.0 and higher (i.e. Volta and later GPU architectures).
+ *
+ * @{
+ */
+
+/**
+ * \brief ID for a metric.
+ *
+ * A metric provides a measure of some aspect of the device.
+ */
+typedef uint32_t CUpti_MetricID;
+
+/**
+ * \brief A metric category.
+ *
+ * Each metric is assigned to a category that represents the general
+ * type of the metric. A metric's category is accessed using \ref
+ * cuptiMetricGetAttribute and the CUPTI_METRIC_ATTR_CATEGORY
+ * attribute.
+ */
+typedef enum {
+  /**
+   * A memory related metric.
+   */
+  CUPTI_METRIC_CATEGORY_MEMORY          = 0,
+  /**
+   * An instruction related metric.
+   */
+  CUPTI_METRIC_CATEGORY_INSTRUCTION     = 1,
+  /**
+   * A multiprocessor related metric.
+   */
+  CUPTI_METRIC_CATEGORY_MULTIPROCESSOR  = 2,
+  /**
+   * A cache related metric.
+   */
+  CUPTI_METRIC_CATEGORY_CACHE           = 3,
+  /**
+   * A texture related metric.
+   */
+  CUPTI_METRIC_CATEGORY_TEXTURE         = 4,
+  /**
+   *A Nvlink related metric.
+   */
+  CUPTI_METRIC_CATEGORY_NVLINK          = 5,
+  /**
+   *A PCIe related metric.
+   */
+  CUPTI_METRIC_CATEGORY_PCIE           = 6,
+  CUPTI_METRIC_CATEGORY_FORCE_INT                         = 0x7fffffff,
+} CUpti_MetricCategory;
+
+/**
+ * \brief A metric evaluation mode.
+ *
+ * A metric can be evaluated per hardware instance to know the load balancing
+ * across instances of a domain or the metric can be evaluated in aggregate mode
+ * when the events involved in metric evaluation are from different event
+ * domains. It might be possible to evaluate some metrics in both
+ * modes for convenience. A metric's evaluation mode is accessed using \ref
+ * CUpti_MetricEvaluationMode and the CUPTI_METRIC_ATTR_EVALUATION_MODE
+ * attribute.
+ */
+typedef enum {
+  /**
+   * If this bit is set, the metric can be profiled for each instance of the
+   * domain. The event values passed to \ref cuptiMetricGetValue can contain
+   * values for one instance of the domain. And \ref cuptiMetricGetValue can
+   * be called for each instance.
+   */
+  CUPTI_METRIC_EVALUATION_MODE_PER_INSTANCE         = 1,
+  /**
+   * If this bit is set, the metric can be profiled over all instances. The
+   * event values passed to \ref cuptiMetricGetValue can be aggregated values
+   * of events for all instances of the domain.
+   */
+  CUPTI_METRIC_EVALUATION_MODE_AGGREGATE            = 1 << 1,
+  CUPTI_METRIC_EVALUATION_MODE_FORCE_INT            = 0x7fffffff,
+} CUpti_MetricEvaluationMode;
+
+/**
+ * \brief Kinds of metric values.
+ *
+ * Metric values can be one of several different kinds. Corresponding
+ * to each kind is a member of the CUpti_MetricValue union. The metric
+ * value returned by \ref cuptiMetricGetValue should be accessed using
+ * the appropriate member of that union based on its value kind.
+ */
+typedef enum {
+  /**
+   * The metric value is a 64-bit double.
+   */
+  CUPTI_METRIC_VALUE_KIND_DOUBLE            = 0,
+  /**
+   * The metric value is a 64-bit unsigned integer.
+   */
+  CUPTI_METRIC_VALUE_KIND_UINT64            = 1,
+  /**
+   * The metric value is a percentage represented by a 64-bit
+   * double. For example, 57.5% is represented by the value 57.5.
+   */
+  CUPTI_METRIC_VALUE_KIND_PERCENT           = 2,
+  /**
+   * The metric value is a throughput represented by a 64-bit
+   * integer. The unit for throughput values is bytes/second.
+   */
+  CUPTI_METRIC_VALUE_KIND_THROUGHPUT        = 3,
+  /**
+   * The metric value is a 64-bit signed integer.
+   */
+  CUPTI_METRIC_VALUE_KIND_INT64             = 4,
+  /**
+   * The metric value is a utilization level, as represented by
+   * CUpti_MetricValueUtilizationLevel.
+   */
+  CUPTI_METRIC_VALUE_KIND_UTILIZATION_LEVEL = 5,
+
+  CUPTI_METRIC_VALUE_KIND_FORCE_INT  = 0x7fffffff
+} CUpti_MetricValueKind;
+
+/**
+ * \brief Enumeration of utilization levels for metrics values of kind
+ * CUPTI_METRIC_VALUE_KIND_UTILIZATION_LEVEL. Utilization values can
+ * vary from IDLE (0) to MAX (10) but the enumeration only provides
+ * specific names for a few values.
+ */
+typedef enum {
+  CUPTI_METRIC_VALUE_UTILIZATION_IDLE      = 0,
+  CUPTI_METRIC_VALUE_UTILIZATION_LOW       = 2,
+  CUPTI_METRIC_VALUE_UTILIZATION_MID       = 5,
+  CUPTI_METRIC_VALUE_UTILIZATION_HIGH      = 8,
+  CUPTI_METRIC_VALUE_UTILIZATION_MAX       = 10,
+  CUPTI_METRIC_VALUE_UTILIZATION_FORCE_INT = 0x7fffffff
+} CUpti_MetricValueUtilizationLevel;
+
+/**
+ * \brief Metric attributes.
+ *
+ * Metric attributes describe properties of a metric. These attributes
+ * can be read using \ref cuptiMetricGetAttribute.
+ */
+typedef enum {
+  /**
+   * Metric name. Value is a null terminated const c-string.
+   */
+  CUPTI_METRIC_ATTR_NAME              = 0,
+  /**
+   * Short description of metric. Value is a null terminated const c-string.
+   */
+  CUPTI_METRIC_ATTR_SHORT_DESCRIPTION = 1,
+  /**
+   * Long description of metric. Value is a null terminated const c-string.
+   */
+  CUPTI_METRIC_ATTR_LONG_DESCRIPTION  = 2,
+  /**
+   * Category of the metric. Value is of type CUpti_MetricCategory.
+   */
+  CUPTI_METRIC_ATTR_CATEGORY          = 3,
+  /**
+   * Value type of the metric. Value is of type CUpti_MetricValueKind.
+   */
+  CUPTI_METRIC_ATTR_VALUE_KIND          = 4,
+  /**
+   * Metric evaluation mode. Value is of type CUpti_MetricEvaluationMode.
+   */
+  CUPTI_METRIC_ATTR_EVALUATION_MODE     = 5,
+  CUPTI_METRIC_ATTR_FORCE_INT         = 0x7fffffff,
+} CUpti_MetricAttribute;
+
+/**
+ * \brief A metric value.
+ *
+ * Metric values can be one of several different kinds. Corresponding
+ * to each kind is a member of the CUpti_MetricValue union. The metric
+ * value returned by \ref cuptiMetricGetValue should be accessed using
+ * the appropriate member of that union based on its value kind.
+ */
+typedef union {
+  /*
+   * Value for CUPTI_METRIC_VALUE_KIND_DOUBLE.
+   */
+  double metricValueDouble;
+  /*
+   * Value for CUPTI_METRIC_VALUE_KIND_UINT64.
+   */
+  uint64_t metricValueUint64;
+  /*
+   * Value for CUPTI_METRIC_VALUE_KIND_INT64.
+   */
+  int64_t metricValueInt64;
+  /*
+   * Value for CUPTI_METRIC_VALUE_KIND_PERCENT. For example, 57.5% is
+   * represented by the value 57.5.
+   */
+  double metricValuePercent;
+  /*
+   * Value for CUPTI_METRIC_VALUE_KIND_THROUGHPUT.  The unit for
+   * throughput values is bytes/second.
+   */
+  uint64_t metricValueThroughput;
+  /*
+   * Value for CUPTI_METRIC_VALUE_KIND_UTILIZATION_LEVEL.
+   */
+  CUpti_MetricValueUtilizationLevel metricValueUtilizationLevel;
+} CUpti_MetricValue;
+
+/**
+ * \brief Device class.
+ *
+ * Enumeration of device classes for metric property
+ * CUPTI_METRIC_PROPERTY_DEVICE_CLASS.
+ */
+typedef enum {
+  CUPTI_METRIC_PROPERTY_DEVICE_CLASS_TESLA          = 0,
+  CUPTI_METRIC_PROPERTY_DEVICE_CLASS_QUADRO         = 1,
+  CUPTI_METRIC_PROPERTY_DEVICE_CLASS_GEFORCE        = 2,
+  CUPTI_METRIC_PROPERTY_DEVICE_CLASS_TEGRA          = 3,
+} CUpti_MetricPropertyDeviceClass;
+
+/**
+ * \brief Metric device properties.
+ *
+ * Metric device properties describe device properties which are needed for a metric.
+ * Some of these properties can be collected using cuDeviceGetAttribute.
+ */
+typedef enum {
+  /*
+   * Number of multiprocessors on a device.  This can be collected
+   * using value of \param CU_DEVICE_ATTRIBUTE_MULTIPROCESSOR_COUNT of
+   * cuDeviceGetAttribute.
+   */
+  CUPTI_METRIC_PROPERTY_MULTIPROCESSOR_COUNT,
+  /*
+   * Maximum number of warps on a multiprocessor. This can be
+   * collected using ratio of value of \param
+   * CU_DEVICE_ATTRIBUTE_MAX_THREADS_PER_MULTIPROCESSOR and \param
+   * CU_DEVICE_ATTRIBUTE_WARP_SIZE of cuDeviceGetAttribute.
+   */
+  CUPTI_METRIC_PROPERTY_WARPS_PER_MULTIPROCESSOR,
+  /*
+   * GPU Time for kernel in ns. This should be profiled using CUPTI
+   * Activity API.
+   */
+  CUPTI_METRIC_PROPERTY_KERNEL_GPU_TIME,
+  /*
+   * Clock rate for device in KHz.  This should be collected using
+   * value of \param CU_DEVICE_ATTRIBUTE_CLOCK_RATE of
+   * cuDeviceGetAttribute.
+   */
+  CUPTI_METRIC_PROPERTY_CLOCK_RATE,
+  /*
+   * Number of Frame buffer units for device. This should be collected
+   * using value of \param CUPTI_DEVICE_ATTRIBUTE_MAX_FRAME_BUFFERS of
+   * cuptiDeviceGetAttribute.
+   */
+  CUPTI_METRIC_PROPERTY_FRAME_BUFFER_COUNT,
+  /*
+   * Global memory bandwidth in KBytes/sec. This should be collected
+   * using value of \param CUPTI_DEVICE_ATTR_GLOBAL_MEMORY_BANDWIDTH
+   * of cuptiDeviceGetAttribute.
+   */
+  CUPTI_METRIC_PROPERTY_GLOBAL_MEMORY_BANDWIDTH,
+  /*
+   * PCIE link rate in Mega bits/sec. This should be collected using
+   * value of \param CUPTI_DEVICE_ATTR_PCIE_LINK_RATE of
+   * cuptiDeviceGetAttribute.
+   */
+  CUPTI_METRIC_PROPERTY_PCIE_LINK_RATE,
+  /*
+   * PCIE link width for device. This should be collected using
+   * value of \param CUPTI_DEVICE_ATTR_PCIE_LINK_WIDTH of
+   * cuptiDeviceGetAttribute.
+   */
+  CUPTI_METRIC_PROPERTY_PCIE_LINK_WIDTH,
+  /*
+   * PCIE generation for device. This should be collected using
+   * value of \param CUPTI_DEVICE_ATTR_PCIE_GEN of
+   * cuptiDeviceGetAttribute.
+   */
+  CUPTI_METRIC_PROPERTY_PCIE_GEN,
+  /*
+   * The device class. This should be collected using
+   * value of \param CUPTI_DEVICE_ATTR_DEVICE_CLASS of
+   * cuptiDeviceGetAttribute.
+   */
+  CUPTI_METRIC_PROPERTY_DEVICE_CLASS,
+  /*
+   * Peak single precision floating point operations that
+   * can be performed in one cycle by the device.
+   * This should be collected using value of
+   * \param CUPTI_DEVICE_ATTR_FLOP_SP_PER_CYCLE of
+   * cuptiDeviceGetAttribute.
+   */
+  CUPTI_METRIC_PROPERTY_FLOP_SP_PER_CYCLE,
+  /*
+   * Peak double precision floating point operations that
+   * can be performed in one cycle by the device.
+   * This should be collected using value of
+   * \param CUPTI_DEVICE_ATTR_FLOP_DP_PER_CYCLE of
+   * cuptiDeviceGetAttribute.
+   */
+  CUPTI_METRIC_PROPERTY_FLOP_DP_PER_CYCLE,
+  /*
+   * Number of L2 units on a device. This can be collected
+   * using value of \param CUPTI_DEVICE_ATTR_MAX_L2_UNITS of
+   * cuDeviceGetAttribute.
+   */
+  CUPTI_METRIC_PROPERTY_L2_UNITS,
+  /*
+   * Whether ECC support is enabled on the device. This can be
+   * collected using value of \param CU_DEVICE_ATTRIBUTE_ECC_ENABLED of
+   * cuDeviceGetAttribute.
+   */
+  CUPTI_METRIC_PROPERTY_ECC_ENABLED,
+  /*
+   * Peak half precision floating point operations that
+   * can be performed in one cycle by the device.
+   * This should be collected using value of
+   * \param CUPTI_DEVICE_ATTR_FLOP_HP_PER_CYCLE of
+   * cuptiDeviceGetAttribute.
+   */
+  CUPTI_METRIC_PROPERTY_FLOP_HP_PER_CYCLE,
+  /*
+   * NVLINK Bandwitdh for device. This should be collected
+   * using value of \param CUPTI_DEVICE_ATTR_GPU_CPU_NVLINK_BW of
+   * cuptiDeviceGetAttribute.
+   */
+  CUPTI_METRIC_PROPERTY_GPU_CPU_NVLINK_BANDWIDTH,
+} CUpti_MetricPropertyID;
+
+/**
+ * \brief Get the total number of metrics available on any device.
+ *
+ * Returns the total number of metrics available on any CUDA-capable
+ * devices.
+ *
+ * \param numMetrics Returns the number of metrics
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p numMetrics is NULL
+*/
+CUptiResult CUPTIAPI cuptiGetNumMetrics(uint32_t *numMetrics);
+
+/**
+ * \brief Get all the metrics available on any device.
+ *
+ * Returns the metric IDs in \p metricArray for all CUDA-capable
+ * devices.  The size of the \p metricArray buffer is given by \p
+ * *arraySizeBytes. The size of the \p metricArray buffer must be at
+ * least \p numMetrics * sizeof(CUpti_MetricID) or all metric IDs will
+ * not be returned. The value returned in \p *arraySizeBytes contains
+ * the number of bytes returned in \p metricArray.
+ *
+ * \param arraySizeBytes The size of \p metricArray in bytes, and
+ * returns the number of bytes written to \p metricArray
+ * \param metricArray Returns the IDs of the metrics
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p arraySizeBytes or
+ * \p metricArray are NULL
+*/
+CUptiResult CUPTIAPI cuptiEnumMetrics(size_t *arraySizeBytes,
+                                      CUpti_MetricID *metricArray);
+
+/**
+ * \brief Get the number of metrics for a device.
+ *
+ * Returns the number of metrics available for a device.
+ *
+ * \param device The CUDA device
+ * \param numMetrics Returns the number of metrics available for the
+ * device
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_NOT_INITIALIZED
+ * \retval CUPTI_ERROR_INVALID_DEVICE
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p numMetrics is NULL
+ */
+CUptiResult CUPTIAPI cuptiDeviceGetNumMetrics(CUdevice device,
+                                              uint32_t *numMetrics);
+
+/**
+ * \brief Get the metrics for a device.
+ *
+ * Returns the metric IDs in \p metricArray for a device.  The size of
+ * the \p metricArray buffer is given by \p *arraySizeBytes. The size
+ * of the \p metricArray buffer must be at least \p numMetrics *
+ * sizeof(CUpti_MetricID) or else all metric IDs will not be
+ * returned. The value returned in \p *arraySizeBytes contains the
+ * number of bytes returned in \p metricArray.
+ *
+ * \param device The CUDA device
+ * \param arraySizeBytes The size of \p metricArray in bytes, and
+ * returns the number of bytes written to \p metricArray
+ * \param metricArray Returns the IDs of the metrics for the device
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_NOT_INITIALIZED
+ * \retval CUPTI_ERROR_INVALID_DEVICE
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p arraySizeBytes or
+ * \p metricArray are NULL
+ */
+CUptiResult CUPTIAPI cuptiDeviceEnumMetrics(CUdevice device,
+                                            size_t *arraySizeBytes,
+                                            CUpti_MetricID *metricArray);
+
+/**
+ * \brief Get a metric attribute.
+ *
+ * Returns a metric attribute in \p *value. The size of the \p
+ * value buffer is given by \p *valueSize. The value returned in \p
+ * *valueSize contains the number of bytes returned in \p value.
+ *
+ * If the attribute value is a c-string that is longer than \p
+ * *valueSize, then only the first \p *valueSize characters will be
+ * returned and there will be no terminating null byte.
+ *
+ * \param metric ID of the metric
+ * \param attrib The metric attribute to read
+ * \param valueSize The size of the \p value buffer in bytes, and
+ * returns the number of bytes written to \p value
+ * \param value Returns the attribute's value
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_NOT_INITIALIZED
+ * \retval CUPTI_ERROR_INVALID_METRIC_ID
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p valueSize or \p value
+ * is NULL, or if \p attrib is not a metric attribute
+ * \retval CUPTI_ERROR_PARAMETER_SIZE_NOT_SUFFICIENT For non-c-string
+ * attribute values, indicates that the \p value buffer is too small
+ * to hold the attribute value.
+ */
+CUptiResult CUPTIAPI cuptiMetricGetAttribute(CUpti_MetricID metric,
+                                             CUpti_MetricAttribute attrib,
+                                             size_t *valueSize,
+                                             void *value);
+
+/**
+ * \brief Find an metric by name.
+ *
+ * Find a metric by name and return the metric ID in \p *metric.
+ *
+ * \param device The CUDA device
+ * \param metricName The name of metric to find
+ * \param metric Returns the ID of the found metric or undefined if
+ * unable to find the metric
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_NOT_INITIALIZED
+ * \retval CUPTI_ERROR_INVALID_DEVICE
+ * \retval CUPTI_ERROR_INVALID_METRIC_NAME if unable to find a metric
+ * with name \p metricName. In this case \p *metric is undefined
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p metricName or \p
+ * metric are NULL.
+ */
+CUptiResult CUPTIAPI cuptiMetricGetIdFromName(CUdevice device,
+                                              const char *metricName,
+                                              CUpti_MetricID *metric);
+
+/**
+ * \brief Get number of events required to calculate a metric.
+ *
+ * Returns the number of events in \p numEvents that are required to
+ * calculate a metric.
+ *
+ * \param metric ID of the metric
+ * \param numEvents Returns the number of events required for the metric
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_NOT_INITIALIZED
+ * \retval CUPTI_ERROR_INVALID_METRIC_ID
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p numEvents is NULL
+ */
+CUptiResult CUPTIAPI cuptiMetricGetNumEvents(CUpti_MetricID metric,
+                                             uint32_t *numEvents);
+
+/**
+ * \brief Get the events required to calculating a metric.
+ *
+ * Gets the event IDs in \p eventIdArray required to calculate a \p
+ * metric. The size of the \p eventIdArray buffer is given by \p
+ * *eventIdArraySizeBytes and must be at least \p numEvents *
+ * sizeof(CUpti_EventID) or all events will not be returned. The value
+ * returned in \p *eventIdArraySizeBytes contains the number of bytes
+ * returned in \p eventIdArray.
+ *
+ * \param metric ID of the metric
+ * \param eventIdArraySizeBytes The size of \p eventIdArray in bytes,
+ * and returns the number of bytes written to \p eventIdArray
+ * \param eventIdArray Returns the IDs of the events required to
+ * calculate \p metric
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_NOT_INITIALIZED
+ * \retval CUPTI_ERROR_INVALID_METRIC_ID
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p eventIdArraySizeBytes or \p
+ * eventIdArray are NULL.
+ */
+CUptiResult CUPTIAPI cuptiMetricEnumEvents(CUpti_MetricID metric,
+                                           size_t *eventIdArraySizeBytes,
+                                           CUpti_EventID *eventIdArray);
+
+/**
+ * \brief Get number of properties required to calculate a metric.
+ *
+ * Returns the number of properties in \p numProp that are required to
+ * calculate a metric.
+ *
+ * \param metric ID of the metric
+ * \param numProp Returns the number of properties required for the
+ * metric
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_NOT_INITIALIZED
+ * \retval CUPTI_ERROR_INVALID_METRIC_ID
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p numProp is NULL
+ */
+CUptiResult CUPTIAPI cuptiMetricGetNumProperties(CUpti_MetricID metric,
+                                                 uint32_t *numProp);
+
+/**
+ * \brief Get the properties required to calculating a metric.
+ *
+ * Gets the property IDs in \p propIdArray required to calculate a \p
+ * metric. The size of the \p propIdArray buffer is given by \p
+ * *propIdArraySizeBytes and must be at least \p numProp *
+ * sizeof(CUpti_DeviceAttribute) or all properties will not be
+ * returned. The value returned in \p *propIdArraySizeBytes contains
+ * the number of bytes returned in \p propIdArray.
+ *
+ * \param metric ID of the metric
+ * \param propIdArraySizeBytes The size of \p propIdArray in bytes,
+ * and returns the number of bytes written to \p propIdArray
+ * \param propIdArray Returns the IDs of the properties required to
+ * calculate \p metric
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_NOT_INITIALIZED
+ * \retval CUPTI_ERROR_INVALID_METRIC_ID
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p propIdArraySizeBytes or \p
+ * propIdArray are NULL.
+ */
+CUptiResult CUPTIAPI cuptiMetricEnumProperties(CUpti_MetricID metric,
+                                               size_t *propIdArraySizeBytes,
+                                               CUpti_MetricPropertyID *propIdArray);
+
+
+/**
+ * \brief For a metric get the groups of events that must be collected
+ * in the same pass.
+ *
+ * For a metric get the groups of events that must be collected in the
+ * same pass to ensure that the metric is calculated correctly. If the
+ * events are not collected as specified then the metric value may be
+ * inaccurate.
+ *
+ * The function returns NULL if a metric does not have any required
+ * event group. In this case the events needed for the metric can be
+ * grouped in any manner for collection.
+ *
+ * \param context The context for event collection
+ * \param metric The metric ID
+ * \param eventGroupSets Returns a CUpti_EventGroupSets object that
+ * indicates the events that must be collected in the same pass to
+ * ensure the metric is calculated correctly.  Returns NULL if no
+ * grouping is required for metric
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_NOT_INITIALIZED
+ * \retval CUPTI_ERROR_INVALID_METRIC_ID
+ */
+CUptiResult CUPTIAPI cuptiMetricGetRequiredEventGroupSets(CUcontext context,
+                                                          CUpti_MetricID metric,
+                                                          CUpti_EventGroupSets **eventGroupSets);
+
+/**
+ * \brief For a set of metrics, get the grouping that indicates the
+ * number of passes and the event groups necessary to collect the
+ * events required for those metrics.
+ *
+ * For a set of metrics, get the grouping that indicates the number of
+ * passes and the event groups necessary to collect the events
+ * required for those metrics.
+ *
+ * \see cuptiEventGroupSetsCreate for details on event group set
+ * creation.
+ *
+ * \param context The context for event collection
+ * \param metricIdArraySizeBytes Size of the metricIdArray in bytes
+ * \param metricIdArray Array of metric IDs
+ * \param eventGroupPasses Returns a CUpti_EventGroupSets object that
+ * indicates the number of passes required to collect the events and
+ * the events to collect on each pass
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_NOT_INITIALIZED
+ * \retval CUPTI_ERROR_INVALID_CONTEXT
+ * \retval CUPTI_ERROR_INVALID_METRIC_ID
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p metricIdArray or
+ * \p eventGroupPasses is NULL
+ */
+CUptiResult CUPTIAPI cuptiMetricCreateEventGroupSets(CUcontext context,
+                                                     size_t metricIdArraySizeBytes,
+                                                     CUpti_MetricID *metricIdArray,
+                                                     CUpti_EventGroupSets **eventGroupPasses);
+
+/**
+ * \brief Calculate the value for a metric.
+ *
+ * Use the events collected for a metric to calculate the metric
+ * value. Metric value evaluation depends on the evaluation mode
+ * \ref CUpti_MetricEvaluationMode that the metric supports.
+ * If a metric has evaluation mode as CUPTI_METRIC_EVALUATION_MODE_PER_INSTANCE,
+ * then it assumes that the input event value is for one domain instance.
+ * If a metric has evaluation mode as CUPTI_METRIC_EVALUATION_MODE_AGGREGATE,
+ * it assumes that input event values are
+ * normalized to represent all domain instances on a device. For the
+ * most accurate metric collection, the events required for the metric
+ * should be collected for all profiled domain instances. For example,
+ * to collect all instances of an event, set the
+ * CUPTI_EVENT_GROUP_ATTR_PROFILE_ALL_DOMAIN_INSTANCES attribute on
+ * the group containing the event to 1. The normalized value for the
+ * event is then: (\p sum_event_values * \p totalInstanceCount) / \p
+ * instanceCount, where \p sum_event_values is the summation of the
+ * event values across all profiled domain instances, \p
+ * totalInstanceCount is obtained from querying
+ * CUPTI_EVENT_DOMAIN_ATTR_TOTAL_INSTANCE_COUNT and \p instanceCount
+ * is obtained from querying CUPTI_EVENT_GROUP_ATTR_INSTANCE_COUNT (or
+ * CUPTI_EVENT_DOMAIN_ATTR_INSTANCE_COUNT).
+ *
+ * \param device The CUDA device that the metric is being calculated for
+ * \param metric The metric ID
+ * \param eventIdArraySizeBytes The size of \p eventIdArray in bytes
+ * \param eventIdArray The event IDs required to calculate \p metric
+ * \param eventValueArraySizeBytes The size of \p eventValueArray in bytes
+ * \param eventValueArray The normalized event values required to
+ * calculate \p metric. The values must be order to match the order of
+ * events in \p eventIdArray
+ * \param timeDuration The duration over which the events were
+ * collected, in ns
+ * \param metricValue Returns the value for the metric
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_NOT_INITIALIZED
+ * \retval CUPTI_ERROR_INVALID_METRIC_ID
+ * \retval CUPTI_ERROR_INVALID_OPERATION
+ * \retval CUPTI_ERROR_PARAMETER_SIZE_NOT_SUFFICIENT if the
+ * eventIdArray does not contain all the events needed for metric
+ * \retval CUPTI_ERROR_INVALID_EVENT_VALUE if any of the
+ * event values required for the metric is CUPTI_EVENT_OVERFLOW
+ * \retval CUPTI_ERROR_INVALID_METRIC_VALUE if the computed metric value
+ * cannot be represented in the metric's value type. For example,
+ * if the metric value type is unsigned and the computed metric value is negative
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p metricValue,
+ * \p eventIdArray or \p eventValueArray is NULL
+ */
+CUptiResult CUPTIAPI cuptiMetricGetValue(CUdevice device,
+                                         CUpti_MetricID metric,
+                                         size_t eventIdArraySizeBytes,
+                                         CUpti_EventID *eventIdArray,
+                                         size_t eventValueArraySizeBytes,
+                                         uint64_t *eventValueArray,
+                                         uint64_t timeDuration,
+                                         CUpti_MetricValue *metricValue);
+
+/**
+ * \brief Calculate the value for a metric.
+ *
+ * Use the events and properties collected for a metric to calculate
+ * the metric value. Metric value evaluation depends on the evaluation
+ * mode \ref CUpti_MetricEvaluationMode that the metric supports.  If
+ * a metric has evaluation mode as
+ * CUPTI_METRIC_EVALUATION_MODE_PER_INSTANCE, then it assumes that the
+ * input event value is for one domain instance.  If a metric has
+ * evaluation mode as CUPTI_METRIC_EVALUATION_MODE_AGGREGATE, it
+ * assumes that input event values are normalized to represent all
+ * domain instances on a device. For the most accurate metric
+ * collection, the events required for the metric should be collected
+ * for all profiled domain instances. For example, to collect all
+ * instances of an event, set the
+ * CUPTI_EVENT_GROUP_ATTR_PROFILE_ALL_DOMAIN_INSTANCES attribute on
+ * the group containing the event to 1. The normalized value for the
+ * event is then: (\p sum_event_values * \p totalInstanceCount) / \p
+ * instanceCount, where \p sum_event_values is the summation of the
+ * event values across all profiled domain instances, \p
+ * totalInstanceCount is obtained from querying
+ * CUPTI_EVENT_DOMAIN_ATTR_TOTAL_INSTANCE_COUNT and \p instanceCount
+ * is obtained from querying CUPTI_EVENT_GROUP_ATTR_INSTANCE_COUNT (or
+ * CUPTI_EVENT_DOMAIN_ATTR_INSTANCE_COUNT).
+ *
+ * \param metric The metric ID
+ * \param eventIdArraySizeBytes The size of \p eventIdArray in bytes
+ * \param eventIdArray The event IDs required to calculate \p metric
+ * \param eventValueArraySizeBytes The size of \p eventValueArray in bytes
+ * \param eventValueArray The normalized event values required to
+ * calculate \p metric. The values must be order to match the order of
+ * events in \p eventIdArray
+ * \param propIdArraySizeBytes The size of \p propIdArray in bytes
+ * \param propIdArray The metric property IDs required to calculate \p metric
+ * \param propValueArraySizeBytes The size of \p propValueArray in bytes
+ * \param propValueArray The metric property values required to
+ * calculate \p metric. The values must be order to match the order of
+ * metric properties in \p propIdArray
+ * \param metricValue Returns the value for the metric
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_NOT_INITIALIZED
+ * \retval CUPTI_ERROR_INVALID_METRIC_ID
+ * \retval CUPTI_ERROR_INVALID_OPERATION
+ * \retval CUPTI_ERROR_PARAMETER_SIZE_NOT_SUFFICIENT if the
+ * eventIdArray does not contain all the events needed for metric
+ * \retval CUPTI_ERROR_INVALID_EVENT_VALUE if any of the
+ * event values required for the metric is CUPTI_EVENT_OVERFLOW
+ * \retval CUPTI_ERROR_NOT_COMPATIBLE if the computed metric value
+ * cannot be represented in the metric's value type. For example,
+ * if the metric value type is unsigned and the computed metric value is negative
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p metricValue,
+ * \p eventIdArray or \p eventValueArray is NULL
+ */
+CUptiResult CUPTIAPI cuptiMetricGetValue2(CUpti_MetricID metric,
+                                          size_t eventIdArraySizeBytes,
+                                          CUpti_EventID *eventIdArray,
+                                          size_t eventValueArraySizeBytes,
+                                          uint64_t *eventValueArray,
+                                          size_t propIdArraySizeBytes,
+                                          CUpti_MetricPropertyID *propIdArray,
+                                          size_t propValueArraySizeBytes,
+                                          uint64_t *propValueArray,
+                                          CUpti_MetricValue *metricValue);
+
+/** @} */ /* END CUPTI_METRIC_API */
+
+#if defined(__GNUC__) && defined(CUPTI_LIB)
+    #pragma GCC visibility pop
+#endif
+
+#if defined(__cplusplus)
+}
+#endif
+
+#endif /*_CUPTI_METRIC_H_*/
+
+

Prism/LLaDA/LLaDA_Prism/.venv/lib/python3.12/site-packages/nvidia/cuda_cupti/include/cupti_nvtx_cbid.h

@@ -0,0 +1,111 @@
+/*
+ * Copyright 2013-2017 NVIDIA Corporation.  All rights reserved.
+ *
+ * NOTICE TO LICENSEE:
+ *
+ * This source code and/or documentation ("Licensed Deliverables") are
+ * subject to NVIDIA intellectual property rights under U.S. and
+ * international Copyright laws.
+ *
+ * These Licensed Deliverables contained herein is PROPRIETARY and
+ * CONFIDENTIAL to NVIDIA and is being provided under the terms and
+ * conditions of a form of NVIDIA software license agreement by and
+ * between NVIDIA and Licensee ("License Agreement") or electronically
+ * accepted by Licensee.  Notwithstanding any terms or conditions to
+ * the contrary in the License Agreement, reproduction or disclosure
+ * of the Licensed Deliverables to any third party without the express
+ * written consent of NVIDIA is prohibited.
+ *
+ * NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE
+ * LICENSE AGREEMENT, NVIDIA MAKES NO REPRESENTATION ABOUT THE
+ * SUITABILITY OF THESE LICENSED DELIVERABLES FOR ANY PURPOSE.  IT IS
+ * PROVIDED "AS IS" WITHOUT EXPRESS OR IMPLIED WARRANTY OF ANY KIND.
+ * NVIDIA DISCLAIMS ALL WARRANTIES WITH REGARD TO THESE LICENSED
+ * DELIVERABLES, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY,
+ * NONINFRINGEMENT, AND FITNESS FOR A PARTICULAR PURPOSE.
+ * NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE
+ * LICENSE AGREEMENT, IN NO EVENT SHALL NVIDIA BE LIABLE FOR ANY
+ * SPECIAL, INDIRECT, INCIDENTAL, OR CONSEQUENTIAL DAMAGES, OR ANY
+ * DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS,
+ * WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS
+ * ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE
+ * OF THESE LICENSED DELIVERABLES.
+ *
+ * U.S. Government End Users.  These Licensed Deliverables are a
+ * "commercial item" as that term is defined at 48 C.F.R. 2.101 (OCT
+ * 1995), consisting of "commercial computer software" and "commercial
+ * computer software documentation" as such terms are used in 48
+ * C.F.R. 12.212 (SEPT 1995) and is provided to the U.S. Government
+ * only as a commercial end item.  Consistent with 48 C.F.R.12.212 and
+ * 48 C.F.R. 227.7202-1 through 227.7202-4 (JUNE 1995), all
+ * U.S. Government End Users acquire the Licensed Deliverables with
+ * only those rights set forth herein.
+ *
+ * Any use of the Licensed Deliverables in individual and commercial
+ * software must include, in the user documentation and internal
+ * comments to the code, the above Disclaimer and U.S. Government End
+ * Users Notice.
+ */
+
+#if defined(__GNUC__) && defined(CUPTI_LIB)
+    #pragma GCC visibility push(default)
+#endif
+
+typedef enum {
+  CUPTI_CBID_NVTX_INVALID                               = 0,
+  CUPTI_CBID_NVTX_nvtxMarkA                             = 1,
+  CUPTI_CBID_NVTX_nvtxMarkW                             = 2,
+  CUPTI_CBID_NVTX_nvtxMarkEx                            = 3,
+  CUPTI_CBID_NVTX_nvtxRangeStartA                       = 4,
+  CUPTI_CBID_NVTX_nvtxRangeStartW                       = 5,
+  CUPTI_CBID_NVTX_nvtxRangeStartEx                      = 6,
+  CUPTI_CBID_NVTX_nvtxRangeEnd                          = 7,
+  CUPTI_CBID_NVTX_nvtxRangePushA                        = 8,
+  CUPTI_CBID_NVTX_nvtxRangePushW                        = 9,
+  CUPTI_CBID_NVTX_nvtxRangePushEx                       = 10,
+  CUPTI_CBID_NVTX_nvtxRangePop                          = 11,
+  CUPTI_CBID_NVTX_nvtxNameCategoryA                     = 12,
+  CUPTI_CBID_NVTX_nvtxNameCategoryW                     = 13,
+  CUPTI_CBID_NVTX_nvtxNameOsThreadA                     = 14,
+  CUPTI_CBID_NVTX_nvtxNameOsThreadW                     = 15,
+  CUPTI_CBID_NVTX_nvtxNameCuDeviceA                     = 16,
+  CUPTI_CBID_NVTX_nvtxNameCuDeviceW                     = 17,
+  CUPTI_CBID_NVTX_nvtxNameCuContextA                    = 18,
+  CUPTI_CBID_NVTX_nvtxNameCuContextW                    = 19,
+  CUPTI_CBID_NVTX_nvtxNameCuStreamA                     = 20,
+  CUPTI_CBID_NVTX_nvtxNameCuStreamW                     = 21,
+  CUPTI_CBID_NVTX_nvtxNameCuEventA                      = 22,
+  CUPTI_CBID_NVTX_nvtxNameCuEventW                      = 23,
+  CUPTI_CBID_NVTX_nvtxNameCudaDeviceA                   = 24,
+  CUPTI_CBID_NVTX_nvtxNameCudaDeviceW                   = 25,
+  CUPTI_CBID_NVTX_nvtxNameCudaStreamA                   = 26,
+  CUPTI_CBID_NVTX_nvtxNameCudaStreamW                   = 27,
+  CUPTI_CBID_NVTX_nvtxNameCudaEventA                    = 28,
+  CUPTI_CBID_NVTX_nvtxNameCudaEventW                    = 29,
+  CUPTI_CBID_NVTX_nvtxDomainMarkEx                      = 30,
+  CUPTI_CBID_NVTX_nvtxDomainRangeStartEx                = 31,
+  CUPTI_CBID_NVTX_nvtxDomainRangeEnd                    = 32,
+  CUPTI_CBID_NVTX_nvtxDomainRangePushEx                 = 33,
+  CUPTI_CBID_NVTX_nvtxDomainRangePop                    = 34,
+  CUPTI_CBID_NVTX_nvtxDomainResourceCreate              = 35,
+  CUPTI_CBID_NVTX_nvtxDomainResourceDestroy             = 36,
+  CUPTI_CBID_NVTX_nvtxDomainNameCategoryA               = 37,
+  CUPTI_CBID_NVTX_nvtxDomainNameCategoryW               = 38,
+  CUPTI_CBID_NVTX_nvtxDomainRegisterStringA             = 39,
+  CUPTI_CBID_NVTX_nvtxDomainRegisterStringW             = 40,
+  CUPTI_CBID_NVTX_nvtxDomainCreateA                     = 41,
+  CUPTI_CBID_NVTX_nvtxDomainCreateW                     = 42,
+  CUPTI_CBID_NVTX_nvtxDomainDestroy                     = 43,
+  CUPTI_CBID_NVTX_nvtxDomainSyncUserCreate              = 44,
+  CUPTI_CBID_NVTX_nvtxDomainSyncUserDestroy             = 45,
+  CUPTI_CBID_NVTX_nvtxDomainSyncUserAcquireStart        = 46,
+  CUPTI_CBID_NVTX_nvtxDomainSyncUserAcquireFailed       = 47,
+  CUPTI_CBID_NVTX_nvtxDomainSyncUserAcquireSuccess      = 48,
+  CUPTI_CBID_NVTX_nvtxDomainSyncUserReleasing           = 49,
+  CUPTI_CBID_NVTX_SIZE,
+  CUPTI_CBID_NVTX_FORCE_INT                             = 0x7fffffff
+} CUpti_nvtx_api_trace_cbid;
+
+#if defined(__GNUC__) && defined(CUPTI_LIB)
+    #pragma GCC visibility pop
+#endif    

Prism/LLaDA/LLaDA_Prism/.venv/lib/python3.12/site-packages/nvidia/cuda_cupti/include/cupti_pcsampling.h

@@ -0,0 +1,936 @@
+/*
+ * Copyright 2020-2022 NVIDIA Corporation.  All rights reserved.
+ *
+ * NOTICE TO LICENSEE:
+ *
+ * This source code and/or documentation ("Licensed Deliverables") are
+ * subject to NVIDIA intellectual property rights under U.S. and
+ * international Copyright laws.
+ *
+ * These Licensed Deliverables contained herein is PROPRIETARY and
+ * CONFIDENTIAL to NVIDIA and is being provided under the terms and
+ * conditions of a form of NVIDIA software license agreement by and
+ * between NVIDIA and Licensee ("License Agreement") or electronically
+ * accepted by Licensee.  Notwithstanding any terms or conditions to
+ * the contrary in the License Agreement, reproduction or disclosure
+ * of the Licensed Deliverables to any third party without the express
+ * written consent of NVIDIA is prohibited.
+ *
+ * NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE
+ * LICENSE AGREEMENT, NVIDIA MAKES NO REPRESENTATION ABOUT THE
+ * SUITABILITY OF THESE LICENSED DELIVERABLES FOR ANY PURPOSE.  IT IS
+ * PROVIDED "AS IS" WITHOUT EXPRESS OR IMPLIED WARRANTY OF ANY KIND.
+ * NVIDIA DISCLAIMS ALL WARRANTIES WITH REGARD TO THESE LICENSED
+ * DELIVERABLES, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY,
+ * NONINFRINGEMENT, AND FITNESS FOR A PARTICULAR PURPOSE.
+ * NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE
+ * LICENSE AGREEMENT, IN NO EVENT SHALL NVIDIA BE LIABLE FOR ANY
+ * SPECIAL, INDIRECT, INCIDENTAL, OR CONSEQUENTIAL DAMAGES, OR ANY
+ * DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS,
+ * WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS
+ * ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE
+ * OF THESE LICENSED DELIVERABLES.
+ *
+ * U.S. Government End Users.  These Licensed Deliverables are a
+ * "commercial item" as that term is defined at 48 C.F.R. 2.101 (OCT
+ * 1995), consisting of "commercial computer software" and "commercial
+ * computer software documentation" as such terms are used in 48
+ * C.F.R. 12.212 (SEPT 1995) and is provided to the U.S. Government
+ * only as a commercial end item.  Consistent with 48 C.F.R.12.212 and
+ * 48 C.F.R. 227.7202-1 through 227.7202-4 (JUNE 1995), all
+ * U.S. Government End Users acquire the Licensed Deliverables with
+ * only those rights set forth herein.
+ *
+ * Any use of the Licensed Deliverables in individual and commercial
+ * software must include, in the user documentation and internal
+ * comments to the code, the above Disclaimer and U.S. Government End
+ * Users Notice.
+ */
+
+#if !defined(_CUPTI_PCSAMPLING_H_)
+#define _CUPTI_PCSAMPLING_H_
+
+#include <cuda.h>
+#include <stdint.h>
+#include <stddef.h>
+#include "cupti_result.h"
+#include "cupti_common.h"
+
+
+#if defined(__cplusplus)
+extern "C" {
+#endif
+
+#if defined(__GNUC__) && defined(CUPTI_LIB)
+    #pragma GCC visibility push(default)
+#endif
+
+/**
+ * \defgroup CUPTI_PCSAMPLING_API CUPTI PC Sampling API
+ * Functions, types, and enums that implement the CUPTI PC Sampling API.
+ * @{
+ */
+
+#ifndef CUPTI_PCSAMPLING_STRUCT_SIZE
+#define CUPTI_PCSAMPLING_STRUCT_SIZE(type_, lastfield_)                     (offsetof(type_, lastfield_) + sizeof(((type_*)0)->lastfield_))
+#endif
+
+#ifndef CUPTI_STALL_REASON_STRING_SIZE
+#define CUPTI_STALL_REASON_STRING_SIZE                                            128
+#endif
+
+/**
+ * \brief PC Sampling collection mode
+ */
+typedef enum
+{
+  /**
+   * INVALID Value
+   */
+  CUPTI_PC_SAMPLING_COLLECTION_MODE_INVALID                   = 0,
+  /**
+   * Continuous mode. Kernels are not serialized in this mode.
+   */
+  CUPTI_PC_SAMPLING_COLLECTION_MODE_CONTINUOUS                = 1,
+  /**
+   * Serialized mode. Kernels are serialized in this mode.
+   */
+  CUPTI_PC_SAMPLING_COLLECTION_MODE_KERNEL_SERIALIZED         = 2,
+} CUpti_PCSamplingCollectionMode;
+
+/**
+ * \brief PC Sampling stall reasons
+ */
+typedef struct PACKED_ALIGNMENT
+{
+  /**
+   * [r] Collected stall reason index
+   */
+  uint32_t pcSamplingStallReasonIndex;
+  /**
+   * [r] Number of times the PC was sampled with the stallReason.
+   */
+  uint32_t samples;
+} CUpti_PCSamplingStallReason;
+
+/**
+ * \brief PC Sampling data
+ */
+typedef struct PACKED_ALIGNMENT
+{
+  /**
+   * [w] Size of the data structure.
+   * CUPTI client should set the size of the structure. It will be used in CUPTI to check what fields are
+   * available in the structure. Used to preserve backward compatibility.
+   */
+  size_t size;
+  /**
+   * [r] Unique cubin id
+   */
+  uint64_t cubinCrc;
+  /**
+   * [r] PC offset
+   */
+  uint64_t pcOffset;
+  /**
+   * The function's unique symbol index in the module.
+   */
+  uint32_t functionIndex;
+  /**
+   * Padding
+   */
+  uint32_t pad;
+  /**
+   * [r] The function name. This name string might be shared across all the records
+   * including records from activity APIs representing the same function, and so it should not be
+   * modified or freed until post processing of all the records is done. Once done, it is user’s responsibility to
+   * free the memory using free() function.
+   */
+  char* functionName;
+  /**
+   * [r] Collected stall reason count
+   */
+  size_t stallReasonCount;
+  /**
+   * [r] Stall reason id
+   * Total samples
+   */
+  CUpti_PCSamplingStallReason *stallReason;
+  /**
+   * The correlation ID of the kernel to which this result is associated. Only valid for serialized mode of pc sampling collection.
+   * For continous mode of collection the correlationId will be set to 0.
+   */
+  uint32_t correlationId;
+} CUpti_PCSamplingPCData;
+
+/**
+ * \brief PC Sampling output data format
+ */
+typedef enum
+{
+    CUPTI_PC_SAMPLING_OUTPUT_DATA_FORMAT_INVALID          = 0,
+  /**
+   * HW buffer data will be parsed during collection of data
+   */
+    CUPTI_PC_SAMPLING_OUTPUT_DATA_FORMAT_PARSED           = 1,
+} CUpti_PCSamplingOutputDataFormat;
+
+/**
+ * \brief Collected PC Sampling data
+ *
+ */
+typedef struct PACKED_ALIGNMENT
+{
+  /**
+   * [w] Size of the data structure.
+   * CUPTI client should set the size of the structure. It will be used in CUPTI to check what fields are
+   * available in the structure. Used to preserve backward compatibility.
+   */
+  size_t size;
+  /**
+   * [w] Number of PCs to be collected
+   */
+  size_t collectNumPcs;
+  /**
+   * [r] Number of samples collected across all PCs.
+   * It includes samples for user modules, samples for non-user kernels and dropped samples.
+   * It includes counts for all non selected stall reasons.
+   * CUPTI does not provide PC records for non-user kernels.
+   * CUPTI does not provide PC records for instructions for which all selected stall reason metrics counts are zero.
+   */
+  uint64_t totalSamples;
+  /**
+   * [r] Number of samples that were dropped by hardware due to backpressure/overflow.
+   */
+  uint64_t droppedSamples;
+  /**
+   * [r] Number of PCs collected
+   */
+  size_t totalNumPcs;
+  /**
+   * [r] Number of PCs available for collection
+   */
+  size_t remainingNumPcs;
+  /**
+   * [r] Unique identifier for each range.
+   * Data collected across multiple ranges in multiple buffers can be identified using range id.
+   */
+  uint64_t rangeId;
+  /**
+   * [r] Profiled PC data
+   * This data struct should have enough memory to collect number of PCs mentioned in \brief collectNumPcs
+   */
+  CUpti_PCSamplingPCData *pPcData;
+  /**
+   * [r] Number of samples collected across all non user kernels PCs.
+   * It includes samples for non-user kernels.
+   * It includes counts for all non selected stall reasons as well.
+   * CUPTI does not provide PC records for non-user kernels.
+   */
+  uint64_t nonUsrKernelsTotalSamples;
+
+  /**
+   * [r] Status of the hardware buffer.
+   * CUPTI returns the error code CUPTI_ERROR_OUT_OF_MEMORY when hardware buffer is full.
+   * When hardware buffer is full, user will get pc data as 0. To mitigate this issue, one or more of the below options can be tried:
+   * 1. Increase the hardware buffer size using the attribute CUPTI_PC_SAMPLING_CONFIGURATION_ATTR_TYPE_HARDWARE_BUFFER_SIZE
+   * 2. Decrease the thread sleep span using the attribute CUPTI_PC_SAMPLING_CONFIGURATION_ATTR_TYPE_WORKER_THREAD_PERIODIC_SLEEP_SPAN
+   * 3. Decrease the sampling frequency using the attribute CUPTI_PC_SAMPLING_CONFIGURATION_ATTR_TYPE_SAMPLING_PERIOD
+   */
+  uint8_t hardwareBufferFull;
+} CUpti_PCSamplingData;
+
+/**
+ * \brief PC Sampling configuration attributes
+ *
+ * PC Sampling configuration attribute types. These attributes can be read
+ * using \ref cuptiPCSamplingGetConfigurationAttribute and can be written
+ * using \ref cuptiPCSamplingSetConfigurationAttribute. Attributes marked
+ * [r] can only be read using \ref cuptiPCSamplingGetConfigurationAttribute
+ * [w] can only be written using \ref cuptiPCSamplingSetConfigurationAttribute
+ * [rw] can be read using \ref cuptiPCSamplingGetConfigurationAttribute and
+ * written using \ref cuptiPCSamplingSetConfigurationAttribute
+ */
+typedef enum
+{
+  CUPTI_PC_SAMPLING_CONFIGURATION_ATTR_TYPE_INVALID                            = 0,
+  /**
+   * [rw] Sampling period for PC Sampling.
+   * DEFAULT - CUPTI defined value based on number of SMs
+   * Valid values for the sampling
+   * periods are between 5 to 31 both inclusive. This will set the
+   * sampling period to (2^samplingPeriod) cycles.
+   * For e.g. for sampling period = 5 to 31, cycles = 32, 64, 128,..., 2^31
+   * Value is a uint32_t
+   */
+  CUPTI_PC_SAMPLING_CONFIGURATION_ATTR_TYPE_SAMPLING_PERIOD                    = 1,
+  /**
+   * [w] Number of stall reasons to collect.
+   * DEFAULT - All stall reasons will be collected
+   * Value is a size_t
+   * [w] Stall reasons to collect
+   * DEFAULT - All stall reasons will be collected
+   * Input value should be a pointer pointing to array of stall reason indexes
+   * containing all the stall reason indexes to collect.
+   */
+  CUPTI_PC_SAMPLING_CONFIGURATION_ATTR_TYPE_STALL_REASON                       = 2,
+  /**
+   * [rw] Size of SW buffer for raw PC counter data downloaded from HW buffer
+   * DEFAULT - 1 MB, which can accommodate approximately 5500 PCs
+   * with all stall reasons
+   * Approximately it takes 16 Bytes (and some fixed size memory)
+   * to accommodate one PC with one stall reason
+   * For e.g. 1 PC with 1 stall reason = 32 Bytes
+   *          1 PC with 2 stall reason = 48 Bytes
+   *          1 PC with 4 stall reason = 96 Bytes
+   * Value is a size_t
+   */
+  CUPTI_PC_SAMPLING_CONFIGURATION_ATTR_TYPE_SCRATCH_BUFFER_SIZE                = 3,
+  /**
+   * [rw] Size of HW buffer in bytes
+   * DEFAULT - 512 MB
+   * If sampling period is too less, HW buffer can overflow
+   * and drop PC data
+   * Value is a size_t
+   */
+  CUPTI_PC_SAMPLING_CONFIGURATION_ATTR_TYPE_HARDWARE_BUFFER_SIZE               = 4,
+  /**
+   * [rw] PC Sampling collection mode
+   * DEFAULT - CUPTI_PC_SAMPLING_COLLECTION_MODE_CONTINUOUS
+   * Input value should be of type \ref CUpti_PCSamplingCollectionMode.
+   */
+  CUPTI_PC_SAMPLING_CONFIGURATION_ATTR_TYPE_COLLECTION_MODE                    = 5,
+  /**
+   * [rw] Control over PC Sampling data collection range
+   * Default - 0
+   * 1 - Allows user to start and stop PC Sampling using APIs -
+   * \ref cuptiPCSamplingStart() - Start PC Sampling
+   * \ref cuptiPCSamplingStop() - Stop PC Sampling
+   * Value is a uint32_t
+   */
+  CUPTI_PC_SAMPLING_CONFIGURATION_ATTR_TYPE_ENABLE_START_STOP_CONTROL          = 6,
+  /**
+   * [w] Value for output data format
+   * Default - CUPTI_PC_SAMPLING_OUTPUT_DATA_FORMAT_PARSED
+   * Input value should be of type \ref CUpti_PCSamplingOutputDataFormat.
+   */
+  CUPTI_PC_SAMPLING_CONFIGURATION_ATTR_TYPE_OUTPUT_DATA_FORMAT                 = 7,
+  /**
+   * [w] Data buffer to hold collected PC Sampling data PARSED_DATA
+   * Default - none.
+   * Buffer type is void * which can point to PARSED_DATA
+   * Refer \ref CUpti_PCSamplingData for buffer format for PARSED_DATA
+   */
+  CUPTI_PC_SAMPLING_CONFIGURATION_ATTR_TYPE_SAMPLING_DATA_BUFFER               = 8,
+  /**
+   * [rw] Control sleep time of the worker threads created by CUPTI for various PC sampling operations.
+   * CUPTI creates multiple worker threads to offload certain operations to these threads. This includes decoding of HW data to
+   * the CUPTI PC sampling data and correlating PC data to SASS instructions. CUPTI wakes up these threads periodically.
+   * Default - 100 milliseconds.
+   * Value is a uint32_t
+   */
+  CUPTI_PC_SAMPLING_CONFIGURATION_ATTR_TYPE_WORKER_THREAD_PERIODIC_SLEEP_SPAN  = 9,
+  CUPTI_PC_SAMPLING_CONFIGURATION_ATTR_TYPE_FORCE_INT                          = 0x7fffffff,
+} CUpti_PCSamplingConfigurationAttributeType;
+
+/**
+ * \brief PC sampling configuration information structure
+ *
+ * This structure provides \ref CUpti_PCSamplingConfigurationAttributeType which can be configured
+ * or queried for PC sampling configuration
+ */
+typedef struct
+{
+  /**
+   * Refer \ref CUpti_PCSamplingConfigurationAttributeType for all supported attribute types
+   */
+  CUpti_PCSamplingConfigurationAttributeType attributeType;
+  /*
+   * Configure or query status for \p attributeType
+   * CUPTI_SUCCESS for valid \p attributeType and \p attributeData
+   * CUPTI_ERROR_INVALID_OPERATION if \p attributeData is not valid
+   * CUPTI_ERROR_INVALID_PARAMETER if \p attributeType is not valid
+   */
+  CUptiResult attributeStatus;
+  union
+  {
+    /**
+     * Invalid Value
+     */
+    struct
+    {
+      uint64_t data[3];
+    } invalidData;
+    /**
+     * Refer \ref CUPTI_PC_SAMPLING_CONFIGURATION_ATTR_TYPE_SAMPLING_PERIOD
+     */
+    struct
+    {
+      uint32_t samplingPeriod;
+    } samplingPeriodData;
+    /**
+     * Refer \ref CUPTI_PC_SAMPLING_CONFIGURATION_ATTR_TYPE_STALL_REASON
+     */
+    struct
+    {
+      size_t stallReasonCount;
+      uint32_t *pStallReasonIndex;
+    } stallReasonData;
+    /**
+     * Refer \ref CUPTI_PC_SAMPLING_CONFIGURATION_ATTR_TYPE_SCRATCH_BUFFER_SIZE
+     */
+    struct
+    {
+      size_t scratchBufferSize;
+    } scratchBufferSizeData;
+    /**
+     * Refer \ref CUPTI_PC_SAMPLING_CONFIGURATION_ATTR_TYPE_HARDWARE_BUFFER_SIZE
+     */
+    struct
+    {
+      size_t hardwareBufferSize;
+    } hardwareBufferSizeData;
+    /**
+     * Refer \ref CUPTI_PC_SAMPLING_CONFIGURATION_ATTR_TYPE_COLLECTION_MODE
+     */
+    struct
+    {
+      CUpti_PCSamplingCollectionMode collectionMode;
+    } collectionModeData;
+    /**
+     * Refer \ref CUPTI_PC_SAMPLING_CONFIGURATION_ATTR_TYPE_ENABLE_START_STOP_CONTROL
+     */
+    struct
+    {
+      uint32_t enableStartStopControl;
+    } enableStartStopControlData;
+    /**
+     * Refer \ref CUPTI_PC_SAMPLING_CONFIGURATION_ATTR_TYPE_OUTPUT_DATA_FORMAT
+     */
+    struct
+    {
+      CUpti_PCSamplingOutputDataFormat outputDataFormat;
+    } outputDataFormatData;
+    /**
+     * Refer \ref CUPTI_PC_SAMPLING_CONFIGURATION_ATTR_TYPE_SAMPLING_DATA_BUFFER
+     */
+    struct
+    {
+      void *samplingDataBuffer;
+    } samplingDataBufferData;
+    /**
+     * Refer \ref CUPTI_PC_SAMPLING_CONFIGURATION_ATTR_TYPE_WORKER_THREAD_PERIODIC_SLEEP_SPAN
+     */
+    struct
+    {
+      uint32_t workerThreadPeriodicSleepSpan;
+    } workerThreadPeriodicSleepSpanData;
+    
+  } attributeData;
+} CUpti_PCSamplingConfigurationInfo;
+
+/**
+ * \brief PC sampling configuration structure
+ *
+ * This structure configures PC sampling using \ref cuptiPCSamplingSetConfigurationAttribute
+ * and queries PC sampling default configuration using \ref cuptiPCSamplingGetConfigurationAttribute
+ */
+typedef struct
+{
+  /**
+   * [w] Size of the data structure i.e. CUpti_PCSamplingConfigurationInfoParamsSize
+   * CUPTI client should set the size of the structure. It will be used in CUPTI to check what fields are
+   * available in the structure. Used to preserve backward compatibility.
+   */
+  size_t size;
+  /**
+   * [w] Assign to NULL
+   */
+  void* pPriv;
+  /**
+   * [w] CUcontext
+   */
+  CUcontext ctx;
+  /**
+   * [w] Number of attributes to configure using \ref cuptiPCSamplingSetConfigurationAttribute or query
+   * using \ref cuptiPCSamplingGetConfigurationAttribute
+   */
+  size_t numAttributes;
+  /**
+   * Refer \ref CUpti_PCSamplingConfigurationInfo
+   */
+  CUpti_PCSamplingConfigurationInfo *pPCSamplingConfigurationInfo;
+} CUpti_PCSamplingConfigurationInfoParams;
+#define CUpti_PCSamplingConfigurationInfoParamsSize                 CUPTI_PCSAMPLING_STRUCT_SIZE(CUpti_PCSamplingConfigurationInfoParams,pPCSamplingConfigurationInfo)
+
+/**
+ * \brief Write PC Sampling configuration attribute.
+ *
+ * \param pParams A pointer to \ref CUpti_PCSamplingConfigurationInfoParams
+ * containing PC sampling configuration.
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_OPERATION if this API is called with
+ * some invalid \p attrib.
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if attribute \p value is not valid
+ * or any \p pParams is not valid
+ * \retval CUPTI_ERROR_NOT_SUPPORTED indicates that the system/device
+ * does not support the API
+ */
+CUptiResult CUPTIAPI cuptiPCSamplingSetConfigurationAttribute(CUpti_PCSamplingConfigurationInfoParams *pParams);
+
+/**
+ * \brief Read PC Sampling configuration attribute.
+ *
+ * \param pParams A pointer to \ref CUpti_PCSamplingConfigurationInfoParams
+ * containing PC sampling configuration.
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_OPERATION if this API is called with
+ * some invalid attribute.
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p attrib is not valid
+ * or any \p pParams is not valid
+ * \retval CUPTI_ERROR_PARAMETER_SIZE_NOT_SUFFICIENT indicates that
+ * the \p value buffer is too small to hold the attribute value
+ * \retval CUPTI_ERROR_NOT_SUPPORTED indicates that the system/device
+ * does not support the API
+ */
+CUptiResult CUPTIAPI cuptiPCSamplingGetConfigurationAttribute(CUpti_PCSamplingConfigurationInfoParams *pParams);
+
+/**
+ * \brief Params for cuptiPCSamplingEnable
+ */
+typedef struct
+{
+  /**
+   * [w] Size of the data structure i.e. CUpti_PCSamplingGetDataParamsSize
+   * CUPTI client should set the size of the structure. It will be used in CUPTI to check what fields are
+   * available in the structure. Used to preserve backward compatibility.
+   */
+  size_t size;
+  /**
+   * [w] Assign to NULL
+   */
+  void* pPriv;
+  /**
+   * [w] CUcontext
+   */
+  CUcontext ctx;
+  /**
+   * \param pcSamplingData Data buffer to hold collected PC Sampling data PARSED_DATA
+   * Buffer type is void * which can point to PARSED_DATA
+   * Refer \ref CUpti_PCSamplingData for buffer format for PARSED_DATA
+   */
+  void *pcSamplingData;
+} CUpti_PCSamplingGetDataParams;
+#define CUpti_PCSamplingGetDataParamsSize                           CUPTI_PCSAMPLING_STRUCT_SIZE(CUpti_PCSamplingGetDataParams, pcSamplingData)
+/**
+ * \brief Flush GPU PC sampling data periodically.
+ *
+ * Flushing of GPU PC Sampling data is required at following point to maintain uniqueness of PCs:
+ * For \brief CUPTI_PC_SAMPLING_COLLECTION_MODE_CONTINUOUS, after every module load-unload-load
+ * For \brief CUPTI_PC_SAMPLING_COLLECTION_MODE_KERNEL_SERIALIZED, after every kernel ends
+ * If configuration option \brief CUPTI_PC_SAMPLING_CONFIGURATION_ATTR_TYPE_ENABLE_START_STOP_CONTROL
+ * is enabled, then after every range end i.e. \brief cuptiPCSamplingStop()
+ *
+ * If application is profiled in \brief CUPTI_PC_SAMPLING_COLLECTION_MODE_CONTINUOUS, with disabled
+ * \brief CUPTI_PC_SAMPLING_CONFIGURATION_ATTR_TYPE_ENABLE_START_STOP_CONTROL, and there is no module unload,
+ * user can collect data in two ways:
+ * Use \brief cuptiPCSamplingGetData() API periodically
+ * Use \brief cuptiPCSamplingDisable() on application exit and read GPU PC sampling data from sampling
+ * data buffer passed during configuration.
+ * Note: In case, \brief cuptiPCSamplingGetData() API is not called periodically, then sampling data buffer
+ * passed during configuration should be large enough to hold all PCs data.
+ *       \brief cuptiPCSamplingGetData() API never does device synchronization.
+ *       It is possible that when the API is called there is some unconsumed data from the HW buffer. In this case
+ * CUPTI provides only the data available with it at that moment.
+ *
+ * \param pParams A pointer to \ref CUpti_PCSamplingGetDataParams
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_OPERATION if this API is called without
+ * enabling PC sampling.
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_NOT_SUPPORTED indicates that the system/device
+ * \retval CUPTI_ERROR_OUT_OF_MEMORY indicates that the HW buffer is full
+ * does not support the API
+ */
+CUptiResult CUPTIAPI cuptiPCSamplingGetData(CUpti_PCSamplingGetDataParams *pParams);
+
+/**
+ * \brief Params for cuptiPCSamplingEnable
+ */
+typedef struct
+{
+  /**
+   * [w] Size of the data structure i.e. CUpti_PCSamplingEnableParamsSize
+   * CUPTI client should set the size of the structure. It will be used in CUPTI to check what fields are
+   * available in the structure. Used to preserve backward compatibility.
+   */
+  size_t size;
+  /**
+   * [w] Assign to NULL
+   */
+  void* pPriv;
+  /**
+   * [w] CUcontext
+   */
+  CUcontext ctx;
+} CUpti_PCSamplingEnableParams;
+#define CUpti_PCSamplingEnableParamsSize                           CUPTI_PCSAMPLING_STRUCT_SIZE(CUpti_PCSamplingEnableParams, ctx)
+
+/**
+ * \brief Enable PC sampling.
+ *
+ * \param pParams A pointer to \ref CUpti_PCSamplingEnableParams
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_NOT_SUPPORTED indicates that the system/device
+ * does not support the API
+ */
+CUptiResult CUPTIAPI cuptiPCSamplingEnable(CUpti_PCSamplingEnableParams *pParams);
+
+/**
+ * \brief Params for cuptiPCSamplingDisable
+ */
+typedef struct
+{
+  /**
+   * [w] Size of the data structure i.e. CUpti_PCSamplingDisableParamsSize
+   * CUPTI client should set the size of the structure. It will be used in CUPTI to check what fields are
+   * available in the structure. Used to preserve backward compatibility.
+   */
+  size_t size;
+  /**
+   * [w] Assign to NULL
+   */
+  void* pPriv;
+  /**
+   * [w] CUcontext
+   */
+  CUcontext ctx;
+} CUpti_PCSamplingDisableParams;
+#define CUpti_PCSamplingDisableParamsSize                           CUPTI_PCSAMPLING_STRUCT_SIZE(CUpti_PCSamplingDisableParams, ctx)
+
+/**
+ * \brief Disable PC sampling.
+ *
+ * For application which doesn't destroy the CUDA context explicitly,
+ * this API does the PC Sampling tear-down, joins threads and copies PC records in the buffer provided
+ * during the PC sampling configuration. PC records which can't be accommodated in the buffer are discarded.
+ *
+ * \param pParams A pointer to \ref CUpti_PCSamplingDisableParams
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_NOT_SUPPORTED indicates that the system/device
+ * does not support the API
+ */
+CUptiResult CUPTIAPI cuptiPCSamplingDisable(CUpti_PCSamplingDisableParams *pParams);
+
+/**
+ * \brief Params for cuptiPCSamplingStart
+ */
+typedef struct
+{
+  /**
+   * [w] Size of the data structure i.e. CUpti_PCSamplingStartParamsSize
+   * CUPTI client should set the size of the structure. It will be used in CUPTI to check what fields are
+   * available in the structure. Used to preserve backward compatibility.
+   */
+  size_t size;
+  /**
+   * [w] Assign to NULL
+   */
+  void* pPriv;
+  /**
+   * [w] CUcontext
+   */
+  CUcontext ctx;
+} CUpti_PCSamplingStartParams;
+#define CUpti_PCSamplingStartParamsSize                             CUPTI_PCSAMPLING_STRUCT_SIZE(CUpti_PCSamplingStartParams, ctx)
+
+/**
+ * \brief Start PC sampling.
+ *
+ * User can collect PC Sampling data for user-defined range specified by Start/Stop APIs.
+ * This API can be used to mark starting of range. Set configuration option
+ * \brief CUPTI_PC_SAMPLING_CONFIGURATION_ATTR_TYPE_ENABLE_START_STOP_CONTROL to use this API.
+ *
+ * \param pParams A pointer to \ref CUpti_PCSamplingStartParams
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_OPERATION if this API is called with
+ * incorrect PC Sampling configuration.
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_NOT_SUPPORTED indicates that the system/device
+ * does not support the API
+ */
+CUptiResult CUPTIAPI cuptiPCSamplingStart(CUpti_PCSamplingStartParams *pParams);
+
+/**
+ * \brief Params for cuptiPCSamplingStop
+ */
+typedef struct
+{
+  /**
+   * [w] Size of the data structure i.e. CUpti_PCSamplingStopParamsSize
+   * CUPTI client should set the size of the structure. It will be used in CUPTI to check what fields are
+   * available in the structure. Used to preserve backward compatibility.
+   */
+  size_t size;
+  /**
+   * [w] Assign to NULL
+   */
+  void* pPriv;
+  /**
+   * [w] CUcontext
+   */
+  CUcontext ctx;
+} CUpti_PCSamplingStopParams;
+#define CUpti_PCSamplingStopParamsSize                              CUPTI_PCSAMPLING_STRUCT_SIZE(CUpti_PCSamplingStopParams, ctx)
+
+/**
+ * \brief Stop PC sampling.
+ *
+ * User can collect PC Sampling data for user-defined range specified by Start/Stop APIs.
+ * This API can be used to mark end of range. Set configuration option
+ * \brief CUPTI_PC_SAMPLING_CONFIGURATION_ATTR_TYPE_ENABLE_START_STOP_CONTROL to use this API.
+ *
+ * \param pParams A pointer to \ref CUpti_PCSamplingStopParams
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_OPERATION if this API is called with
+ * incorrect PC Sampling configuration.
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_NOT_SUPPORTED indicates that the system/device
+ * does not support the API
+ */
+CUptiResult CUPTIAPI cuptiPCSamplingStop(CUpti_PCSamplingStopParams *pParams);
+
+/**
+ * \brief Params for cuptiPCSamplingGetNumStallReasons
+ */
+typedef struct
+{
+  /**
+   * [w] Size of the data structure i.e. CUpti_PCSamplingGetNumStallReasonsParamsSize
+   * CUPTI client should set the size of the structure. It will be used in CUPTI to check what fields are
+   * available in the structure. Used to preserve backward compatibility.
+   */
+  size_t size;
+  /**
+   * [w] Assign to NULL
+   */
+  void* pPriv;
+  /**
+   * [w] CUcontext
+   */
+  CUcontext ctx;
+  /**
+   * [r] Number of stall reasons
+   */
+  size_t *numStallReasons;
+} CUpti_PCSamplingGetNumStallReasonsParams;
+#define CUpti_PCSamplingGetNumStallReasonsParamsSize                CUPTI_PCSAMPLING_STRUCT_SIZE(CUpti_PCSamplingGetNumStallReasonsParams, numStallReasons)
+
+/**
+ * \brief Get PC sampling stall reason count.
+ *
+ * \param pParams A pointer to \ref CUpti_PCSamplingGetNumStallReasonsParams
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_NOT_SUPPORTED indicates that the system/device
+ * does not support the API
+ */
+CUptiResult CUPTIAPI cuptiPCSamplingGetNumStallReasons(CUpti_PCSamplingGetNumStallReasonsParams *pParams);
+
+/**
+ * \brief Params for cuptiPCSamplingGetStallReasons
+ */
+typedef struct
+{
+  /**
+   * [w] Size of the data structure i.e. CUpti_PCSamplingGetStallReasonsParamsSize
+   * CUPTI client should set the size of the structure. It will be used in CUPTI to check what fields are
+   * available in the structure. Used to preserve backward compatibility.
+   */
+  size_t size;
+  /**
+   * [w] Assign to NULL
+   */
+  void* pPriv;
+  /**
+   * [w] CUcontext
+   */
+  CUcontext ctx;
+  /**
+   * [w] Number of stall reasons
+   */
+  size_t numStallReasons;
+  /**
+   * [r] Stall reason index
+   */
+  uint32_t *stallReasonIndex;
+  /**
+   * [r] Stall reasons name
+   */
+  char **stallReasons;
+} CUpti_PCSamplingGetStallReasonsParams;
+#define CUpti_PCSamplingGetStallReasonsParamsSize                   CUPTI_PCSAMPLING_STRUCT_SIZE(CUpti_PCSamplingGetStallReasonsParams, stallReasons)
+
+/**
+ * \brief Get PC sampling stall reasons.
+ *
+ * \param pParams A pointer to \ref CUpti_PCSamplingGetStallReasonsParams
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_NOT_SUPPORTED indicates that the system/device
+ * does not support the API
+ */
+CUptiResult CUPTIAPI cuptiPCSamplingGetStallReasons(CUpti_PCSamplingGetStallReasonsParams *pParams);
+
+
+/**
+ * \brief Params for cuptiGetSassToSourceCorrelation
+ */
+typedef struct CUpti_GetSassToSourceCorrelationParams {
+  /**
+   * [w] Size of the data structure i.e. CUpti_GetSassToSourceCorrelationParamsSize
+   * CUPTI client should set the size of the structure. It will be used in CUPTI to check what fields are
+   * available in the structure. Used to preserve backward compatibility.
+   */
+  size_t size;
+  /**
+   * [w] Pointer to cubin binary where function belongs.
+   */
+  const void* cubin;
+  /**
+   * [w] Function name to which PC belongs.
+   */
+  const char *functionName;
+  /**
+   * [w] Size of cubin binary.
+   */
+  size_t cubinSize;
+  /**
+   * [r] Line number in the source code.
+   */
+  uint32_t lineNumber;
+  /**
+   * [w] PC offset
+   */
+  uint64_t pcOffset;
+  /**
+   * [r] Path for the source file.
+   */
+  char *fileName;
+  /**
+   * [r] Path for the directory of source file.
+   */
+  char *dirName;
+} CUpti_GetSassToSourceCorrelationParams;
+
+#define CUpti_GetSassToSourceCorrelationParamsSize     CUPTI_PCSAMPLING_STRUCT_SIZE(CUpti_GetSassToSourceCorrelationParams, dirName)
+
+/**
+ * \brief SASS to Source correlation.
+ *
+ * \param pParams A pointer to \ref CUpti_GetSassToSourceCorrelationParams
+ *
+ * It is expected from user to free allocated memory for fileName and dirName after use.
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if either of the parameters cubin or functionName
+ * is NULL or cubinSize is zero or size field is not set correctly.
+ * \retval CUPTI_ERROR_INVALID_MODULE provided cubin is invalid.
+ * \retval CUPTI_ERROR_UNKNOWN an internal error occurred.
+ * This error code is also used for cases when the function is not present in the module.
+ * A better error code will be returned in the future release.
+ */
+CUptiResult CUPTIAPI cuptiGetSassToSourceCorrelation(CUpti_GetSassToSourceCorrelationParams *pParams);
+
+/**
+ * \brief Params for cuptiGetCubinCrc
+ */
+typedef struct {
+  /**
+   * [w] Size of configuration structure.
+   * CUPTI client should set the size of the structure. It will be used in CUPTI to check what fields are
+   * available in the structure. Used to preserve backward compatibility.
+   */
+  size_t size;
+  /**
+   * [w] Size of cubin binary.
+   */
+  size_t cubinSize;
+  /**
+   * [w] Pointer to cubin binary
+   */
+  const void* cubin;
+  /**
+   * [r] Computed CRC will be stored in it.
+   */
+  uint64_t cubinCrc;
+} CUpti_GetCubinCrcParams;
+#define CUpti_GetCubinCrcParamsSize     CUPTI_PCSAMPLING_STRUCT_SIZE(CUpti_GetCubinCrcParams, cubinCrc)
+
+/**
+ * \brief Get the CRC of cubin.
+ *
+ * This function returns the CRC of provided cubin binary.
+ *
+ * \param pParams A pointer to \ref CUpti_GetCubinCrcParams
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if parameter cubin is NULL or
+ * provided cubinSize is zero or size field is not set.
+ */
+CUptiResult CUPTIAPI cuptiGetCubinCrc(CUpti_GetCubinCrcParams *pParams);
+
+/**
+ * \brief Function type for callback used by CUPTI to request crc of
+ * loaded module.
+ *
+ * This callback function ask for crc of provided module in function.
+ * The provided crc will be stored in PC sampling records i.e. in the field 'cubinCrc' of the PC sampling
+ * struct CUpti_PCSamplingPCData. The CRC is uses during the offline source correlation to uniquely identify the module.
+ *
+ * \param cubin The pointer to cubin binary
+ * \param cubinSize The size of cubin binary.
+ * \param cubinCrc Returns the computed crc of cubin.
+ */
+typedef void (CUPTIAPI *CUpti_ComputeCrcCallbackFunc)(
+    const void* cubin,
+    size_t cubinSize,
+    uint64_t *cubinCrc);
+
+/**
+ * \brief Register callback function with CUPTI to use
+ * your own algorithm to compute cubin crc.
+ *
+ * This function registers a callback function and it gets called
+ * from CUPTI when a CUDA module is loaded.
+ *
+ * \param funcComputeCubinCrc callback is invoked when a CUDA module
+ * is loaded.
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p funcComputeCubinCrc is NULL.
+ */
+CUptiResult CUPTIAPI cuptiRegisterComputeCrcCallback(CUpti_ComputeCrcCallbackFunc funcComputeCubinCrc);
+
+/** @} */ /* END CUPTI_PCSAMPLING_API */
+
+#if defined(__GNUC__) && defined(CUPTI_LIB)
+    #pragma GCC visibility pop
+#endif
+
+#if defined(__cplusplus)
+}
+#endif
+
+#endif /*_CUPTI_PCSAMPLING_H_*/

Prism/LLaDA/LLaDA_Prism/.venv/lib/python3.12/site-packages/nvidia/cuda_cupti/include/cupti_pcsampling_util.h

@@ -0,0 +1,402 @@
+#if !defined(_CUPTI_PCSAMPLING_UTIL_H_)
+#define _CUPTI_PCSAMPLING_UTIL_H_
+
+#include <cupti_pcsampling.h>
+#include <fstream>
+
+#include <cupti_common.h>
+
+#ifndef CUPTI_UTIL_STRUCT_SIZE
+#define CUPTI_UTIL_STRUCT_SIZE(type_, lastfield_)                     (offsetof(type_, lastfield_) + sizeof(((type_*)0)->lastfield_))
+#endif
+
+#ifndef CHECK_PC_SAMPLING_STRUCT_FIELD_EXISTS
+#define CHECK_PC_SAMPLING_STRUCT_FIELD_EXISTS(type, member, structSize)    \
+    (offsetof(type, member) < structSize)
+#endif
+
+#if defined(__cplusplus)
+extern "C" {
+#endif
+
+#if defined(__GNUC__)
+    #pragma GCC visibility push(default)
+#endif
+
+namespace CUPTI { namespace PcSamplingUtil {
+
+/**
+ * \defgroup CUPTI_PCSAMPLING_UTILITY CUPTI PC Sampling Utility API
+ * Functions, types, and enums that implement the CUPTI PC Sampling Utility API.
+ * @{
+ */
+
+/**
+ * \brief Header info will be stored in file.
+ */
+typedef struct PACKED_ALIGNMENT {
+  /**
+   * Version of file format.
+   */
+  uint32_t version;
+  /**
+   * Total number of buffers present in the file.
+   */
+  uint32_t totalBuffers;
+} Header;
+
+/**
+ * \brief BufferInfo will be stored in the file for every buffer
+ *  i.e for every call of UtilDumpPcSamplingBufferInFile() API.
+ */
+typedef struct PACKED_ALIGNMENT {
+  /**
+   * Total number of PC records.
+   */
+  uint64_t recordCount;
+  /**
+   * Count of all stall reasons supported on the GPU
+   */
+  size_t numStallReasons;
+  /**
+   * Total number of stall reasons in single record.
+   */
+  uint64_t numSelectedStallReasons;
+  /**
+   * Buffer size in Bytes.
+   */
+  uint64_t bufferByteSize;
+} BufferInfo;
+
+/**
+ * \brief All available stall reasons name and respective indexes
+ * will be stored in it.
+ */
+typedef struct PACKED_ALIGNMENT {
+  /**
+   * Number of all available stall reasons
+   */
+  size_t numStallReasons;
+  /**
+   * Stall reasons names of all available stall reasons
+   */
+  char **stallReasons;
+  /**
+   * Stall reason index of all available stall reasons
+   */
+  uint32_t *stallReasonIndex;
+} PcSamplingStallReasons;
+
+/**
+ * \brief CUPTI PC sampling buffer types.
+ *
+ */
+typedef enum {
+  /**
+   * Invalid buffer type.
+   */
+  PC_SAMPLING_BUFFER_INVALID             = 0,
+  /**
+   * Refers to CUpti_PCSamplingData buffer.
+   */
+  PC_SAMPLING_BUFFER_PC_TO_COUNTER_DATA  = 1
+} PcSamplingBufferType;
+
+/**
+ * \brief CUPTI PC sampling utility API result codes.
+ *
+ * Error and result codes returned by CUPTI PC sampling utility API.
+ */
+typedef enum {
+  /**
+   * No error
+   */
+  CUPTI_UTIL_SUCCESS                                       = 0,
+  /**
+   * One or more of the parameters are invalid.
+   */
+  CUPTI_UTIL_ERROR_INVALID_PARAMETER                       = 1,
+  /**
+   * Unable to create a new file
+   */
+  CUPTI_UTIL_ERROR_UNABLE_TO_CREATE_FILE                   = 2,
+  /**
+   * Unable to open a file
+   */
+  CUPTI_UTIL_ERROR_UNABLE_TO_OPEN_FILE                     = 3,
+  /**
+   * Read or write operation failed
+   */
+  CUPTI_UTIL_ERROR_READ_WRITE_OPERATION_FAILED             = 4,
+  /**
+   * Provided file handle is corrupted.
+   */
+  CUPTI_UTIL_ERROR_FILE_HANDLE_CORRUPTED                   = 5,
+  /**
+   * seek operation failed.
+   */
+  CUPTI_UTIL_ERROR_SEEK_OPERATION_FAILED                   = 6,
+  /**
+   * Unable to allocate enough memory to perform the requested
+   * operation.
+   */
+  CUPTI_UTIL_ERROR_OUT_OF_MEMORY                           = 7,
+  /**
+   * An unknown internal error has occurred.
+   */
+  CUPTI_UTIL_ERROR_UNKNOWN                                 = 999,
+  CUPTI_UTIL_ERROR_FORCE_INT                               = 0x7fffffff
+} CUptiUtilResult;
+
+/**
+ * \brief Params for \ref CuptiUtilPutPcSampData
+ */
+typedef struct {
+  /**
+   * Size of the data structure i.e. CUpti_PCSamplingDisableParamsSize
+   * CUPTI client should set the size of the structure. It will be used in CUPTI to check what fields are
+   * available in the structure. Used to preserve backward compatibility.
+   */
+  size_t size;
+  /**
+   * Type of buffer to store in file
+   */
+  PcSamplingBufferType bufferType;
+  /**
+   * PC sampling buffer.
+   */
+  void *pSamplingData;
+  /**
+   * Number of configured attributes
+   */
+  size_t numAttributes;
+  /**
+   * Refer \ref CUpti_PCSamplingConfigurationInfo
+   * It is expected to provide configuration details of at least
+   * CUPTI_PC_SAMPLING_CONFIGURATION_ATTR_TYPE_STALL_REASON attribute.
+   */
+  CUpti_PCSamplingConfigurationInfo *pPCSamplingConfigurationInfo;
+  /**
+   * Refer \ref PcSamplingStallReasons.
+   */
+  PcSamplingStallReasons *pPcSamplingStallReasons;
+  /**
+   * File name to store buffer into it.
+   */
+  const char* fileName;
+} CUptiUtil_PutPcSampDataParams;
+#define CUptiUtil_PutPcSampDataParamsSize                   CUPTI_UTIL_STRUCT_SIZE(CUptiUtil_PutPcSampDataParams, fileName)
+
+/**
+ * \brief Dump PC sampling data into the file.
+ *
+ * This API can be called multiple times.
+ * It will append buffer in the file.
+ * For every buffer it will store BufferInfo
+ * so that before retrieving data it will help to allocate buffer
+ * to store retrieved data.
+ * This API creates file if file does not present.
+ * If stallReasonIndex or stallReasons pointer of \ref CUptiUtil_PutPcSampDataParams is NULL
+ * then stall reasons data  will not be stored in file.
+ * It is expected to store all available stall reason data at least once to refer it during
+ * offline correlation.
+ *
+ * \retval CUPTI_UTIL_SUCCESS
+ * \retval CUPTI_UTIL_ERROR_INVALID_PARAMETER error out if buffer type is invalid
+ * or if either of pSamplingData, pParams pointer is NULL or stall reason configuration details not provided
+ * or filename is empty.
+ * \retval CUPTI_UTIL_ERROR_UNABLE_TO_CREATE_FILE
+ * \retval CUPTI_UTIL_ERROR_UNABLE_TO_OPEN_FILE
+ * \retval CUPTI_UTIL_ERROR_READ_WRITE_OPERATION_FAILED
+ */
+CUptiUtilResult CUPTIUTILAPI CuptiUtilPutPcSampData(CUptiUtil_PutPcSampDataParams *pParams);
+
+/**
+ * \brief Params for \ref CuptiUtilGetHeaderData
+ */
+typedef struct {
+  /**
+   * Size of the data structure i.e. CUpti_PCSamplingDisableParamsSize
+   * CUPTI client should set the size of the structure. It will be used in CUPTI to check what fields are
+   * available in the structure. Used to preserve backward compatibility.
+   */
+  size_t size;
+  /**
+   * File handle.
+   */
+  std::ifstream *fileHandler;
+  /**
+   * Header Info.
+   */
+  Header headerInfo;
+
+} CUptiUtil_GetHeaderDataParams;
+#define CUptiUtil_GetHeaderDataParamsSize                   CUPTI_UTIL_STRUCT_SIZE(CUptiUtil_GetHeaderDataParams, headerInfo)
+
+/**
+ * \brief Get header data of file.
+ *
+ * This API must be called once initially while retrieving data from file.
+ * \ref Header structure, it gives info about total number
+ * of buffers present in the file.
+ *
+ * \retval CUPTI_UTIL_SUCCESS
+ * \retval CUPTI_UTIL_ERROR_INVALID_PARAMETER error out if either of pParam or fileHandle is NULL or param struct size is incorrect.
+ * \retval CUPTI_UTIL_ERROR_FILE_HANDLE_CORRUPTED file handle is not in good state to read data from file
+ * \retval CUPTI_UTIL_ERROR_READ_WRITE_OPERATION_FAILED  failed to read data from file.
+ */
+CUptiUtilResult CUPTIUTILAPI CuptiUtilGetHeaderData(CUptiUtil_GetHeaderDataParams *pParams);
+
+/**
+ * \brief Params for \ref CuptiUtilGetBufferInfo
+ */
+typedef struct {
+  /**
+   * Size of the data structure i.e. CUpti_PCSamplingDisableParamsSize
+   * CUPTI client should set the size of the structure. It will be used in CUPTI to check what fields are
+   * available in the structure. Used to preserve backward compatibility.
+   */
+  size_t size;
+  /**
+   * File handle.
+   */
+  std::ifstream *fileHandler;
+  /**
+   * Buffer Info.
+   */
+  BufferInfo bufferInfoData;
+} CUptiUtil_GetBufferInfoParams;
+#define CUptiUtil_GetBufferInfoParamsSize                   CUPTI_UTIL_STRUCT_SIZE(CUptiUtil_GetBufferInfoParams, bufferInfoData)
+
+/**
+ * \brief Get buffer info data of file.
+ *
+ * This API must be called every time before calling CuptiUtilGetPcSampData API.
+ * \ref BufferInfo structure, it gives info about recordCount and stallReasonCount
+ * of every record in the buffer. This will help to allocate exact buffer to retrieve data into it.
+ *
+ * \retval CUPTI_UTIL_SUCCESS
+ * \retval CUPTI_UTIL_ERROR_INVALID_PARAMETER error out if either of pParam or fileHandle is NULL or param struct size is incorrect.
+ * \retval CUPTI_UTIL_ERROR_FILE_HANDLE_CORRUPTED file handle is not in good state to read data from file.
+ * \retval CUPTI_UTIL_ERROR_READ_WRITE_OPERATION_FAILED failed to read data from file.
+ */
+CUptiUtilResult CUPTIUTILAPI CuptiUtilGetBufferInfo(CUptiUtil_GetBufferInfoParams *pParams);
+
+/**
+ * \brief Params for \ref CuptiUtilGetPcSampData
+ */
+typedef struct {
+  /**
+   * Size of the data structure i.e. CUpti_PCSamplingDisableParamsSize
+   * CUPTI client should set the size of the structure. It will be used in CUPTI to check what fields are
+   * available in the structure. Used to preserve backward compatibility.
+   */
+  size_t size;
+  /**
+   * File handle.
+   */
+  std::ifstream *fileHandler;
+  /**
+   * Type of buffer to store in file
+   */
+  PcSamplingBufferType bufferType;
+  /**
+   * Pointer to collected buffer info using \ref CuptiUtilGetBufferInfo
+   */
+  BufferInfo *pBufferInfoData;
+  /**
+   * Pointer to allocated memory to store retrieved data from file.
+   */
+  void *pSamplingData;
+  /**
+   * Number of configuration attributes
+   */
+  size_t numAttributes;
+  /**
+   * Refer \ref CUpti_PCSamplingConfigurationInfo
+   */
+  CUpti_PCSamplingConfigurationInfo *pPCSamplingConfigurationInfo;
+  /**
+   * Refer \ref PcSamplingStallReasons.
+   * For stallReasons field of \ref PcSamplingStallReasons it is expected to
+   * allocate memory for each string element of array.
+   */
+  PcSamplingStallReasons *pPcSamplingStallReasons;
+} CUptiUtil_GetPcSampDataParams;
+#define CUptiUtil_GetPcSampDataParamsSize                   CUPTI_UTIL_STRUCT_SIZE(CUptiUtil_GetPcSampDataParams, pPcSamplingStallReasons)
+
+/**
+ * \brief Retrieve PC sampling data from file into allocated buffer.
+ *
+ * This API must be called after CuptiUtilGetBufferInfo API.
+ * It will retrieve data from file into allocated buffer.
+ *
+ * \retval CUPTI_UTIL_SUCCESS
+ * \retval CUPTI_UTIL_ERROR_INVALID_PARAMETER error out if buffer type is invalid
+ * or if either of pSampData, pParams is NULL. If pPcSamplingStallReasons is not NULL then
+ * error out if either of stallReasonIndex, stallReasons or stallReasons array element pointer is NULL.
+ * or filename is empty.
+ * \retval CUPTI_UTIL_ERROR_READ_WRITE_OPERATION_FAILED
+ * \retval CUPTI_UTIL_ERROR_FILE_HANDLE_CORRUPTED file handle is not in good state to read data from file.
+ */
+CUptiUtilResult CUPTIUTILAPI CuptiUtilGetPcSampData(CUptiUtil_GetPcSampDataParams *pParams);
+
+/**
+ * \brief Params for \ref CuptiUtilMergePcSampData
+ */
+typedef struct
+{
+  /**
+   * Size of the data structure i.e. CUpti_PCSamplingDisableParamsSize
+   * CUPTI client should set the size of the structure. It will be used in CUPTI to check what fields are
+   * available in the structure. Used to preserve backward compatibility.
+   */
+  size_t size;
+  /**
+   * Number of buffers to merge.
+   */
+  size_t numberOfBuffers;
+  /**
+   * Pointer to array of buffers to merge
+   */
+  CUpti_PCSamplingData *PcSampDataBuffer;
+  /**
+   * Pointer to array of merged buffers as per the range id.
+   */
+  CUpti_PCSamplingData **MergedPcSampDataBuffers;
+  /**
+   * Number of merged buffers.
+   */
+  size_t *numMergedBuffer;
+} CUptiUtil_MergePcSampDataParams;
+#define CUptiUtil_MergePcSampDataParamsSize                   CUPTI_UTIL_STRUCT_SIZE(CUptiUtil_MergePcSampDataParams, numMergedBuffer)
+
+/**
+ * \brief Merge PC sampling data range id wise.
+ *
+ * This API merge PC sampling data range id wise.
+ * It allocates memory for merged data and fill data in it
+ * and provide buffer pointer in MergedPcSampDataBuffers field.
+ * It is expected from user to free merge data buffers after use.
+ *
+ * \retval CUPTI_UTIL_SUCCESS
+ * \retval CUPTI_UTIL_ERROR_INVALID_PARAMETER error out if param struct size is invalid
+ * or count of buffers to merge is invalid i.e less than 1
+ * or either of PcSampDataBuffer, MergedPcSampDataBuffers, numMergedBuffer is NULL
+ * \retval CUPTI_UTIL_ERROR_OUT_OF_MEMORY Unable to allocate memory for merged buffer.
+ */
+CUptiUtilResult CUPTIUTILAPI CuptiUtilMergePcSampData(CUptiUtil_MergePcSampDataParams *pParams);
+
+/** @} */ /* END CUPTI_PCSAMPLING_UTILITY */
+
+} }
+
+#if defined(__GNUC__)
+    #pragma GCC visibility pop
+#endif
+
+#if defined(__cplusplus)
+}
+#endif
+
+#endif

Prism/LLaDA/LLaDA_Prism/.venv/lib/python3.12/site-packages/nvidia/cuda_cupti/include/cupti_pmsampling.h

@@ -0,0 +1,490 @@
+/*
+ * Copyright 2024 NVIDIA Corporation. All rights reserved.
+ *
+ * NOTICE TO LICENSEE:
+ *
+ * This source code and/or documentation ("Licensed Deliverables") are
+ * subject to NVIDIA intellectual property rights under U.S. and
+ * international Copyright laws.
+ *
+ * These Licensed Deliverables contained herein is PROPRIETARY and
+ * CONFIDENTIAL to NVIDIA and is being provided under the terms and
+ * conditions of a form of NVIDIA software license agreement by and
+ * between NVIDIA and Licensee ("License Agreement") or electronically
+ * accepted by Licensee.  Notwithstanding any terms or conditions to
+ * the contrary in the License Agreement, reproduction or disclosure
+ * of the Licensed Deliverables to any third party without the express
+ * written consent of NVIDIA is prohibited.
+ *
+ * NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE
+ * LICENSE AGREEMENT, NVIDIA MAKES NO REPRESENTATION ABOUT THE
+ * SUITABILITY OF THESE LICENSED DELIVERABLES FOR ANY PURPOSE.  IT IS
+ * PROVIDED "AS IS" WITHOUT EXPRESS OR IMPLIED WARRANTY OF ANY KIND.
+ * NVIDIA DISCLAIMS ALL WARRANTIES WITH REGARD TO THESE LICENSED
+ * DELIVERABLES, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY,
+ * NONINFRINGEMENT, AND FITNESS FOR A PARTICULAR PURPOSE.
+ * NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE
+ * LICENSE AGREEMENT, IN NO EVENT SHALL NVIDIA BE LIABLE FOR ANY
+ * SPECIAL, INDIRECT, INCIDENTAL, OR CONSEQUENTIAL DAMAGES, OR ANY
+ * DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS,
+ * WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS
+ * ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE
+ * OF THESE LICENSED DELIVERABLES.
+ *
+ * U.S. Government End Users.  These Licensed Deliverables are a
+ * "commercial item" as that term is defined at 48 C.F.R. 2.101 (OCT
+ * 1995), consisting of "commercial computer software" and "commercial
+ * computer software documentation" as such terms are used in 48
+ * C.F.R. 12.212 (SEPT 1995) and is provided to the U.S. Government
+ * only as a commercial end item.  Consistent with 48 C.F.R.12.212 and
+ * 48 C.F.R. 227.7202-1 through 227.7202-4 (JUNE 1995), all
+ * U.S. Government End Users acquire the Licensed Deliverables with
+ * only those rights set forth herein.
+ *
+ * Any use of the Licensed Deliverables in individual and commercial
+ * software must include, in the user documentation and internal
+ * comments to the code, the above Disclaimer and U.S. Government End
+ * Users Notice.
+ */
+
+#if !defined(_CUPTI_PMSAMPLING_H_)
+#define _CUPTI_PMSAMPLING_H_
+
+#include <cuda.h>
+#include <cupti_result.h>
+#include <stddef.h>
+#include <stdint.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#if defined(__GNUC__) && defined(CUPTI_LIB)
+    #pragma GCC visibility push(default)
+#endif
+
+#ifndef CUPTI_PROFILER_STRUCT_SIZE
+#define CUPTI_PROFILER_STRUCT_SIZE(type_, lastfield_)                     (offsetof(type_, lastfield_) + sizeof(((type_*)0)->lastfield_))
+#endif
+
+/* CUPTI PM sampling APIs */
+/**
+ * \defgroup CUPTI_PM_SAMPLING_API CUPTI PM Sampling API
+ * Functions to enable, disable, start, stop, and decode PM sampling.
+ * @{
+ */
+typedef struct CUpti_PmSampling_Object CUpti_PmSampling_Object;
+
+typedef enum CUpti_PmSampling_TriggerMode
+{
+    /// The trigger is based off of the SYSCLK frequency, note SYS frequency by default is variable.
+    /// the sample interval (set in the struct CUpti_PmSampling_SetConfig_Params) is in terms of clocks.
+    CUPTI_PM_SAMPLING_TRIGGER_MODE_GPU_SYSCLK_INTERVAL = 0,
+    /// The trigger is based off of a fixed frequency source.
+    /// The sample interval (set in the struct CUpti_PmSampling_SetConfig_Params) is in terms of nanoseconds.
+    /// Note: This trigger mode is not supported on Turing GPU architecture and GA100 GPU.
+    /// It is supported on Ampere GA10x and later GPU architectures.
+    CUPTI_PM_SAMPLING_TRIGGER_MODE_GPU_TIME_INTERVAL = 1,
+    CUPTI_PM_SAMPLING_TRIGGER_MODE_COUNT
+} CUpti_PmSampling_TriggerMode;
+
+typedef enum CUpti_PmSampling_DecodeStopReason
+{
+    CUPTI_PM_SAMPLING_DECODE_STOP_REASON_OTHER = 0,
+    /// Counter data image is full.
+    CUPTI_PM_SAMPLING_DECODE_STOP_REASON_COUNTER_DATA_FULL,
+    /// All the records in the hardware buffer is decoded.
+    CUPTI_PM_SAMPLING_DECODE_STOP_REASON_END_OF_RECORDS,
+    CUPTI_PM_SAMPLING_DECODE_STOP_REASON_COUNT
+} CUpti_PmSampling_DecodeStopReason;
+
+typedef enum CUpti_PmSampling_HardwareBuffer_AppendMode
+{
+    /// Keep the oldest records in the hardware buffer.
+    /// CUPTI will report error for overflow in case hardware buffer is getting filled up.
+    CUPTI_PM_SAMPLING_HARDWARE_BUFFER_APPEND_MODE_KEEP_OLDEST = 0,
+    /// Keep the latest records in the hardware buffer.
+    /// Note: This mode is not supported on Turing GPU architecture.
+    /// It is supported on Ampere and later GPU architectures.
+    CUPTI_PM_SAMPLING_HARDWARE_BUFFER_APPEND_MODE_KEEP_LATEST = 1
+} CUpti_PmSampling_HardwareBuffer_AppendMode;
+
+/**
+ * \brief Params for cuptiPmSamplingSetConfig
+ */
+typedef struct CUpti_PmSampling_SetConfig_Params
+{
+    /// [in] Size of the data structure.
+    size_t structSize;
+    /// [in] Set to NULL.
+    void* pPriv;
+    /// [in] PM sampling object.
+    CUpti_PmSampling_Object* pPmSamplingObject;
+    /// [in] Size of the config image.
+    size_t configSize;
+    /// [in] Config image.
+    const uint8_t* pConfig;
+    /// [in] The hardware buffer size in which raw PM sampling data
+    /// will be stored. These samples will be decoded to counter data
+    /// image with \ref cuptiPmSamplingDecodeData call.
+    size_t hardwareBufferSize;
+    /// [in] For the trigger mode `CUPTI_PM_SAMPLING_TRIGGER_MODE_GPU_SYSCLK_INTERVAL`, sampling interval
+    /// is the number of sys clock cycles. For the trigger mode `CUPTI_PM_SAMPLING_TRIGGER_MODE_GPU_TIME_INTERVAL`,
+    /// sampling interval is in nanoseconds.
+    uint64_t samplingInterval;
+    /// [in] Trigger mode.
+    /// Note: CUPTI_PM_SAMPLING_TRIGGER_MODE_GPU_TIME_INTERVAL is not supported in Turing and GA100.
+    /// Supported from GA10x onwards.
+    CUpti_PmSampling_TriggerMode triggerMode;
+    /// [in] Append mode for the records in hardware buffer.
+    /// For KEEP_OLDEST mode, all the records will be kept in the buffer and in case hardware buffer is getting filled up.
+    /// overflow will be set to 1 in \ref CUpti_PmSampling_DecodeData_Params. For KEEP_LATEST mode, the new records will
+    /// overwrite the oldest records in the buffer in case of filled buffer.
+    CUpti_PmSampling_HardwareBuffer_AppendMode hwBufferAppendMode;
+} CUpti_PmSampling_SetConfig_Params;
+
+#define CUpti_PmSampling_SetConfig_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_PmSampling_SetConfig_Params, hwBufferAppendMode)
+
+/**
+ * \brief Set the configuration for PM sampling like sampling interval, maximum number of samples
+ * filled in HW buffer, trigger mode and the config image which has scheduling info for metric collection.
+ *
+ * \param pParams A pointer to \ref CUpti_PmSampling_SetConfig_Params
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_NOT_SUPPORTED for config image which require multiple passes for data collection
+ */
+CUptiResult CUPTIAPI cuptiPmSamplingSetConfig(CUpti_PmSampling_SetConfig_Params* pParams);
+
+/**
+ * \brief Params for cuptiPmSamplingEnable
+ */
+typedef struct CUpti_PmSampling_Enable_Params
+{
+    /// [in] Size of the data structure.
+    size_t structSize;
+    /// [in] Set to NULL.
+    void* pPriv;
+    /// [in] Device index.
+    size_t deviceIndex;
+    /// [out] PM sampling object.
+    CUpti_PmSampling_Object* pPmSamplingObject;
+} CUpti_PmSampling_Enable_Params;
+
+#define CUpti_PmSampling_Enable_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_PmSampling_Enable_Params, pPmSamplingObject)
+
+/**
+ * \brief Create a PM sampling object and enable PM sampling on the CUDA device.
+ *
+ * \param pParams A pointer to \ref CUpti_PmSampling_Enable_Params
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_OUT_OF_MEMORY if memory allocation fails while creating the PM sampling object
+ * \retval CUPTI_ERROR_INVALID_OPERATION if PM sampling is already enabled on the device
+ * \retval CUPTI_ERROR_INSUFFICIENT_PRIVILEGES if the user does not have sufficient privileges to perform the operation
+ * \retval CUPTI_ERROR_UNKNOWN for any internal error
+ */
+CUptiResult CUPTIAPI cuptiPmSamplingEnable(CUpti_PmSampling_Enable_Params* pParams);
+
+/**
+ * \brief Params for cuptiPmSamplingDisable
+ */
+typedef struct CUpti_PmSampling_Disable_Params
+{
+    /// [in] Size of the data structure.
+    size_t structSize;
+    /// [in] Set to NULL.
+    void* pPriv;
+    /// [in] PM sampling object.
+    CUpti_PmSampling_Object* pPmSamplingObject;
+} CUpti_PmSampling_Disable_Params;
+
+#define CUpti_PmSampling_Disable_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_PmSampling_Disable_Params, pPmSamplingObject)
+
+/**
+ * \brief Disable PM sampling on the CUDA device and destroy the PM sampling object.
+ *
+ * \param pParams A pointer to \ref CUpti_PmSampling_Disable_Params
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_UNKNOWN for any internal error
+ */
+CUptiResult CUPTIAPI cuptiPmSamplingDisable(CUpti_PmSampling_Disable_Params* pParams);
+
+/**
+ * \brief Params for cuptiPmSamplingStart
+ */
+typedef struct CUpti_PmSampling_Start_Params
+{
+    /// [in] Size of the data structure.
+    size_t structSize;
+    /// [in] Set to NULL.
+    void* pPriv;
+    /// [in] PM sampling object.
+    CUpti_PmSampling_Object* pPmSamplingObject;
+} CUpti_PmSampling_Start_Params;
+
+#define CUpti_PmSampling_Start_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_PmSampling_Start_Params, pPmSamplingObject)
+
+/**
+ * \brief Start the PM sampling. The GPU will start collecting the metrics data
+ * periodically based on trigger type and sampling interval passed in CUpti_PmSampling_SetConfig_Params.
+ * The collected data will be stored in the hardware buffer.
+ *
+ * \param pParams A pointer to \ref CUpti_PmSampling_Start_Params
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_INVALID_OPERATION if PM sampling Start is called without enabling PM sampling,
+ * and PM sampling is already started
+ * \retval CUPTI_ERROR_UNKNOWN for any internal error
+ */
+CUptiResult CUPTIAPI cuptiPmSamplingStart(CUpti_PmSampling_Start_Params* pParams);
+
+/**
+ * \brief Params for cuptiPmSamplingStop
+ */
+typedef struct CUpti_PmSampling_Stop_Params
+{
+    /// [in] Size of the data structure.
+    size_t structSize;
+    /// [in] Set to NULL.
+    void* pPriv;
+    /// [in] PM sampling object.
+    CUpti_PmSampling_Object* pPmSamplingObject;
+} CUpti_PmSampling_Stop_Params;
+
+#define CUpti_PmSampling_Stop_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_PmSampling_Stop_Params, pPmSamplingObject)
+
+/**
+ * \brief Stop the PM sampling. The GPU will stop collecting the metrics data.
+ *
+ * \param pParams A pointer to \ref CUpti_PmSampling_Stop_Params
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_INVALID_OPERATION if PM sampling Stop is called without enabling PM sampling,
+ * and PM sampling is already stopped
+ * \retval CUPTI_ERROR_UNKNOWN for any internal error
+ */
+CUptiResult CUPTIAPI cuptiPmSamplingStop(CUpti_PmSampling_Stop_Params* pParams);
+
+/**
+ * \brief Params for cuptiPmSamplingDecodeData
+ */
+typedef struct CUpti_PmSampling_DecodeData_Params
+{
+    /// [in] Size of the data structure.
+    size_t structSize;
+    /// [in] Set to NULL.
+    void* pPriv;
+    /// [in] PM sampling object.
+    CUpti_PmSampling_Object* pPmSamplingObject;
+    /// [in] Counter data image.
+    uint8_t* pCounterDataImage;
+    /// [in] Size of the counter data image.
+    size_t counterDataImageSize;
+    /// [out] decode stop reason
+    CUpti_PmSampling_DecodeStopReason decodeStopReason;
+    /// [out] overflow status for hardware buffer.
+    /// To avoid overflow, either increase the maxSamples values in
+    /// \ref CUpti_PmSampling_SetConfig_Params or reduce the sampling interval.
+    uint8_t overflow;
+} CUpti_PmSampling_DecodeData_Params;
+
+#define CUpti_PmSampling_DecodeData_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_PmSampling_DecodeData_Params, overflow)
+
+/**
+ * \brief Decode the metrics data stored in the hardware buffer to the counter data image.
+ *
+ *
+ * \param pParams A pointer to \ref CUpti_PmSampling_DecodeData_Params
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_INVALID_OPERATION if PM sampling DecodeData is called without enabling PM sampling
+ * \retval CUPTI_ERROR_OUT_OF_MEMORY if there is record overflow in the hardware buffer
+ * \retval CUPTI_ERROR_UNKNOWN for any internal error
+ */
+CUptiResult CUPTIAPI cuptiPmSamplingDecodeData(CUpti_PmSampling_DecodeData_Params* pParams);
+
+/**
+ * \brief Params for cuptiPmSamplingGetCounterData
+ */
+typedef struct CUpti_PmSampling_GetCounterAvailability_Params
+{
+    /// [in] Size of the data structure.
+    size_t structSize;
+    /// [in] Set to NULL.
+    void* pPriv;
+    /// [in] Device index.
+    size_t deviceIndex;
+    /// [inout] Size of the counter availability image. When pCounterAvailabilityImage is NULL,
+    /// this field is used to return the size of the counter availability image.
+    size_t counterAvailabilityImageSize;
+    /// [out] Counter availability image.
+    uint8_t* pCounterAvailabilityImage;
+} CUpti_PmSampling_GetCounterAvailability_Params;
+#define CUpti_PmSampling_GetCounterAvailability_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_PmSampling_GetCounterAvailability_Params, pCounterAvailabilityImage)
+
+/**
+ * \brief Query counter availibility information in a buffer which can be used to filter unavailable raw metrics on host.
+ * Note: This API may fail, if any profiling or sampling session is active on the specified device.
+ *
+ * \param pParams A pointer to \ref CUpti_PmSampling_GetCounterAvailability_Params
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_INSUFFICIENT_PRIVILEGES if the user does not have sufficient privileges to perform the operation
+ * \retval CUPTI_ERROR_UNKNOWN for any internal error
+ */
+CUptiResult CUPTIAPI cuptiPmSamplingGetCounterAvailability(CUpti_PmSampling_GetCounterAvailability_Params* pParams);
+
+/**
+ * \brief Params for cuptiPmSamplingGetCounterDataSize
+ */
+typedef struct CUpti_PmSampling_GetCounterDataSize_Params
+{
+    /// [in] Size of the data structure.
+    size_t structSize;
+    /// [in] Set to NULL.
+    void* pPriv;
+    /// [in] PM sampling object.
+    CUpti_PmSampling_Object* pPmSamplingObject;
+    /// [in] Names of the metrics to be collected.
+    const char** pMetricNames;
+    /// [in] Number of metrics to be collected.
+    size_t numMetrics;
+    /// [in] Maximum number of samples to be stored in the counter data image.
+    uint32_t maxSamples;
+    /// [out] Size of the counter data image.
+    size_t counterDataSize;
+} CUpti_PmSampling_GetCounterDataSize_Params;
+#define CUpti_PmSampling_GetCounterDataSize_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_PmSampling_GetCounterDataSize_Params, counterDataSize)
+
+/**
+ * \brief Query the size of the counter data image which will be used to store the metrics data.
+ * User need to allocate the memory for the counter data image based on the size returned by this API.
+ *
+ * \param pParams A pointer to \ref CUpti_PmSampling_GetCounterDataSize_Params
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_INVALID_OPERATION if PM sampling GetCounterDataSize is called without enabling PM sampling
+ * \retval CUPTI_ERROR_UNKNOWN for any internal error
+ */
+CUptiResult CUPTIAPI cuptiPmSamplingGetCounterDataSize(CUpti_PmSampling_GetCounterDataSize_Params* pParams);
+
+/**
+ * \brief Params for cuptiPmSamplingCounterDataImageInitialize
+ */
+typedef struct CUpti_PmSampling_CounterDataImage_Initialize_Params
+{
+    /// [in] Size of the data structure.
+    size_t structSize;
+    /// [in] Set to NULL.
+    void* pPriv;
+    /// [in] PM sampling object.
+    CUpti_PmSampling_Object* pPmSamplingObject;
+    /// [in] Size of the counter data image.
+    size_t counterDataSize;
+    /// [in] Counter data image.
+    uint8_t* pCounterData;
+} CUpti_PmSampling_CounterDataImage_Initialize_Params;
+#define CUpti_PmSampling_CounterDataImage_Initialize_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_PmSampling_CounterDataImage_Initialize_Params, pCounterData)
+
+/**
+ * \brief Initialize the counter data to CUPTI record format for storing the metric data.
+ *
+ * \param pParams A pointer to \ref CUpti_PmSampling_CounterDataImage_Initialize_Params
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_INVALID_OPERATION if PM sampling CounterDataInitialize is called without enabling PM sampling
+ * \retval CUPTI_ERROR_UNKNOWN for any internal error
+ */
+CUptiResult CUPTIAPI cuptiPmSamplingCounterDataImageInitialize(CUpti_PmSampling_CounterDataImage_Initialize_Params* pParams);
+
+/**
+ * \brief Params for cuptiPmSamplingGetCounterDataInfo
+ */
+typedef struct CUpti_PmSampling_GetCounterDataInfo_Params
+{
+    /// [in] Size of the data structure.
+    size_t structSize;
+    /// [in] Set to NULL.
+    void* pPriv;
+    /// [in] Counter data image.
+    const uint8_t* pCounterDataImage;
+    /// [in] Size of the counter data image.
+    size_t counterDataImageSize;
+    /// [out] Number of samples in the counter data image.
+    size_t numTotalSamples;
+    /// [out] Number of populated samples.
+    size_t numPopulatedSamples;
+    /// [out] Number of samples that have been completed.
+    size_t numCompletedSamples;
+} CUpti_PmSampling_GetCounterDataInfo_Params;
+#define CUpti_PmSampling_GetCounterDataInfo_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_PmSampling_GetCounterDataInfo_Params, numCompletedSamples)
+
+/**
+ * \brief Get the counter data info like number of samples, number of populated
+ * samples and number of completed samples in a counter data image.
+ *
+ * \param pParams A pointer to \ref CUpti_PmSampling_GetCounterDataInfo_Params
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_UNKNOWN for any internal error
+ */
+CUptiResult CUPTIAPI cuptiPmSamplingGetCounterDataInfo(CUpti_PmSampling_GetCounterDataInfo_Params* pParams);
+
+/**
+ * \brief Params for cuptiPmSamplingCounterDataGetSampleInfo
+ */
+typedef struct CUpti_PmSampling_CounterData_GetSampleInfo_Params
+{
+    /// [in] Size of the data structure.
+    size_t structSize;
+    /// [in] Set to NULL.
+    void* pPriv;
+    /// [in] PM sampling object.
+    CUpti_PmSampling_Object* pPmSamplingObject;
+    /// [in] Counter data image.
+    const uint8_t* pCounterDataImage;
+    /// [in] Size of the counter data image.
+    size_t counterDataImageSize;
+    /// [in] Index of the sample.
+    size_t sampleIndex;
+    /// [out] Start time of the sample.
+    uint64_t startTimestamp;
+    /// [out] End time of the sample.
+    uint64_t endTimestamp;
+} CUpti_PmSampling_CounterData_GetSampleInfo_Params;
+#define CUpti_PmSampling_CounterData_GetSampleInfo_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_PmSampling_CounterData_GetSampleInfo_Params, endTimestamp)
+
+/**
+ * \brief Get the sample info (start and end time stamp) for the given sample index.
+ * Each sample is distinguished by the start and end time stamp.
+ *
+ * \param pParams A pointer to \ref CUpti_PmSampling_CounterData_GetSampleInfo_Params
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_UNKNOWN for any internal error
+ */
+CUptiResult CUPTIAPI cuptiPmSamplingCounterDataGetSampleInfo(CUpti_PmSampling_CounterData_GetSampleInfo_Params* pParams);
+
+/** @} */ /* END CUPTI_PMSAMPLING_API */
+#if defined(__GNUC__) && defined(CUPTI_LIB)
+    #pragma GCC visibility pop
+#endif
+
+#ifdef __cplusplus
+} /* extern "C" */
+#endif
+
+#endif // _CUPTI_PMSAMPLING_H_

Prism/LLaDA/LLaDA_Prism/.venv/lib/python3.12/site-packages/nvidia/cuda_cupti/include/cupti_profiler_host.h

@@ -0,0 +1,541 @@
+/*
+ * Copyright 2024 NVIDIA Corporation. All rights reserved.
+ *
+ * NOTICE TO LICENSEE:
+ *
+ * This source code and/or documentation ("Licensed Deliverables") are
+ * subject to NVIDIA intellectual property rights under U.S. and
+ * international Copyright laws.
+ *
+ * These Licensed Deliverables contained herein is PROPRIETARY and
+ * CONFIDENTIAL to NVIDIA and is being provided under the terms and
+ * conditions of a form of NVIDIA software license agreement by and
+ * between NVIDIA and Licensee ("License Agreement") or electronically
+ * accepted by Licensee.  Notwithstanding any terms or conditions to
+ * the contrary in the License Agreement, reproduction or disclosure
+ * of the Licensed Deliverables to any third party without the express
+ * written consent of NVIDIA is prohibited.
+ *
+ * NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE
+ * LICENSE AGREEMENT, NVIDIA MAKES NO REPRESENTATION ABOUT THE
+ * SUITABILITY OF THESE LICENSED DELIVERABLES FOR ANY PURPOSE.  IT IS
+ * PROVIDED "AS IS" WITHOUT EXPRESS OR IMPLIED WARRANTY OF ANY KIND.
+ * NVIDIA DISCLAIMS ALL WARRANTIES WITH REGARD TO THESE LICENSED
+ * DELIVERABLES, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY,
+ * NONINFRINGEMENT, AND FITNESS FOR A PARTICULAR PURPOSE.
+ * NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE
+ * LICENSE AGREEMENT, IN NO EVENT SHALL NVIDIA BE LIABLE FOR ANY
+ * SPECIAL, INDIRECT, INCIDENTAL, OR CONSEQUENTIAL DAMAGES, OR ANY
+ * DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS,
+ * WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS
+ * ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE
+ * OF THESE LICENSED DELIVERABLES.
+ *
+ * U.S. Government End Users.  These Licensed Deliverables are a
+ * "commercial item" as that term is defined at 48 C.F.R. 2.101 (OCT
+ * 1995), consisting of "commercial computer software" and "commercial
+ * computer software documentation" as such terms are used in 48
+ * C.F.R. 12.212 (SEPT 1995) and is provided to the U.S. Government
+ * only as a commercial end item.  Consistent with 48 C.F.R.12.212 and
+ * 48 C.F.R. 227.7202-1 through 227.7202-4 (JUNE 1995), all
+ * U.S. Government End Users acquire the Licensed Deliverables with
+ * only those rights set forth herein.
+ *
+ * Any use of the Licensed Deliverables in individual and commercial
+ * software must include, in the user documentation and internal
+ * comments to the code, the above Disclaimer and U.S. Government End
+ * Users Notice.
+ */
+
+#if !defined(_CUPTI_PROFILER_HOST_H_)
+#define _CUPTI_PROFILER_HOST_H_
+
+/*
+CUPTI profiler host API's
+This file contains the CUPTI profiling host API's.
+*/
+#include <cupti_result.h>
+#include <stdint.h>
+#include <stddef.h>
+#include <string>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#if defined(__GNUC__) && defined(CUPTI_LIB)
+    #pragma GCC visibility push(default)
+#endif
+
+/**
+ * \defgroup CUPTI_PROFILER_HOST_API CUPTI Profiler Host API
+ * Functions, types, and enums that implement the CUPTI Profiler Host API.
+ * @{
+ */
+#ifndef CUPTI_PROFILER_STRUCT_SIZE
+#define CUPTI_PROFILER_STRUCT_SIZE(type_, lastfield_)                     (offsetof(type_, lastfield_) + sizeof(((type_*)0)->lastfield_))
+#endif
+
+typedef enum CUpti_MetricType
+{
+    CUPTI_METRIC_TYPE_COUNTER = 0,
+    CUPTI_METRIC_TYPE_RATIO,
+    CUPTI_METRIC_TYPE_THROUGHPUT,
+    CUPTI_METRIC_TYPE__COUNT
+} CUpti_MetricType;
+
+typedef enum CUpti_ProfilerType
+{
+    CUPTI_PROFILER_TYPE_RANGE_PROFILER,
+    CUPTI_PROFILER_TYPE_PM_SAMPLING,
+    CUPTI_PROFILER_TYPE_PROFILER_INVALID
+} CUpti_ProfilerType;
+
+typedef struct CUpti_Profiler_Host_Object CUpti_Profiler_Host_Object;
+
+/**
+ * \brief Params for cuptiProfilerHostInitialize
+ */
+typedef struct CUpti_Profiler_Host_Initialize_Params
+{
+    /// [in] Size of the data structure.
+    /// CUPTI client should set the size of the structure. It will be used in CUPTI to check what fields are
+    /// available in the structure. Used to preserve backward compatibility.
+    size_t structSize;
+    /// [in] Assign to NULL
+    void* pPriv;
+    /// [in] the profiler kind one from CUpti_ProfilerType
+    CUpti_ProfilerType profilerType;
+    /// [in] accepted for chips supported at the time-of-release.
+    const char* pChipName;
+    /// [in] buffer with counter availability image - required for future chip support
+    const uint8_t* pCounterAvailabilityImage;
+    /// [out] binary blob allocated by CUPTI and operations associated with this object.
+    CUpti_Profiler_Host_Object* pHostObject;
+} CUpti_Profiler_Host_Initialize_Params;
+
+#define CUpti_Profiler_Host_Initialize_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_Profiler_Host_Initialize_Params, pHostObject)
+
+/**
+ * \brief Create and initialize the profiler host object (CUpti_Profiler_Host_Object).
+ * 
+ * \param pParams A pointer to \ref CUpti_Profiler_Host_Initialize_Params
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_UNKNOWN for any internal error
+ */
+CUptiResult CUPTIAPI cuptiProfilerHostInitialize(CUpti_Profiler_Host_Initialize_Params* pParams);
+
+/**
+ * \brief Params for cuptiProfilerHostDeinitialize
+ */
+typedef struct CUpti_Profiler_Host_Deinitialize_Params
+{
+    /// [in] Size of the data structure.
+    /// CUPTI client should set the size of the structure. It will be used in CUPTI to check what fields are
+    /// available in the structure. Used to preserve backward compatibility.
+    size_t structSize;
+    /// [in] Assign to NULL
+    void* pPriv;
+    /// [in] reference to the profiler host object allocated by CUPTI in cuptiProfilerHostInitialize
+    struct CUpti_Profiler_Host_Object* pHostObject;
+} CUpti_Profiler_Host_Deinitialize_Params;
+
+#define CUpti_Profiler_Host_Deinitialize_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_Profiler_Host_Deinitialize_Params, pHostObject)
+
+/**
+ * \brief Deinitialize and destroy the profiler host object (CUpti_Profiler_Host_Object).
+ * 
+ * \param pParams A pointer to \ref CUpti_Profiler_Host_Deinitialize_Params
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_UNKNOWN for any internal error
+ */
+CUptiResult CUPTIAPI cuptiProfilerHostDeinitialize(CUpti_Profiler_Host_Deinitialize_Params* pParams);
+
+/**
+ * \brief Params for cuptiProfilerHostGetSupportedChips
+ */
+typedef struct CUpti_Profiler_Host_GetSupportedChips_Params
+{
+    /// [in] Size of the data structure.
+    size_t structSize;
+    /// [in] Assign to NULL
+    void* pPriv;
+    /// [out] number of supported chips
+    size_t numChips;
+    /// [out] list of supported chips
+    const char* const* ppChipNames;
+} CUpti_Profiler_Host_GetSupportedChips_Params;
+
+#define CUpti_Profiler_Host_GetSupportedChips_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_Profiler_Host_GetSupportedChips_Params, ppChipNames)
+
+/**
+ * \brief Get the list of supported chips.
+ * 
+ * \param pParams A pointer to \ref CUpti_Profiler_Host_GetSupportedChips_Params
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_UNKNOWN for any internal error
+ */
+CUptiResult CUPTIAPI cuptiProfilerHostGetSupportedChips(CUpti_Profiler_Host_GetSupportedChips_Params* pParams);
+
+/**
+ * \brief Params for cuptiProfilerHostGetSupportedMetrics
+ */
+typedef struct CUpti_Profiler_Host_GetBaseMetrics_Params
+{
+    /// [in] Size of the data structure.
+    size_t structSize;
+    /// [in] Assign to NULL
+    void* pPriv;
+    /// [in] reference to the profiler host object allocated by CUPTI in cuptiProfilerHostInitialize
+    struct CUpti_Profiler_Host_Object* pHostObject;
+    /// [in] metric type (counter, ratio, throughput)
+    CUpti_MetricType metricType;
+    /// [out] list of base metrics supported of queried metric type for the chip
+    const char** ppMetricNames;
+    /// [out] number of metrics
+    size_t numMetrics;
+} CUpti_Profiler_Host_GetBaseMetrics_Params;
+
+#define CUpti_Profiler_Host_GetBaseMetrics_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_Profiler_Host_GetBaseMetrics_Params, numMetrics)
+
+/**
+ * \brief Get the list of supported base metrics for the chip.
+ * 
+ * \param pParams A pointer to \ref CUpti_Profiler_Host_GetBaseMetrics_Params
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_UNKNOWN for any internal error
+ */
+CUptiResult CUPTIAPI cuptiProfilerHostGetBaseMetrics(CUpti_Profiler_Host_GetBaseMetrics_Params* pParams);
+
+/**
+ * \brief Params for cuptiProfilerHostGetSubMetrics
+ */
+typedef struct CUpti_Profiler_Host_GetSubMetrics_Params
+{
+    /// [in] Size of the data structure.
+    size_t structSize;
+    /// [in] Assign to NULL
+    void* pPriv;
+    /// [in] reference to the profiler host object allocated by CUPTI in cuptiProfilerHostInitialize
+    CUpti_Profiler_Host_Object* pHostObject;
+    /// [in] the metric type for queried metric
+    CUpti_MetricType metricType;
+    /// [in] metric name for which sub-metric will be listed
+    const char* pMetricName;
+    /// [out] number of submetrics supported
+    size_t numOfSubmetrics;
+    /// [out] list of submetrics supported for the metric.
+    const char** ppSubMetrics;
+} CUpti_Profiler_Host_GetSubMetrics_Params;
+
+#define CUpti_Profiler_Host_GetSubMetrics_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_Profiler_Host_GetSubMetrics_Params, ppSubMetrics)
+
+/**
+ * \brief Get the list of supported sub-metrics for the metric.
+ * 
+ * \param pParams A pointer to \ref CUpti_Profiler_Host_GetSubMetrics_Params
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_INVALID_METRIC_NAME if the metric name is not valid or not supported for the chip
+ * \retval CUPTI_ERROR_UNKNOWN for any internal error
+ */
+CUptiResult CUPTIAPI cuptiProfilerHostGetSubMetrics(CUpti_Profiler_Host_GetSubMetrics_Params* pParams);
+
+/**
+ * \brief Params for cuptiProfilerHostGetMetricProperties
+ */
+typedef struct CUpti_Profiler_Host_GetMetricProperties_Params
+{
+    /// [in] Size of the data structure.
+    size_t structSize;
+    /// [in] Assign to NULL
+    void* pPriv;
+    /// [in] reference to the profiler host object allocated by CUPTI in cuptiProfilerHostInitialize
+    CUpti_Profiler_Host_Object* pHostObject;
+    /// [in] metric name for which its properties will be listed
+    const char* pMetricName;
+    /// [out] a short description about the metric
+    const char* pDescription;
+    /// [out] associated hw unit for the metric
+    const char* pHwUnit;
+    /// [out] the dimension of the metric values
+    const char* pDimUnit;
+    /// [out] the metric type (counter, ratio or throughput)
+    CUpti_MetricType metricType;
+} CUpti_Profiler_Host_GetMetricProperties_Params;
+
+#define CUpti_Profiler_Host_GetMetricProperties_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_Profiler_Host_GetMetricProperties_Params, metricType)
+
+/**
+ * \brief Get the properties of the metric.
+ * 
+ * \param pParams A pointer to \ref CUpti_Profiler_Host_GetMetricProperties_Params
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_INVALID_METRIC_NAME if the metric name is not valid or not supported for the chip
+ * \retval CUPTI_ERROR_UNKNOWN for any internal error
+ */
+CUptiResult CUPTIAPI cuptiProfilerHostGetMetricProperties(CUpti_Profiler_Host_GetMetricProperties_Params* pParams);
+
+/**
+ * \brief Params for cuptiProfilerHostGetRangeName
+ */
+typedef struct CUpti_Profiler_Host_GetRangeName_Params
+{
+    /// [in] Size of the data structure.
+    size_t structSize;
+    /// [in] Assign to NULL
+    void* pPriv;
+    /// [in] the counter data image where profiling data has been decoded
+    const uint8_t* pCounterDataImage;
+    /// [in] size of counter data image
+    size_t counterDataImageSize;
+    /// [in] range index for which the range name will be queried
+    size_t rangeIndex;
+    /// [in] used in case of nested ranges, default="/". Range1<delimiter>Range2 
+    const char* delimiter;
+    /// [out] the range name.
+    /// Note: that the CUPTI allocate the memory internal and
+    /// its user responsibility to free up the allocated memory
+    const char* pRangeName;
+} CUpti_Profiler_Host_GetRangeName_Params;
+
+#define CUpti_Profiler_Host_GetRangeName_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_Profiler_Host_GetRangeName_Params, pRangeName)
+
+/**
+ * \brief Get the range name for the range index stored in the counter data.
+ * In Range profiler, for Auto range mode the range name will be numeric value
+ * assigned to the kernel based on execution order. For user range mode, the 
+ * name of range will be based on the range name provided by the user using
+ * Push range API.
+ * 
+ * \param pParams A pointer to \ref CUpti_Profiler_Host_GetRangeName_Params
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_UNKNOWN for any internal error
+ */
+CUptiResult CUPTIAPI cuptiProfilerHostGetRangeName(CUpti_Profiler_Host_GetRangeName_Params* pParams);
+
+/**
+ * \brief Params for cuptiProfilerHostEvaluateToGpuValues
+ */
+typedef struct CUpti_Profiler_Host_EvaluateToGpuValues_Params
+{
+    /// [in] Size of the data structure.
+    size_t structSize;
+    /// [in] Assign to NULL
+    void* pPriv;
+    /// [in] reference to the profiler host object allocated by CUPTI in cuptiProfilerHostInitialize
+    CUpti_Profiler_Host_Object* pHostObject;
+    /// [in] the counter data image where profiling data has been decoded
+    const uint8_t* pCounterDataImage;
+    /// [in] size of counter data image
+    size_t counterDataImageSize;
+    /// [in] range index for which the range name will be queried
+    size_t rangeIndex;
+    /// [in] the metrics for which GPU values will be evaluated for the range
+    const char** ppMetricNames;
+    /// [in] number of metrics
+    size_t numMetrics;
+    /// [out] output value for given metric and range index
+    double* pMetricValues;
+} CUpti_Profiler_Host_EvaluateToGpuValues_Params;
+
+#define CUpti_Profiler_Host_EvaluateToGpuValues_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_Profiler_Host_EvaluateToGpuValues_Params, pMetricValues)
+
+/**
+ * \brief Evaluate the metric values for the range index stored in the counter data.
+ * 
+ * \param pParams A pointer to \ref CUpti_Profiler_Host_EvaluateToGpuValues_Params
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_INVALID_METRIC_NAME if the metric name is not valid or not supported for the chip
+ * \retval CUPTI_ERROR_UNKNOWN for any internal error
+ */
+CUptiResult CUPTIAPI cuptiProfilerHostEvaluateToGpuValues(CUpti_Profiler_Host_EvaluateToGpuValues_Params* pParams);
+
+/**
+ * \brief Params for cuptiProfilerHostConfigAddMetrics
+ */
+typedef struct CUpti_Profiler_Host_ConfigAddMetrics_Params
+{
+    /// [in] Size of the data structure.
+    size_t structSize;
+    /// [in] Assign to NULL
+    void* pPriv;
+    /// [in] reference to the profiler host object allocated by CUPTI in cuptiProfilerHostInitialize
+    struct CUpti_Profiler_Host_Object* pHostObject;
+    /// [in] metric names for which config image will be generated
+    const char** ppMetricNames;
+    /// [in] number of metrics
+    size_t numMetrics;
+} CUpti_Profiler_Host_ConfigAddMetrics_Params;
+
+#define CUpti_Profiler_Host_ConfigAddMetrics_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_Profiler_Host_ConfigAddMetrics_Params, numMetrics)
+
+/**
+ * \brief Add the metrics to the profiler host object for generating the config image.
+ * The config image will have the required information to schedule the metrics for
+ * collecting the profiling data. 
+ * Note: PM sampling only supports single pass config image.
+ * 
+ * \param pParams A pointer to \ref CUpti_Profiler_Host_ConfigAddMetrics_Params
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_INVALID_METRIC_NAME if the metric name is not valid or not supported for the chip
+ * \retval CUPTI_ERROR_UNKNOWN for any internal error
+ */
+CUptiResult CUPTIAPI cuptiProfilerHostConfigAddMetrics(CUpti_Profiler_Host_ConfigAddMetrics_Params* pParams);
+
+/**
+ * \brief Params for cuptiProfilerHostGetConfigImageSize
+ */
+typedef struct CUpti_Profiler_Host_GetConfigImageSize_Params
+{
+    /// [in] Size of the data structure.
+    size_t structSize;
+    /// [in] Assign to NULL
+    void* pPriv;
+    /// [in] reference to the profiler host object allocated by CUPTI in cuptiProfilerHostInitialize
+    CUpti_Profiler_Host_Object* pHostObject;
+    /// [out] the size of config image, users need to allocate the buffer for storing
+    size_t configImageSize;
+} CUpti_Profiler_Host_GetConfigImageSize_Params;
+
+#define CUpti_Profiler_Host_GetConfigImageSize_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_Profiler_Host_GetConfigImageSize_Params, configImageSize)
+
+/**
+ * \brief Get the size of the config image for the metrics added to the profiler host object.
+ * Users need to allocate the buffer for storing the config image.
+ * 
+ * \param pParams A pointer to \ref CUpti_Profiler_Host_GetConfigImageSize_Params
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_UNKNOWN for any internal error
+ */
+CUptiResult CUPTIAPI cuptiProfilerHostGetConfigImageSize(CUpti_Profiler_Host_GetConfigImageSize_Params* pParams);
+
+/**
+ * \brief Params for cuptiProfilerHostGetConfigImage
+ */
+typedef struct CUpti_Profiler_Host_GetConfigImage_Params
+{
+    /// [in] Size of the data structure.
+    size_t structSize;
+    /// [in] Assign to NULL
+    void* pPriv;
+    /// [in] reference to the profiler host object allocated by CUPTI in cuptiProfilerHostInitialize
+    CUpti_Profiler_Host_Object* pHostObject;
+    /// [in] Number of bytes allocated for pBuffer
+    size_t configImageSize;
+    /// [out] Buffer receiving the config image
+    uint8_t* pConfigImage;
+} CUpti_Profiler_Host_GetConfigImage_Params;
+
+#define CUpti_Profiler_Host_GetConfigImage_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_Profiler_Host_GetConfigImage_Params, pConfigImage)
+
+/**
+ * \brief Get the config image for the metrics added to the profiler host object.
+ * User will pass the allocated buffer to store the config image.
+ * 
+ * \param pParams A pointer to \ref CUpti_Profiler_Host_GetConfigImage_Params
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_UNKNOWN for any internal error
+ */
+CUptiResult CUPTIAPI cuptiProfilerHostGetConfigImage(CUpti_Profiler_Host_GetConfigImage_Params* pParams);
+
+/**
+ * \brief Params for cuptiProfilerHostGetNumOfPasses
+ */
+typedef struct CUpti_Profiler_Host_GetNumOfPasses_Params
+{
+    /// [in] Size of the data structure.
+    size_t structSize;
+    /// [in] Assign to NULL
+    void* pPriv;
+    /// [in] Number of bytes allocated for pConfigImage
+    size_t configImageSize;
+    /// [in] the config image buffer
+    uint8_t* pConfigImage;
+    /// [out] number of passes required for profiling scheduled metrics in the config image
+    size_t numOfPasses;
+} CUpti_Profiler_Host_GetNumOfPasses_Params;
+
+#define CUpti_Profiler_Host_GetNumOfPasses_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_Profiler_Host_GetNumOfPasses_Params, numOfPasses)
+
+/**
+ * \brief Get the number of passes required for profiling the scheduled metrics in the config image.
+ * 
+ * \param pParams A pointer to \ref CUpti_Profiler_Host_GetNumOfPasses_Params
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_UNKNOWN for any internal error
+ */
+CUptiResult CUPTIAPI cuptiProfilerHostGetNumOfPasses(CUpti_Profiler_Host_GetNumOfPasses_Params* pParams);
+
+/**
+ * \brief Params for cuptiProfilerHostGetMaxNumHardwareMetricsPerPass
+ */
+typedef struct CUpti_Profiler_Host_GetMaxNumHardwareMetricsPerPass_Params
+{
+    /// [in] Size of the data structure.
+    size_t structSize;
+    /// [in] Assign to NULL
+    void* pPriv;
+    /// [in] the profiler kind one from CUpti_ProfilerType
+    CUpti_ProfilerType profilerType;
+    /// [in] accepted for chips supported at the time-of-release.
+    const char* pChipName;
+    /// [in] buffer with counter availability image - required for future chip support
+    uint8_t* pCounterAvailabilityImage;
+    /// [out] maximum number of metrics that can be scheduled in a pass
+    size_t maxMetricsPerPass;
+} CUpti_Profiler_Host_GetMaxNumHardwareMetricsPerPass_Params;
+
+#define CUpti_Profiler_Host_GetMaxNumHardwareMetricsPerPass_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_Profiler_Host_GetMaxNumHardwareMetricsPerPass_Params, maxMetricsPerPass)
+
+/**
+ * \brief Get the maximum number of hardware metrics (metric names which doesn't include _sass_ keyword)
+ * that can be scheduled in a single pass for a chip. While this represents a theoretical upper limit,
+ * practical constraints may prevent reaching this threshold for a specific set of metrics. Furthermore,
+ * the maximum achievable value is contingent upon the characteristics and architecture of the chip in question.
+ * 
+ * Use cuptiProfilerHostGetNumOfPasses API for getting the actual number of passes required for the
+ * for collecting the profiling data for the scheduled metrics in a config image.
+ * 
+ * \param pParams A pointer to \ref CUpti_Profiler_Host_GetMaxNumHardwareMetricsPerPass_Params
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_UNKNOWN for any internal error
+ */
+CUptiResult CUPTIAPI cuptiProfilerHostGetMaxNumHardwareMetricsPerPass(CUpti_Profiler_Host_GetMaxNumHardwareMetricsPerPass_Params* pParams);
+
+/** @} */ /* END CUPTI_METRIC_API */
+#if defined(__GNUC__) && defined(CUPTI_LIB)
+    #pragma GCC visibility pop
+#endif
+
+
+#ifdef __cplusplus
+} /* extern "C" */
+#endif
+
+#endif

Prism/LLaDA/LLaDA_Prism/.venv/lib/python3.12/site-packages/nvidia/cuda_cupti/include/cupti_profiler_target.h

@@ -0,0 +1,602 @@
+/*
+ * Copyright 2011-2023   NVIDIA Corporation.  All rights reserved.
+ *
+ * NOTICE TO LICENSEE:
+ *
+ * This source code and/or documentation ("Licensed Deliverables") are
+ * subject to NVIDIA intellectual property rights under U.S. and
+ * international Copyright laws.
+ *
+ * These Licensed Deliverables contained herein is PROPRIETARY and
+ * CONFIDENTIAL to NVIDIA and is being provided under the terms and
+ * conditions of a form of NVIDIA software license agreement by and
+ * between NVIDIA and Licensee ("License Agreement") or electronically
+ * accepted by Licensee.  Notwithstanding any terms or conditions to
+ * the contrary in the License Agreement, reproduction or disclosure
+ * of the Licensed Deliverables to any third party without the express
+ * written consent of NVIDIA is prohibited.
+ *
+ * NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE
+ * LICENSE AGREEMENT, NVIDIA MAKES NO REPRESENTATION ABOUT THE
+ * SUITABILITY OF THESE LICENSED DELIVERABLES FOR ANY PURPOSE.  IT IS
+ * PROVIDED "AS IS" WITHOUT EXPRESS OR IMPLIED WARRANTY OF ANY KIND.
+ * NVIDIA DISCLAIMS ALL WARRANTIES WITH REGARD TO THESE LICENSED
+ * DELIVERABLES, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY,
+ * NONINFRINGEMENT, AND FITNESS FOR A PARTICULAR PURPOSE.
+ * NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE
+ * LICENSE AGREEMENT, IN NO EVENT SHALL NVIDIA BE LIABLE FOR ANY
+ * SPECIAL, INDIRECT, INCIDENTAL, OR CONSEQUENTIAL DAMAGES, OR ANY
+ * DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS,
+ * WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS
+ * ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE
+ * OF THESE LICENSED DELIVERABLES.
+ *
+ * U.S. Government End Users.  These Licensed Deliverables are a
+ * "commercial item" as that term is defined at 48 C.F.R. 2.101 (OCT
+ * 1995), consisting of "commercial computer software" and "commercial
+ * computer software documentation" as such terms are used in 48
+ * C.F.R. 12.212 (SEPT 1995) and is provided to the U.S. Government
+ * only as a commercial end item.  Consistent with 48 C.F.R.12.212 and
+ * 48 C.F.R. 227.7202-1 through 227.7202-4 (JUNE 1995), all
+ * U.S. Government End Users acquire the Licensed Deliverables with
+ * only those rights set forth herein.
+ *
+ * Any use of the Licensed Deliverables in individual and commercial
+ * software must include, in the user documentation and internal
+ * comments to the code, the above Disclaimer and U.S. Government End
+ * Users Notice.
+ */
+
+#if !defined(_CUPTI_PROFILER_TARGET_H_)
+#define _CUPTI_PROFILER_TARGET_H_
+
+#include <cuda.h>
+#include <cupti_result.h>
+#include <stddef.h>
+#include <stdint.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#if defined(__GNUC__) && defined(CUPTI_LIB)
+    #pragma GCC visibility push(default)
+#endif
+
+/**
+ * \defgroup CUPTI_PROFILER_API CUPTI Profiling API
+ * Functions, types, and enums that implement the CUPTI Profiling API.
+ * @{
+ */
+#ifndef CUPTI_PROFILER_STRUCT_SIZE
+#define CUPTI_PROFILER_STRUCT_SIZE(type_, lastfield_)                     (offsetof(type_, lastfield_) + sizeof(((type_*)0)->lastfield_))
+#endif
+
+/**
+ * \brief Profiler range attribute
+ *
+ * A metric enabled in the session's configuration is collected separately per unique range-stack in the pass.
+ * This is an attribute to collect metrics around each kernel in a profiling session or in an user defined range.
+ */
+typedef enum
+{
+    /**
+     * Invalid value
+     */
+    CUPTI_Range_INVALID,
+    /**
+     * Ranges are auto defined around each kernel in a profiling session
+     */
+    CUPTI_AutoRange,
+    /**
+     * A range in which metric data to be collected is defined by the user
+     */
+    CUPTI_UserRange,
+    /**
+     * Range count
+     */
+    CUPTI_Range_COUNT,
+} CUpti_ProfilerRange;
+
+/**
+ * \brief Profiler replay attribute
+ *
+ * For metrics which require multipass collection, a replay of the GPU kernel(s) is required.
+ * This is an attribute which specify how the replay of the kernel(s) to be measured is done.
+ */
+typedef enum
+{
+    /**
+     * Invalid Value
+     */
+    CUPTI_Replay_INVALID,
+    /**
+     * Replay is done by CUPTI user around the process
+     */
+    CUPTI_ApplicationReplay,
+    /**
+     * Replay is done around kernel implicitly by CUPTI
+     */
+    CUPTI_KernelReplay,
+    /**
+     * Replay is done by CUPTI user within a process
+     */
+    CUPTI_UserReplay,
+    /**
+     * Replay count
+     */
+    CUPTI_Replay_COUNT,
+} CUpti_ProfilerReplayMode;
+
+/**
+ * \brief Default parameter for cuptiProfilerInitialize
+ */
+typedef struct CUpti_Profiler_Initialize_Params
+{
+    size_t structSize;                                      //!< [in] CUpti_Profiler_Initialize_Params_STRUCT_SIZE
+    void* pPriv;                                            //!< [in] assign to NULL
+
+} CUpti_Profiler_Initialize_Params;
+#define CUpti_Profiler_Initialize_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_Profiler_Initialize_Params, pPriv)
+
+/**
+ * \brief Default parameter for cuptiProfilerDeInitialize
+ */
+typedef struct CUpti_Profiler_DeInitialize_Params
+{
+    size_t structSize;                                      //!< [in] CUpti_Profiler_DeInitialize_Params_STRUCT_SIZE
+    void* pPriv;                                            //!< [in] assign to NULL
+
+} CUpti_Profiler_DeInitialize_Params;
+#define CUpti_Profiler_DeInitialize_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_Profiler_DeInitialize_Params, pPriv)
+
+/**
+ * \brief Initializes the profiler interface
+ *
+ * Loads the required libraries in the process address space.
+ * Sets up the hooks with the CUDA driver.
+ */
+CUptiResult CUPTIAPI cuptiProfilerInitialize(CUpti_Profiler_Initialize_Params *pParams);
+
+/**
+ * \brief DeInitializes the profiler interface
+ */
+CUptiResult CUPTIAPI cuptiProfilerDeInitialize(CUpti_Profiler_DeInitialize_Params *pParams);
+
+/**
+ * \brief Input parameter to define the counterDataImage
+ */
+typedef struct CUpti_Profiler_CounterDataImageOptions
+{
+    size_t structSize;                                          //!< [in] CUpti_Profiler_CounterDataImageOptions_Params_STRUCT_SIZE
+    void* pPriv;                                                //!< [in] assign to NULL
+
+    const uint8_t* pCounterDataPrefix;                          /**< [in] Address of CounterDataPrefix generated from NVPW_CounterDataBuilder_GetCounterDataPrefix().
+                                                                    Must be align(8).*/
+    size_t counterDataPrefixSize;                               //!< [in] Size of CounterDataPrefix generated from NVPW_CounterDataBuilder_GetCounterDataPrefix().
+    uint32_t maxNumRanges;                                      //!< [in] Maximum number of ranges that can be profiled
+    uint32_t maxNumRangeTreeNodes;                              //!< [in] Maximum number of RangeTree nodes; must be >= maxNumRanges
+    uint32_t maxRangeNameLength;                                //!< [in] Maximum string length of each RangeName, including the trailing NULL character
+} CUpti_Profiler_CounterDataImageOptions;
+#define CUpti_Profiler_CounterDataImageOptions_STRUCT_SIZE                       CUPTI_PROFILER_STRUCT_SIZE(CUpti_Profiler_CounterDataImageOptions, maxRangeNameLength)
+
+/**
+ * \brief Params for cuptiProfilerCounterDataImageCalculateSize
+ */
+typedef struct CUpti_Profiler_CounterDataImage_CalculateSize_Params
+{
+    size_t structSize;                                          //!< [in] CUpti_Profiler_CounterDataImage_CalculateSize_Params_STRUCT_SIZE
+    void* pPriv;                                                //!< [in] assign to NULL
+
+    size_t sizeofCounterDataImageOptions;                       //!< [in] CUpti_Profiler_CounterDataImageOptions_STRUCT_SIZE
+    const CUpti_Profiler_CounterDataImageOptions* pOptions;     //!< [in] Pointer to Counter Data Image Options
+    size_t counterDataImageSize;                                //!< [out]
+} CUpti_Profiler_CounterDataImage_CalculateSize_Params;
+#define CUpti_Profiler_CounterDataImage_CalculateSize_Params_STRUCT_SIZE         CUPTI_PROFILER_STRUCT_SIZE(CUpti_Profiler_CounterDataImage_CalculateSize_Params, counterDataImageSize)
+
+/**
+ * \brief Params for cuptiProfilerCounterDataImageInitialize
+ */
+typedef struct CUpti_Profiler_CounterDataImage_Initialize_Params
+{
+    size_t structSize;                                          //!< [in] CUpti_Profiler_CounterDataImage_Initialize_Params_STRUCT_SIZE
+    void* pPriv;                                                //!< [in] assign to NULL
+
+    size_t sizeofCounterDataImageOptions;                       //!< [in] CUpti_Profiler_CounterDataImageOptions_STRUCT_SIZE
+    const CUpti_Profiler_CounterDataImageOptions* pOptions;     //!< [in] Pointer to Counter Data Image Options
+    size_t counterDataImageSize;                                //!< [in] Size calculated from cuptiProfilerCounterDataImageCalculateSize
+    uint8_t* pCounterDataImage;                                 //!< [in] The buffer to be initialized.
+} CUpti_Profiler_CounterDataImage_Initialize_Params;
+#define CUpti_Profiler_CounterDataImage_Initialize_Params_STRUCT_SIZE            CUPTI_PROFILER_STRUCT_SIZE(CUpti_Profiler_CounterDataImage_Initialize_Params, pCounterDataImage)
+
+/**
+ * \brief A CounterData image allocates space for values for each counter for each range.
+ *
+ * User borne the resposibility of managing the counterDataImage allocations.
+ * CounterDataPrefix contains meta data about the metrics that will be stored in counterDataImage.
+ * Use these APIs to calculate the allocation size and initialize counterData image.
+ */
+CUptiResult CUPTIAPI cuptiProfilerCounterDataImageCalculateSize(CUpti_Profiler_CounterDataImage_CalculateSize_Params* pParams);
+CUptiResult CUPTIAPI cuptiProfilerCounterDataImageInitialize(CUpti_Profiler_CounterDataImage_Initialize_Params* pParams);
+
+/**
+ * \brief Params for cuptiProfilerCounterDataImageCalculateScratchBufferSize
+ */
+typedef struct CUpti_Profiler_CounterDataImage_CalculateScratchBufferSize_Params
+{
+    size_t structSize;                                      //!< [in] CUpti_Profiler_CounterDataImage_CalculateScratchBufferSize_Params_STRUCT_SIZE
+    void* pPriv;                                            //!< [in] assign to NULL
+
+    size_t counterDataImageSize;                            //!< [in] size calculated from cuptiProfilerCounterDataImageCalculateSize
+    uint8_t* pCounterDataImage;                             //!< [in]
+    size_t counterDataScratchBufferSize;                    //!< [out]
+} CUpti_Profiler_CounterDataImage_CalculateScratchBufferSize_Params;
+#define CUpti_Profiler_CounterDataImage_CalculateScratchBufferSize_Params_STRUCT_SIZE    CUPTI_PROFILER_STRUCT_SIZE(CUpti_Profiler_CounterDataImage_CalculateScratchBufferSize_Params, counterDataScratchBufferSize)
+
+/**
+ * \brief Params for cuptiProfilerCounterDataImageInitializeScratchBuffer
+ */
+typedef struct CUpti_Profiler_CounterDataImage_InitializeScratchBuffer_Params
+{
+    size_t structSize;                                      //!< [in] CUpti_Profiler_CounterDataImage_InitializeScratchBuffer_Params_STRUCT_SIZE
+    void* pPriv;                                            //!< [in] assign to NULL
+
+    size_t counterDataImageSize;                            //!< [in] size calculated from cuptiProfilerCounterDataImageCalculateSize
+    uint8_t* pCounterDataImage;                             //!< [in]
+    size_t counterDataScratchBufferSize;                    //!< [in] size calculated using cuptiProfilerCounterDataImageCalculateScratchBufferSize
+    uint8_t* pCounterDataScratchBuffer;                     //!< [in] the scratch buffer to be initialized.
+} CUpti_Profiler_CounterDataImage_InitializeScratchBuffer_Params;
+#define CUpti_Profiler_CounterDataImage_InitializeScratchBuffer_Params_STRUCT_SIZE       CUPTI_PROFILER_STRUCT_SIZE(CUpti_Profiler_CounterDataImage_InitializeScratchBuffer_Params, pCounterDataScratchBuffer)
+
+/**
+ * \brief A temporary storage for CounterData image needed for internal operations
+ *
+ * Use these APIs to calculate the allocation size and initialize counterData image scratch buffer.
+ */
+CUptiResult CUPTIAPI cuptiProfilerCounterDataImageCalculateScratchBufferSize(CUpti_Profiler_CounterDataImage_CalculateScratchBufferSize_Params* pParams);
+CUptiResult CUPTIAPI cuptiProfilerCounterDataImageInitializeScratchBuffer(CUpti_Profiler_CounterDataImage_InitializeScratchBuffer_Params* pParams);
+
+/**
+ * \brief Params for cuptiProfilerBeginSession
+ */
+typedef struct CUpti_Profiler_BeginSession_Params
+{
+    size_t structSize;                                      //!< [in] CUpti_Profiler_BeginSession_Params_STRUCT_SIZE
+    void* pPriv;                                            //!< [in] assign to NULL
+
+    CUcontext ctx;                                          //!< [in] if NULL, the current CUcontext is used
+    size_t counterDataImageSize;                            //!< [in] size calculated from cuptiProfilerCounterDataImageCalculateSize
+    uint8_t* pCounterDataImage;                             //!< [in] address of CounterDataImage
+    size_t counterDataScratchBufferSize;                    //!< [in] size calculated from cuptiProfilerCounterDataImageInitializeScratchBuffer
+    uint8_t* pCounterDataScratchBuffer;                     //!< [in] address of CounterDataImage scratch buffer
+    uint8_t bDumpCounterDataInFile;                          //!< [in] [optional]
+    const char* pCounterDataFilePath;                        //!< [in] [optional]
+    CUpti_ProfilerRange range;                               //!< [in] CUpti_ProfilerRange
+    CUpti_ProfilerReplayMode replayMode;                     //!< [in] CUpti_ProfilerReplayMode
+    /* Replay options, required when replay is done by cupti user */
+    size_t maxRangesPerPass;                                //!< [in] Maximum number of ranges that can be recorded in a single pass.
+    size_t maxLaunchesPerPass;                              //!< [in] Maximum number of kernel launches that can be recorded in a single pass; must be >= maxRangesPerPass.
+
+} CUpti_Profiler_BeginSession_Params;
+#define CUpti_Profiler_BeginSession_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_Profiler_BeginSession_Params, maxLaunchesPerPass)
+/**
+ * \brief Params for cuptiProfilerEndSession
+ */
+typedef struct CUpti_Profiler_EndSession_Params
+{
+    size_t structSize;                                      //!< [in] CUpti_Profiler_EndSession_Params_STRUCT_SIZE
+    void* pPriv;                                            //!< [in] assign to NULL
+
+    CUcontext ctx;                                          //!< [in] if NULL, the current CUcontext is used
+} CUpti_Profiler_EndSession_Params;
+#define CUpti_Profiler_EndSession_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_Profiler_EndSession_Params, ctx)
+
+/**
+ * \brief Begin profiling session sets up the profiling on the device
+ *
+ * Although, it doesn't start the profiling but GPU resources needed for profiling are allocated.
+ * Outside of a session, the GPU will return to its normal operating state.
+ */
+CUptiResult CUPTIAPI cuptiProfilerBeginSession(CUpti_Profiler_BeginSession_Params* pParams);
+/**
+ * \brief Ends profiling session
+ *
+ * Frees up the GPU resources acquired for profiling.
+ * Outside of a session, the GPU will return to it's normal operating state.
+ */
+CUptiResult CUPTIAPI cuptiProfilerEndSession(CUpti_Profiler_EndSession_Params* pParams);
+
+/**
+ * \brief Params for cuptiProfilerSetConfig
+ */
+typedef struct CUpti_Profiler_SetConfig_Params
+{
+    size_t structSize;                                      //!< [in] CUpti_Profiler_SetConfig_Params_STRUCT_SIZE
+    void* pPriv;                                            //!< [in] assign to NULL
+
+    CUcontext ctx;                                          //!< [in] if NULL, the current CUcontext is used
+    const uint8_t* pConfig;                                 //!< [in] Config created by NVPW_RawMetricsConfig_GetConfigImage(). Must be align(8).
+    size_t configSize;                                      //!< [in] size of config
+    uint16_t minNestingLevel;                               //!< [in] the lowest nesting level to be profiled; must be >= 1
+    uint16_t numNestingLevels;                              //!< [in] the number of nesting levels to profile; must be >= 1
+    size_t passIndex;                                       //!< [in] Set this to zero for in-app replay; set this to the output of EndPass() for application replay
+    uint16_t targetNestingLevel;                            //!< [in] Set this to minNestingLevel for in-app replay; set this to the output of EndPass() for application
+} CUpti_Profiler_SetConfig_Params;
+
+#define CUpti_Profiler_SetConfig_Params_STRUCT_SIZE                    CUPTI_PROFILER_STRUCT_SIZE(CUpti_Profiler_SetConfig_Params, targetNestingLevel)
+
+/**
+ * \brief Params for cuptiProfilerUnsetConfig
+ */
+typedef struct CUpti_Profiler_UnsetConfig_Params
+{
+    size_t structSize;                                      //!< [in] CUpti_Profiler_UnsetConfig_Params_STRUCT_SIZE
+    void* pPriv;                                            //!< [in] assign to NULL
+
+    CUcontext ctx;                                          //!< [in] if NULL, the current CUcontext is used
+} CUpti_Profiler_UnsetConfig_Params;
+#define CUpti_Profiler_UnsetConfig_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_Profiler_UnsetConfig_Params, ctx)
+
+/**
+ * \brief Set metrics configuration to be profiled
+ *
+ * Use these APIs to set the config to profile in a session. It can be used for advanced cases such as where multiple
+ * configurations are collected into a single CounterData Image on the need basis, without restarting the session.
+ */
+CUptiResult CUPTIAPI cuptiProfilerSetConfig(CUpti_Profiler_SetConfig_Params* pParams);
+/**
+ * \brief Unset metrics configuration profiled
+ *
+ */
+CUptiResult CUPTIAPI cuptiProfilerUnsetConfig(CUpti_Profiler_UnsetConfig_Params* pParams);
+
+/**
+ * \brief Params for cuptiProfilerBeginPass
+ */
+typedef struct CUpti_Profiler_BeginPass_Params
+{
+    size_t structSize;                                      //!< [in] CUpti_Profiler_BeginPass_Params_STRUCT_SIZE
+    void* pPriv;                                            //!< [in] assign to NULL
+
+    CUcontext ctx;                                          //!< [in] if NULL, the current CUcontext is used
+} CUpti_Profiler_BeginPass_Params;
+#define CUpti_Profiler_BeginPass_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_Profiler_BeginPass_Params, ctx)
+
+/**
+ * \brief Params for cuptiProfilerEndPass
+ */
+typedef struct CUpti_Profiler_EndPass_Params
+{
+    size_t structSize;                                      //!< [in] CUpti_Profiler_EndPass_Params_STRUCT_SIZE
+    void* pPriv;                                            //!< [in] assign to NULL
+
+    CUcontext ctx;                                          //!< [in] if NULL, the current CUcontext is used
+    uint16_t targetNestingLevel;                            //!  [out] The targetNestingLevel that will be collected by the *next* BeginPass.
+    size_t passIndex;                                       //!< [out] The passIndex that will be collected by the *next* BeginPass
+    uint8_t allPassesSubmitted;                             //!< [out] becomes true when the last pass has been queued to the GPU
+} CUpti_Profiler_EndPass_Params;
+#define CUpti_Profiler_EndPass_Params_STRUCT_SIZE                    CUPTI_PROFILER_STRUCT_SIZE(CUpti_Profiler_EndPass_Params, allPassesSubmitted)
+
+/**
+ * \brief Replay API: used for multipass collection.
+
+ * These APIs are used if user chooses to replay by itself \ref CUPTI_UserReplay or \ref CUPTI_ApplicationReplay
+ * for multipass collection of the metrics configurations.
+ * It's a no-op in case of \ref CUPTI_KernelReplay.
+ */
+CUptiResult CUPTIAPI cuptiProfilerBeginPass(CUpti_Profiler_BeginPass_Params* pParams);
+
+/**
+ * \brief Replay API: used for multipass collection.
+
+ * These APIs are used if user chooses to replay by itself \ref CUPTI_UserReplay or \ref CUPTI_ApplicationReplay
+ * for multipass collection of the metrics configurations.
+ * Its a no-op in case of \ref CUPTI_KernelReplay.
+ * Returns information for next pass.
+ */
+CUptiResult CUPTIAPI cuptiProfilerEndPass(CUpti_Profiler_EndPass_Params* pParams);
+
+/**
+ * \brief Params for cuptiProfilerEnableProfiling
+ */
+typedef struct CUpti_Profiler_EnableProfiling_Params
+{
+    size_t structSize;                                      //!< [in] CUpti_Profiler_EnableProfiling_Params_STRUCT_SIZE
+    void* pPriv;                                            //!< [in] assign to NULL
+
+    CUcontext ctx;                                          //!< [in] if NULL, the current CUcontext is used
+} CUpti_Profiler_EnableProfiling_Params;
+#define CUpti_Profiler_EnableProfiling_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_Profiler_EnableProfiling_Params, ctx)
+
+/**
+ * \brief Params for cuptiProfilerDisableProfiling
+ */
+typedef struct CUpti_Profiler_DisableProfiling_Params
+{
+    size_t structSize;                                      //!< [in] CUpti_Profiler_DisableProfiling_Params_STRUCT_SIZE
+    void* pPriv;                                            //!< [in] assign to NULL
+
+    CUcontext ctx;                                          //!< [in] if NULL, the current CUcontext is used
+} CUpti_Profiler_DisableProfiling_Params;
+#define CUpti_Profiler_DisableProfiling_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_Profiler_DisableProfiling_Params, ctx)
+
+/**
+ * \brief Enables Profiling
+ *
+ * In \ref CUPTI_AutoRange, these APIs are used to enable/disable profiling for the kernels to be executed in
+ * a profiling session.
+ */
+CUptiResult CUPTIAPI cuptiProfilerEnableProfiling(CUpti_Profiler_EnableProfiling_Params* pParams);
+
+/**
+ * \brief Disable Profiling
+ *
+ * In \ref CUPTI_AutoRange, these APIs are used to enable/disable profiling for the kernels to be executed in
+ * a profiling session.
+ */
+CUptiResult CUPTIAPI cuptiProfilerDisableProfiling(CUpti_Profiler_DisableProfiling_Params* pParams);
+
+/**
+ * \brief Params for cuptiProfilerIsPassCollected
+ */
+typedef struct CUpti_Profiler_IsPassCollected_Params
+{
+    size_t structSize;                                      //!< [in] CUpti_Profiler_IsPassCollected_Params_STRUCT_SIZE
+    void* pPriv;                                            //!< [in] assign to NULL
+
+    CUcontext ctx;                                          //!< [in] if NULL, the current CUcontext is used
+    size_t numRangesDropped;                                //!< [out] number of ranges whose data was dropped in the processed pass
+    size_t numTraceBytesDropped;                            //!< [out] number of bytes not written to TraceBuffer due to buffer full
+    uint8_t onePassCollected;                               //!< [out] true if a pass was successfully decoded
+    uint8_t allPassesCollected;                             //!< [out] becomes true when the last pass has been decoded
+} CUpti_Profiler_IsPassCollected_Params;
+#define CUpti_Profiler_IsPassCollected_Params_STRUCT_SIZE            CUPTI_PROFILER_STRUCT_SIZE(CUpti_Profiler_IsPassCollected_Params, allPassesCollected)
+
+/**
+ * \brief Asynchronous call to query if the submitted pass to GPU is collected
+ *
+ */
+CUptiResult CUPTIAPI cuptiProfilerIsPassCollected(CUpti_Profiler_IsPassCollected_Params* pParams);
+
+/**
+ * \brief Params for cuptiProfilerFlushCounterData
+ */
+typedef struct CUpti_Profiler_FlushCounterData_Params
+{
+    size_t structSize;                                      //!< [in] CUpti_Profiler_FlushCounterData_Params_STRUCT_SIZE
+    void* pPriv;                                            //!< [in] assign to NULL
+
+    CUcontext ctx;                                          //!< [in] if NULL, the current CUcontext is used
+    size_t numRangesDropped;                                //!< [out] number of ranges whose data was dropped in the processed passes
+    size_t numTraceBytesDropped;                            //!< [out] number of bytes not written to TraceBuffer due to buffer full
+} CUpti_Profiler_FlushCounterData_Params;
+#define CUpti_Profiler_FlushCounterData_Params_STRUCT_SIZE           CUPTI_PROFILER_STRUCT_SIZE(CUpti_Profiler_FlushCounterData_Params, numTraceBytesDropped)
+
+/**
+ * \brief Decode all the submitted passes
+ *
+ * Flush Counter data API to ensure every pass is decoded into the counterDataImage passed at beginSession.
+ * This will cause the CPU/GPU sync to collect all the undecoded pass.
+ */
+CUptiResult CUPTIAPI cuptiProfilerFlushCounterData(CUpti_Profiler_FlushCounterData_Params* pParams);
+
+typedef struct CUpti_Profiler_PushRange_Params
+{
+    size_t structSize;                                      //!< [in] CUpti_Profiler_PushRange_Params_STRUCT_SIZE
+    void* pPriv;                                            //!< [in] assign to NULL
+
+    CUcontext ctx;                                          //!< [in] if NULL, the current CUcontext is used
+    const char* pRangeName;                                 //!< [in] specifies the range for subsequent launches; must not be NULL
+    size_t rangeNameLength;                                 //!< [in] assign to strlen(pRangeName) if known; if set to zero, the library will call strlen()
+} CUpti_Profiler_PushRange_Params;
+#define CUpti_Profiler_PushRange_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_Profiler_PushRange_Params, rangeNameLength)
+
+typedef struct CUpti_Profiler_PopRange_Params
+{
+    size_t structSize;                                      //!< [in] CUpti_Profiler_PopRange_Params_STRUCT_SIZE
+    void* pPriv;                                            //!< [in] assign to NULL
+
+    CUcontext ctx;                                          //!< [in] if NULL, the current CUcontext is used
+} CUpti_Profiler_PopRange_Params;
+#define CUpti_Profiler_PopRange_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_Profiler_PopRange_Params, ctx)
+
+
+/**
+ * \brief Range API's : Push user range
+ *
+ * Counter data is collected per unique range-stack. Identified by a string label passsed by the user.
+ * It's an invalid operation in case of \ref CUPTI_AutoRange.
+ */
+CUptiResult CUPTIAPI cuptiProfilerPushRange(CUpti_Profiler_PushRange_Params *pParams);
+
+/**
+ * \brief Range API's : Pop user range
+ *
+ * Counter data is collected per unique range-stack. Identified by a string label passsed by the user.
+ * It's an invalid operation in case of \ref CUPTI_AutoRange.
+ */
+CUptiResult CUPTIAPI cuptiProfilerPopRange(CUpti_Profiler_PopRange_Params *pParams);
+
+/**
+ * \brief Params for cuptiProfilerGetCounterAvailability
+ */
+typedef struct CUpti_Profiler_GetCounterAvailability_Params
+{
+    size_t structSize;                                  //!< [in] CUpti_Profiler_GetCounterAvailability_Params_STRUCT_SIZE
+    void* pPriv;                                        //!< [in] assign to NULL
+    CUcontext ctx;                                      //!< [in] if NULL, the current CUcontext is used
+    size_t counterAvailabilityImageSize;                //!< [in/out] If `pCounterAvailabilityImage` is NULL, then the required size is returned in
+                                                        //!< `counterAvailabilityImageSize`, otherwise `counterAvailabilityImageSize` should be set to the size of
+                                                        //!< `pCounterAvailabilityImage`, and on return it would be overwritten with number of actual bytes copied
+    uint8_t* pCounterAvailabilityImage;                 //!< [in] buffer receiving counter availability image, may be NULL
+} CUpti_Profiler_GetCounterAvailability_Params;
+#define CUpti_Profiler_GetCounterAvailability_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_Profiler_GetCounterAvailability_Params, pCounterAvailabilityImage)
+
+/**
+ * \brief Query counter availibility
+ *
+ * Use this API to query counter availability information in a buffer which can be used to filter unavailable raw metrics on host.
+ * Note: This API may fail, if any profiling or sampling session is active on the specified context or its device.
+ */
+CUptiResult CUPTIAPI cuptiProfilerGetCounterAvailability(CUpti_Profiler_GetCounterAvailability_Params *pParams);
+
+/// Generic support level enum for CUPTI
+typedef enum
+{
+    CUPTI_PROFILER_CONFIGURATION_UNKNOWN = 0, //!< Configuration support level unknown - either detection code errored out before setting this value, or unable to determine it
+    CUPTI_PROFILER_CONFIGURATION_UNSUPPORTED, //!< Profiling is unavailable.  For specific feature fields, this means that the current configuration of this feature does not work with profiling.  For instance, SLI-enabled devices do not support profiling, and this value would be returned for SLI on an SLI-enabled device.
+    CUPTI_PROFILER_CONFIGURATION_DISABLED,    //!< Profiling would be available for this configuration, but was disabled by the system
+    CUPTI_PROFILER_CONFIGURATION_SUPPORTED    //!< Profiling is supported.  For specific feature fields, this means that the current configuration of this feature works with profiling.  For instance, SLI-enabled devices do not support profiling, and this value would only be returned for devices which are not SLI-enabled.
+} CUpti_Profiler_Support_Level;
+
+/**
+ * \brief Profiler API types
+ */
+typedef enum
+{
+    CUPTI_PROFILER_RANGE_PROFILING = 0,       //!< CUPTI APIs for range based profiling (cuptiProfiler*)
+    CUPTI_PROFILER_PC_SAMPLING,               //!< CUPTI APIs collecting pc sampling data (cuptiPcSampling*)
+    CUPTI_PROFILER_SASS_METRICS,              //!< CUPTI APIs collecting SASS metrics data (cuptiSassMetrics*)
+    CUPTI_PROFILER_PM_SAMPLING,               //!< CUPTI APIs collecting PM Sampling data (cuptiPmSampling*)
+    CUPTI_PROFILER_UNKNOWN
+} CUpti_Profiler_API;
+
+/**
+ * \brief Params for cuptiProfilerDeviceSupported
+ */
+typedef struct
+{
+    size_t structSize;                                //!< [in] Must be CUpti_Profiler_DeviceSupported_Params_STRUCT_SIZE
+    void *pPriv;                                      //!< [in] assign to NULL
+    CUdevice cuDevice;                                //!< [in] if NULL, the current CUcontext is used
+
+    CUpti_Profiler_Support_Level isSupported;         //!< [out] overall SUPPORTED / UNSUPPORTED flag representing whether Profiling and PC Sampling APIs work on the given device and configuration. SUPPORTED if all following flags are SUPPORTED, UNSUPPORTED otherwise.
+
+    CUpti_Profiler_Support_Level architecture;        //!< [out] SUPPORTED if the device architecture level supports the Profiling API (Compute Capability >= 7.0), UNSUPPORTED otherwise
+    CUpti_Profiler_Support_Level sli;                 //!< [out] SUPPORTED if SLI is not enabled, UNSUPPORTED otherwise
+    CUpti_Profiler_Support_Level vGpu;                //!< [out] SUPPORTED if vGPU is supported and profiling is enabled, DISABLED if profiling is supported but not enabled, UNSUPPORTED otherwise
+    CUpti_Profiler_Support_Level confidentialCompute; //!< [out] SUPPORTED if confidential compute is not enabled, UNSUPPORTED otherwise
+    CUpti_Profiler_Support_Level cmp;                 //!< [out] SUPPORTED if not NVIDIA Crypto Mining Processors (CMP), UNSUPPORTED otherwise
+    CUpti_Profiler_Support_Level wsl;                 //!< [out] SUPPORTED if WSL supported, UNSUPPORTED otherwise
+    CUpti_Profiler_API     api;                       //!< [in] the CUPTI API type for which device support will be checked
+} CUpti_Profiler_DeviceSupported_Params;
+#define CUpti_Profiler_DeviceSupported_Params_STRUCT_SIZE CUPTI_PROFILER_STRUCT_SIZE(CUpti_Profiler_DeviceSupported_Params, api)
+
+/**
+ * \brief Query device compatibility with Profiling API
+ *
+ * Use this call to determine whether a compute device and configuration are compatible with the Profiling API.
+ * If the configuration does not support profiling, one of several flags will indicate why.
+ */
+CUptiResult CUPTIAPI cuptiProfilerDeviceSupported(CUpti_Profiler_DeviceSupported_Params *pParams);
+
+/** @} */ /* END CUPTI_METRIC_API */
+#if defined(__GNUC__) && defined(CUPTI_LIB)
+    #pragma GCC visibility pop
+#endif
+
+#ifdef __cplusplus
+} /* extern "C" */
+#endif
+
+#endif /*_CUPTI_PROFILER_TARGET_H_*/

Prism/LLaDA/LLaDA_Prism/.venv/lib/python3.12/site-packages/nvidia/cuda_cupti/include/cupti_range_profiler.h

@@ -0,0 +1,465 @@
+/*
+ * Copyright 2024 NVIDIA Corporation.  All rights reserved.
+ *
+ * NOTICE TO LICENSEE:
+ *
+ * This source code and/or documentation ("Licensed Deliverables") are
+ * subject to NVIDIA intellectual property rights under U.S. and
+ * international Copyright laws.
+ *
+ * These Licensed Deliverables contained herein is PROPRIETARY and
+ * CONFIDENTIAL to NVIDIA and is being provided under the terms and
+ * conditions of a form of NVIDIA software license agreement by and
+ * between NVIDIA and Licensee ("License Agreement") or electronically
+ * accepted by Licensee.  Notwithstanding any terms or conditions to
+ * the contrary in the License Agreement, reproduction or disclosure
+ * of the Licensed Deliverables to any third party without the express
+ * written consent of NVIDIA is prohibited.
+ *
+ * NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE
+ * LICENSE AGREEMENT, NVIDIA MAKES NO REPRESENTATION ABOUT THE
+ * SUITABILITY OF THESE LICENSED DELIVERABLES FOR ANY PURPOSE.  IT IS
+ * PROVIDED "AS IS" WITHOUT EXPRESS OR IMPLIED WARRANTY OF ANY KIND.
+ * NVIDIA DISCLAIMS ALL WARRANTIES WITH REGARD TO THESE LICENSED
+ * DELIVERABLES, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY,
+ * NONINFRINGEMENT, AND FITNESS FOR A PARTICULAR PURPOSE.
+ * NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE
+ * LICENSE AGREEMENT, IN NO EVENT SHALL NVIDIA BE LIABLE FOR ANY
+ * SPECIAL, INDIRECT, INCIDENTAL, OR CONSEQUENTIAL DAMAGES, OR ANY
+ * DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS,
+ * WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS
+ * ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE
+ * OF THESE LICENSED DELIVERABLES.
+ *
+ * U.S. Government End Users.  These Licensed Deliverables are a
+ * "commercial item" as that term is defined at 48 C.F.R. 2.101 (OCT
+ * 1995), consisting of "commercial computer software" and "commercial
+ * computer software documentation" as such terms are used in 48
+ * C.F.R. 12.212 (SEPT 1995) and is provided to the U.S. Government
+ * only as a commercial end item.  Consistent with 48 C.F.R.12.212 and
+ * 48 C.F.R. 227.7202-1 through 227.7202-4 (JUNE 1995), all
+ * U.S. Government End Users acquire the Licensed Deliverables with
+ * only those rights set forth herein.
+ *
+ * Any use of the Licensed Deliverables in individual and commercial
+ * software must include, in the user documentation and internal
+ * comments to the code, the above Disclaimer and U.S. Government End
+ * Users Notice.
+ */
+
+#if !defined(_CUPTI_RANGE_PROFILER_H_)
+#define _CUPTI_RANGE_PROFILER_H_
+
+#include <cuda.h>
+#include <cupti_result.h>
+#include <cupti_profiler_target.h>
+#include <stddef.h>
+#include <stdint.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#if defined(__GNUC__) && defined(CUPTI_LIB)
+    #pragma GCC visibility push(default)
+#endif
+
+/**
+ * \defgroup CUPTI_RANGE_PROFILER_API CUPTI Range Profiling API
+ * Functions, types, and enums that implement the CUPTI Range Profiling API.
+ * @{
+ */
+#ifndef CUPTI_PROFILER_STRUCT_SIZE
+#define CUPTI_PROFILER_STRUCT_SIZE(type_, lastfield_)                     (offsetof(type_, lastfield_) + sizeof(((type_*)0)->lastfield_))
+#endif
+
+
+typedef struct CUpti_RangeProfiler_Object CUpti_RangeProfiler_Object;
+
+/**
+ * \brief Params for cuptiRangeProfilerSetConfig
+ */
+typedef struct CUpti_RangeProfiler_SetConfig_Params
+{
+    /// [in] Size of the data structure.
+    size_t structSize;
+    /// [in] Set to NULL.
+    void* pPriv;
+    /// [in] Range Profiler Object.
+    CUpti_RangeProfiler_Object* pRangeProfilerObject;
+    /// [in] Size of the config image.
+    size_t configSize;
+    /// [in] Config image.
+    const uint8_t* pConfig;
+    /// [in] Size of the counter data image.
+    size_t counterDataImageSize;
+    /// [in] Counter data image.
+    uint8_t* pCounterDataImage;
+    /// [in] Profiling Range mode.
+    CUpti_ProfilerRange range;
+    /// [in] Replay mode.
+    CUpti_ProfilerReplayMode replayMode;
+    /// [in] Maximum number of ranges that can be profiled in a pass.
+    size_t maxRangesPerPass;
+    /// [in] number of nesting level to be profiled. For Auto range mode, this should be set to 1.
+    uint16_t numNestingLevels;
+    /// [in] minimum nesting level to be profiled.
+    uint16_t minNestingLevel;
+    /// [in] Pass index for the replay session.
+    size_t passIndex;
+    /// [in] Target nesting level for the replay session.
+    uint16_t targetNestingLevel;
+} CUpti_RangeProfiler_SetConfig_Params;
+
+#define CUpti_RangeProfiler_SetConfig_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_RangeProfiler_SetConfig_Params, targetNestingLevel)
+
+/**
+ * \brief Set the configuration for range profiler like maximum number of ranges per pass, number of nesting levels,
+ * range and replay mode and the config image which has scheduling info for metric collection.
+ * 
+ * \param pParams A pointer to \ref CUpti_RangeProfiler_SetConfig_Params
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ */
+CUptiResult CUPTIAPI cuptiRangeProfilerSetConfig(CUpti_RangeProfiler_SetConfig_Params* pParams);
+
+/**
+ * \brief Params for cuptiRangeProfilerEnable
+ */
+typedef struct CUpti_RangeProfiler_Enable_Params
+{
+    /// [in] Size of the data structure.
+    size_t structSize;
+    /// [in] Set to NULL.
+    void* pPriv;
+    /// [in] Context to be used for profiling.
+    CUcontext ctx;
+    /// [out] Range Profiler Object.
+    CUpti_RangeProfiler_Object* pRangeProfilerObject;
+} CUpti_RangeProfiler_Enable_Params;
+#define CUpti_RangeProfiler_Enable_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_RangeProfiler_Enable_Params, pRangeProfilerObject)
+
+/**
+ * \brief Create a range profiler object and enable range profiling on the CUDA context.
+ * 
+ * \param pParams A pointer to \ref CUpti_RangeProfiler_Enable_Params
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_OUT_OF_MEMORY if memory allocation fails while creating the PM sampling object
+ * \retval CUPTI_ERROR_INSUFFICIENT_PRIVILEGES if the user does not have sufficient privileges to perform the operation
+ * \retval CUPTI_ERROR_UNKNOWN for any internal error
+ */
+CUptiResult CUPTIAPI cuptiRangeProfilerEnable(CUpti_RangeProfiler_Enable_Params* pParams);
+
+/**
+ * \brief Params for cuptiRangeProfilerDisable
+ */
+typedef struct CUpti_RangeProfiler_Disable_Params
+{
+    /// [in] Size of the data structure.
+    size_t structSize;
+    /// [in] Set to NULL.
+    void* pPriv;
+    /// [in] Range Profiler Object.
+    CUpti_RangeProfiler_Object* pRangeProfilerObject;
+} CUpti_RangeProfiler_Disable_Params;
+#define CUpti_RangeProfiler_Disable_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_RangeProfiler_Disable_Params, pRangeProfilerObject)
+
+/**
+ * \brief Disable the range profiler on the CUDA context and destroy the range profiler object.
+ * 
+ * \param pParams A pointer to \ref CUpti_RangeProfiler_Disable_Params
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ */
+CUptiResult CUPTIAPI cuptiRangeProfilerDisable(CUpti_RangeProfiler_Disable_Params* pParams);
+
+/**
+ * \brief Params for cuptiRangeProfilerStart
+ */
+typedef struct CUpti_RangeProfiler_Start_Params
+{
+    /// [in] Size of the data structure.
+    size_t structSize;
+    /// [in] Set to NULL.
+    void* pPriv;
+    /// [in] Range Profiler Object.
+    CUpti_RangeProfiler_Object* pRangeProfilerObject;
+} CUpti_RangeProfiler_Start_Params;
+#define CUpti_RangeProfiler_Start_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_RangeProfiler_Start_Params, pRangeProfilerObject)
+
+/**
+ * \brief Start the range profiler.
+ * 
+ * \param pParams A pointer to \ref CUpti_RangeProfiler_Start_Params
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_INVALID_OPERATION if range profiler Start is called without enabling range profiler
+ * \retval CUPTI_ERROR_UNKNOWN for any internal error
+ */
+CUptiResult CUPTIAPI cuptiRangeProfilerStart(CUpti_RangeProfiler_Start_Params* pParams);
+
+/**
+ * \brief Params for cuptiRangeProfilerStop
+ */
+typedef struct CUpti_RangeProfiler_Stop_Params
+{
+    /// [in] Size of the data structure.
+    size_t structSize;
+    /// [in] Set to NULL.
+    void* pPriv;
+    /// [in] Range Profiler Object.
+    CUpti_RangeProfiler_Object* pRangeProfilerObject;
+    /// [out] pass index for the replay session.
+    size_t passIndex;
+    /// [out] target nesting level for the replay session.
+    size_t targetNestingLevel;
+    /// [out] 1 if all passes are submitted to GPU for collection, 0 otherwise.
+    uint8_t isAllPassSubmitted;
+} CUpti_RangeProfiler_Stop_Params;
+#define CUpti_RangeProfiler_Stop_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_RangeProfiler_Stop_Params, isAllPassSubmitted)
+
+/**
+ * \brief Stop the range profiler.
+ * 
+ * \param pParams A pointer to \ref CUpti_RangeProfiler_Stop_Params
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_INVALID_OPERATION if range profiler Stop is called without enabling range profiler
+ * \retval CUPTI_ERROR_UNKNOWN for any internal error
+ */
+CUptiResult CUPTIAPI cuptiRangeProfilerStop(CUpti_RangeProfiler_Stop_Params* pParams);
+
+/**
+ * \brief Params for cuptiRangeProfilerPushRange
+ */
+typedef struct CUpti_RangeProfiler_PushRange_Params
+{
+    /// [in] Size of the data structure.
+    size_t structSize;
+    /// [in] Set to NULL.
+    void* pPriv;
+    /// [in] Range Profiler Object.
+    CUpti_RangeProfiler_Object* pRangeProfilerObject;
+    /// [in] Name of the range to be profiled (only valid for User range mode).
+    const char* pRangeName;
+} CUpti_RangeProfiler_PushRange_Params;
+#define CUpti_RangeProfiler_PushRange_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_RangeProfiler_PushRange_Params, pRangeName)
+
+/**
+ * \brief Add a new range to the Range Profiler with a given range name.
+ * For nested ranges, this API should be called again for the innermost range. For profiling the nested
+ * range, users need to set the values for minNestingLevel and numNestingLevels in the SetConfig API.
+ * 
+ * \param pParams A pointer to \ref CUpti_RangeProfiler_PushRange_Params
+ * 
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_INVALID_OPERATION if range profiler PushRange is called without enabling range profiler
+ * \retval CUPTI_ERROR_UNKNOWN for any internal error
+*/
+CUptiResult CUPTIAPI cuptiRangeProfilerPushRange(CUpti_RangeProfiler_PushRange_Params* pParams);
+
+/**
+ * \brief Params for cuptiRangeProfilerPopRange
+ */
+typedef struct CUpti_RangeProfiler_PopRange_Params
+{
+    /// [in] Size of the data structure.
+    size_t structSize;
+    /// [in] Set to NULL.
+    void* pPriv;
+    /// [in] Range Profiler Object.
+    CUpti_RangeProfiler_Object* pRangeProfilerObject;
+} CUpti_RangeProfiler_PopRange_Params;
+#define CUpti_RangeProfiler_PopRange_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_RangeProfiler_PopRange_Params, pRangeProfilerObject)
+
+/**
+ * \brief pop the current range to the Range Profiler.
+ * The number of pop range API call should be same as number of push ranges in the same order.
+ * 
+ * \param pParams A pointer to \ref CUpti_RangeProfiler_PopRange_Params
+ * 
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_INVALID_OPERATION if range profiler PopRange is called without enabling range profiler
+ * \retval CUPTI_ERROR_UNKNOWN for any internal error
+*/
+CUptiResult CUPTIAPI cuptiRangeProfilerPopRange(CUpti_RangeProfiler_PopRange_Params* pParams);
+
+/**
+ * \brief Params for cuptiRangeProfilerDecodeData
+ */
+typedef struct CUpti_RangeProfiler_DecodeData_Params
+{
+    /// [in] Size of the data structure.
+    size_t structSize;
+    /// [in] Set to NULL.
+    void* pPriv;
+    /// [in] Range Profiler Object.
+    CUpti_RangeProfiler_Object* pRangeProfilerObject;
+    /// [out] Number of ranges dropped in the processed passes.
+    size_t numOfRangeDropped;
+} CUpti_RangeProfiler_DecodeData_Params;
+#define CUpti_RangeProfiler_DecodeData_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_RangeProfiler_DecodeData_Params, numOfRangeDropped)
+
+/**
+ * \brief Decode the profiling data stored in the hardware to the counter data image passed in the
+ * SetConfig API. This API should be called after cuptiRangeProfilerStop. The counter data image
+ * will be updated with the profiling data for the ranges profiled.
+ * 
+ * For the cases where the number of ranges counter data image can store is less than the number of ranges
+ * profiled (= maxRangesPerPass in SetConfig API), the counter data image will report dropped ranges.
+ * 
+ * \param pParams A pointer to \ref CUpti_RangeProfiler_DecodeData_Params
+ * 
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_INVALID_OPERATION if range profiler DecodeData is called without enabling range profiler
+ * \retval CUPTI_ERROR_UNKNOWN for any internal error
+*/
+CUptiResult CUPTIAPI cuptiRangeProfilerDecodeData(CUpti_RangeProfiler_DecodeData_Params* pParams);
+
+/**
+ * \brief Params for cuptiRangeProfilerGetCounterDataSize
+ */
+typedef struct CUpti_RangeProfiler_GetCounterDataSize_Params
+{
+    /// [in] Size of the data structure.
+    size_t structSize;
+    /// [in] Set to NULL.
+    void* pPriv;
+    /// [in] Periodic sampler object.
+    CUpti_RangeProfiler_Object* pRangeProfilerObject;
+    /// [in] Names of the metrics to be collected.
+    const char** pMetricNames;
+    /// [in] Number of metrics to be collected.
+    size_t numMetrics;
+    /// [in] Maximum number of ranges to be stored in the counter data image.
+    size_t maxNumOfRanges;
+    /// [in] Maximum number of RangeTree nodes; must be >= maxNumOfRanges
+    uint32_t maxNumRangeTreeNodes;
+    /// [out] Size of the counter data image.
+    size_t counterDataSize;
+} CUpti_RangeProfiler_GetCounterDataSize_Params;
+#define CUpti_RangeProfiler_GetCounterDataSize_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_RangeProfiler_GetCounterDataSize_Params, counterDataSize)
+
+/**
+ * \brief Get the size of the counter data image required to store the profiling data for the ranges profiled.
+ * 
+ * \param pParams A pointer to \ref CUpti_RangeProfiler_GetCounterDataSize_Params
+ * 
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_INVALID_OPERATION if range profiler GetCounterDataSize is called without enabling range profiler
+ * \retval CUPTI_ERROR_UNKNOWN for any internal error
+*/
+CUptiResult CUPTIAPI cuptiRangeProfilerGetCounterDataSize(CUpti_RangeProfiler_GetCounterDataSize_Params* pParams);
+
+/**
+ * \brief Params for cuptiRangeProfilerCounterDataImageInitialize
+ */
+typedef struct CUpti_RangeProfiler_CounterDataImage_Initialize_Params
+{
+    /// [in] Size of the data structure.
+    size_t structSize;
+    /// [in] Set to NULL.
+    void* pPriv;
+    /// [in] Periodic sampler object.
+    CUpti_RangeProfiler_Object* pRangeProfilerObject;
+    /// [in] Size of the counter data image.
+    size_t counterDataSize;
+    /// [in] Counter data image.
+    uint8_t* pCounterData;
+} CUpti_RangeProfiler_CounterDataImage_Initialize_Params;
+#define CUpti_RangeProfiler_CounterDataImage_Initialize_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_RangeProfiler_CounterDataImage_Initialize_Params, pCounterData)
+
+/**
+ * \brief Initialize the counter data image with the profiling data for the ranges profiled.
+ * 
+ * \param pParams A pointer to \ref CUpti_RangeProfiler_CounterDataImage_Initialize_Params
+ * 
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_INVALID_OPERATION if range profiler CounterDataImageInitialize is called without enabling range profiler
+ * \retval CUPTI_ERROR_UNKNOWN for any internal error
+*/
+CUptiResult CUPTIAPI cuptiRangeProfilerCounterDataImageInitialize(CUpti_RangeProfiler_CounterDataImage_Initialize_Params* pParams);
+
+/**
+ * \brief Params for cuptiRangeProfilerGetCounterDataInfo
+ */
+typedef struct CUpti_RangeProfiler_GetCounterDataInfo_Params
+{
+    /// [in] Size of the data structure.
+    size_t structSize;
+    /// [in] Set to NULL.
+    void* pPriv;
+    /// [in] Counter data image.
+    const uint8_t* pCounterDataImage;
+    /// [in] Size of the counter data image.
+    size_t counterDataImageSize;
+    /// [out] Number of ranges in the counter data image.
+    size_t numTotalRanges;
+} CUpti_RangeProfiler_GetCounterDataInfo_Params;
+#define CUpti_RangeProfiler_GetCounterDataInfo_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_RangeProfiler_GetCounterDataInfo_Params, numTotalRanges)
+
+/**
+ * \brief Get the number of ranges stored in the counter data image.
+ * 
+ * \param pParams A pointer to \ref CUpti_RangeProfiler_GetCounterDataInfo_Params
+ * 
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_UNKNOWN for any internal error
+*/
+CUptiResult CUPTIAPI cuptiRangeProfilerGetCounterDataInfo(CUpti_RangeProfiler_GetCounterDataInfo_Params* pParams);
+
+/**
+ * \brief Params for cuptiRangeProfilerCounterDataGetRangeInfo
+ */
+typedef struct CUpti_RangeProfiler_CounterData_GetRangeInfo_Params
+{
+    /// [in] Size of the data structure.
+    size_t structSize;
+    /// [in] Set to NULL.
+    void* pPriv;
+    /// [in] Counter data image.
+    const uint8_t* pCounterDataImage;
+    /// [in] Size of the counter data image.
+    size_t counterDataImageSize;
+    /// [in] Index of the sample.
+    size_t rangeIndex;
+    /// [in] range delimiter.
+    const char* rangeDelimiter;
+    /// [out] RangeName;
+    const char* rangeName;
+} CUpti_RangeProfiler_CounterData_GetRangeInfo_Params;
+#define CUpti_RangeProfiler_CounterData_GetRangeInfo_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_RangeProfiler_CounterData_GetRangeInfo_Params, rangeName)
+
+/**
+ * \brief Get the range name for the given range index.
+ * 
+ * \param pParams A pointer to \ref CUpti_RangeProfiler_CounterData_GetRangeInfo_Params
+ * 
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_UNKNOWN for any internal error
+*/
+CUptiResult CUPTIAPI cuptiRangeProfilerCounterDataGetRangeInfo(CUpti_RangeProfiler_CounterData_GetRangeInfo_Params* pParams);
+
+
+#if defined(__GNUC__) && defined(CUPTI_LIB)
+    #pragma GCC visibility pop
+#endif
+
+#ifdef __cplusplus
+} /* extern "C" */
+#endif
+
+#endif /*_CUPTI_RANGE_PROFILER_H_*/

Prism/LLaDA/LLaDA_Prism/.venv/lib/python3.12/site-packages/nvidia/cuda_cupti/include/cupti_result.h

@@ -0,0 +1,360 @@
+/*
+ * Copyright 2010-2024 NVIDIA Corporation.  All rights reserved.
+ *
+ * NOTICE TO LICENSEE:
+ *
+ * This source code and/or documentation ("Licensed Deliverables") are
+ * subject to NVIDIA intellectual property rights under U.S. and
+ * international Copyright laws.
+ *
+ * These Licensed Deliverables contained herein is PROPRIETARY and
+ * CONFIDENTIAL to NVIDIA and is being provided under the terms and
+ * conditions of a form of NVIDIA software license agreement by and
+ * between NVIDIA and Licensee ("License Agreement") or electronically
+ * accepted by Licensee.  Notwithstanding any terms or conditions to
+ * the contrary in the License Agreement, reproduction or disclosure
+ * of the Licensed Deliverables to any third party without the express
+ * written consent of NVIDIA is prohibited.
+ *
+ * NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE
+ * LICENSE AGREEMENT, NVIDIA MAKES NO REPRESENTATION ABOUT THE
+ * SUITABILITY OF THESE LICENSED DELIVERABLES FOR ANY PURPOSE.  IT IS
+ * PROVIDED "AS IS" WITHOUT EXPRESS OR IMPLIED WARRANTY OF ANY KIND.
+ * NVIDIA DISCLAIMS ALL WARRANTIES WITH REGARD TO THESE LICENSED
+ * DELIVERABLES, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY,
+ * NONINFRINGEMENT, AND FITNESS FOR A PARTICULAR PURPOSE.
+ * NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE
+ * LICENSE AGREEMENT, IN NO EVENT SHALL NVIDIA BE LIABLE FOR ANY
+ * SPECIAL, INDIRECT, INCIDENTAL, OR CONSEQUENTIAL DAMAGES, OR ANY
+ * DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS,
+ * WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS
+ * ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE
+ * OF THESE LICENSED DELIVERABLES.
+ *
+ * U.S. Government End Users.  These Licensed Deliverables are a
+ * "commercial item" as that term is defined at 48 C.F.R. 2.101 (OCT
+ * 1995), consisting of "commercial computer software" and "commercial
+ * computer software documentation" as such terms are used in 48
+ * C.F.R. 12.212 (SEPT 1995) and is provided to the U.S. Government
+ * only as a commercial end item.  Consistent with 48 C.F.R.12.212 and
+ * 48 C.F.R. 227.7202-1 through 227.7202-4 (JUNE 1995), all
+ * U.S. Government End Users acquire the Licensed Deliverables with
+ * only those rights set forth herein.
+ *
+ * Any use of the Licensed Deliverables in individual and commercial
+ * software must include, in the user documentation and internal
+ * comments to the code, the above Disclaimer and U.S. Government End
+ * Users Notice.
+ */
+
+#if !defined(_CUPTI_RESULT_H_)
+#define _CUPTI_RESULT_H_
+
+#ifndef CUPTIAPI
+#ifdef _WIN32
+#define CUPTIAPI __stdcall
+#else
+#define CUPTIAPI
+#endif
+#endif
+
+#if defined(__cplusplus)
+extern "C" {
+#endif
+
+#if defined(__GNUC__) && defined(CUPTI_LIB)
+    #pragma GCC visibility push(default)
+#endif
+
+/**
+ * \defgroup CUPTI_RESULT_API CUPTI Result Codes
+ * Error and result codes returned by CUPTI functions.
+ * @{
+ */
+
+/**
+ * \brief CUPTI result codes.
+ *
+ * Error and result codes returned by CUPTI functions.
+ */
+typedef enum {
+    /**
+     * No error.
+     */
+    CUPTI_SUCCESS                                       = 0,
+    /**
+     * One or more of the parameters is invalid.
+     */
+    CUPTI_ERROR_INVALID_PARAMETER                       = 1,
+    /**
+     * The device does not correspond to a valid CUDA device.
+     */
+    CUPTI_ERROR_INVALID_DEVICE                          = 2,
+    /**
+     * The context is NULL or not valid.
+     */
+    CUPTI_ERROR_INVALID_CONTEXT                         = 3,
+    /**
+     * The event domain id is invalid.
+     */
+    CUPTI_ERROR_INVALID_EVENT_DOMAIN_ID                 = 4,
+    /**
+     * The event id is invalid.
+     */
+    CUPTI_ERROR_INVALID_EVENT_ID                        = 5,
+    /**
+     * The event name is invalid.
+     */
+    CUPTI_ERROR_INVALID_EVENT_NAME                      = 6,
+    /**
+     * The current operation cannot be performed due to dependency on
+     * other factors.
+     */
+    CUPTI_ERROR_INVALID_OPERATION                       = 7,
+    /**
+     * Unable to allocate enough memory to perform the requested
+     * operation.
+     */
+    CUPTI_ERROR_OUT_OF_MEMORY                           = 8,
+    /**
+     * An error occurred on the performance monitoring hardware.
+     */
+    CUPTI_ERROR_HARDWARE                                = 9,
+    /**
+     * The output buffer size is not sufficient to return all
+     * requested data.
+     */
+    CUPTI_ERROR_PARAMETER_SIZE_NOT_SUFFICIENT           = 10,
+    /**
+     * API is not implemented.
+     */
+    CUPTI_ERROR_API_NOT_IMPLEMENTED                     = 11,
+    /**
+     * The maximum limit is reached.
+     */
+    CUPTI_ERROR_MAX_LIMIT_REACHED                       = 12,
+    /**
+     * The object is not yet ready to perform the requested operation.
+     */
+    CUPTI_ERROR_NOT_READY                               = 13,
+    /**
+     * The current operation is not compatible with the current state
+     * of the object
+     */
+    CUPTI_ERROR_NOT_COMPATIBLE                          = 14,
+    /**
+     * CUPTI is unable to initialize its connection to the CUDA
+     * driver.
+     */
+    CUPTI_ERROR_NOT_INITIALIZED                         = 15,
+    /**
+     * The metric id is invalid.
+     */
+    CUPTI_ERROR_INVALID_METRIC_ID                        = 16,
+    /**
+     * The metric name is invalid.
+     */
+    CUPTI_ERROR_INVALID_METRIC_NAME                      = 17,
+    /**
+     * The queue is empty.
+     */
+    CUPTI_ERROR_QUEUE_EMPTY                              = 18,
+    /**
+     * Invalid handle (internal?).
+     */
+    CUPTI_ERROR_INVALID_HANDLE                           = 19,
+    /**
+     * Invalid stream.
+     */
+    CUPTI_ERROR_INVALID_STREAM                           = 20,
+    /**
+     * Invalid kind.
+     */
+    CUPTI_ERROR_INVALID_KIND                             = 21,
+    /**
+     * Invalid event value.
+     */
+    CUPTI_ERROR_INVALID_EVENT_VALUE                      = 22,
+    /**
+     * CUPTI is disabled due to conflicts with other enabled profilers
+     */
+    CUPTI_ERROR_DISABLED                                 = 23,
+    /**
+     * Invalid module.
+     */
+    CUPTI_ERROR_INVALID_MODULE                           = 24,
+    /**
+     * Invalid metric value.
+     */
+    CUPTI_ERROR_INVALID_METRIC_VALUE                     = 25,
+    /**
+     * The performance monitoring hardware is in use by other client.
+     */
+    CUPTI_ERROR_HARDWARE_BUSY                            = 26,
+    /**
+     * The attempted operation is not supported on the current
+     * system or device.
+     */
+    CUPTI_ERROR_NOT_SUPPORTED                            = 27,
+    /**
+     * Unified memory profiling is not supported on the system.
+     * Potential reason could be unsupported OS or architecture.
+     */
+    CUPTI_ERROR_UM_PROFILING_NOT_SUPPORTED               = 28,
+    /**
+     * Unified memory profiling is not supported on the device
+     */
+    CUPTI_ERROR_UM_PROFILING_NOT_SUPPORTED_ON_DEVICE     = 29,
+    /**
+     * Unified memory profiling is not supported on a multi-GPU
+     * configuration without P2P support between any pair of devices
+     */
+    CUPTI_ERROR_UM_PROFILING_NOT_SUPPORTED_ON_NON_P2P_DEVICES = 30,
+    /**
+     * Unified memory profiling is not supported under the
+     * Multi-Process Service (MPS) environment. CUDA 7.5 removes this
+     * restriction.
+     */
+    CUPTI_ERROR_UM_PROFILING_NOT_SUPPORTED_WITH_MPS      = 31,
+    /**
+     * In CUDA 9.0, devices with compute capability 7.0 don't
+     * support CDP tracing
+     */
+    CUPTI_ERROR_CDP_TRACING_NOT_SUPPORTED                = 32,
+    /**
+     * Profiling on virtualized GPU is not supported.
+     */
+    CUPTI_ERROR_VIRTUALIZED_DEVICE_NOT_SUPPORTED         = 33,
+    /**
+     * Profiling results might be incorrect for CUDA applications
+     * compiled with nvcc version older than 9.0 for devices with
+     * compute capability 6.0 and 6.1.
+     * Profiling session will continue and CUPTI will notify it using this error code.
+     * User is advised to recompile the application code with nvcc version 9.0 or later.
+     * Ignore this warning if code is already compiled with the recommended nvcc version.
+     */
+    CUPTI_ERROR_CUDA_COMPILER_NOT_COMPATIBLE             = 34,
+    /**
+     * User doesn't have sufficient privileges which are required to
+     * start the profiling session.
+     * One possible reason for this may be that the NVIDIA driver or your system
+     * administrator may have restricted access to the NVIDIA GPU performance counters.
+     * To learn how to resolve this issue and find more information, please visit
+     * https://developer.nvidia.com/CUPTI_ERROR_INSUFFICIENT_PRIVILEGES
+     */
+    CUPTI_ERROR_INSUFFICIENT_PRIVILEGES                  = 35,
+    /**
+     * Legacy CUPTI Profiling API i.e. event API from the header cupti_events.h and
+     * metric API from the header cupti_metrics.h are not compatible with the
+     * Profiling API in the header cupti_profiler_target.h and Perfworks metrics API
+     * in the headers nvperf_host.h and nvperf_target.h.
+     */
+    CUPTI_ERROR_OLD_PROFILER_API_INITIALIZED             = 36,
+    /**
+     * Missing definition of the OpenACC API routine in the linked OpenACC library.
+     *
+     * One possible reason is that OpenACC library is linked statically in the
+     * user application, which might not have the definition of all the OpenACC
+     * API routines needed for the OpenACC profiling, as compiler might ignore
+     * definitions for the functions not used in the application. This issue
+     * can be mitigated by linking the OpenACC library dynamically.
+     */
+    CUPTI_ERROR_OPENACC_UNDEFINED_ROUTINE                = 37,
+    /**
+     * Legacy CUPTI Profiling API i.e. event API from the header cupti_events.h and
+     * metric API from the header cupti_metrics.h are not supported on devices with
+     * compute capability 7.5 and higher (i.e. Turing and later GPU architectures).
+     * These APIs are deprecated in the CUDA 12.8 release and will be removed in a future CUDA release.
+     * These are replaced by the host profiling API in the header cupti_profiler_host.h and
+     * target profiling API in the header cupti_range_profiler.h which are supported on
+     * devices with compute capability 7.0 and higher (i.e. Volta and later GPU
+     * architectures).
+     */
+    CUPTI_ERROR_LEGACY_PROFILER_NOT_SUPPORTED            = 38,
+    /**
+     * CUPTI doesn't allow multiple callback subscribers. Only a single subscriber
+     * can be registered at a time.
+     * Same error code is used when application is launched using NVIDIA tools
+     * like nvprof, Visual Profiler, Nsight Systems, Nsight Compute, cuda-gdb and
+     * cuda-memcheck.
+     */
+    CUPTI_ERROR_MULTIPLE_SUBSCRIBERS_NOT_SUPPORTED       = 39,
+    /**
+     * Profiling on virtualized GPU is not allowed by hypervisor.
+     */
+    CUPTI_ERROR_VIRTUALIZED_DEVICE_INSUFFICIENT_PRIVILEGES = 40,
+    /**
+     * Profiling and tracing are not allowed when confidential computing mode
+     * is enabled.
+     */
+    CUPTI_ERROR_CONFIDENTIAL_COMPUTING_NOT_SUPPORTED = 41,
+    /**
+     * CUPTI does not support NVIDIA Crypto Mining Processors (CMP).
+     * For more information, please visit https://developer.nvidia.com/ERR_NVCMPGPU
+    */
+    CUPTI_ERROR_CMP_DEVICE_NOT_SUPPORTED = 42,
+    /**
+     * Profiling on Multi-instance GPU (MIG) is not supported.
+     */
+    CUPTI_ERROR_MIG_DEVICE_NOT_SUPPORTED = 43,
+    /**
+     * Profiling on SLI device is not supported.
+     */
+    CUPTI_ERROR_SLI_DEVICE_NOT_SUPPORTED = 44,
+    /**
+     * Profiling on WSL device is not supported.
+     */
+    CUPTI_ERROR_WSL_DEVICE_NOT_SUPPORTED = 45,
+    /**
+     * An unknown internal error has occurred.
+     */
+    CUPTI_ERROR_UNKNOWN                                  = 999,
+    CUPTI_ERROR_FORCE_INT                                = 0x7fffffff
+} CUptiResult;
+
+/**
+ * \brief Get the descriptive string for a CUptiResult.
+ *
+ * Return the descriptive string for a CUptiResult in \p *str.
+ * \note \b Thread-safety: this function is thread safe.
+ *
+ * \param result The result to get the string for
+ * \param str Returns the string
+ *
+ * \retval CUPTI_SUCCESS on success
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p str is NULL or \p
+ * result is not a valid CUptiResult
+ */
+CUptiResult CUPTIAPI cuptiGetResultString(CUptiResult result, const char **str);
+
+/**
+ * @brief Get the descriptive message corresponding to error codes returned
+ * by CUPTI.
+ * 
+ * Return the descriptive error message for a CUptiResult in \p *str.
+ * \note \b Thread-safety: this function is thread safe.
+ * 
+ * \param result The result to get the descriptive error message for
+ * \param str Returns the error message string
+ * 
+ * \retval CUPTI_SUCCESS on success
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p str is NULL or \p
+ * result is not a valid CUptiResult
+ * 
+ */
+
+CUptiResult CUPTIAPI cuptiGetErrorMessage(CUptiResult result, const char **str);
+
+/** @} */ /* END CUPTI_RESULT_API */
+
+#if defined(__GNUC__) && defined(CUPTI_LIB)
+    #pragma GCC visibility pop
+#endif
+
+#if defined(__cplusplus)
+}
+#endif
+
+#endif /*_CUPTI_RESULT_H_*/
+
+

Prism/LLaDA/LLaDA_Prism/.venv/lib/python3.12/site-packages/nvidia/cuda_cupti/include/cupti_runtime_cbid.h

@@ -0,0 +1,504 @@
+
+// *************************************************************************
+//      Definitions of indices for API functions, unique across entire API
+// *************************************************************************
+
+// This file is generated.  Any changes you make will be lost during the next clean build.
+// CUDA public interface, for type definitions and cu* function prototypes
+
+#if !defined(_CUPTI_RUNTIME_CBID_H)
+#define _CUPTI_RUNTIME_CBID_H
+
+typedef enum CUpti_runtime_api_trace_cbid_enum {
+    CUPTI_RUNTIME_TRACE_CBID_INVALID                                                       = 0,
+    CUPTI_RUNTIME_TRACE_CBID_cudaDriverGetVersion_v3020                                    = 1,
+    CUPTI_RUNTIME_TRACE_CBID_cudaRuntimeGetVersion_v3020                                   = 2,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGetDeviceCount_v3020                                      = 3,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGetDeviceProperties_v3020                                 = 4,
+    CUPTI_RUNTIME_TRACE_CBID_cudaChooseDevice_v3020                                        = 5,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGetChannelDesc_v3020                                      = 6,
+    CUPTI_RUNTIME_TRACE_CBID_cudaCreateChannelDesc_v3020                                   = 7,
+    CUPTI_RUNTIME_TRACE_CBID_cudaConfigureCall_v3020                                       = 8,
+    CUPTI_RUNTIME_TRACE_CBID_cudaSetupArgument_v3020                                       = 9,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGetLastError_v3020                                        = 10,
+    CUPTI_RUNTIME_TRACE_CBID_cudaPeekAtLastError_v3020                                     = 11,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGetErrorString_v3020                                      = 12,
+    CUPTI_RUNTIME_TRACE_CBID_cudaLaunch_v3020                                              = 13,
+    CUPTI_RUNTIME_TRACE_CBID_cudaFuncSetCacheConfig_v3020                                  = 14,
+    CUPTI_RUNTIME_TRACE_CBID_cudaFuncGetAttributes_v3020                                   = 15,
+    CUPTI_RUNTIME_TRACE_CBID_cudaSetDevice_v3020                                           = 16,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGetDevice_v3020                                           = 17,
+    CUPTI_RUNTIME_TRACE_CBID_cudaSetValidDevices_v3020                                     = 18,
+    CUPTI_RUNTIME_TRACE_CBID_cudaSetDeviceFlags_v3020                                      = 19,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMalloc_v3020                                              = 20,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMallocPitch_v3020                                         = 21,
+    CUPTI_RUNTIME_TRACE_CBID_cudaFree_v3020                                                = 22,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMallocArray_v3020                                         = 23,
+    CUPTI_RUNTIME_TRACE_CBID_cudaFreeArray_v3020                                           = 24,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMallocHost_v3020                                          = 25,
+    CUPTI_RUNTIME_TRACE_CBID_cudaFreeHost_v3020                                            = 26,
+    CUPTI_RUNTIME_TRACE_CBID_cudaHostAlloc_v3020                                           = 27,
+    CUPTI_RUNTIME_TRACE_CBID_cudaHostGetDevicePointer_v3020                                = 28,
+    CUPTI_RUNTIME_TRACE_CBID_cudaHostGetFlags_v3020                                        = 29,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemGetInfo_v3020                                          = 30,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpy_v3020                                              = 31,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpy2D_v3020                                            = 32,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpyToArray_v3020                                       = 33,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpy2DToArray_v3020                                     = 34,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpyFromArray_v3020                                     = 35,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpy2DFromArray_v3020                                   = 36,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpyArrayToArray_v3020                                  = 37,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpy2DArrayToArray_v3020                                = 38,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpyToSymbol_v3020                                      = 39,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpyFromSymbol_v3020                                    = 40,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpyAsync_v3020                                         = 41,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpyToArrayAsync_v3020                                  = 42,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpyFromArrayAsync_v3020                                = 43,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpy2DAsync_v3020                                       = 44,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpy2DToArrayAsync_v3020                                = 45,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpy2DFromArrayAsync_v3020                              = 46,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpyToSymbolAsync_v3020                                 = 47,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpyFromSymbolAsync_v3020                               = 48,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemset_v3020                                              = 49,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemset2D_v3020                                            = 50,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemsetAsync_v3020                                         = 51,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemset2DAsync_v3020                                       = 52,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGetSymbolAddress_v3020                                    = 53,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGetSymbolSize_v3020                                       = 54,
+    CUPTI_RUNTIME_TRACE_CBID_cudaBindTexture_v3020                                         = 55,
+    CUPTI_RUNTIME_TRACE_CBID_cudaBindTexture2D_v3020                                       = 56,
+    CUPTI_RUNTIME_TRACE_CBID_cudaBindTextureToArray_v3020                                  = 57,
+    CUPTI_RUNTIME_TRACE_CBID_cudaUnbindTexture_v3020                                       = 58,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGetTextureAlignmentOffset_v3020                           = 59,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGetTextureReference_v3020                                 = 60,
+    CUPTI_RUNTIME_TRACE_CBID_cudaBindSurfaceToArray_v3020                                  = 61,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGetSurfaceReference_v3020                                 = 62,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGLSetGLDevice_v3020                                       = 63,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGLRegisterBufferObject_v3020                              = 64,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGLMapBufferObject_v3020                                   = 65,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGLUnmapBufferObject_v3020                                 = 66,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGLUnregisterBufferObject_v3020                            = 67,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGLSetBufferObjectMapFlags_v3020                           = 68,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGLMapBufferObjectAsync_v3020                              = 69,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGLUnmapBufferObjectAsync_v3020                            = 70,
+    CUPTI_RUNTIME_TRACE_CBID_cudaWGLGetDevice_v3020                                        = 71,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphicsGLRegisterImage_v3020                             = 72,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphicsGLRegisterBuffer_v3020                            = 73,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphicsUnregisterResource_v3020                          = 74,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphicsResourceSetMapFlags_v3020                         = 75,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphicsMapResources_v3020                                = 76,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphicsUnmapResources_v3020                              = 77,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphicsResourceGetMappedPointer_v3020                    = 78,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphicsSubResourceGetMappedArray_v3020                   = 79,
+    CUPTI_RUNTIME_TRACE_CBID_cudaVDPAUGetDevice_v3020                                      = 80,
+    CUPTI_RUNTIME_TRACE_CBID_cudaVDPAUSetVDPAUDevice_v3020                                 = 81,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphicsVDPAURegisterVideoSurface_v3020                   = 82,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphicsVDPAURegisterOutputSurface_v3020                  = 83,
+    CUPTI_RUNTIME_TRACE_CBID_cudaD3D11GetDevice_v3020                                      = 84,
+    CUPTI_RUNTIME_TRACE_CBID_cudaD3D11GetDevices_v3020                                     = 85,
+    CUPTI_RUNTIME_TRACE_CBID_cudaD3D11SetDirect3DDevice_v3020                              = 86,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphicsD3D11RegisterResource_v3020                       = 87,
+    CUPTI_RUNTIME_TRACE_CBID_cudaD3D10GetDevice_v3020                                      = 88,
+    CUPTI_RUNTIME_TRACE_CBID_cudaD3D10GetDevices_v3020                                     = 89,
+    CUPTI_RUNTIME_TRACE_CBID_cudaD3D10SetDirect3DDevice_v3020                              = 90,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphicsD3D10RegisterResource_v3020                       = 91,
+    CUPTI_RUNTIME_TRACE_CBID_cudaD3D10RegisterResource_v3020                               = 92,
+    CUPTI_RUNTIME_TRACE_CBID_cudaD3D10UnregisterResource_v3020                             = 93,
+    CUPTI_RUNTIME_TRACE_CBID_cudaD3D10MapResources_v3020                                   = 94,
+    CUPTI_RUNTIME_TRACE_CBID_cudaD3D10UnmapResources_v3020                                 = 95,
+    CUPTI_RUNTIME_TRACE_CBID_cudaD3D10ResourceSetMapFlags_v3020                            = 96,
+    CUPTI_RUNTIME_TRACE_CBID_cudaD3D10ResourceGetSurfaceDimensions_v3020                   = 97,
+    CUPTI_RUNTIME_TRACE_CBID_cudaD3D10ResourceGetMappedArray_v3020                         = 98,
+    CUPTI_RUNTIME_TRACE_CBID_cudaD3D10ResourceGetMappedPointer_v3020                       = 99,
+    CUPTI_RUNTIME_TRACE_CBID_cudaD3D10ResourceGetMappedSize_v3020                          = 100,
+    CUPTI_RUNTIME_TRACE_CBID_cudaD3D10ResourceGetMappedPitch_v3020                         = 101,
+    CUPTI_RUNTIME_TRACE_CBID_cudaD3D9GetDevice_v3020                                       = 102,
+    CUPTI_RUNTIME_TRACE_CBID_cudaD3D9GetDevices_v3020                                      = 103,
+    CUPTI_RUNTIME_TRACE_CBID_cudaD3D9SetDirect3DDevice_v3020                               = 104,
+    CUPTI_RUNTIME_TRACE_CBID_cudaD3D9GetDirect3DDevice_v3020                               = 105,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphicsD3D9RegisterResource_v3020                        = 106,
+    CUPTI_RUNTIME_TRACE_CBID_cudaD3D9RegisterResource_v3020                                = 107,
+    CUPTI_RUNTIME_TRACE_CBID_cudaD3D9UnregisterResource_v3020                              = 108,
+    CUPTI_RUNTIME_TRACE_CBID_cudaD3D9MapResources_v3020                                    = 109,
+    CUPTI_RUNTIME_TRACE_CBID_cudaD3D9UnmapResources_v3020                                  = 110,
+    CUPTI_RUNTIME_TRACE_CBID_cudaD3D9ResourceSetMapFlags_v3020                             = 111,
+    CUPTI_RUNTIME_TRACE_CBID_cudaD3D9ResourceGetSurfaceDimensions_v3020                    = 112,
+    CUPTI_RUNTIME_TRACE_CBID_cudaD3D9ResourceGetMappedArray_v3020                          = 113,
+    CUPTI_RUNTIME_TRACE_CBID_cudaD3D9ResourceGetMappedPointer_v3020                        = 114,
+    CUPTI_RUNTIME_TRACE_CBID_cudaD3D9ResourceGetMappedSize_v3020                           = 115,
+    CUPTI_RUNTIME_TRACE_CBID_cudaD3D9ResourceGetMappedPitch_v3020                          = 116,
+    CUPTI_RUNTIME_TRACE_CBID_cudaD3D9Begin_v3020                                           = 117,
+    CUPTI_RUNTIME_TRACE_CBID_cudaD3D9End_v3020                                             = 118,
+    CUPTI_RUNTIME_TRACE_CBID_cudaD3D9RegisterVertexBuffer_v3020                            = 119,
+    CUPTI_RUNTIME_TRACE_CBID_cudaD3D9UnregisterVertexBuffer_v3020                          = 120,
+    CUPTI_RUNTIME_TRACE_CBID_cudaD3D9MapVertexBuffer_v3020                                 = 121,
+    CUPTI_RUNTIME_TRACE_CBID_cudaD3D9UnmapVertexBuffer_v3020                               = 122,
+    CUPTI_RUNTIME_TRACE_CBID_cudaThreadExit_v3020                                          = 123,
+    CUPTI_RUNTIME_TRACE_CBID_cudaSetDoubleForDevice_v3020                                  = 124,
+    CUPTI_RUNTIME_TRACE_CBID_cudaSetDoubleForHost_v3020                                    = 125,
+    CUPTI_RUNTIME_TRACE_CBID_cudaThreadSynchronize_v3020                                   = 126,
+    CUPTI_RUNTIME_TRACE_CBID_cudaThreadGetLimit_v3020                                      = 127,
+    CUPTI_RUNTIME_TRACE_CBID_cudaThreadSetLimit_v3020                                      = 128,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamCreate_v3020                                        = 129,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamDestroy_v3020                                       = 130,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamSynchronize_v3020                                   = 131,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamQuery_v3020                                         = 132,
+    CUPTI_RUNTIME_TRACE_CBID_cudaEventCreate_v3020                                         = 133,
+    CUPTI_RUNTIME_TRACE_CBID_cudaEventCreateWithFlags_v3020                                = 134,
+    CUPTI_RUNTIME_TRACE_CBID_cudaEventRecord_v3020                                         = 135,
+    CUPTI_RUNTIME_TRACE_CBID_cudaEventDestroy_v3020                                        = 136,
+    CUPTI_RUNTIME_TRACE_CBID_cudaEventSynchronize_v3020                                    = 137,
+    CUPTI_RUNTIME_TRACE_CBID_cudaEventQuery_v3020                                          = 138,
+    CUPTI_RUNTIME_TRACE_CBID_cudaEventElapsedTime_v3020                                    = 139,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMalloc3D_v3020                                            = 140,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMalloc3DArray_v3020                                       = 141,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemset3D_v3020                                            = 142,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemset3DAsync_v3020                                       = 143,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpy3D_v3020                                            = 144,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpy3DAsync_v3020                                       = 145,
+    CUPTI_RUNTIME_TRACE_CBID_cudaThreadSetCacheConfig_v3020                                = 146,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamWaitEvent_v3020                                     = 147,
+    CUPTI_RUNTIME_TRACE_CBID_cudaD3D11GetDirect3DDevice_v3020                              = 148,
+    CUPTI_RUNTIME_TRACE_CBID_cudaD3D10GetDirect3DDevice_v3020                              = 149,
+    CUPTI_RUNTIME_TRACE_CBID_cudaThreadGetCacheConfig_v3020                                = 150,
+    CUPTI_RUNTIME_TRACE_CBID_cudaPointerGetAttributes_v4000                                = 151,
+    CUPTI_RUNTIME_TRACE_CBID_cudaHostRegister_v4000                                        = 152,
+    CUPTI_RUNTIME_TRACE_CBID_cudaHostUnregister_v4000                                      = 153,
+    CUPTI_RUNTIME_TRACE_CBID_cudaDeviceCanAccessPeer_v4000                                 = 154,
+    CUPTI_RUNTIME_TRACE_CBID_cudaDeviceEnablePeerAccess_v4000                              = 155,
+    CUPTI_RUNTIME_TRACE_CBID_cudaDeviceDisablePeerAccess_v4000                             = 156,
+    CUPTI_RUNTIME_TRACE_CBID_cudaPeerRegister_v4000                                        = 157,
+    CUPTI_RUNTIME_TRACE_CBID_cudaPeerUnregister_v4000                                      = 158,
+    CUPTI_RUNTIME_TRACE_CBID_cudaPeerGetDevicePointer_v4000                                = 159,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpyPeer_v4000                                          = 160,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpyPeerAsync_v4000                                     = 161,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpy3DPeer_v4000                                        = 162,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpy3DPeerAsync_v4000                                   = 163,
+    CUPTI_RUNTIME_TRACE_CBID_cudaDeviceReset_v3020                                         = 164,
+    CUPTI_RUNTIME_TRACE_CBID_cudaDeviceSynchronize_v3020                                   = 165,
+    CUPTI_RUNTIME_TRACE_CBID_cudaDeviceGetLimit_v3020                                      = 166,
+    CUPTI_RUNTIME_TRACE_CBID_cudaDeviceSetLimit_v3020                                      = 167,
+    CUPTI_RUNTIME_TRACE_CBID_cudaDeviceGetCacheConfig_v3020                                = 168,
+    CUPTI_RUNTIME_TRACE_CBID_cudaDeviceSetCacheConfig_v3020                                = 169,
+    CUPTI_RUNTIME_TRACE_CBID_cudaProfilerInitialize_v4000                                  = 170,
+    CUPTI_RUNTIME_TRACE_CBID_cudaProfilerStart_v4000                                       = 171,
+    CUPTI_RUNTIME_TRACE_CBID_cudaProfilerStop_v4000                                        = 172,
+    CUPTI_RUNTIME_TRACE_CBID_cudaDeviceGetByPCIBusId_v4010                                 = 173,
+    CUPTI_RUNTIME_TRACE_CBID_cudaDeviceGetPCIBusId_v4010                                   = 174,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGLGetDevices_v4010                                        = 175,
+    CUPTI_RUNTIME_TRACE_CBID_cudaIpcGetEventHandle_v4010                                   = 176,
+    CUPTI_RUNTIME_TRACE_CBID_cudaIpcOpenEventHandle_v4010                                  = 177,
+    CUPTI_RUNTIME_TRACE_CBID_cudaIpcGetMemHandle_v4010                                     = 178,
+    CUPTI_RUNTIME_TRACE_CBID_cudaIpcOpenMemHandle_v4010                                    = 179,
+    CUPTI_RUNTIME_TRACE_CBID_cudaIpcCloseMemHandle_v4010                                   = 180,
+    CUPTI_RUNTIME_TRACE_CBID_cudaArrayGetInfo_v4010                                        = 181,
+    CUPTI_RUNTIME_TRACE_CBID_cudaFuncSetSharedMemConfig_v4020                              = 182,
+    CUPTI_RUNTIME_TRACE_CBID_cudaDeviceGetSharedMemConfig_v4020                            = 183,
+    CUPTI_RUNTIME_TRACE_CBID_cudaDeviceSetSharedMemConfig_v4020                            = 184,
+    CUPTI_RUNTIME_TRACE_CBID_cudaCreateTextureObject_v5000                                 = 185,
+    CUPTI_RUNTIME_TRACE_CBID_cudaDestroyTextureObject_v5000                                = 186,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGetTextureObjectResourceDesc_v5000                        = 187,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGetTextureObjectTextureDesc_v5000                         = 188,
+    CUPTI_RUNTIME_TRACE_CBID_cudaCreateSurfaceObject_v5000                                 = 189,
+    CUPTI_RUNTIME_TRACE_CBID_cudaDestroySurfaceObject_v5000                                = 190,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGetSurfaceObjectResourceDesc_v5000                        = 191,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMallocMipmappedArray_v5000                                = 192,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGetMipmappedArrayLevel_v5000                              = 193,
+    CUPTI_RUNTIME_TRACE_CBID_cudaFreeMipmappedArray_v5000                                  = 194,
+    CUPTI_RUNTIME_TRACE_CBID_cudaBindTextureToMipmappedArray_v5000                         = 195,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphicsResourceGetMappedMipmappedArray_v5000             = 196,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamAddCallback_v5000                                   = 197,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamCreateWithFlags_v5000                               = 198,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGetTextureObjectResourceViewDesc_v5000                    = 199,
+    CUPTI_RUNTIME_TRACE_CBID_cudaDeviceGetAttribute_v5000                                  = 200,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamDestroy_v5050                                       = 201,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamCreateWithPriority_v5050                            = 202,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamGetPriority_v5050                                   = 203,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamGetFlags_v5050                                      = 204,
+    CUPTI_RUNTIME_TRACE_CBID_cudaDeviceGetStreamPriorityRange_v5050                        = 205,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMallocManaged_v6000                                       = 206,
+    CUPTI_RUNTIME_TRACE_CBID_cudaOccupancyMaxActiveBlocksPerMultiprocessor_v6000           = 207,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamAttachMemAsync_v6000                                = 208,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGetErrorName_v6050                                        = 209,
+    CUPTI_RUNTIME_TRACE_CBID_cudaOccupancyMaxActiveBlocksPerMultiprocessor_v6050           = 210,
+    CUPTI_RUNTIME_TRACE_CBID_cudaLaunchKernel_v7000                                        = 211,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGetDeviceFlags_v7000                                      = 212,
+    CUPTI_RUNTIME_TRACE_CBID_cudaLaunch_ptsz_v7000                                         = 213,
+    CUPTI_RUNTIME_TRACE_CBID_cudaLaunchKernel_ptsz_v7000                                   = 214,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpy_ptds_v7000                                         = 215,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpy2D_ptds_v7000                                       = 216,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpyToArray_ptds_v7000                                  = 217,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpy2DToArray_ptds_v7000                                = 218,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpyFromArray_ptds_v7000                                = 219,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpy2DFromArray_ptds_v7000                              = 220,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpyArrayToArray_ptds_v7000                             = 221,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpy2DArrayToArray_ptds_v7000                           = 222,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpyToSymbol_ptds_v7000                                 = 223,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpyFromSymbol_ptds_v7000                               = 224,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpyAsync_ptsz_v7000                                    = 225,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpyToArrayAsync_ptsz_v7000                             = 226,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpyFromArrayAsync_ptsz_v7000                           = 227,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpy2DAsync_ptsz_v7000                                  = 228,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpy2DToArrayAsync_ptsz_v7000                           = 229,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpy2DFromArrayAsync_ptsz_v7000                         = 230,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpyToSymbolAsync_ptsz_v7000                            = 231,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpyFromSymbolAsync_ptsz_v7000                          = 232,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemset_ptds_v7000                                         = 233,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemset2D_ptds_v7000                                       = 234,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemsetAsync_ptsz_v7000                                    = 235,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemset2DAsync_ptsz_v7000                                  = 236,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamGetPriority_ptsz_v7000                              = 237,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamGetFlags_ptsz_v7000                                 = 238,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamSynchronize_ptsz_v7000                              = 239,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamQuery_ptsz_v7000                                    = 240,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamAttachMemAsync_ptsz_v7000                           = 241,
+    CUPTI_RUNTIME_TRACE_CBID_cudaEventRecord_ptsz_v7000                                    = 242,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemset3D_ptds_v7000                                       = 243,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemset3DAsync_ptsz_v7000                                  = 244,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpy3D_ptds_v7000                                       = 245,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpy3DAsync_ptsz_v7000                                  = 246,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamWaitEvent_ptsz_v7000                                = 247,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamAddCallback_ptsz_v7000                              = 248,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpy3DPeer_ptds_v7000                                   = 249,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpy3DPeerAsync_ptsz_v7000                              = 250,
+    CUPTI_RUNTIME_TRACE_CBID_cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags_v7000  = 251,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemPrefetchAsync_v8000                                    = 252,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemPrefetchAsync_ptsz_v8000                               = 253,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemAdvise_v8000                                           = 254,
+    CUPTI_RUNTIME_TRACE_CBID_cudaDeviceGetP2PAttribute_v8000                               = 255,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphicsEGLRegisterImage_v7000                            = 256,
+    CUPTI_RUNTIME_TRACE_CBID_cudaEGLStreamConsumerConnect_v7000                            = 257,
+    CUPTI_RUNTIME_TRACE_CBID_cudaEGLStreamConsumerDisconnect_v7000                         = 258,
+    CUPTI_RUNTIME_TRACE_CBID_cudaEGLStreamConsumerAcquireFrame_v7000                       = 259,
+    CUPTI_RUNTIME_TRACE_CBID_cudaEGLStreamConsumerReleaseFrame_v7000                       = 260,
+    CUPTI_RUNTIME_TRACE_CBID_cudaEGLStreamProducerConnect_v7000                            = 261,
+    CUPTI_RUNTIME_TRACE_CBID_cudaEGLStreamProducerDisconnect_v7000                         = 262,
+    CUPTI_RUNTIME_TRACE_CBID_cudaEGLStreamProducerPresentFrame_v7000                       = 263,
+    CUPTI_RUNTIME_TRACE_CBID_cudaEGLStreamProducerReturnFrame_v7000                        = 264,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphicsResourceGetMappedEglFrame_v7000                   = 265,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemRangeGetAttribute_v8000                                = 266,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemRangeGetAttributes_v8000                               = 267,
+    CUPTI_RUNTIME_TRACE_CBID_cudaEGLStreamConsumerConnectWithFlags_v7000                   = 268,
+    CUPTI_RUNTIME_TRACE_CBID_cudaLaunchCooperativeKernel_v9000                             = 269,
+    CUPTI_RUNTIME_TRACE_CBID_cudaLaunchCooperativeKernel_ptsz_v9000                        = 270,
+    CUPTI_RUNTIME_TRACE_CBID_cudaEventCreateFromEGLSync_v9000                              = 271,
+    CUPTI_RUNTIME_TRACE_CBID_cudaLaunchCooperativeKernelMultiDevice_v9000                  = 272,
+    CUPTI_RUNTIME_TRACE_CBID_cudaFuncSetAttribute_v9000                                    = 273,
+    CUPTI_RUNTIME_TRACE_CBID_cudaImportExternalMemory_v10000                               = 274,
+    CUPTI_RUNTIME_TRACE_CBID_cudaExternalMemoryGetMappedBuffer_v10000                      = 275,
+    CUPTI_RUNTIME_TRACE_CBID_cudaExternalMemoryGetMappedMipmappedArray_v10000              = 276,
+    CUPTI_RUNTIME_TRACE_CBID_cudaDestroyExternalMemory_v10000                              = 277,
+    CUPTI_RUNTIME_TRACE_CBID_cudaImportExternalSemaphore_v10000                            = 278,
+    CUPTI_RUNTIME_TRACE_CBID_cudaSignalExternalSemaphoresAsync_v10000                      = 279,
+    CUPTI_RUNTIME_TRACE_CBID_cudaSignalExternalSemaphoresAsync_ptsz_v10000                 = 280,
+    CUPTI_RUNTIME_TRACE_CBID_cudaWaitExternalSemaphoresAsync_v10000                        = 281,
+    CUPTI_RUNTIME_TRACE_CBID_cudaWaitExternalSemaphoresAsync_ptsz_v10000                   = 282,
+    CUPTI_RUNTIME_TRACE_CBID_cudaDestroyExternalSemaphore_v10000                           = 283,
+    CUPTI_RUNTIME_TRACE_CBID_cudaLaunchHostFunc_v10000                                     = 284,
+    CUPTI_RUNTIME_TRACE_CBID_cudaLaunchHostFunc_ptsz_v10000                                = 285,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphCreate_v10000                                        = 286,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphKernelNodeGetParams_v10000                           = 287,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphKernelNodeSetParams_v10000                           = 288,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphAddKernelNode_v10000                                 = 289,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphAddMemcpyNode_v10000                                 = 290,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphMemcpyNodeGetParams_v10000                           = 291,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphMemcpyNodeSetParams_v10000                           = 292,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphAddMemsetNode_v10000                                 = 293,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphMemsetNodeGetParams_v10000                           = 294,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphMemsetNodeSetParams_v10000                           = 295,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphAddHostNode_v10000                                   = 296,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphHostNodeGetParams_v10000                             = 297,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphAddChildGraphNode_v10000                             = 298,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphChildGraphNodeGetGraph_v10000                        = 299,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphAddEmptyNode_v10000                                  = 300,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphClone_v10000                                         = 301,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphNodeFindInClone_v10000                               = 302,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphNodeGetType_v10000                                   = 303,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphGetRootNodes_v10000                                  = 304,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphNodeGetDependencies_v10000                           = 305,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphNodeGetDependentNodes_v10000                         = 306,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphAddDependencies_v10000                               = 307,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphRemoveDependencies_v10000                            = 308,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphDestroyNode_v10000                                   = 309,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphInstantiate_v10000                                   = 310,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphLaunch_v10000                                        = 311,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphLaunch_ptsz_v10000                                   = 312,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphExecDestroy_v10000                                   = 313,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphDestroy_v10000                                       = 314,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamBeginCapture_v10000                                 = 315,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamBeginCapture_ptsz_v10000                            = 316,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamIsCapturing_v10000                                  = 317,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamIsCapturing_ptsz_v10000                             = 318,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamEndCapture_v10000                                   = 319,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamEndCapture_ptsz_v10000                              = 320,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphHostNodeSetParams_v10000                             = 321,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphGetNodes_v10000                                      = 322,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphGetEdges_v10000                                      = 323,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamGetCaptureInfo_v10010                               = 324,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamGetCaptureInfo_ptsz_v10010                          = 325,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphExecKernelNodeSetParams_v10010                       = 326,
+    CUPTI_RUNTIME_TRACE_CBID_cudaThreadExchangeStreamCaptureMode_v10010                    = 327,
+    CUPTI_RUNTIME_TRACE_CBID_cudaDeviceGetNvSciSyncAttributes_v10020                       = 328,
+    CUPTI_RUNTIME_TRACE_CBID_cudaOccupancyAvailableDynamicSMemPerBlock_v10200              = 329,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamSetFlags_v10200                                     = 330,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamSetFlags_ptsz_v10200                                = 331,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphExecMemcpyNodeSetParams_v10020                       = 332,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphExecMemsetNodeSetParams_v10020                       = 333,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphExecHostNodeSetParams_v10020                         = 334,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphExecUpdate_v10020                                    = 335,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGetFuncBySymbol_v11000                                    = 336,
+    CUPTI_RUNTIME_TRACE_CBID_cudaCtxResetPersistingL2Cache_v11000                          = 337,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphKernelNodeCopyAttributes_v11000                      = 338,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphKernelNodeGetAttribute_v11000                        = 339,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphKernelNodeSetAttribute_v11000                        = 340,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamCopyAttributes_v11000                               = 341,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamCopyAttributes_ptsz_v11000                          = 342,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamGetAttribute_v11000                                 = 343,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamGetAttribute_ptsz_v11000                            = 344,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamSetAttribute_v11000                                 = 345,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamSetAttribute_ptsz_v11000                            = 346,
+    CUPTI_RUNTIME_TRACE_CBID_cudaDeviceGetTexture1DLinearMaxWidth_v11010                   = 347,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphUpload_v10000                                        = 348,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphUpload_ptsz_v10000                                   = 349,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphAddMemcpyNodeToSymbol_v11010                         = 350,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphAddMemcpyNodeFromSymbol_v11010                       = 351,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphAddMemcpyNode1D_v11010                               = 352,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphMemcpyNodeSetParamsToSymbol_v11010                   = 353,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphMemcpyNodeSetParamsFromSymbol_v11010                 = 354,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphMemcpyNodeSetParams1D_v11010                         = 355,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphExecMemcpyNodeSetParamsToSymbol_v11010               = 356,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphExecMemcpyNodeSetParamsFromSymbol_v11010             = 357,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphExecMemcpyNodeSetParams1D_v11010                     = 358,
+    CUPTI_RUNTIME_TRACE_CBID_cudaArrayGetSparseProperties_v11010                           = 359,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMipmappedArrayGetSparseProperties_v11010                  = 360,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphExecChildGraphNodeSetParams_v11010                   = 361,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphAddEventRecordNode_v11010                            = 362,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphEventRecordNodeGetEvent_v11010                       = 363,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphEventRecordNodeSetEvent_v11010                       = 364,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphAddEventWaitNode_v11010                              = 365,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphEventWaitNodeGetEvent_v11010                         = 366,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphEventWaitNodeSetEvent_v11010                         = 367,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphExecEventRecordNodeSetEvent_v11010                   = 368,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphExecEventWaitNodeSetEvent_v11010                     = 369,
+    CUPTI_RUNTIME_TRACE_CBID_cudaEventRecordWithFlags_v11010                               = 370,
+    CUPTI_RUNTIME_TRACE_CBID_cudaEventRecordWithFlags_ptsz_v11010                          = 371,
+    CUPTI_RUNTIME_TRACE_CBID_cudaDeviceGetDefaultMemPool_v11020                            = 372,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMallocAsync_v11020                                        = 373,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMallocAsync_ptsz_v11020                                   = 374,
+    CUPTI_RUNTIME_TRACE_CBID_cudaFreeAsync_v11020                                          = 375,
+    CUPTI_RUNTIME_TRACE_CBID_cudaFreeAsync_ptsz_v11020                                     = 376,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemPoolTrimTo_v11020                                      = 377,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemPoolSetAttribute_v11020                                = 378,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemPoolGetAttribute_v11020                                = 379,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemPoolSetAccess_v11020                                   = 380,
+    CUPTI_RUNTIME_TRACE_CBID_cudaArrayGetPlane_v11020                                      = 381,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemPoolGetAccess_v11020                                   = 382,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemPoolCreate_v11020                                      = 383,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemPoolDestroy_v11020                                     = 384,
+    CUPTI_RUNTIME_TRACE_CBID_cudaDeviceSetMemPool_v11020                                   = 385,
+    CUPTI_RUNTIME_TRACE_CBID_cudaDeviceGetMemPool_v11020                                   = 386,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemPoolExportToShareableHandle_v11020                     = 387,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemPoolImportFromShareableHandle_v11020                   = 388,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemPoolExportPointer_v11020                               = 389,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemPoolImportPointer_v11020                               = 390,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMallocFromPoolAsync_v11020                                = 391,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMallocFromPoolAsync_ptsz_v11020                           = 392,
+    CUPTI_RUNTIME_TRACE_CBID_cudaSignalExternalSemaphoresAsync_v2_v11020                   = 393,
+    CUPTI_RUNTIME_TRACE_CBID_cudaSignalExternalSemaphoresAsync_v2_ptsz_v11020              = 394,
+    CUPTI_RUNTIME_TRACE_CBID_cudaWaitExternalSemaphoresAsync_v2_v11020                     = 395,
+    CUPTI_RUNTIME_TRACE_CBID_cudaWaitExternalSemaphoresAsync_v2_ptsz_v11020                = 396,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphAddExternalSemaphoresSignalNode_v11020               = 397,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphExternalSemaphoresSignalNodeGetParams_v11020         = 398,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphExternalSemaphoresSignalNodeSetParams_v11020         = 399,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphAddExternalSemaphoresWaitNode_v11020                 = 400,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphExternalSemaphoresWaitNodeGetParams_v11020           = 401,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphExternalSemaphoresWaitNodeSetParams_v11020           = 402,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphExecExternalSemaphoresSignalNodeSetParams_v11020     = 403,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphExecExternalSemaphoresWaitNodeSetParams_v11020       = 404,
+    CUPTI_RUNTIME_TRACE_CBID_cudaDeviceFlushGPUDirectRDMAWrites_v11030                     = 405,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGetDriverEntryPoint_v11030                                = 406,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGetDriverEntryPoint_ptsz_v11030                           = 407,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphDebugDotPrint_v11030                                 = 408,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamGetCaptureInfo_v2_v11030                            = 409,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamGetCaptureInfo_v2_ptsz_v11030                       = 410,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamUpdateCaptureDependencies_v11030                    = 411,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamUpdateCaptureDependencies_ptsz_v11030               = 412,
+    CUPTI_RUNTIME_TRACE_CBID_cudaUserObjectCreate_v11030                                   = 413,
+    CUPTI_RUNTIME_TRACE_CBID_cudaUserObjectRetain_v11030                                   = 414,
+    CUPTI_RUNTIME_TRACE_CBID_cudaUserObjectRelease_v11030                                  = 415,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphRetainUserObject_v11030                              = 416,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphReleaseUserObject_v11030                             = 417,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphInstantiateWithFlags_v11040                          = 418,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphAddMemAllocNode_v11040                               = 419,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphMemAllocNodeGetParams_v11040                         = 420,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphAddMemFreeNode_v11040                                = 421,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphMemFreeNodeGetParams_v11040                          = 422,
+    CUPTI_RUNTIME_TRACE_CBID_cudaDeviceGraphMemTrim_v11040                                 = 423,
+    CUPTI_RUNTIME_TRACE_CBID_cudaDeviceGetGraphMemAttribute_v11040                         = 424,
+    CUPTI_RUNTIME_TRACE_CBID_cudaDeviceSetGraphMemAttribute_v11040                         = 425,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphNodeSetEnabled_v11060                                = 426,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphNodeGetEnabled_v11060                                = 427,
+    CUPTI_RUNTIME_TRACE_CBID_cudaArrayGetMemoryRequirements_v11060                         = 428,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMipmappedArrayGetMemoryRequirements_v11060                = 429,
+    CUPTI_RUNTIME_TRACE_CBID_cudaLaunchKernelExC_v11060                                    = 430,
+    CUPTI_RUNTIME_TRACE_CBID_cudaLaunchKernelExC_ptsz_v11060                               = 431,
+    CUPTI_RUNTIME_TRACE_CBID_cudaOccupancyMaxPotentialClusterSize_v11070                   = 432,
+    CUPTI_RUNTIME_TRACE_CBID_cudaOccupancyMaxActiveClusters_v11070                         = 433,
+    CUPTI_RUNTIME_TRACE_CBID_cudaCreateTextureObject_v2_v11080                             = 434,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGetTextureObjectTextureDesc_v2_v11080                     = 435,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphInstantiateWithParams_v12000                         = 436,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphInstantiateWithParams_ptsz_v12000                    = 437,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphExecGetFlags_v12000                                  = 438,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGetKernel_v12000                                          = 439,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGetDeviceProperties_v2_v12000                             = 440,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamGetId_v12000                                        = 441,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamGetId_ptsz_v12000                                   = 442,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphInstantiate_v12000                                   = 443,
+    CUPTI_RUNTIME_TRACE_CBID_cudaInitDevice_v12000                                         = 444,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphAddNode_v12020                                       = 445,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphNodeSetParams_v12020                                 = 446,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphExecNodeSetParams_v12020                             = 447,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemAdvise_v2_v12020                                       = 448,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemPrefetchAsync_v2_v12020                                = 449,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemPrefetchAsync_v2_ptsz_v12020                           = 450,
+    CUPTI_RUNTIME_TRACE_CBID_cudaFuncGetName_v12030                                        = 451,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamBeginCaptureToGraph_v12030                          = 452,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamBeginCaptureToGraph_ptsz_v12030                     = 453,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphConditionalHandleCreate_v12030                       = 454,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphGetEdges_v2_v12030                                   = 455,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphNodeGetDependencies_v2_v12030                        = 456,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphNodeGetDependentNodes_v2_v12030                      = 457,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphAddDependencies_v2_v12030                            = 458,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphRemoveDependencies_v2_v12030                         = 459,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGraphAddNode_v2_v12030                                    = 460,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamGetCaptureInfo_v3_v12030                            = 461,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamGetCaptureInfo_v3_ptsz_v12030                       = 462,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamUpdateCaptureDependencies_v2_v12030                 = 463,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamUpdateCaptureDependencies_v2_ptsz_v12030            = 464,
+    CUPTI_RUNTIME_TRACE_CBID_cudaDeviceRegisterAsyncNotification_v12040                    = 465,
+    CUPTI_RUNTIME_TRACE_CBID_cudaDeviceUnregisterAsyncNotification_v12040                  = 466,
+    CUPTI_RUNTIME_TRACE_CBID_cudaFuncGetParamInfo_v12040                                   = 467,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGetDriverEntryPointByVersion_v12050                       = 468,
+    CUPTI_RUNTIME_TRACE_CBID_cudaGetDriverEntryPointByVersion_ptsz_v12050                  = 469,
+    CUPTI_RUNTIME_TRACE_CBID_cuda470_v12060                                                = 470,
+    CUPTI_RUNTIME_TRACE_CBID_cuda471_v12060                                                = 471,
+    CUPTI_RUNTIME_TRACE_CBID_cuda472_v12060                                                = 472,
+    CUPTI_RUNTIME_TRACE_CBID_cuda473_v12060                                                = 473,
+    CUPTI_RUNTIME_TRACE_CBID_cuda474_v12060                                                = 474,
+    CUPTI_RUNTIME_TRACE_CBID_cuda475_v12060                                                = 475,
+    CUPTI_RUNTIME_TRACE_CBID_cuda476_v12060                                                = 476,
+    CUPTI_RUNTIME_TRACE_CBID_cuda477_v12060                                                = 477,
+    CUPTI_RUNTIME_TRACE_CBID_cuda478_v12060                                                = 478,
+    CUPTI_RUNTIME_TRACE_CBID_cuda479_v12060                                                = 479,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamGetDevice_v12080                                    = 480,
+    CUPTI_RUNTIME_TRACE_CBID_cudaStreamGetDevice_ptsz_v12080                               = 481,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpyBatchAsync_v12080                                   = 482,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpyBatchAsync_ptsz_v12080                              = 483,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpy3DBatchAsync_v12080                                 = 484,
+    CUPTI_RUNTIME_TRACE_CBID_cudaMemcpy3DBatchAsync_ptsz_v12080                            = 485,
+    CUPTI_RUNTIME_TRACE_CBID_cudaEventElapsedTime_v2_v12080                                = 486,
+    CUPTI_RUNTIME_TRACE_CBID_SIZE                                                          = 487,
+    CUPTI_RUNTIME_TRACE_CBID_FORCE_INT                                                     = 0x7fffffff
+} CUpti_runtime_api_trace_cbid;
+
+#endif

Prism/LLaDA/LLaDA_Prism/.venv/lib/python3.12/site-packages/nvidia/cuda_cupti/include/cupti_sass_metrics.h

@@ -0,0 +1,436 @@
+/*
+ * Copyright 2023 NVIDIA Corporation. All rights reserved.
+ *
+ * NOTICE TO LICENSEE:
+ *
+ * This source code and/or documentation ("Licensed Deliverables") are
+ * subject to NVIDIA intellectual property rights under U.S. and
+ * international Copyright laws.
+ *
+ * These Licensed Deliverables contained herein is PROPRIETARY and
+ * CONFIDENTIAL to NVIDIA and is being provided under the terms and
+ * conditions of a form of NVIDIA software license agreement by and
+ * between NVIDIA and Licensee ("License Agreement") or electronically
+ * accepted by Licensee.  Notwithstanding any terms or conditions to
+ * the contrary in the License Agreement, reproduction or disclosure
+ * of the Licensed Deliverables to any third party without the express
+ * written consent of NVIDIA is prohibited.
+ *
+ * NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE
+ * LICENSE AGREEMENT, NVIDIA MAKES NO REPRESENTATION ABOUT THE
+ * SUITABILITY OF THESE LICENSED DELIVERABLES FOR ANY PURPOSE.  IT IS
+ * PROVIDED "AS IS" WITHOUT EXPRESS OR IMPLIED WARRANTY OF ANY KIND.
+ * NVIDIA DISCLAIMS ALL WARRANTIES WITH REGARD TO THESE LICENSED
+ * DELIVERABLES, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY,
+ * NONINFRINGEMENT, AND FITNESS FOR A PARTICULAR PURPOSE.
+ * NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE
+ * LICENSE AGREEMENT, IN NO EVENT SHALL NVIDIA BE LIABLE FOR ANY
+ * SPECIAL, INDIRECT, INCIDENTAL, OR CONSEQUENTIAL DAMAGES, OR ANY
+ * DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS,
+ * WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS
+ * ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE
+ * OF THESE LICENSED DELIVERABLES.
+ *
+ * U.S. Government End Users.  These Licensed Deliverables are a
+ * "commercial item" as that term is defined at 48 C.F.R. 2.101 (OCT
+ * 1995), consisting of "commercial computer software" and "commercial
+ * computer software documentation" as such terms are used in 48
+ * C.F.R. 12.212 (SEPT 1995) and is provided to the U.S. Government
+ * only as a commercial end item.  Consistent with 48 C.F.R.12.212 and
+ * 48 C.F.R. 227.7202-1 through 227.7202-4 (JUNE 1995), all
+ * U.S. Government End Users acquire the Licensed Deliverables with
+ * only those rights set forth herein.
+ *
+ * Any use of the Licensed Deliverables in individual and commercial
+ * software must include, in the user documentation and internal
+ * comments to the code, the above Disclaimer and U.S. Government End
+ * Users Notice.
+ */
+
+#if !defined(_CUPTI_SASS_METRICS_H_)
+#define _CUPTI_SASS_METRICS_H_
+
+#include <cuda.h>
+#include <cupti_result.h>
+#include <cupti_profiler_target.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#if defined(__GNUC__) && defined(CUPTI_LIB)
+    #pragma GCC visibility push(default)
+#endif
+
+/**
+ * \defgroup CUPTI_SASS_METRICS_API CUPTI SASS Metrics API
+ * Functions, types, and enums that implement the CUPTI SASS Metrics API.
+ * @{
+ */
+
+typedef enum
+{
+    /// SASS metric data will be collected at GPU level. 
+    /// In CUpti_SassMetricsGetDataProperties_Params struct the numOfInstances will be equal to 1
+    CUPTI_SASS_METRICS_OUTPUT_GRANULARITY_GPU = 0,
+
+    /// SASS metric data will be collected at SM level
+    /// In CUpti_SassMetricsGetDataProperties_Params struct the numOfInstances will be equal to number of SMs in the GPU
+    CUPTI_SASS_METRICS_OUTPUT_GRANULARITY_SM = 1,
+
+    /// SASS metric data will be collected at SM sub-partition level
+    /// In CUpti_SassMetricsGetDataProperties_Params struct the numOfInstances will be equal to number of SM sub-partitions in the GPU
+    CUPTI_SASS_METRICS_OUTPUT_GRANULARITY_SMSP = 2,
+
+    CUPTI_SASS_METRICS_OUTPUT_GRANULARITY_INVALID
+} CUpti_SassMetrics_OutputGranularity;
+
+typedef struct CUpti_SassMetrics_MetricDetails
+{
+    /// unique ID for the SASS metric
+    uint64_t metricId;
+    /// metric name
+    const char* pMetricName;
+    /// metric description
+    const char* pMetricDescription;
+} CUpti_SassMetrics_MetricDetails;
+
+/**
+ * \brief Params for cuptiSassMetricsGetNumOfMetrics
+ */
+typedef struct CUpti_SassMetrics_GetNumOfMetrics_Params
+{
+    /// [in] should be equal to CUpti_SassMetrics_GetNumOfMetrics_Params_STRUCT_SIZE
+    size_t structSize;
+    /// [in] assign to NULL
+    void* pPriv;
+    /// [in] chip name for which metrics will be queried
+    const char* pChipName;
+    /// [out] number of metrics supported for the queried chip
+    size_t numOfMetrics;
+} CUpti_SassMetrics_GetNumOfMetrics_Params;
+
+#define CUpti_SassMetrics_GetNumOfMetrics_Params_STRUCT_SIZE                 CUPTI_PROFILER_STRUCT_SIZE(CUpti_SassMetrics_GetNumOfMetrics_Params, numOfMetrics)
+
+/**
+ * \brief Get the number of supported SASS metrics for the chip.
+ * 
+ * \param pParams A pointer to \ref CUpti_SassMetrics_GetNumOfMetrics_Params
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_NOT_SUPPORTED indicates that the system/device doesn't support SASS metric collection
+ */
+CUptiResult CUPTIAPI cuptiSassMetricsGetNumOfMetrics(CUpti_SassMetrics_GetNumOfMetrics_Params* pParams);
+
+/**
+ * \brief Params for cuptiSassMetricsGetMetrics
+ */
+typedef struct CUpti_SassMetrics_GetMetrics_Params
+{
+    /// [in] should be equal to CUpti_SassMetrics_GetMetrics_Params_STRUCT_SIZE
+    size_t structSize;
+    /// [in] assign to NULL
+    void* pPriv;
+    /// [in] chip name for which metrics will be queried
+    const char* pChipName;
+    /// [in] number of metrics supported for the queried chip (can be queried using cuptiSassMetricsGetNumOfMetrics())
+    size_t numOfMetrics;
+    /// [out] list of metrics supported for queried chip
+    CUpti_SassMetrics_MetricDetails* pMetricsList;
+} CUpti_SassMetrics_GetMetrics_Params;
+#define CUpti_SassMetrics_GetMetrics_Params_STRUCT_SIZE                 CUPTI_PROFILER_STRUCT_SIZE(CUpti_SassMetrics_GetMetrics_Params, pMetricsList)
+
+/**
+ * \brief Get the list of all supported SASS metrics for the chip.
+ * 
+ * \param pParams A pointer to \ref CUpti_SassMetrics_GetMetrics_Params
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_NOT_SUPPORTED indicates that the system/device doesn't support SASS metric collection
+ */
+CUptiResult CUPTIAPI cuptiSassMetricsGetMetrics(CUpti_SassMetrics_GetMetrics_Params* pParams);
+
+/**
+ * \brief Params for cuptiSassMetricsGetProperties
+ */
+typedef struct CUpti_SassMetrics_GetProperties_Params
+{
+    /// [in] should be equal to CUpti_SassMetrics_GetProperties_Params_STRUCT_SIZE
+    size_t structSize;
+    /// [in] assign to NULL
+    void* pPriv;
+    /// [in] chip name for which metric will be queried
+    const char* pChipName;
+    /// [in] metric name
+    const char* pMetricName;
+    /// [out] returns the metric ID and the metric description
+    CUpti_SassMetrics_MetricDetails metric;
+} CUpti_SassMetrics_GetProperties_Params;
+#define CUpti_SassMetrics_GetProperties_Params_STRUCT_SIZE        CUPTI_PROFILER_STRUCT_SIZE(CUpti_SassMetrics_GetProperties_Params, metric)
+
+/**
+ * \brief Get metric properties for the queried metric.
+ * For a given metric the results will be put in CUpti_SassMetrics_MetricDetails which
+ * stores metric ID, description of the metric.
+ * 
+ * \param pParams A pointer to \ref CUpti_SassMetrics_GetProperties_Params
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_NOT_SUPPORTED indicates that the system/device doesn't support SASS metric data collection
+ */
+CUptiResult CUPTIAPI cuptiSassMetricsGetProperties(CUpti_SassMetrics_GetProperties_Params *pParams);
+
+typedef struct CUpti_SassMetrics_Config
+{
+    /// [in] unique id for the SASS metric, can be queried using cuptiSassMetricsGetProperties()
+    uint64_t metricId;
+    /// [in] CUpti_SassMetrics_OutputGranularity
+    uint8_t outputGranularity;
+} CUpti_SassMetrics_Config;
+
+/**
+ * \brief Params for cuptiSassMetricsSetConfig
+ */
+typedef struct CUpti_SassMetricsSetConfig_Params
+{
+    /// [in] equal to CUpti_SassMetricsSetConfig_Params_STRUCT_SIZE
+    size_t structSize;
+    /// [in] assign to NULL
+    void* pPriv;
+    /// [in] num of metric configs, will be equal to number of metrics queried
+    size_t numOfMetricConfig;
+    /// [in] list of metric config generated for given sass metrics
+    CUpti_SassMetrics_Config* pConfigs;
+    /// [in] device index for which config will be set, user can call this once for
+    /// the device on which the the SASS metric data will be collected
+    uint32_t deviceIndex;
+} CUpti_SassMetricsSetConfig_Params;
+#define CUpti_SassMetricsSetConfig_Params_STRUCT_SIZE                    CUPTI_PROFILER_STRUCT_SIZE(CUpti_SassMetricsSetConfig_Params, deviceIndex)
+
+/**
+ * \brief Set config for the SASS metric data collection for a device.
+ * User need to call this API before calling any of the SASS metric data collection APIs.
+ * Each set config API call need to be followed by cuptiSassPatchingUnSetConfig API
+ * before calling the cuptiSassMetricsSetConfig() API again for the same device.
+ * 
+ * \param pParams A pointer to \ref CUpti_SassMetricsSetConfig_Params
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_INVALID_CONTEXT if any cuda context has not been created prior to this API call
+ * \retval CUPTI_ERROR_INVALID_OPERATION if this is called multiple times for the device without calling unset config API
+ * \retval CUPTI_ERROR_NOT_SUPPORTED indicates that the system/device doesn't support SASS metric data collection
+ */
+CUptiResult CUPTIAPI cuptiSassMetricsSetConfig(CUpti_SassMetricsSetConfig_Params *pParams);
+
+/**
+ * \brief Params for cuptiSassMetricsUnsetConfig
+ */
+typedef struct CUpti_SassMetricsUnsetConfig_Params
+{
+    /// [in] equal to CUpti_SassMetricsUnsetConfig_Params_STRUCT_SIZE
+    size_t structSize;
+    /// [in] assign to NULL
+    void* pPriv;
+    /// [in] device index for which SASS metric data collection config will get reset, user need to call this API for
+    /// all the devices on which the the SASS metric data collection have been configured.
+    uint32_t deviceIndex;
+} CUpti_SassMetricsUnsetConfig_Params;
+#define CUpti_SassMetricsUnsetConfig_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_SassMetricsUnsetConfig_Params, deviceIndex)
+
+/**
+ * \brief Unset config API will reset the SASS metric data collection configuration for the device.
+ * Once this API called CUPTI will deallocate all the memory allocated and remove all
+ * the configuration for SASS metric data collection. User can only call this API for a device where
+ * cuptiSassMetricsSetConfig() API has been called earlier for the device.
+ * 
+ * \param pParams A pointer to \ref CUpti_SassMetricsSetConfig_Params
+ * 
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_INVALID_CONTEXT if any cuda context has not been created prior to this API call
+ * \retval CUPTI_ERROR_INVALID_OPERATION if this is called multiple times for the device without calling set config API
+ * \retval CUPTI_ERROR_NOT_SUPPORTED indicates that the system/device doesn't support SASS metric data collection
+ */
+CUptiResult CUPTIAPI cuptiSassMetricsUnsetConfig(CUpti_SassMetricsUnsetConfig_Params *pParams);
+
+/**
+ * \brief Params for cuptiSassMetricsEnable
+ */
+typedef struct CUpti_SassMetricsEnable_Params
+{
+    /// [in] equal to CUpti_SassMetricsEnable_Params_STRUCT_SIZE
+    size_t structSize;
+    /// [in] assign to NULL
+    void* pPriv;
+    /// [in] CUDA context on which SASS metric data collection will be enabled.
+    /// If set NULL, default context will be consider for SASS metric data collection.
+    CUcontext ctx;
+    /// [in] if false, all the functions will patched regardless of their execution with cuptiSassMetricsEnable() API call.
+    /// when this parameter is set to true, metric data collection for the function will be done at the very first execution in the enable/disble
+    /// range.
+    uint8_t enableLazyPatching;
+} CUpti_SassMetricsEnable_Params;
+#define CUpti_SassMetricsEnable_Params_STRUCT_SIZE                       CUPTI_PROFILER_STRUCT_SIZE(CUpti_SassMetricsEnable_Params, enableLazyPatching)
+
+/**
+ * \brief Sass metric data collection enable API will mark the start of a range, between which kernel
+ *  will be profiled for SASS metrics.
+ *
+ * \param pParams A pointer to \ref CUpti_SassMetricsEnable_Params
+ * 
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_NOT_SUPPORTED indicates that the system/device doesn't support SASS metric data collection
+ * \retval CUPTI_ERROR_INVALID_CONTEXT if any cuda context has not been created prior to this API call
+ * \retval CUPTI_ERROR_INVALID_OPERATION if this API is called multiple times for a cuda context without calling 
+ * cuptiSassMetricsDisable() API or called before cuptiSassMetricsSetConfig() API call.
+ */
+CUptiResult CUPTIAPI cuptiSassMetricsEnable(CUpti_SassMetricsEnable_Params* pParams);
+
+/**
+ * \brief Params for cuptiSassMetricsDisable
+ */
+typedef struct CUpti_SassMetricsDisable_Params
+{
+    /// [in] equal to CUpti_SassMetricsDisable_Params_STRUCT_SIZE
+    size_t structSize;
+    /// [in] assign to NULL
+    void* pPriv;
+    /// [in] CUDA context on which SASS metric data collection will be disabled.
+    /// If set NULL, default context will be consider for SASS metric data collection.
+    CUcontext ctx;
+    /// [out] Num of dropped SASS records will be equal to numOfPatchedInstructions * numOfInstances.
+    /// Number of dropped records will be zero when data is flushed prior to calling the disable API.
+    size_t numOfDroppedRecords;
+} CUpti_SassMetricsDisable_Params;
+#define CUpti_SassMetricsDisable_Params_STRUCT_SIZE                      CUPTI_PROFILER_STRUCT_SIZE(CUpti_SassMetricsDisable_Params, numOfDroppedRecords)
+
+/**
+ * \brief SASS metric data collection disable API will mark the end of a range, any kernel launched after this
+ * API call will not be profiled for the SASS metrics.
+ *
+ * \param pParams A pointer to \ref CUpti_SassMetricsDisable_Params
+ * 
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_NOT_SUPPORTED indicates that the system/device doesn't support SASS metric data collection
+ * \retval CUPTI_ERROR_INVALID_CONTEXT if any cuda context has not been created prior to this API call
+ * \retval CUPTI_ERROR_INVALID_OPERATION if this API is called multiple times for a cuda context without calling 
+ * cuptiSassMetricsEnable() API or called before cuptiSassMetricsSetConfig() API call.
+ */
+CUptiResult CUPTIAPI cuptiSassMetricsDisable(CUpti_SassMetricsDisable_Params* pParams);
+
+/**
+ * \brief Params for cuptiSassMetricsGetDataProperties
+ */
+typedef struct CUpti_SassMetricsGetDataProperties_Params
+{
+    /// [in] equal to CUpti_SassMetricsGetDataProperties_Params_STRUCT_SIZE
+    size_t structSize;
+    /// [in] assign to NULL
+    void* pPriv;
+    /// [in] CUDA context on which SASS metric data collection was enabled.
+    /// If set NULL, default context will be consider for SASS metric data collection.
+    CUcontext ctx;
+    /// [out] total number of SASS records has been collected
+    size_t numOfPatchedInstructionRecords;
+    /// [out] number of instances for each metric value per instruction.
+    /// This will depend on CUpti_SassPatching_OutputGranularity level set for the metric config.
+    size_t numOfInstances;
+} CUpti_SassMetricsGetDataProperties_Params;
+
+#define CUpti_SassMetricsGetDataProperties_Params_STRUCT_SIZE           CUPTI_PROFILER_STRUCT_SIZE(CUpti_SassMetricsGetDataProperties_Params, numOfInstances)
+/**
+ * \brief SASS metric data properties API will give the data regarding number of instances of a metric
+ * value and number of SASS instruction data has been collected. The number of instances of a metric
+ * will vary as per user set the output granularity level with CUpti_SassMetrics_OutputGranularity value.
+ * User need to allocate memory for retriving the SASS data using cuptiSassMetricsFlushData() API.
+ * 
+ * \param pParams A pointer to \ref CUpti_SassMetricsGetDataProperties_Params
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_NOT_SUPPORTED indicates that the system/device doesn't support SASS metric data collection
+ * \retval CUPTI_ERROR_INVALID_OPERATION if this API is called outside the enable/disable range.
+ */
+CUptiResult CUPTIAPI cuptiSassMetricsGetDataProperties(CUpti_SassMetricsGetDataProperties_Params* pParams);
+
+typedef struct CUpti_SassMetrics_InstanceValue
+{
+    // unique id of the metric
+    uint64_t metricId;
+    // metric value 
+    uint64_t value;
+} CUpti_SassMetrics_InstanceValue;
+#define CUpti_SassMetrics_InstanceValue_STRUCT_SIZE                      CUPTI_PROFILER_STRUCT_SIZE(CUpti_SassMetrics_InstanceValue, value)
+
+typedef struct CUpti_SassMetrics_Data
+{
+    /// [in] equal to CUpti_SassMetricsFlushData_Params_STRUCT_SIZE
+    size_t structSize;
+    /// [in] assign to NULL
+    void* pPriv;
+    /// [out] Unique cubin id
+    uint32_t cubinCrc;
+    /// [out] function's unique symbol index in the module.
+    uint32_t functionIndex;
+    /// [out] The function name
+    const char* functionName;
+    /// [out] pc offset for the function in a module
+    uint32_t pcOffset;
+    /// [out] array of size equal to number of instances per metric, which contains the metric ID and metric value.
+    CUpti_SassMetrics_InstanceValue* pInstanceValues;
+} CUpti_SassMetrics_Data;
+
+/**
+ * \brief Params for cuptiSassMetricsFlushData
+ */
+typedef struct CUpti_SassMetricsFlushData_Params
+{
+    /// [in] equal to CUpti_SassMetricsFlushData_Params_STRUCT_SIZE
+    size_t structSize;
+    /// [in] assign to NULL
+    void* pPriv;
+    /// [in] CUDA context on which SASS metric data collection was enabled.
+    /// If set NULL, default context will be consider for SASS metric data collection.
+    CUcontext ctx;
+    /// [in] number of patched instruction record will be retrived, user can call cuptiSassMetricsGetDataProperties()
+    /// for getting total number of records available.
+    size_t numOfPatchedInstructionRecords;
+    /// [in] number of patched instruction record instances for a metric, user can call cuptiSassMetricsGetDataProperties()
+    /// for getting total number of instances for each record per metric available.
+    size_t numOfInstances;
+    /// [out] 
+    CUpti_SassMetrics_Data* pMetricsData;
+} CUpti_SassMetricsFlushData_Params;
+#define CUpti_SassMetricsFlushData_Params_STRUCT_SIZE                      CUPTI_PROFILER_STRUCT_SIZE(CUpti_SassMetricsFlushData_Params, numOfInstances)
+
+/**
+ * \brief Flush SASS metrics data from CUPTI internal buffer to the user buffer.
+ * User needs to allocate the buffer for retrieving the data. The number of records collected
+ * can be queried using the API cuptiSassMetricsGetDataProperties().
+ *
+ * \param pParams A pointer to \ref CUpti_SassMetricsFlushData_Params
+ *
+ * \retval CUPTI_SUCCESS
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if any \p pParams is not valid
+ * \retval CUPTI_ERROR_NOT_SUPPORTED indicates that the system/device doesn't support SASS metric data collection.
+ * \retval CUPTI_ERROR_INVALID_OPERATION if this API is called outside the enable/disable range.
+ */
+CUptiResult CUPTIAPI cuptiSassMetricsFlushData(CUpti_SassMetricsFlushData_Params* pParams);
+
+/** @} */ /* END CUPTI_SASS_METRICS_API */
+
+#if defined(__GNUC__) && defined(CUPTI_LIB)
+    #pragma GCC visibility pop
+#endif
+
+#ifdef __cplusplus
+} /* extern "C" */
+#endif
+
+#endif // _CUPTI_SASS_METRICS_H_

Prism/LLaDA/LLaDA_Prism/.venv/lib/python3.12/site-packages/nvidia/cuda_cupti/include/cupti_target.h

@@ -0,0 +1,43 @@
+#if !defined(_CUPTI_TARGET_H_)
+#define _CUPTI_TARGET_H_
+
+/*
+CUPTI profiler target API's
+This file contains the CUPTI profiling API's.
+*/
+#include <cupti_result.h>
+#include <stddef.h>
+#include <stdint.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#if defined(__GNUC__) && defined(CUPTI_LIB)
+    #pragma GCC visibility push(default)
+#endif
+
+#ifndef CUPTI_PROFILER_STRUCT_SIZE
+#define CUPTI_PROFILER_STRUCT_SIZE(type_, lastfield_)                     (offsetof(type_, lastfield_) + sizeof(((type_*)0)->lastfield_))
+#endif
+
+typedef struct CUpti_Device_GetChipName_Params
+{
+    size_t structSize;                                      //!< [in]
+    void* pPriv;                                            //!< [in] assign to NULL
+
+    size_t deviceIndex;                                     //!< [in]
+    const char* pChipName;                                  //!< [out]
+} CUpti_Device_GetChipName_Params;
+
+#define CUpti_Device_GetChipName_Params_STRUCT_SIZE                  CUPTI_PROFILER_STRUCT_SIZE(CUpti_Device_GetChipName_Params, pChipName)
+CUptiResult CUPTIAPI cuptiDeviceGetChipName(CUpti_Device_GetChipName_Params *pParams);
+
+#if defined(__GNUC__) && defined(CUPTI_LIB)
+    #pragma GCC visibility pop
+#endif
+
+#ifdef __cplusplus
+} /* extern "C" */
+#endif
+#endif

Prism/LLaDA/LLaDA_Prism/.venv/lib/python3.12/site-packages/nvidia/cuda_cupti/include/cupti_version.h

@@ -0,0 +1,137 @@
+/*
+ * Copyright 2010-2024 NVIDIA Corporation.  All rights reserved.
+ *
+ * NOTICE TO LICENSEE:
+ *
+ * This source code and/or documentation ("Licensed Deliverables") are
+ * subject to NVIDIA intellectual property rights under U.S. and
+ * international Copyright laws.
+ *
+ * These Licensed Deliverables contained herein is PROPRIETARY and
+ * CONFIDENTIAL to NVIDIA and is being provided under the terms and
+ * conditions of a form of NVIDIA software license agreement by and
+ * between NVIDIA and Licensee ("License Agreement") or electronically
+ * accepted by Licensee.  Notwithstanding any terms or conditions to
+ * the contrary in the License Agreement, reproduction or disclosure
+ * of the Licensed Deliverables to any third party without the express
+ * written consent of NVIDIA is prohibited.
+ *
+ * NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE
+ * LICENSE AGREEMENT, NVIDIA MAKES NO REPRESENTATION ABOUT THE
+ * SUITABILITY OF THESE LICENSED DELIVERABLES FOR ANY PURPOSE.  IT IS
+ * PROVIDED "AS IS" WITHOUT EXPRESS OR IMPLIED WARRANTY OF ANY KIND.
+ * NVIDIA DISCLAIMS ALL WARRANTIES WITH REGARD TO THESE LICENSED
+ * DELIVERABLES, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY,
+ * NONINFRINGEMENT, AND FITNESS FOR A PARTICULAR PURPOSE.
+ * NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE
+ * LICENSE AGREEMENT, IN NO EVENT SHALL NVIDIA BE LIABLE FOR ANY
+ * SPECIAL, INDIRECT, INCIDENTAL, OR CONSEQUENTIAL DAMAGES, OR ANY
+ * DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS,
+ * WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS
+ * ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE
+ * OF THESE LICENSED DELIVERABLES.
+ *
+ * U.S. Government End Users.  These Licensed Deliverables are a
+ * "commercial item" as that term is defined at 48 C.F.R. 2.101 (OCT
+ * 1995), consisting of "commercial computer software" and "commercial
+ * computer software documentation" as such terms are used in 48
+ * C.F.R. 12.212 (SEPT 1995) and is provided to the U.S. Government
+ * only as a commercial end item.  Consistent with 48 C.F.R.12.212 and
+ * 48 C.F.R. 227.7202-1 through 227.7202-4 (JUNE 1995), all
+ * U.S. Government End Users acquire the Licensed Deliverables with
+ * only those rights set forth herein.
+ *
+ * Any use of the Licensed Deliverables in individual and commercial
+ * software must include, in the user documentation and internal
+ * comments to the code, the above Disclaimer and U.S. Government End
+ * Users Notice.
+ */
+
+#if !defined(_CUPTI_VERSION_H_)
+#define _CUPTI_VERSION_H_
+
+#include <cuda_stdint.h>
+#include <cupti_result.h>
+
+#ifndef CUPTIAPI
+#ifdef _WIN32
+#define CUPTIAPI __stdcall
+#else
+#define CUPTIAPI
+#endif
+#endif
+
+#if defined(__cplusplus)
+extern "C" {
+#endif
+
+#if defined(__GNUC__) && defined(CUPTI_LIB)
+    #pragma GCC visibility push(default)
+#endif
+
+/**
+ * \defgroup CUPTI_VERSION_API CUPTI Version
+ * Function and macro to determine the CUPTI version.
+ * @{
+ */
+
+/**
+ * \brief The API version for this implementation of CUPTI.
+ *
+ * The API version for this implementation of CUPTI. This define along
+ * with \ref cuptiGetVersion can be used to dynamically detect if the
+ * version of CUPTI compiled against matches the version of the loaded
+ * CUPTI library.
+ *
+ * v1 : CUDAToolsSDK 4.0
+ * v2 : CUDAToolsSDK 4.1
+ * v3 : CUDA Toolkit 5.0
+ * v4 : CUDA Toolkit 5.5
+ * v5 : CUDA Toolkit 6.0
+ * v6 : CUDA Toolkit 6.5
+ * v7 : CUDA Toolkit 6.5(with sm_52 support)
+ * v8 : CUDA Toolkit 7.0
+ * v9 : CUDA Toolkit 8.0
+ * v10 : CUDA Toolkit 9.0
+ * v11 : CUDA Toolkit 9.1
+ * v12 : CUDA Toolkit 10.0, 10.1 and 10.2
+ * v13 : CUDA Toolkit 11.0
+ * v14 : CUDA Toolkit 11.1
+ * v15 : CUDA Toolkit 11.2, 11.3 and 11.4
+ * v16 : CUDA Toolkit 11.5
+ * v17 : CUDA Toolkit 11.6
+ * v18 : CUDA Toolkit 11.8
+ * v19 : CUDA Toolkit 12.0
+ * v20 : CUDA Toolkit 12.2
+ * v21 : CUDA Toolkit 12.3
+ * v22 : CUDA Toolkit 12.4
+ * v23 : CUDA Toolkit 12.5
+ * v24 : CUDA Toolkit 12.6
+ * v26 : CUDA Toolkit 12.8
+ */
+#define CUPTI_API_VERSION 26
+
+/**
+ * \brief Get the CUPTI API version.
+ *
+ * Return the API version in \p *version.
+ *
+ * \param version Returns the version
+ *
+ * \retval CUPTI_SUCCESS on success
+ * \retval CUPTI_ERROR_INVALID_PARAMETER if \p version is NULL
+ * \sa CUPTI_API_VERSION
+ */
+CUptiResult CUPTIAPI cuptiGetVersion(uint32_t *version);
+
+/** @} */ /* END CUPTI_VERSION_API */
+
+#if defined(__GNUC__) && defined(CUPTI_LIB)
+    #pragma GCC visibility pop
+#endif
+
+#if defined(__cplusplus)
+}
+#endif
+
+#endif /*_CUPTI_VERSION_H_*/

Prism/LLaDA/LLaDA_Prism/.venv/lib/python3.12/site-packages/nvidia/cuda_cupti/include/generated_cudaGL_meta.h

@@ -0,0 +1,116 @@
+// This file is generated.  Any changes you make will be lost during the next clean build.
+
+// Dependent includes
+#ifdef __APPLE__
+#include <OpenGL/gl.h>
+#else
+#include <GL/gl.h>
+#endif
+
+// CUDA public interface, for type definitions and cu* function prototypes
+#include "cudaGL.h"
+
+
+// *************************************************************************
+//      Definitions of structs to hold parameters for each function
+// *************************************************************************
+
+typedef struct cuGraphicsGLRegisterBuffer_params_st {
+    CUgraphicsResource *pCudaResource;
+    GLuint buffer;
+    unsigned int Flags;
+} cuGraphicsGLRegisterBuffer_params;
+
+typedef struct cuGraphicsGLRegisterImage_params_st {
+    CUgraphicsResource *pCudaResource;
+    GLuint image;
+    GLenum target;
+    unsigned int Flags;
+} cuGraphicsGLRegisterImage_params;
+
+typedef struct cuGLGetDevices_v2_params_st {
+    unsigned int *pCudaDeviceCount;
+    CUdevice *pCudaDevices;
+    unsigned int cudaDeviceCount;
+    CUGLDeviceList deviceList;
+} cuGLGetDevices_v2_params;
+
+typedef struct cuGLCtxCreate_v2_params_st {
+    CUcontext *pCtx;
+    unsigned int Flags;
+    CUdevice device;
+} cuGLCtxCreate_v2_params;
+
+typedef struct cuGLRegisterBufferObject_params_st {
+    GLuint buffer;
+} cuGLRegisterBufferObject_params;
+
+typedef struct cuGLMapBufferObject_v2_ptds_params_st {
+    CUdeviceptr *dptr;
+    size_t *size;
+    GLuint buffer;
+} cuGLMapBufferObject_v2_ptds_params;
+
+typedef struct cuGLUnmapBufferObject_params_st {
+    GLuint buffer;
+} cuGLUnmapBufferObject_params;
+
+typedef struct cuGLUnregisterBufferObject_params_st {
+    GLuint buffer;
+} cuGLUnregisterBufferObject_params;
+
+typedef struct cuGLSetBufferObjectMapFlags_params_st {
+    GLuint buffer;
+    unsigned int Flags;
+} cuGLSetBufferObjectMapFlags_params;
+
+typedef struct cuGLMapBufferObjectAsync_v2_ptsz_params_st {
+    CUdeviceptr *dptr;
+    size_t *size;
+    GLuint buffer;
+    CUstream hStream;
+} cuGLMapBufferObjectAsync_v2_ptsz_params;
+
+typedef struct cuGLUnmapBufferObjectAsync_params_st {
+    GLuint buffer;
+    CUstream hStream;
+} cuGLUnmapBufferObjectAsync_params;
+
+typedef struct cuGLGetDevices_params_st {
+    unsigned int *pCudaDeviceCount;
+    CUdevice *pCudaDevices;
+    unsigned int cudaDeviceCount;
+    CUGLDeviceList deviceList;
+} cuGLGetDevices_params;
+
+typedef struct cuGLMapBufferObject_v2_params_st {
+    CUdeviceptr *dptr;
+    size_t *size;
+    GLuint buffer;
+} cuGLMapBufferObject_v2_params;
+
+typedef struct cuGLMapBufferObjectAsync_v2_params_st {
+    CUdeviceptr *dptr;
+    size_t *size;
+    GLuint buffer;
+    CUstream hStream;
+} cuGLMapBufferObjectAsync_v2_params;
+
+typedef struct cuGLCtxCreate_params_st {
+    CUcontext *pCtx;
+    unsigned int Flags;
+    CUdevice device;
+} cuGLCtxCreate_params;
+
+typedef struct cuGLMapBufferObject_params_st {
+    CUdeviceptr_v1 *dptr;
+    unsigned int *size;
+    GLuint buffer;
+} cuGLMapBufferObject_params;
+
+typedef struct cuGLMapBufferObjectAsync_params_st {
+    CUdeviceptr_v1 *dptr;
+    unsigned int *size;
+    GLuint buffer;
+    CUstream hStream;
+} cuGLMapBufferObjectAsync_params;

Prism/LLaDA/LLaDA_Prism/.venv/lib/python3.12/site-packages/nvidia/cuda_cupti/include/generated_cudaVDPAU_meta.h

@@ -0,0 +1,46 @@
+// This file is generated.  Any changes you make will be lost during the next clean build.
+
+// Dependent includes
+#include <vdpau/vdpau.h>
+
+// CUDA public interface, for type definitions and cu* function prototypes
+#include "cudaVDPAU.h"
+
+
+// *************************************************************************
+//      Definitions of structs to hold parameters for each function
+// *************************************************************************
+
+typedef struct cuVDPAUGetDevice_params_st {
+    CUdevice *pDevice;
+    VdpDevice vdpDevice;
+    VdpGetProcAddress *vdpGetProcAddress;
+} cuVDPAUGetDevice_params;
+
+typedef struct cuVDPAUCtxCreate_v2_params_st {
+    CUcontext *pCtx;
+    unsigned int flags;
+    CUdevice device;
+    VdpDevice vdpDevice;
+    VdpGetProcAddress *vdpGetProcAddress;
+} cuVDPAUCtxCreate_v2_params;
+
+typedef struct cuGraphicsVDPAURegisterVideoSurface_params_st {
+    CUgraphicsResource *pCudaResource;
+    VdpVideoSurface vdpSurface;
+    unsigned int flags;
+} cuGraphicsVDPAURegisterVideoSurface_params;
+
+typedef struct cuGraphicsVDPAURegisterOutputSurface_params_st {
+    CUgraphicsResource *pCudaResource;
+    VdpOutputSurface vdpSurface;
+    unsigned int flags;
+} cuGraphicsVDPAURegisterOutputSurface_params;
+
+typedef struct cuVDPAUCtxCreate_params_st {
+    CUcontext *pCtx;
+    unsigned int flags;
+    CUdevice device;
+    VdpDevice vdpDevice;
+    VdpGetProcAddress *vdpGetProcAddress;
+} cuVDPAUCtxCreate_params;

Prism/LLaDA/LLaDA_Prism/.venv/lib/python3.12/site-packages/nvidia/cuda_cupti/include/generated_cuda_gl_interop_meta.h

@@ -0,0 +1,71 @@
+// This file is generated.  Any changes you make will be lost during the next clean build.
+
+// CUDA public interface, for type definitions and api function prototypes
+#include "cuda_gl_interop.h"
+
+// *************************************************************************
+//      Definitions of structs to hold parameters for each function
+// *************************************************************************
+
+// Currently used parameter trace structures
+typedef struct cudaGLGetDevices_v4010_params_st {
+    unsigned int *pCudaDeviceCount;
+    int *pCudaDevices;
+    unsigned int cudaDeviceCount;
+    enum cudaGLDeviceList deviceList;
+} cudaGLGetDevices_v4010_params;
+
+typedef struct cudaGraphicsGLRegisterImage_v3020_params_st {
+    struct cudaGraphicsResource **resource;
+    GLuint image;
+    GLenum target;
+    unsigned int flags;
+} cudaGraphicsGLRegisterImage_v3020_params;
+
+typedef struct cudaGraphicsGLRegisterBuffer_v3020_params_st {
+    struct cudaGraphicsResource **resource;
+    GLuint buffer;
+    unsigned int flags;
+} cudaGraphicsGLRegisterBuffer_v3020_params;
+
+typedef struct cudaGLSetGLDevice_v3020_params_st {
+    int device;
+} cudaGLSetGLDevice_v3020_params;
+
+typedef struct cudaGLRegisterBufferObject_v3020_params_st {
+    GLuint bufObj;
+} cudaGLRegisterBufferObject_v3020_params;
+
+typedef struct cudaGLMapBufferObject_v3020_params_st {
+    void **devPtr;
+    GLuint bufObj;
+} cudaGLMapBufferObject_v3020_params;
+
+typedef struct cudaGLUnmapBufferObject_v3020_params_st {
+    GLuint bufObj;
+} cudaGLUnmapBufferObject_v3020_params;
+
+typedef struct cudaGLUnregisterBufferObject_v3020_params_st {
+    GLuint bufObj;
+} cudaGLUnregisterBufferObject_v3020_params;
+
+typedef struct cudaGLSetBufferObjectMapFlags_v3020_params_st {
+    GLuint bufObj;
+    unsigned int flags;
+} cudaGLSetBufferObjectMapFlags_v3020_params;
+
+typedef struct cudaGLMapBufferObjectAsync_v3020_params_st {
+    void **devPtr;
+    GLuint bufObj;
+    cudaStream_t stream;
+} cudaGLMapBufferObjectAsync_v3020_params;
+
+typedef struct cudaGLUnmapBufferObjectAsync_v3020_params_st {
+    GLuint bufObj;
+    cudaStream_t stream;
+} cudaGLUnmapBufferObjectAsync_v3020_params;
+
+// Parameter trace structures for removed functions
+
+
+// End of parameter trace structures

Prism/LLaDA/LLaDA_Prism/.venv/lib/python3.12/site-packages/nvidia/cuda_cupti/include/generated_cuda_runtime_api_meta.h

@@ -0,0 +1,2372 @@
+// This file is generated.  Any changes you make will be lost during the next clean build.
+
+// CUDA public interface, for type definitions and api function prototypes
+#include "cuda_runtime_api.h"
+
+// *************************************************************************
+//      Definitions of structs to hold parameters for each function
+// *************************************************************************
+
+// Currently used parameter trace structures
+typedef struct cudaDeviceSetLimit_v3020_params_st {
+    enum cudaLimit limit;
+    size_t value;
+} cudaDeviceSetLimit_v3020_params;
+
+typedef struct cudaDeviceGetLimit_v3020_params_st {
+    size_t *pValue;
+    enum cudaLimit limit;
+} cudaDeviceGetLimit_v3020_params;
+
+typedef struct cudaDeviceGetTexture1DLinearMaxWidth_v11010_params_st {
+    size_t *maxWidthInElements;
+    const struct cudaChannelFormatDesc *fmtDesc;
+    int device;
+} cudaDeviceGetTexture1DLinearMaxWidth_v11010_params;
+
+typedef struct cudaDeviceGetCacheConfig_v3020_params_st {
+    enum cudaFuncCache *pCacheConfig;
+} cudaDeviceGetCacheConfig_v3020_params;
+
+typedef struct cudaDeviceGetStreamPriorityRange_v5050_params_st {
+    int *leastPriority;
+    int *greatestPriority;
+} cudaDeviceGetStreamPriorityRange_v5050_params;
+
+typedef struct cudaDeviceSetCacheConfig_v3020_params_st {
+    enum cudaFuncCache cacheConfig;
+} cudaDeviceSetCacheConfig_v3020_params;
+
+typedef struct cudaDeviceGetByPCIBusId_v4010_params_st {
+    int *device;
+    const char *pciBusId;
+} cudaDeviceGetByPCIBusId_v4010_params;
+
+typedef struct cudaDeviceGetPCIBusId_v4010_params_st {
+    char *pciBusId;
+    int len;
+    int device;
+} cudaDeviceGetPCIBusId_v4010_params;
+
+typedef struct cudaIpcGetEventHandle_v4010_params_st {
+    cudaIpcEventHandle_t *handle;
+    cudaEvent_t event;
+} cudaIpcGetEventHandle_v4010_params;
+
+typedef struct cudaIpcOpenEventHandle_v4010_params_st {
+    cudaEvent_t *event;
+    cudaIpcEventHandle_t handle;
+} cudaIpcOpenEventHandle_v4010_params;
+
+typedef struct cudaIpcGetMemHandle_v4010_params_st {
+    cudaIpcMemHandle_t *handle;
+    void *devPtr;
+} cudaIpcGetMemHandle_v4010_params;
+
+typedef struct cudaIpcOpenMemHandle_v4010_params_st {
+    void **devPtr;
+    cudaIpcMemHandle_t handle;
+    unsigned int flags;
+} cudaIpcOpenMemHandle_v4010_params;
+
+typedef struct cudaIpcCloseMemHandle_v4010_params_st {
+    void *devPtr;
+} cudaIpcCloseMemHandle_v4010_params;
+
+typedef struct cudaDeviceFlushGPUDirectRDMAWrites_v11030_params_st {
+    enum cudaFlushGPUDirectRDMAWritesTarget target;
+    enum cudaFlushGPUDirectRDMAWritesScope scope;
+} cudaDeviceFlushGPUDirectRDMAWrites_v11030_params;
+
+typedef struct cudaDeviceRegisterAsyncNotification_v12040_params_st {
+    int device;
+    cudaAsyncCallback callbackFunc;
+    void *userData;
+    cudaAsyncCallbackHandle_t *callback;
+} cudaDeviceRegisterAsyncNotification_v12040_params;
+
+typedef struct cudaDeviceUnregisterAsyncNotification_v12040_params_st {
+    int device;
+    cudaAsyncCallbackHandle_t callback;
+} cudaDeviceUnregisterAsyncNotification_v12040_params;
+
+typedef struct cudaDeviceGetSharedMemConfig_v4020_params_st {
+    enum cudaSharedMemConfig *pConfig;
+} cudaDeviceGetSharedMemConfig_v4020_params;
+
+typedef struct cudaDeviceSetSharedMemConfig_v4020_params_st {
+    enum cudaSharedMemConfig config;
+} cudaDeviceSetSharedMemConfig_v4020_params;
+
+typedef struct cudaGetErrorName_v6050_params_st {
+    cudaError_t error;
+} cudaGetErrorName_v6050_params;
+
+typedef struct cudaGetErrorString_v3020_params_st {
+    cudaError_t error;
+} cudaGetErrorString_v3020_params;
+
+typedef struct cudaGetDeviceCount_v3020_params_st {
+    int *count;
+} cudaGetDeviceCount_v3020_params;
+
+typedef struct cudaGetDeviceProperties_v2_v12000_params_st {
+    struct cudaDeviceProp *prop;
+    int device;
+} cudaGetDeviceProperties_v2_v12000_params;
+
+typedef struct cudaDeviceGetAttribute_v5000_params_st {
+    int *value;
+    enum cudaDeviceAttr attr;
+    int device;
+} cudaDeviceGetAttribute_v5000_params;
+
+typedef struct cudaDeviceGetDefaultMemPool_v11020_params_st {
+    cudaMemPool_t *memPool;
+    int device;
+} cudaDeviceGetDefaultMemPool_v11020_params;
+
+typedef struct cudaDeviceSetMemPool_v11020_params_st {
+    int device;
+    cudaMemPool_t memPool;
+} cudaDeviceSetMemPool_v11020_params;
+
+typedef struct cudaDeviceGetMemPool_v11020_params_st {
+    cudaMemPool_t *memPool;
+    int device;
+} cudaDeviceGetMemPool_v11020_params;
+
+typedef struct cudaDeviceGetNvSciSyncAttributes_v10020_params_st {
+    void *nvSciSyncAttrList;
+    int device;
+    int flags;
+} cudaDeviceGetNvSciSyncAttributes_v10020_params;
+
+typedef struct cudaDeviceGetP2PAttribute_v8000_params_st {
+    int *value;
+    enum cudaDeviceP2PAttr attr;
+    int srcDevice;
+    int dstDevice;
+} cudaDeviceGetP2PAttribute_v8000_params;
+
+typedef struct cudaChooseDevice_v3020_params_st {
+    int *device;
+    const struct cudaDeviceProp *prop;
+} cudaChooseDevice_v3020_params;
+
+typedef struct cudaInitDevice_v12000_params_st {
+    int device;
+    unsigned int deviceFlags;
+    unsigned int flags;
+} cudaInitDevice_v12000_params;
+
+typedef struct cudaSetDevice_v3020_params_st {
+    int device;
+} cudaSetDevice_v3020_params;
+
+typedef struct cudaGetDevice_v3020_params_st {
+    int *device;
+} cudaGetDevice_v3020_params;
+
+typedef struct cudaSetValidDevices_v3020_params_st {
+    int *device_arr;
+    int len;
+} cudaSetValidDevices_v3020_params;
+
+typedef struct cudaSetDeviceFlags_v3020_params_st {
+    unsigned int flags;
+} cudaSetDeviceFlags_v3020_params;
+
+typedef struct cudaGetDeviceFlags_v7000_params_st {
+    unsigned int *flags;
+} cudaGetDeviceFlags_v7000_params;
+
+typedef struct cudaStreamCreate_v3020_params_st {
+    cudaStream_t *pStream;
+} cudaStreamCreate_v3020_params;
+
+typedef struct cudaStreamCreateWithFlags_v5000_params_st {
+    cudaStream_t *pStream;
+    unsigned int flags;
+} cudaStreamCreateWithFlags_v5000_params;
+
+typedef struct cudaStreamCreateWithPriority_v5050_params_st {
+    cudaStream_t *pStream;
+    unsigned int flags;
+    int priority;
+} cudaStreamCreateWithPriority_v5050_params;
+
+typedef struct cudaStreamGetPriority_ptsz_v7000_params_st {
+    cudaStream_t hStream;
+    int *priority;
+} cudaStreamGetPriority_ptsz_v7000_params;
+
+typedef struct cudaStreamGetFlags_ptsz_v7000_params_st {
+    cudaStream_t hStream;
+    unsigned int *flags;
+} cudaStreamGetFlags_ptsz_v7000_params;
+
+typedef struct cudaStreamGetId_ptsz_v12000_params_st {
+    cudaStream_t hStream;
+    unsigned long long *streamId;
+} cudaStreamGetId_ptsz_v12000_params;
+
+typedef struct cudaStreamGetDevice_ptsz_v12080_params_st {
+    cudaStream_t hStream;
+    int *device;
+} cudaStreamGetDevice_ptsz_v12080_params;
+
+typedef struct cudaStreamCopyAttributes_ptsz_v11000_params_st {
+    cudaStream_t dst;
+    cudaStream_t src;
+} cudaStreamCopyAttributes_ptsz_v11000_params;
+
+typedef struct cudaStreamGetAttribute_ptsz_v11000_params_st {
+    cudaStream_t hStream;
+    cudaStreamAttrID attr;
+    cudaStreamAttrValue *value_out;
+} cudaStreamGetAttribute_ptsz_v11000_params;
+
+typedef struct cudaStreamSetAttribute_ptsz_v11000_params_st {
+    cudaStream_t hStream;
+    cudaStreamAttrID attr;
+    const cudaStreamAttrValue *value;
+} cudaStreamSetAttribute_ptsz_v11000_params;
+
+typedef struct cudaStreamDestroy_v5050_params_st {
+    cudaStream_t stream;
+} cudaStreamDestroy_v5050_params;
+
+typedef struct cudaStreamWaitEvent_ptsz_v7000_params_st {
+    cudaStream_t stream;
+    cudaEvent_t event;
+    unsigned int flags;
+} cudaStreamWaitEvent_ptsz_v7000_params;
+
+typedef struct cudaStreamAddCallback_ptsz_v7000_params_st {
+    cudaStream_t stream;
+    cudaStreamCallback_t callback;
+    void *userData;
+    unsigned int flags;
+} cudaStreamAddCallback_ptsz_v7000_params;
+
+typedef struct cudaStreamSynchronize_ptsz_v7000_params_st {
+    cudaStream_t stream;
+} cudaStreamSynchronize_ptsz_v7000_params;
+
+typedef struct cudaStreamQuery_ptsz_v7000_params_st {
+    cudaStream_t stream;
+} cudaStreamQuery_ptsz_v7000_params;
+
+typedef struct cudaStreamAttachMemAsync_ptsz_v7000_params_st {
+    cudaStream_t stream;
+    void *devPtr;
+    size_t length;
+    unsigned int flags;
+} cudaStreamAttachMemAsync_ptsz_v7000_params;
+
+typedef struct cudaStreamBeginCapture_ptsz_v10000_params_st {
+    cudaStream_t stream;
+    enum cudaStreamCaptureMode mode;
+} cudaStreamBeginCapture_ptsz_v10000_params;
+
+typedef struct cudaStreamBeginCaptureToGraph_ptsz_v12030_params_st {
+    cudaStream_t stream;
+    cudaGraph_t graph;
+    const cudaGraphNode_t *dependencies;
+    const cudaGraphEdgeData *dependencyData;
+    size_t numDependencies;
+    enum cudaStreamCaptureMode mode;
+} cudaStreamBeginCaptureToGraph_ptsz_v12030_params;
+
+typedef struct cudaThreadExchangeStreamCaptureMode_v10010_params_st {
+    enum cudaStreamCaptureMode *mode;
+} cudaThreadExchangeStreamCaptureMode_v10010_params;
+
+typedef struct cudaStreamEndCapture_ptsz_v10000_params_st {
+    cudaStream_t stream;
+    cudaGraph_t *pGraph;
+} cudaStreamEndCapture_ptsz_v10000_params;
+
+typedef struct cudaStreamIsCapturing_ptsz_v10000_params_st {
+    cudaStream_t stream;
+    enum cudaStreamCaptureStatus *pCaptureStatus;
+} cudaStreamIsCapturing_ptsz_v10000_params;
+
+typedef struct cudaStreamGetCaptureInfo_v2_ptsz_v11030_params_st {
+    cudaStream_t stream;
+    enum cudaStreamCaptureStatus *captureStatus_out;
+    unsigned long long *id_out;
+    cudaGraph_t *graph_out;
+    const cudaGraphNode_t **dependencies_out;
+    size_t *numDependencies_out;
+} cudaStreamGetCaptureInfo_v2_ptsz_v11030_params;
+
+typedef struct cudaStreamGetCaptureInfo_v3_ptsz_v12030_params_st {
+    cudaStream_t stream;
+    enum cudaStreamCaptureStatus *captureStatus_out;
+    unsigned long long *id_out;
+    cudaGraph_t *graph_out;
+    const cudaGraphNode_t **dependencies_out;
+    const cudaGraphEdgeData **edgeData_out;
+    size_t *numDependencies_out;
+} cudaStreamGetCaptureInfo_v3_ptsz_v12030_params;
+
+typedef struct cudaStreamUpdateCaptureDependencies_ptsz_v11030_params_st {
+    cudaStream_t stream;
+    cudaGraphNode_t *dependencies;
+    size_t numDependencies;
+    unsigned int flags;
+} cudaStreamUpdateCaptureDependencies_ptsz_v11030_params;
+
+typedef struct cudaStreamUpdateCaptureDependencies_v2_ptsz_v12030_params_st {
+    cudaStream_t stream;
+    cudaGraphNode_t *dependencies;
+    const cudaGraphEdgeData *dependencyData;
+    size_t numDependencies;
+    unsigned int flags;
+} cudaStreamUpdateCaptureDependencies_v2_ptsz_v12030_params;
+
+typedef struct cudaEventCreate_v3020_params_st {
+    cudaEvent_t *event;
+} cudaEventCreate_v3020_params;
+
+typedef struct cudaEventCreateWithFlags_v3020_params_st {
+    cudaEvent_t *event;
+    unsigned int flags;
+} cudaEventCreateWithFlags_v3020_params;
+
+typedef struct cudaEventRecord_ptsz_v7000_params_st {
+    cudaEvent_t event;
+    cudaStream_t stream;
+} cudaEventRecord_ptsz_v7000_params;
+
+typedef struct cudaEventRecordWithFlags_ptsz_v11010_params_st {
+    cudaEvent_t event;
+    cudaStream_t stream;
+    unsigned int flags;
+} cudaEventRecordWithFlags_ptsz_v11010_params;
+
+typedef struct cudaEventQuery_v3020_params_st {
+    cudaEvent_t event;
+} cudaEventQuery_v3020_params;
+
+typedef struct cudaEventSynchronize_v3020_params_st {
+    cudaEvent_t event;
+} cudaEventSynchronize_v3020_params;
+
+typedef struct cudaEventDestroy_v3020_params_st {
+    cudaEvent_t event;
+} cudaEventDestroy_v3020_params;
+
+typedef struct cudaEventElapsedTime_v3020_params_st {
+    float *ms;
+    cudaEvent_t start;
+    cudaEvent_t end;
+} cudaEventElapsedTime_v3020_params;
+
+typedef struct cudaEventElapsedTime_v2_v12080_params_st {
+    float *ms;
+    cudaEvent_t start;
+    cudaEvent_t end;
+} cudaEventElapsedTime_v2_v12080_params;
+
+typedef struct cudaImportExternalMemory_v10000_params_st {
+    cudaExternalMemory_t *extMem_out;
+    const struct cudaExternalMemoryHandleDesc *memHandleDesc;
+} cudaImportExternalMemory_v10000_params;
+
+typedef struct cudaExternalMemoryGetMappedBuffer_v10000_params_st {
+    void **devPtr;
+    cudaExternalMemory_t extMem;
+    const struct cudaExternalMemoryBufferDesc *bufferDesc;
+} cudaExternalMemoryGetMappedBuffer_v10000_params;
+
+typedef struct cudaExternalMemoryGetMappedMipmappedArray_v10000_params_st {
+    cudaMipmappedArray_t *mipmap;
+    cudaExternalMemory_t extMem;
+    const struct cudaExternalMemoryMipmappedArrayDesc *mipmapDesc;
+} cudaExternalMemoryGetMappedMipmappedArray_v10000_params;
+
+typedef struct cudaDestroyExternalMemory_v10000_params_st {
+    cudaExternalMemory_t extMem;
+} cudaDestroyExternalMemory_v10000_params;
+
+typedef struct cudaImportExternalSemaphore_v10000_params_st {
+    cudaExternalSemaphore_t *extSem_out;
+    const struct cudaExternalSemaphoreHandleDesc *semHandleDesc;
+} cudaImportExternalSemaphore_v10000_params;
+
+typedef struct cudaSignalExternalSemaphoresAsync_v2_ptsz_v11020_params_st {
+    const cudaExternalSemaphore_t *extSemArray;
+    const struct cudaExternalSemaphoreSignalParams *paramsArray;
+    unsigned int numExtSems;
+    cudaStream_t stream;
+} cudaSignalExternalSemaphoresAsync_v2_ptsz_v11020_params;
+
+typedef struct cudaWaitExternalSemaphoresAsync_v2_ptsz_v11020_params_st {
+    const cudaExternalSemaphore_t *extSemArray;
+    const struct cudaExternalSemaphoreWaitParams *paramsArray;
+    unsigned int numExtSems;
+    cudaStream_t stream;
+} cudaWaitExternalSemaphoresAsync_v2_ptsz_v11020_params;
+
+typedef struct cudaDestroyExternalSemaphore_v10000_params_st {
+    cudaExternalSemaphore_t extSem;
+} cudaDestroyExternalSemaphore_v10000_params;
+
+typedef struct cudaLaunchKernel_ptsz_v7000_params_st {
+    const void *func;
+    dim3 gridDim;
+    dim3 blockDim;
+    void **args;
+    size_t sharedMem;
+    cudaStream_t stream;
+} cudaLaunchKernel_ptsz_v7000_params;
+
+typedef struct cudaLaunchKernelExC_ptsz_v11060_params_st {
+    const cudaLaunchConfig_t *config;
+    const void *func;
+    void **args;
+} cudaLaunchKernelExC_ptsz_v11060_params;
+
+typedef struct cudaLaunchCooperativeKernel_ptsz_v9000_params_st {
+    const void *func;
+    dim3 gridDim;
+    dim3 blockDim;
+    void **args;
+    size_t sharedMem;
+    cudaStream_t stream;
+} cudaLaunchCooperativeKernel_ptsz_v9000_params;
+
+typedef struct cudaLaunchCooperativeKernelMultiDevice_v9000_params_st {
+    struct cudaLaunchParams *launchParamsList;
+    unsigned int numDevices;
+    unsigned int flags;
+} cudaLaunchCooperativeKernelMultiDevice_v9000_params;
+
+typedef struct cudaFuncSetCacheConfig_v3020_params_st {
+    const void *func;
+    enum cudaFuncCache cacheConfig;
+} cudaFuncSetCacheConfig_v3020_params;
+
+typedef struct cudaFuncGetAttributes_v3020_params_st {
+    struct cudaFuncAttributes *attr;
+    const void *func;
+} cudaFuncGetAttributes_v3020_params;
+
+typedef struct cudaFuncSetAttribute_v9000_params_st {
+    const void *func;
+    enum cudaFuncAttribute attr;
+    int value;
+} cudaFuncSetAttribute_v9000_params;
+
+typedef struct cudaFuncGetName_v12030_params_st {
+    const char **name;
+    const void *func;
+} cudaFuncGetName_v12030_params;
+
+typedef struct cudaFuncGetParamInfo_v12040_params_st {
+    const void *func;
+    size_t paramIndex;
+    size_t *paramOffset;
+    size_t *paramSize;
+} cudaFuncGetParamInfo_v12040_params;
+
+typedef struct cudaLaunchHostFunc_ptsz_v10000_params_st {
+    cudaStream_t stream;
+    cudaHostFn_t fn;
+    void *userData;
+} cudaLaunchHostFunc_ptsz_v10000_params;
+
+typedef struct cudaFuncSetSharedMemConfig_v4020_params_st {
+    const void *func;
+    enum cudaSharedMemConfig config;
+} cudaFuncSetSharedMemConfig_v4020_params;
+
+typedef struct cudaOccupancyMaxActiveBlocksPerMultiprocessor_v6050_params_st {
+    int *numBlocks;
+    const void *func;
+    int blockSize;
+    size_t dynamicSMemSize;
+} cudaOccupancyMaxActiveBlocksPerMultiprocessor_v6050_params;
+
+typedef struct cudaOccupancyAvailableDynamicSMemPerBlock_v10200_params_st {
+    size_t *dynamicSmemSize;
+    const void *func;
+    int numBlocks;
+    int blockSize;
+} cudaOccupancyAvailableDynamicSMemPerBlock_v10200_params;
+
+typedef struct cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags_v7000_params_st {
+    int *numBlocks;
+    const void *func;
+    int blockSize;
+    size_t dynamicSMemSize;
+    unsigned int flags;
+} cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags_v7000_params;
+
+typedef struct cudaOccupancyMaxPotentialClusterSize_v11070_params_st {
+    int *clusterSize;
+    const void *func;
+    const cudaLaunchConfig_t *launchConfig;
+} cudaOccupancyMaxPotentialClusterSize_v11070_params;
+
+typedef struct cudaOccupancyMaxActiveClusters_v11070_params_st {
+    int *numClusters;
+    const void *func;
+    const cudaLaunchConfig_t *launchConfig;
+} cudaOccupancyMaxActiveClusters_v11070_params;
+
+typedef struct cudaMallocManaged_v6000_params_st {
+    void **devPtr;
+    size_t size;
+    unsigned int flags;
+} cudaMallocManaged_v6000_params;
+
+typedef struct cudaMalloc_v3020_params_st {
+    void **devPtr;
+    size_t size;
+} cudaMalloc_v3020_params;
+
+typedef struct cudaMallocHost_v3020_params_st {
+    void **ptr;
+    size_t size;
+} cudaMallocHost_v3020_params;
+
+typedef struct cudaMallocPitch_v3020_params_st {
+    void **devPtr;
+    size_t *pitch;
+    size_t width;
+    size_t height;
+} cudaMallocPitch_v3020_params;
+
+typedef struct cudaMallocArray_v3020_params_st {
+    cudaArray_t *array;
+    const struct cudaChannelFormatDesc *desc;
+    size_t width;
+    size_t height;
+    unsigned int flags;
+} cudaMallocArray_v3020_params;
+
+typedef struct cudaFree_v3020_params_st {
+    void *devPtr;
+} cudaFree_v3020_params;
+
+typedef struct cudaFreeHost_v3020_params_st {
+    void *ptr;
+} cudaFreeHost_v3020_params;
+
+typedef struct cudaFreeArray_v3020_params_st {
+    cudaArray_t array;
+} cudaFreeArray_v3020_params;
+
+typedef struct cudaFreeMipmappedArray_v5000_params_st {
+    cudaMipmappedArray_t mipmappedArray;
+} cudaFreeMipmappedArray_v5000_params;
+
+typedef struct cudaHostAlloc_v3020_params_st {
+    void **pHost;
+    size_t size;
+    unsigned int flags;
+} cudaHostAlloc_v3020_params;
+
+typedef struct cudaHostRegister_v4000_params_st {
+    void *ptr;
+    size_t size;
+    unsigned int flags;
+} cudaHostRegister_v4000_params;
+
+typedef struct cudaHostUnregister_v4000_params_st {
+    void *ptr;
+} cudaHostUnregister_v4000_params;
+
+typedef struct cudaHostGetDevicePointer_v3020_params_st {
+    void **pDevice;
+    void *pHost;
+    unsigned int flags;
+} cudaHostGetDevicePointer_v3020_params;
+
+typedef struct cudaHostGetFlags_v3020_params_st {
+    unsigned int *pFlags;
+    void *pHost;
+} cudaHostGetFlags_v3020_params;
+
+typedef struct cudaMalloc3D_v3020_params_st {
+    struct cudaPitchedPtr *pitchedDevPtr;
+    struct cudaExtent extent;
+} cudaMalloc3D_v3020_params;
+
+typedef struct cudaMalloc3DArray_v3020_params_st {
+    cudaArray_t *array;
+    const struct cudaChannelFormatDesc *desc;
+    struct cudaExtent extent;
+    unsigned int flags;
+} cudaMalloc3DArray_v3020_params;
+
+typedef struct cudaMallocMipmappedArray_v5000_params_st {
+    cudaMipmappedArray_t *mipmappedArray;
+    const struct cudaChannelFormatDesc *desc;
+    struct cudaExtent extent;
+    unsigned int numLevels;
+    unsigned int flags;
+} cudaMallocMipmappedArray_v5000_params;
+
+typedef struct cudaGetMipmappedArrayLevel_v5000_params_st {
+    cudaArray_t *levelArray;
+    cudaMipmappedArray_const_t mipmappedArray;
+    unsigned int level;
+} cudaGetMipmappedArrayLevel_v5000_params;
+
+typedef struct cudaMemcpy3D_ptds_v7000_params_st {
+    const struct cudaMemcpy3DParms *p;
+} cudaMemcpy3D_ptds_v7000_params;
+
+typedef struct cudaMemcpy3DPeer_ptds_v7000_params_st {
+    const struct cudaMemcpy3DPeerParms *p;
+} cudaMemcpy3DPeer_ptds_v7000_params;
+
+typedef struct cudaMemcpy3DAsync_ptsz_v7000_params_st {
+    const struct cudaMemcpy3DParms *p;
+    cudaStream_t stream;
+} cudaMemcpy3DAsync_ptsz_v7000_params;
+
+typedef struct cudaMemcpy3DPeerAsync_ptsz_v7000_params_st {
+    const struct cudaMemcpy3DPeerParms *p;
+    cudaStream_t stream;
+} cudaMemcpy3DPeerAsync_ptsz_v7000_params;
+
+typedef struct cudaMemGetInfo_v3020_params_st {
+    size_t *free;
+    size_t *total;
+} cudaMemGetInfo_v3020_params;
+
+typedef struct cudaArrayGetInfo_v4010_params_st {
+    struct cudaChannelFormatDesc *desc;
+    struct cudaExtent *extent;
+    unsigned int *flags;
+    cudaArray_t array;
+} cudaArrayGetInfo_v4010_params;
+
+typedef struct cudaArrayGetPlane_v11020_params_st {
+    cudaArray_t *pPlaneArray;
+    cudaArray_t hArray;
+    unsigned int planeIdx;
+} cudaArrayGetPlane_v11020_params;
+
+typedef struct cudaArrayGetMemoryRequirements_v11060_params_st {
+    struct cudaArrayMemoryRequirements *memoryRequirements;
+    cudaArray_t array;
+    int device;
+} cudaArrayGetMemoryRequirements_v11060_params;
+
+typedef struct cudaMipmappedArrayGetMemoryRequirements_v11060_params_st {
+    struct cudaArrayMemoryRequirements *memoryRequirements;
+    cudaMipmappedArray_t mipmap;
+    int device;
+} cudaMipmappedArrayGetMemoryRequirements_v11060_params;
+
+typedef struct cudaArrayGetSparseProperties_v11010_params_st {
+    struct cudaArraySparseProperties *sparseProperties;
+    cudaArray_t array;
+} cudaArrayGetSparseProperties_v11010_params;
+
+typedef struct cudaMipmappedArrayGetSparseProperties_v11010_params_st {
+    struct cudaArraySparseProperties *sparseProperties;
+    cudaMipmappedArray_t mipmap;
+} cudaMipmappedArrayGetSparseProperties_v11010_params;
+
+typedef struct cudaMemcpy_ptds_v7000_params_st {
+    void *dst;
+    const void *src;
+    size_t count;
+    enum cudaMemcpyKind kind;
+} cudaMemcpy_ptds_v7000_params;
+
+typedef struct cudaMemcpyPeer_v4000_params_st {
+    void *dst;
+    int dstDevice;
+    const void *src;
+    int srcDevice;
+    size_t count;
+} cudaMemcpyPeer_v4000_params;
+
+typedef struct cudaMemcpy2D_ptds_v7000_params_st {
+    void *dst;
+    size_t dpitch;
+    const void *src;
+    size_t spitch;
+    size_t width;
+    size_t height;
+    enum cudaMemcpyKind kind;
+} cudaMemcpy2D_ptds_v7000_params;
+
+typedef struct cudaMemcpy2DToArray_ptds_v7000_params_st {
+    cudaArray_t dst;
+    size_t wOffset;
+    size_t hOffset;
+    const void *src;
+    size_t spitch;
+    size_t width;
+    size_t height;
+    enum cudaMemcpyKind kind;
+} cudaMemcpy2DToArray_ptds_v7000_params;
+
+typedef struct cudaMemcpy2DFromArray_ptds_v7000_params_st {
+    void *dst;
+    size_t dpitch;
+    cudaArray_const_t src;
+    size_t wOffset;
+    size_t hOffset;
+    size_t width;
+    size_t height;
+    enum cudaMemcpyKind kind;
+} cudaMemcpy2DFromArray_ptds_v7000_params;
+
+typedef struct cudaMemcpy2DArrayToArray_ptds_v7000_params_st {
+    cudaArray_t dst;
+    size_t wOffsetDst;
+    size_t hOffsetDst;
+    cudaArray_const_t src;
+    size_t wOffsetSrc;
+    size_t hOffsetSrc;
+    size_t width;
+    size_t height;
+    enum cudaMemcpyKind kind;
+} cudaMemcpy2DArrayToArray_ptds_v7000_params;
+
+typedef struct cudaMemcpyToSymbol_ptds_v7000_params_st {
+    const void *symbol;
+    const void *src;
+    size_t count;
+    size_t offset;
+    enum cudaMemcpyKind kind;
+} cudaMemcpyToSymbol_ptds_v7000_params;
+
+typedef struct cudaMemcpyFromSymbol_ptds_v7000_params_st {
+    void *dst;
+    const void *symbol;
+    size_t count;
+    size_t offset;
+    enum cudaMemcpyKind kind;
+} cudaMemcpyFromSymbol_ptds_v7000_params;
+
+typedef struct cudaMemcpyAsync_ptsz_v7000_params_st {
+    void *dst;
+    const void *src;
+    size_t count;
+    enum cudaMemcpyKind kind;
+    cudaStream_t stream;
+} cudaMemcpyAsync_ptsz_v7000_params;
+
+typedef struct cudaMemcpyPeerAsync_v4000_params_st {
+    void *dst;
+    int dstDevice;
+    const void *src;
+    int srcDevice;
+    size_t count;
+    cudaStream_t stream;
+} cudaMemcpyPeerAsync_v4000_params;
+
+typedef struct cudaMemcpyBatchAsync_ptsz_v12080_params_st {
+    void **dsts;
+    void **srcs;
+    size_t *sizes;
+    size_t count;
+    struct cudaMemcpyAttributes *attrs;
+    size_t *attrsIdxs;
+    size_t numAttrs;
+    size_t *failIdx;
+    cudaStream_t stream;
+} cudaMemcpyBatchAsync_ptsz_v12080_params;
+
+typedef struct cudaMemcpy3DBatchAsync_ptsz_v12080_params_st {
+    size_t numOps;
+    struct cudaMemcpy3DBatchOp *opList;
+    size_t *failIdx;
+    unsigned long long flags;
+    cudaStream_t stream;
+} cudaMemcpy3DBatchAsync_ptsz_v12080_params;
+
+typedef struct cudaMemcpy2DAsync_ptsz_v7000_params_st {
+    void *dst;
+    size_t dpitch;
+    const void *src;
+    size_t spitch;
+    size_t width;
+    size_t height;
+    enum cudaMemcpyKind kind;
+    cudaStream_t stream;
+} cudaMemcpy2DAsync_ptsz_v7000_params;
+
+typedef struct cudaMemcpy2DToArrayAsync_ptsz_v7000_params_st {
+    cudaArray_t dst;
+    size_t wOffset;
+    size_t hOffset;
+    const void *src;
+    size_t spitch;
+    size_t width;
+    size_t height;
+    enum cudaMemcpyKind kind;
+    cudaStream_t stream;
+} cudaMemcpy2DToArrayAsync_ptsz_v7000_params;
+
+typedef struct cudaMemcpy2DFromArrayAsync_ptsz_v7000_params_st {
+    void *dst;
+    size_t dpitch;
+    cudaArray_const_t src;
+    size_t wOffset;
+    size_t hOffset;
+    size_t width;
+    size_t height;
+    enum cudaMemcpyKind kind;
+    cudaStream_t stream;
+} cudaMemcpy2DFromArrayAsync_ptsz_v7000_params;
+
+typedef struct cudaMemcpyToSymbolAsync_ptsz_v7000_params_st {
+    const void *symbol;
+    const void *src;
+    size_t count;
+    size_t offset;
+    enum cudaMemcpyKind kind;
+    cudaStream_t stream;
+} cudaMemcpyToSymbolAsync_ptsz_v7000_params;
+
+typedef struct cudaMemcpyFromSymbolAsync_ptsz_v7000_params_st {
+    void *dst;
+    const void *symbol;
+    size_t count;
+    size_t offset;
+    enum cudaMemcpyKind kind;
+    cudaStream_t stream;
+} cudaMemcpyFromSymbolAsync_ptsz_v7000_params;
+
+typedef struct cudaMemset_ptds_v7000_params_st {
+    void *devPtr;
+    int value;
+    size_t count;
+} cudaMemset_ptds_v7000_params;
+
+typedef struct cudaMemset2D_ptds_v7000_params_st {
+    void *devPtr;
+    size_t pitch;
+    int value;
+    size_t width;
+    size_t height;
+} cudaMemset2D_ptds_v7000_params;
+
+typedef struct cudaMemset3D_ptds_v7000_params_st {
+    struct cudaPitchedPtr pitchedDevPtr;
+    int value;
+    struct cudaExtent extent;
+} cudaMemset3D_ptds_v7000_params;
+
+typedef struct cudaMemsetAsync_ptsz_v7000_params_st {
+    void *devPtr;
+    int value;
+    size_t count;
+    cudaStream_t stream;
+} cudaMemsetAsync_ptsz_v7000_params;
+
+typedef struct cudaMemset2DAsync_ptsz_v7000_params_st {
+    void *devPtr;
+    size_t pitch;
+    int value;
+    size_t width;
+    size_t height;
+    cudaStream_t stream;
+} cudaMemset2DAsync_ptsz_v7000_params;
+
+typedef struct cudaMemset3DAsync_ptsz_v7000_params_st {
+    struct cudaPitchedPtr pitchedDevPtr;
+    int value;
+    struct cudaExtent extent;
+    cudaStream_t stream;
+} cudaMemset3DAsync_ptsz_v7000_params;
+
+typedef struct cudaGetSymbolAddress_v3020_params_st {
+    void **devPtr;
+    const void *symbol;
+} cudaGetSymbolAddress_v3020_params;
+
+typedef struct cudaGetSymbolSize_v3020_params_st {
+    size_t *size;
+    const void *symbol;
+} cudaGetSymbolSize_v3020_params;
+
+typedef struct cudaMemPrefetchAsync_ptsz_v8000_params_st {
+    const void *devPtr;
+    size_t count;
+    int dstDevice;
+    cudaStream_t stream;
+} cudaMemPrefetchAsync_ptsz_v8000_params;
+
+typedef struct cudaMemPrefetchAsync_v2_ptsz_v12020_params_st {
+    const void *devPtr;
+    size_t count;
+    struct cudaMemLocation location;
+    unsigned int flags;
+    cudaStream_t stream;
+} cudaMemPrefetchAsync_v2_ptsz_v12020_params;
+
+typedef struct cudaMemAdvise_v8000_params_st {
+    const void *devPtr;
+    size_t count;
+    enum cudaMemoryAdvise advice;
+    int device;
+} cudaMemAdvise_v8000_params;
+
+typedef struct cudaMemAdvise_v2_v12020_params_st {
+    const void *devPtr;
+    size_t count;
+    enum cudaMemoryAdvise advice;
+    struct cudaMemLocation location;
+} cudaMemAdvise_v2_v12020_params;
+
+typedef struct cudaMemRangeGetAttribute_v8000_params_st {
+    void *data;
+    size_t dataSize;
+    enum cudaMemRangeAttribute attribute;
+    const void *devPtr;
+    size_t count;
+} cudaMemRangeGetAttribute_v8000_params;
+
+typedef struct cudaMemRangeGetAttributes_v8000_params_st {
+    void **data;
+    size_t *dataSizes;
+    enum cudaMemRangeAttribute *attributes;
+    size_t numAttributes;
+    const void *devPtr;
+    size_t count;
+} cudaMemRangeGetAttributes_v8000_params;
+
+typedef struct cudaMemcpyToArray_ptds_v7000_params_st {
+    cudaArray_t dst;
+    size_t wOffset;
+    size_t hOffset;
+    const void *src;
+    size_t count;
+    enum cudaMemcpyKind kind;
+} cudaMemcpyToArray_ptds_v7000_params;
+
+typedef struct cudaMemcpyFromArray_ptds_v7000_params_st {
+    void *dst;
+    cudaArray_const_t src;
+    size_t wOffset;
+    size_t hOffset;
+    size_t count;
+    enum cudaMemcpyKind kind;
+} cudaMemcpyFromArray_ptds_v7000_params;
+
+typedef struct cudaMemcpyArrayToArray_ptds_v7000_params_st {
+    cudaArray_t dst;
+    size_t wOffsetDst;
+    size_t hOffsetDst;
+    cudaArray_const_t src;
+    size_t wOffsetSrc;
+    size_t hOffsetSrc;
+    size_t count;
+    enum cudaMemcpyKind kind;
+} cudaMemcpyArrayToArray_ptds_v7000_params;
+
+typedef struct cudaMemcpyToArrayAsync_ptsz_v7000_params_st {
+    cudaArray_t dst;
+    size_t wOffset;
+    size_t hOffset;
+    const void *src;
+    size_t count;
+    enum cudaMemcpyKind kind;
+    cudaStream_t stream;
+} cudaMemcpyToArrayAsync_ptsz_v7000_params;
+
+typedef struct cudaMemcpyFromArrayAsync_ptsz_v7000_params_st {
+    void *dst;
+    cudaArray_const_t src;
+    size_t wOffset;
+    size_t hOffset;
+    size_t count;
+    enum cudaMemcpyKind kind;
+    cudaStream_t stream;
+} cudaMemcpyFromArrayAsync_ptsz_v7000_params;
+
+typedef struct cudaMallocAsync_ptsz_v11020_params_st {
+    void **devPtr;
+    size_t size;
+    cudaStream_t hStream;
+} cudaMallocAsync_ptsz_v11020_params;
+
+typedef struct cudaFreeAsync_ptsz_v11020_params_st {
+    void *devPtr;
+    cudaStream_t hStream;
+} cudaFreeAsync_ptsz_v11020_params;
+
+typedef struct cudaMemPoolTrimTo_v11020_params_st {
+    cudaMemPool_t memPool;
+    size_t minBytesToKeep;
+} cudaMemPoolTrimTo_v11020_params;
+
+typedef struct cudaMemPoolSetAttribute_v11020_params_st {
+    cudaMemPool_t memPool;
+    enum cudaMemPoolAttr attr;
+    void *value;
+} cudaMemPoolSetAttribute_v11020_params;
+
+typedef struct cudaMemPoolGetAttribute_v11020_params_st {
+    cudaMemPool_t memPool;
+    enum cudaMemPoolAttr attr;
+    void *value;
+} cudaMemPoolGetAttribute_v11020_params;
+
+typedef struct cudaMemPoolSetAccess_v11020_params_st {
+    cudaMemPool_t memPool;
+    const struct cudaMemAccessDesc *descList;
+    size_t count;
+} cudaMemPoolSetAccess_v11020_params;
+
+typedef struct cudaMemPoolGetAccess_v11020_params_st {
+    enum cudaMemAccessFlags *flags;
+    cudaMemPool_t memPool;
+    struct cudaMemLocation *location;
+} cudaMemPoolGetAccess_v11020_params;
+
+typedef struct cudaMemPoolCreate_v11020_params_st {
+    cudaMemPool_t *memPool;
+    const struct cudaMemPoolProps *poolProps;
+} cudaMemPoolCreate_v11020_params;
+
+typedef struct cudaMemPoolDestroy_v11020_params_st {
+    cudaMemPool_t memPool;
+} cudaMemPoolDestroy_v11020_params;
+
+typedef struct cudaMallocFromPoolAsync_ptsz_v11020_params_st {
+    void **ptr;
+    size_t size;
+    cudaMemPool_t memPool;
+    cudaStream_t stream;
+} cudaMallocFromPoolAsync_ptsz_v11020_params;
+
+typedef struct cudaMemPoolExportToShareableHandle_v11020_params_st {
+    void *shareableHandle;
+    cudaMemPool_t memPool;
+    enum cudaMemAllocationHandleType handleType;
+    unsigned int flags;
+} cudaMemPoolExportToShareableHandle_v11020_params;
+
+typedef struct cudaMemPoolImportFromShareableHandle_v11020_params_st {
+    cudaMemPool_t *memPool;
+    void *shareableHandle;
+    enum cudaMemAllocationHandleType handleType;
+    unsigned int flags;
+} cudaMemPoolImportFromShareableHandle_v11020_params;
+
+typedef struct cudaMemPoolExportPointer_v11020_params_st {
+    struct cudaMemPoolPtrExportData *exportData;
+    void *ptr;
+} cudaMemPoolExportPointer_v11020_params;
+
+typedef struct cudaMemPoolImportPointer_v11020_params_st {
+    void **ptr;
+    cudaMemPool_t memPool;
+    struct cudaMemPoolPtrExportData *exportData;
+} cudaMemPoolImportPointer_v11020_params;
+
+typedef struct cudaPointerGetAttributes_v4000_params_st {
+    struct cudaPointerAttributes *attributes;
+    const void *ptr;
+} cudaPointerGetAttributes_v4000_params;
+
+typedef struct cudaDeviceCanAccessPeer_v4000_params_st {
+    int *canAccessPeer;
+    int device;
+    int peerDevice;
+} cudaDeviceCanAccessPeer_v4000_params;
+
+typedef struct cudaDeviceEnablePeerAccess_v4000_params_st {
+    int peerDevice;
+    unsigned int flags;
+} cudaDeviceEnablePeerAccess_v4000_params;
+
+typedef struct cudaDeviceDisablePeerAccess_v4000_params_st {
+    int peerDevice;
+} cudaDeviceDisablePeerAccess_v4000_params;
+
+typedef struct cudaGraphicsUnregisterResource_v3020_params_st {
+    cudaGraphicsResource_t resource;
+} cudaGraphicsUnregisterResource_v3020_params;
+
+typedef struct cudaGraphicsResourceSetMapFlags_v3020_params_st {
+    cudaGraphicsResource_t resource;
+    unsigned int flags;
+} cudaGraphicsResourceSetMapFlags_v3020_params;
+
+typedef struct cudaGraphicsMapResources_v3020_params_st {
+    int count;
+    cudaGraphicsResource_t *resources;
+    cudaStream_t stream;
+} cudaGraphicsMapResources_v3020_params;
+
+typedef struct cudaGraphicsUnmapResources_v3020_params_st {
+    int count;
+    cudaGraphicsResource_t *resources;
+    cudaStream_t stream;
+} cudaGraphicsUnmapResources_v3020_params;
+
+typedef struct cudaGraphicsResourceGetMappedPointer_v3020_params_st {
+    void **devPtr;
+    size_t *size;
+    cudaGraphicsResource_t resource;
+} cudaGraphicsResourceGetMappedPointer_v3020_params;
+
+typedef struct cudaGraphicsSubResourceGetMappedArray_v3020_params_st {
+    cudaArray_t *array;
+    cudaGraphicsResource_t resource;
+    unsigned int arrayIndex;
+    unsigned int mipLevel;
+} cudaGraphicsSubResourceGetMappedArray_v3020_params;
+
+typedef struct cudaGraphicsResourceGetMappedMipmappedArray_v5000_params_st {
+    cudaMipmappedArray_t *mipmappedArray;
+    cudaGraphicsResource_t resource;
+} cudaGraphicsResourceGetMappedMipmappedArray_v5000_params;
+
+typedef struct cudaGetChannelDesc_v3020_params_st {
+    struct cudaChannelFormatDesc *desc;
+    cudaArray_const_t array;
+} cudaGetChannelDesc_v3020_params;
+
+typedef struct cudaCreateChannelDesc_v3020_params_st {
+    int x;
+    int y;
+    int z;
+    int w;
+    enum cudaChannelFormatKind f;
+} cudaCreateChannelDesc_v3020_params;
+
+typedef struct cudaCreateTextureObject_v5000_params_st {
+    cudaTextureObject_t *pTexObject;
+    const struct cudaResourceDesc *pResDesc;
+    const struct cudaTextureDesc *pTexDesc;
+    const struct cudaResourceViewDesc *pResViewDesc;
+} cudaCreateTextureObject_v5000_params;
+
+typedef struct cudaDestroyTextureObject_v5000_params_st {
+    cudaTextureObject_t texObject;
+} cudaDestroyTextureObject_v5000_params;
+
+typedef struct cudaGetTextureObjectResourceDesc_v5000_params_st {
+    struct cudaResourceDesc *pResDesc;
+    cudaTextureObject_t texObject;
+} cudaGetTextureObjectResourceDesc_v5000_params;
+
+typedef struct cudaGetTextureObjectTextureDesc_v5000_params_st {
+    struct cudaTextureDesc *pTexDesc;
+    cudaTextureObject_t texObject;
+} cudaGetTextureObjectTextureDesc_v5000_params;
+
+typedef struct cudaGetTextureObjectResourceViewDesc_v5000_params_st {
+    struct cudaResourceViewDesc *pResViewDesc;
+    cudaTextureObject_t texObject;
+} cudaGetTextureObjectResourceViewDesc_v5000_params;
+
+typedef struct cudaCreateSurfaceObject_v5000_params_st {
+    cudaSurfaceObject_t *pSurfObject;
+    const struct cudaResourceDesc *pResDesc;
+} cudaCreateSurfaceObject_v5000_params;
+
+typedef struct cudaDestroySurfaceObject_v5000_params_st {
+    cudaSurfaceObject_t surfObject;
+} cudaDestroySurfaceObject_v5000_params;
+
+typedef struct cudaGetSurfaceObjectResourceDesc_v5000_params_st {
+    struct cudaResourceDesc *pResDesc;
+    cudaSurfaceObject_t surfObject;
+} cudaGetSurfaceObjectResourceDesc_v5000_params;
+
+typedef struct cudaDriverGetVersion_v3020_params_st {
+    int *driverVersion;
+} cudaDriverGetVersion_v3020_params;
+
+typedef struct cudaRuntimeGetVersion_v3020_params_st {
+    int *runtimeVersion;
+} cudaRuntimeGetVersion_v3020_params;
+
+typedef struct cudaGraphCreate_v10000_params_st {
+    cudaGraph_t *pGraph;
+    unsigned int flags;
+} cudaGraphCreate_v10000_params;
+
+typedef struct cudaGraphAddKernelNode_v10000_params_st {
+    cudaGraphNode_t *pGraphNode;
+    cudaGraph_t graph;
+    const cudaGraphNode_t *pDependencies;
+    size_t numDependencies;
+    const struct cudaKernelNodeParams *pNodeParams;
+} cudaGraphAddKernelNode_v10000_params;
+
+typedef struct cudaGraphKernelNodeGetParams_v10000_params_st {
+    cudaGraphNode_t node;
+    struct cudaKernelNodeParams *pNodeParams;
+} cudaGraphKernelNodeGetParams_v10000_params;
+
+typedef struct cudaGraphKernelNodeSetParams_v10000_params_st {
+    cudaGraphNode_t node;
+    const struct cudaKernelNodeParams *pNodeParams;
+} cudaGraphKernelNodeSetParams_v10000_params;
+
+typedef struct cudaGraphKernelNodeCopyAttributes_v11000_params_st {
+    cudaGraphNode_t hSrc;
+    cudaGraphNode_t hDst;
+} cudaGraphKernelNodeCopyAttributes_v11000_params;
+
+typedef struct cudaGraphKernelNodeGetAttribute_v11000_params_st {
+    cudaGraphNode_t hNode;
+    cudaKernelNodeAttrID attr;
+    cudaKernelNodeAttrValue *value_out;
+} cudaGraphKernelNodeGetAttribute_v11000_params;
+
+typedef struct cudaGraphKernelNodeSetAttribute_v11000_params_st {
+    cudaGraphNode_t hNode;
+    cudaKernelNodeAttrID attr;
+    const cudaKernelNodeAttrValue *value;
+} cudaGraphKernelNodeSetAttribute_v11000_params;
+
+typedef struct cudaGraphAddMemcpyNode_v10000_params_st {
+    cudaGraphNode_t *pGraphNode;
+    cudaGraph_t graph;
+    const cudaGraphNode_t *pDependencies;
+    size_t numDependencies;
+    const struct cudaMemcpy3DParms *pCopyParams;
+} cudaGraphAddMemcpyNode_v10000_params;
+
+typedef struct cudaGraphAddMemcpyNodeToSymbol_v11010_params_st {
+    cudaGraphNode_t *pGraphNode;
+    cudaGraph_t graph;
+    const cudaGraphNode_t *pDependencies;
+    size_t numDependencies;
+    const void *symbol;
+    const void *src;
+    size_t count;
+    size_t offset;
+    enum cudaMemcpyKind kind;
+} cudaGraphAddMemcpyNodeToSymbol_v11010_params;
+
+typedef struct cudaGraphAddMemcpyNodeFromSymbol_v11010_params_st {
+    cudaGraphNode_t *pGraphNode;
+    cudaGraph_t graph;
+    const cudaGraphNode_t *pDependencies;
+    size_t numDependencies;
+    void *dst;
+    const void *symbol;
+    size_t count;
+    size_t offset;
+    enum cudaMemcpyKind kind;
+} cudaGraphAddMemcpyNodeFromSymbol_v11010_params;
+
+typedef struct cudaGraphAddMemcpyNode1D_v11010_params_st {
+    cudaGraphNode_t *pGraphNode;
+    cudaGraph_t graph;
+    const cudaGraphNode_t *pDependencies;
+    size_t numDependencies;
+    void *dst;
+    const void *src;
+    size_t count;
+    enum cudaMemcpyKind kind;
+} cudaGraphAddMemcpyNode1D_v11010_params;
+
+typedef struct cudaGraphMemcpyNodeGetParams_v10000_params_st {
+    cudaGraphNode_t node;
+    struct cudaMemcpy3DParms *pNodeParams;
+} cudaGraphMemcpyNodeGetParams_v10000_params;
+
+typedef struct cudaGraphMemcpyNodeSetParams_v10000_params_st {
+    cudaGraphNode_t node;
+    const struct cudaMemcpy3DParms *pNodeParams;
+} cudaGraphMemcpyNodeSetParams_v10000_params;
+
+typedef struct cudaGraphMemcpyNodeSetParamsToSymbol_v11010_params_st {
+    cudaGraphNode_t node;
+    const void *symbol;
+    const void *src;
+    size_t count;
+    size_t offset;
+    enum cudaMemcpyKind kind;
+} cudaGraphMemcpyNodeSetParamsToSymbol_v11010_params;
+
+typedef struct cudaGraphMemcpyNodeSetParamsFromSymbol_v11010_params_st {
+    cudaGraphNode_t node;
+    void *dst;
+    const void *symbol;
+    size_t count;
+    size_t offset;
+    enum cudaMemcpyKind kind;
+} cudaGraphMemcpyNodeSetParamsFromSymbol_v11010_params;
+
+typedef struct cudaGraphMemcpyNodeSetParams1D_v11010_params_st {
+    cudaGraphNode_t node;
+    void *dst;
+    const void *src;
+    size_t count;
+    enum cudaMemcpyKind kind;
+} cudaGraphMemcpyNodeSetParams1D_v11010_params;
+
+typedef struct cudaGraphAddMemsetNode_v10000_params_st {
+    cudaGraphNode_t *pGraphNode;
+    cudaGraph_t graph;
+    const cudaGraphNode_t *pDependencies;
+    size_t numDependencies;
+    const struct cudaMemsetParams *pMemsetParams;
+} cudaGraphAddMemsetNode_v10000_params;
+
+typedef struct cudaGraphMemsetNodeGetParams_v10000_params_st {
+    cudaGraphNode_t node;
+    struct cudaMemsetParams *pNodeParams;
+} cudaGraphMemsetNodeGetParams_v10000_params;
+
+typedef struct cudaGraphMemsetNodeSetParams_v10000_params_st {
+    cudaGraphNode_t node;
+    const struct cudaMemsetParams *pNodeParams;
+} cudaGraphMemsetNodeSetParams_v10000_params;
+
+typedef struct cudaGraphAddHostNode_v10000_params_st {
+    cudaGraphNode_t *pGraphNode;
+    cudaGraph_t graph;
+    const cudaGraphNode_t *pDependencies;
+    size_t numDependencies;
+    const struct cudaHostNodeParams *pNodeParams;
+} cudaGraphAddHostNode_v10000_params;
+
+typedef struct cudaGraphHostNodeGetParams_v10000_params_st {
+    cudaGraphNode_t node;
+    struct cudaHostNodeParams *pNodeParams;
+} cudaGraphHostNodeGetParams_v10000_params;
+
+typedef struct cudaGraphHostNodeSetParams_v10000_params_st {
+    cudaGraphNode_t node;
+    const struct cudaHostNodeParams *pNodeParams;
+} cudaGraphHostNodeSetParams_v10000_params;
+
+typedef struct cudaGraphAddChildGraphNode_v10000_params_st {
+    cudaGraphNode_t *pGraphNode;
+    cudaGraph_t graph;
+    const cudaGraphNode_t *pDependencies;
+    size_t numDependencies;
+    cudaGraph_t childGraph;
+} cudaGraphAddChildGraphNode_v10000_params;
+
+typedef struct cudaGraphChildGraphNodeGetGraph_v10000_params_st {
+    cudaGraphNode_t node;
+    cudaGraph_t *pGraph;
+} cudaGraphChildGraphNodeGetGraph_v10000_params;
+
+typedef struct cudaGraphAddEmptyNode_v10000_params_st {
+    cudaGraphNode_t *pGraphNode;
+    cudaGraph_t graph;
+    const cudaGraphNode_t *pDependencies;
+    size_t numDependencies;
+} cudaGraphAddEmptyNode_v10000_params;
+
+typedef struct cudaGraphAddEventRecordNode_v11010_params_st {
+    cudaGraphNode_t *pGraphNode;
+    cudaGraph_t graph;
+    const cudaGraphNode_t *pDependencies;
+    size_t numDependencies;
+    cudaEvent_t event;
+} cudaGraphAddEventRecordNode_v11010_params;
+
+typedef struct cudaGraphEventRecordNodeGetEvent_v11010_params_st {
+    cudaGraphNode_t node;
+    cudaEvent_t *event_out;
+} cudaGraphEventRecordNodeGetEvent_v11010_params;
+
+typedef struct cudaGraphEventRecordNodeSetEvent_v11010_params_st {
+    cudaGraphNode_t node;
+    cudaEvent_t event;
+} cudaGraphEventRecordNodeSetEvent_v11010_params;
+
+typedef struct cudaGraphAddEventWaitNode_v11010_params_st {
+    cudaGraphNode_t *pGraphNode;
+    cudaGraph_t graph;
+    const cudaGraphNode_t *pDependencies;
+    size_t numDependencies;
+    cudaEvent_t event;
+} cudaGraphAddEventWaitNode_v11010_params;
+
+typedef struct cudaGraphEventWaitNodeGetEvent_v11010_params_st {
+    cudaGraphNode_t node;
+    cudaEvent_t *event_out;
+} cudaGraphEventWaitNodeGetEvent_v11010_params;
+
+typedef struct cudaGraphEventWaitNodeSetEvent_v11010_params_st {
+    cudaGraphNode_t node;
+    cudaEvent_t event;
+} cudaGraphEventWaitNodeSetEvent_v11010_params;
+
+typedef struct cudaGraphAddExternalSemaphoresSignalNode_v11020_params_st {
+    cudaGraphNode_t *pGraphNode;
+    cudaGraph_t graph;
+    const cudaGraphNode_t *pDependencies;
+    size_t numDependencies;
+    const struct cudaExternalSemaphoreSignalNodeParams *nodeParams;
+} cudaGraphAddExternalSemaphoresSignalNode_v11020_params;
+
+typedef struct cudaGraphExternalSemaphoresSignalNodeGetParams_v11020_params_st {
+    cudaGraphNode_t hNode;
+    struct cudaExternalSemaphoreSignalNodeParams *params_out;
+} cudaGraphExternalSemaphoresSignalNodeGetParams_v11020_params;
+
+typedef struct cudaGraphExternalSemaphoresSignalNodeSetParams_v11020_params_st {
+    cudaGraphNode_t hNode;
+    const struct cudaExternalSemaphoreSignalNodeParams *nodeParams;
+} cudaGraphExternalSemaphoresSignalNodeSetParams_v11020_params;
+
+typedef struct cudaGraphAddExternalSemaphoresWaitNode_v11020_params_st {
+    cudaGraphNode_t *pGraphNode;
+    cudaGraph_t graph;
+    const cudaGraphNode_t *pDependencies;
+    size_t numDependencies;
+    const struct cudaExternalSemaphoreWaitNodeParams *nodeParams;
+} cudaGraphAddExternalSemaphoresWaitNode_v11020_params;
+
+typedef struct cudaGraphExternalSemaphoresWaitNodeGetParams_v11020_params_st {
+    cudaGraphNode_t hNode;
+    struct cudaExternalSemaphoreWaitNodeParams *params_out;
+} cudaGraphExternalSemaphoresWaitNodeGetParams_v11020_params;
+
+typedef struct cudaGraphExternalSemaphoresWaitNodeSetParams_v11020_params_st {
+    cudaGraphNode_t hNode;
+    const struct cudaExternalSemaphoreWaitNodeParams *nodeParams;
+} cudaGraphExternalSemaphoresWaitNodeSetParams_v11020_params;
+
+typedef struct cudaGraphAddMemAllocNode_v11040_params_st {
+    cudaGraphNode_t *pGraphNode;
+    cudaGraph_t graph;
+    const cudaGraphNode_t *pDependencies;
+    size_t numDependencies;
+    struct cudaMemAllocNodeParams *nodeParams;
+} cudaGraphAddMemAllocNode_v11040_params;
+
+typedef struct cudaGraphMemAllocNodeGetParams_v11040_params_st {
+    cudaGraphNode_t node;
+    struct cudaMemAllocNodeParams *params_out;
+} cudaGraphMemAllocNodeGetParams_v11040_params;
+
+typedef struct cudaGraphAddMemFreeNode_v11040_params_st {
+    cudaGraphNode_t *pGraphNode;
+    cudaGraph_t graph;
+    const cudaGraphNode_t *pDependencies;
+    size_t numDependencies;
+    void *dptr;
+} cudaGraphAddMemFreeNode_v11040_params;
+
+typedef struct cudaGraphMemFreeNodeGetParams_v11040_params_st {
+    cudaGraphNode_t node;
+    void *dptr_out;
+} cudaGraphMemFreeNodeGetParams_v11040_params;
+
+typedef struct cudaDeviceGraphMemTrim_v11040_params_st {
+    int device;
+} cudaDeviceGraphMemTrim_v11040_params;
+
+typedef struct cudaDeviceGetGraphMemAttribute_v11040_params_st {
+    int device;
+    enum cudaGraphMemAttributeType attr;
+    void *value;
+} cudaDeviceGetGraphMemAttribute_v11040_params;
+
+typedef struct cudaDeviceSetGraphMemAttribute_v11040_params_st {
+    int device;
+    enum cudaGraphMemAttributeType attr;
+    void *value;
+} cudaDeviceSetGraphMemAttribute_v11040_params;
+
+typedef struct cudaGraphClone_v10000_params_st {
+    cudaGraph_t *pGraphClone;
+    cudaGraph_t originalGraph;
+} cudaGraphClone_v10000_params;
+
+typedef struct cudaGraphNodeFindInClone_v10000_params_st {
+    cudaGraphNode_t *pNode;
+    cudaGraphNode_t originalNode;
+    cudaGraph_t clonedGraph;
+} cudaGraphNodeFindInClone_v10000_params;
+
+typedef struct cudaGraphNodeGetType_v10000_params_st {
+    cudaGraphNode_t node;
+    enum cudaGraphNodeType *pType;
+} cudaGraphNodeGetType_v10000_params;
+
+typedef struct cudaGraphGetNodes_v10000_params_st {
+    cudaGraph_t graph;
+    cudaGraphNode_t *nodes;
+    size_t *numNodes;
+} cudaGraphGetNodes_v10000_params;
+
+typedef struct cudaGraphGetRootNodes_v10000_params_st {
+    cudaGraph_t graph;
+    cudaGraphNode_t *pRootNodes;
+    size_t *pNumRootNodes;
+} cudaGraphGetRootNodes_v10000_params;
+
+typedef struct cudaGraphGetEdges_v10000_params_st {
+    cudaGraph_t graph;
+    cudaGraphNode_t *from;
+    cudaGraphNode_t *to;
+    size_t *numEdges;
+} cudaGraphGetEdges_v10000_params;
+
+typedef struct cudaGraphGetEdges_v2_v12030_params_st {
+    cudaGraph_t graph;
+    cudaGraphNode_t *from;
+    cudaGraphNode_t *to;
+    cudaGraphEdgeData *edgeData;
+    size_t *numEdges;
+} cudaGraphGetEdges_v2_v12030_params;
+
+typedef struct cudaGraphNodeGetDependencies_v10000_params_st {
+    cudaGraphNode_t node;
+    cudaGraphNode_t *pDependencies;
+    size_t *pNumDependencies;
+} cudaGraphNodeGetDependencies_v10000_params;
+
+typedef struct cudaGraphNodeGetDependencies_v2_v12030_params_st {
+    cudaGraphNode_t node;
+    cudaGraphNode_t *pDependencies;
+    cudaGraphEdgeData *edgeData;
+    size_t *pNumDependencies;
+} cudaGraphNodeGetDependencies_v2_v12030_params;
+
+typedef struct cudaGraphNodeGetDependentNodes_v10000_params_st {
+    cudaGraphNode_t node;
+    cudaGraphNode_t *pDependentNodes;
+    size_t *pNumDependentNodes;
+} cudaGraphNodeGetDependentNodes_v10000_params;
+
+typedef struct cudaGraphNodeGetDependentNodes_v2_v12030_params_st {
+    cudaGraphNode_t node;
+    cudaGraphNode_t *pDependentNodes;
+    cudaGraphEdgeData *edgeData;
+    size_t *pNumDependentNodes;
+} cudaGraphNodeGetDependentNodes_v2_v12030_params;
+
+typedef struct cudaGraphAddDependencies_v10000_params_st {
+    cudaGraph_t graph;
+    const cudaGraphNode_t *from;
+    const cudaGraphNode_t *to;
+    size_t numDependencies;
+} cudaGraphAddDependencies_v10000_params;
+
+typedef struct cudaGraphAddDependencies_v2_v12030_params_st {
+    cudaGraph_t graph;
+    const cudaGraphNode_t *from;
+    const cudaGraphNode_t *to;
+    const cudaGraphEdgeData *edgeData;
+    size_t numDependencies;
+} cudaGraphAddDependencies_v2_v12030_params;
+
+typedef struct cudaGraphRemoveDependencies_v10000_params_st {
+    cudaGraph_t graph;
+    const cudaGraphNode_t *from;
+    const cudaGraphNode_t *to;
+    size_t numDependencies;
+} cudaGraphRemoveDependencies_v10000_params;
+
+typedef struct cudaGraphRemoveDependencies_v2_v12030_params_st {
+    cudaGraph_t graph;
+    const cudaGraphNode_t *from;
+    const cudaGraphNode_t *to;
+    const cudaGraphEdgeData *edgeData;
+    size_t numDependencies;
+} cudaGraphRemoveDependencies_v2_v12030_params;
+
+typedef struct cudaGraphDestroyNode_v10000_params_st {
+    cudaGraphNode_t node;
+} cudaGraphDestroyNode_v10000_params;
+
+typedef struct cudaGraphInstantiate_v12000_params_st {
+    cudaGraphExec_t *pGraphExec;
+    cudaGraph_t graph;
+    unsigned long long flags;
+} cudaGraphInstantiate_v12000_params;
+
+typedef struct cudaGraphInstantiateWithFlags_v11040_params_st {
+    cudaGraphExec_t *pGraphExec;
+    cudaGraph_t graph;
+    unsigned long long flags;
+} cudaGraphInstantiateWithFlags_v11040_params;
+
+typedef struct cudaGraphInstantiateWithParams_ptsz_v12000_params_st {
+    cudaGraphExec_t *pGraphExec;
+    cudaGraph_t graph;
+    cudaGraphInstantiateParams *instantiateParams;
+} cudaGraphInstantiateWithParams_ptsz_v12000_params;
+
+typedef struct cudaGraphExecGetFlags_v12000_params_st {
+    cudaGraphExec_t graphExec;
+    unsigned long long *flags;
+} cudaGraphExecGetFlags_v12000_params;
+
+typedef struct cudaGraphExecKernelNodeSetParams_v10010_params_st {
+    cudaGraphExec_t hGraphExec;
+    cudaGraphNode_t node;
+    const struct cudaKernelNodeParams *pNodeParams;
+} cudaGraphExecKernelNodeSetParams_v10010_params;
+
+typedef struct cudaGraphExecMemcpyNodeSetParams_v10020_params_st {
+    cudaGraphExec_t hGraphExec;
+    cudaGraphNode_t node;
+    const struct cudaMemcpy3DParms *pNodeParams;
+} cudaGraphExecMemcpyNodeSetParams_v10020_params;
+
+typedef struct cudaGraphExecMemcpyNodeSetParamsToSymbol_v11010_params_st {
+    cudaGraphExec_t hGraphExec;
+    cudaGraphNode_t node;
+    const void *symbol;
+    const void *src;
+    size_t count;
+    size_t offset;
+    enum cudaMemcpyKind kind;
+} cudaGraphExecMemcpyNodeSetParamsToSymbol_v11010_params;
+
+typedef struct cudaGraphExecMemcpyNodeSetParamsFromSymbol_v11010_params_st {
+    cudaGraphExec_t hGraphExec;
+    cudaGraphNode_t node;
+    void *dst;
+    const void *symbol;
+    size_t count;
+    size_t offset;
+    enum cudaMemcpyKind kind;
+} cudaGraphExecMemcpyNodeSetParamsFromSymbol_v11010_params;
+
+typedef struct cudaGraphExecMemcpyNodeSetParams1D_v11010_params_st {
+    cudaGraphExec_t hGraphExec;
+    cudaGraphNode_t node;
+    void *dst;
+    const void *src;
+    size_t count;
+    enum cudaMemcpyKind kind;
+} cudaGraphExecMemcpyNodeSetParams1D_v11010_params;
+
+typedef struct cudaGraphExecMemsetNodeSetParams_v10020_params_st {
+    cudaGraphExec_t hGraphExec;
+    cudaGraphNode_t node;
+    const struct cudaMemsetParams *pNodeParams;
+} cudaGraphExecMemsetNodeSetParams_v10020_params;
+
+typedef struct cudaGraphExecHostNodeSetParams_v10020_params_st {
+    cudaGraphExec_t hGraphExec;
+    cudaGraphNode_t node;
+    const struct cudaHostNodeParams *pNodeParams;
+} cudaGraphExecHostNodeSetParams_v10020_params;
+
+typedef struct cudaGraphExecChildGraphNodeSetParams_v11010_params_st {
+    cudaGraphExec_t hGraphExec;
+    cudaGraphNode_t node;
+    cudaGraph_t childGraph;
+} cudaGraphExecChildGraphNodeSetParams_v11010_params;
+
+typedef struct cudaGraphExecEventRecordNodeSetEvent_v11010_params_st {
+    cudaGraphExec_t hGraphExec;
+    cudaGraphNode_t hNode;
+    cudaEvent_t event;
+} cudaGraphExecEventRecordNodeSetEvent_v11010_params;
+
+typedef struct cudaGraphExecEventWaitNodeSetEvent_v11010_params_st {
+    cudaGraphExec_t hGraphExec;
+    cudaGraphNode_t hNode;
+    cudaEvent_t event;
+} cudaGraphExecEventWaitNodeSetEvent_v11010_params;
+
+typedef struct cudaGraphExecExternalSemaphoresSignalNodeSetParams_v11020_params_st {
+    cudaGraphExec_t hGraphExec;
+    cudaGraphNode_t hNode;
+    const struct cudaExternalSemaphoreSignalNodeParams *nodeParams;
+} cudaGraphExecExternalSemaphoresSignalNodeSetParams_v11020_params;
+
+typedef struct cudaGraphExecExternalSemaphoresWaitNodeSetParams_v11020_params_st {
+    cudaGraphExec_t hGraphExec;
+    cudaGraphNode_t hNode;
+    const struct cudaExternalSemaphoreWaitNodeParams *nodeParams;
+} cudaGraphExecExternalSemaphoresWaitNodeSetParams_v11020_params;
+
+typedef struct cudaGraphNodeSetEnabled_v11060_params_st {
+    cudaGraphExec_t hGraphExec;
+    cudaGraphNode_t hNode;
+    unsigned int isEnabled;
+} cudaGraphNodeSetEnabled_v11060_params;
+
+typedef struct cudaGraphNodeGetEnabled_v11060_params_st {
+    cudaGraphExec_t hGraphExec;
+    cudaGraphNode_t hNode;
+    unsigned int *isEnabled;
+} cudaGraphNodeGetEnabled_v11060_params;
+
+typedef struct cudaGraphExecUpdate_v10020_params_st {
+    cudaGraphExec_t hGraphExec;
+    cudaGraph_t hGraph;
+    cudaGraphExecUpdateResultInfo *resultInfo;
+} cudaGraphExecUpdate_v10020_params;
+
+typedef struct cudaGraphUpload_ptsz_v10000_params_st {
+    cudaGraphExec_t graphExec;
+    cudaStream_t stream;
+} cudaGraphUpload_ptsz_v10000_params;
+
+typedef struct cudaGraphLaunch_ptsz_v10000_params_st {
+    cudaGraphExec_t graphExec;
+    cudaStream_t stream;
+} cudaGraphLaunch_ptsz_v10000_params;
+
+typedef struct cudaGraphExecDestroy_v10000_params_st {
+    cudaGraphExec_t graphExec;
+} cudaGraphExecDestroy_v10000_params;
+
+typedef struct cudaGraphDestroy_v10000_params_st {
+    cudaGraph_t graph;
+} cudaGraphDestroy_v10000_params;
+
+typedef struct cudaGraphDebugDotPrint_v11030_params_st {
+    cudaGraph_t graph;
+    const char *path;
+    unsigned int flags;
+} cudaGraphDebugDotPrint_v11030_params;
+
+typedef struct cudaUserObjectCreate_v11030_params_st {
+    cudaUserObject_t *object_out;
+    void *ptr;
+    cudaHostFn_t destroy;
+    unsigned int initialRefcount;
+    unsigned int flags;
+} cudaUserObjectCreate_v11030_params;
+
+typedef struct cudaUserObjectRetain_v11030_params_st {
+    cudaUserObject_t object;
+    unsigned int count;
+} cudaUserObjectRetain_v11030_params;
+
+typedef struct cudaUserObjectRelease_v11030_params_st {
+    cudaUserObject_t object;
+    unsigned int count;
+} cudaUserObjectRelease_v11030_params;
+
+typedef struct cudaGraphRetainUserObject_v11030_params_st {
+    cudaGraph_t graph;
+    cudaUserObject_t object;
+    unsigned int count;
+    unsigned int flags;
+} cudaGraphRetainUserObject_v11030_params;
+
+typedef struct cudaGraphReleaseUserObject_v11030_params_st {
+    cudaGraph_t graph;
+    cudaUserObject_t object;
+    unsigned int count;
+} cudaGraphReleaseUserObject_v11030_params;
+
+typedef struct cudaGraphAddNode_v12020_params_st {
+    cudaGraphNode_t *pGraphNode;
+    cudaGraph_t graph;
+    const cudaGraphNode_t *pDependencies;
+    size_t numDependencies;
+    struct cudaGraphNodeParams *nodeParams;
+} cudaGraphAddNode_v12020_params;
+
+typedef struct cudaGraphAddNode_v2_v12030_params_st {
+    cudaGraphNode_t *pGraphNode;
+    cudaGraph_t graph;
+    const cudaGraphNode_t *pDependencies;
+    const cudaGraphEdgeData *dependencyData;
+    size_t numDependencies;
+    struct cudaGraphNodeParams *nodeParams;
+} cudaGraphAddNode_v2_v12030_params;
+
+typedef struct cudaGraphNodeSetParams_v12020_params_st {
+    cudaGraphNode_t node;
+    struct cudaGraphNodeParams *nodeParams;
+} cudaGraphNodeSetParams_v12020_params;
+
+typedef struct cudaGraphExecNodeSetParams_v12020_params_st {
+    cudaGraphExec_t graphExec;
+    cudaGraphNode_t node;
+    struct cudaGraphNodeParams *nodeParams;
+} cudaGraphExecNodeSetParams_v12020_params;
+
+typedef struct cudaGraphConditionalHandleCreate_v12030_params_st {
+    cudaGraphConditionalHandle *pHandle_out;
+    cudaGraph_t graph;
+    unsigned int defaultLaunchValue;
+    unsigned int flags;
+} cudaGraphConditionalHandleCreate_v12030_params;
+
+typedef struct cudaGetDriverEntryPoint_ptsz_v11030_params_st {
+    const char *symbol;
+    void **funcPtr;
+    unsigned long long flags;
+    enum cudaDriverEntryPointQueryResult *driverStatus;
+} cudaGetDriverEntryPoint_ptsz_v11030_params;
+
+typedef struct cudaGetDriverEntryPointByVersion_ptsz_v12050_params_st {
+    const char *symbol;
+    void **funcPtr;
+    unsigned int cudaVersion;
+    unsigned long long flags;
+    enum cudaDriverEntryPointQueryResult *driverStatus;
+} cudaGetDriverEntryPointByVersion_ptsz_v12050_params;
+
+typedef struct cudaGetFuncBySymbol_v11000_params_st {
+    cudaFunction_t *functionPtr;
+    const void *symbolPtr;
+} cudaGetFuncBySymbol_v11000_params;
+
+typedef struct cudaGetKernel_v12000_params_st {
+    cudaKernel_t *kernelPtr;
+    const void *entryFuncAddr;
+} cudaGetKernel_v12000_params;
+
+typedef struct cudaMemcpy_v3020_params_st {
+    void *dst;
+    const void *src;
+    size_t count;
+    enum cudaMemcpyKind kind;
+} cudaMemcpy_v3020_params;
+
+typedef struct cudaMemcpyToSymbol_v3020_params_st {
+    const void *symbol;
+    const void *src;
+    size_t count;
+    size_t offset;
+    enum cudaMemcpyKind kind;
+} cudaMemcpyToSymbol_v3020_params;
+
+typedef struct cudaMemcpyFromSymbol_v3020_params_st {
+    void *dst;
+    const void *symbol;
+    size_t count;
+    size_t offset;
+    enum cudaMemcpyKind kind;
+} cudaMemcpyFromSymbol_v3020_params;
+
+typedef struct cudaMemcpy2D_v3020_params_st {
+    void *dst;
+    size_t dpitch;
+    const void *src;
+    size_t spitch;
+    size_t width;
+    size_t height;
+    enum cudaMemcpyKind kind;
+} cudaMemcpy2D_v3020_params;
+
+typedef struct cudaMemcpyToArray_v3020_params_st {
+    cudaArray_t dst;
+    size_t wOffset;
+    size_t hOffset;
+    const void *src;
+    size_t count;
+    enum cudaMemcpyKind kind;
+} cudaMemcpyToArray_v3020_params;
+
+typedef struct cudaMemcpy2DToArray_v3020_params_st {
+    cudaArray_t dst;
+    size_t wOffset;
+    size_t hOffset;
+    const void *src;
+    size_t spitch;
+    size_t width;
+    size_t height;
+    enum cudaMemcpyKind kind;
+} cudaMemcpy2DToArray_v3020_params;
+
+typedef struct cudaMemcpyFromArray_v3020_params_st {
+    void *dst;
+    cudaArray_const_t src;
+    size_t wOffset;
+    size_t hOffset;
+    size_t count;
+    enum cudaMemcpyKind kind;
+} cudaMemcpyFromArray_v3020_params;
+
+typedef struct cudaMemcpy2DFromArray_v3020_params_st {
+    void *dst;
+    size_t dpitch;
+    cudaArray_const_t src;
+    size_t wOffset;
+    size_t hOffset;
+    size_t width;
+    size_t height;
+    enum cudaMemcpyKind kind;
+} cudaMemcpy2DFromArray_v3020_params;
+
+typedef struct cudaMemcpyArrayToArray_v3020_params_st {
+    cudaArray_t dst;
+    size_t wOffsetDst;
+    size_t hOffsetDst;
+    cudaArray_const_t src;
+    size_t wOffsetSrc;
+    size_t hOffsetSrc;
+    size_t count;
+    enum cudaMemcpyKind kind;
+} cudaMemcpyArrayToArray_v3020_params;
+
+typedef struct cudaMemcpy2DArrayToArray_v3020_params_st {
+    cudaArray_t dst;
+    size_t wOffsetDst;
+    size_t hOffsetDst;
+    cudaArray_const_t src;
+    size_t wOffsetSrc;
+    size_t hOffsetSrc;
+    size_t width;
+    size_t height;
+    enum cudaMemcpyKind kind;
+} cudaMemcpy2DArrayToArray_v3020_params;
+
+typedef struct cudaMemcpy3D_v3020_params_st {
+    const struct cudaMemcpy3DParms *p;
+} cudaMemcpy3D_v3020_params;
+
+typedef struct cudaMemcpy3DPeer_v4000_params_st {
+    const struct cudaMemcpy3DPeerParms *p;
+} cudaMemcpy3DPeer_v4000_params;
+
+typedef struct cudaMemcpyBatchAsync_v12080_params_st {
+    void **dsts;
+    void **srcs;
+    size_t *sizes;
+    size_t count;
+    struct cudaMemcpyAttributes *attrs;
+    size_t *attrsIdxs;
+    size_t numAttrs;
+    size_t *failIdx;
+    cudaStream_t stream;
+} cudaMemcpyBatchAsync_v12080_params;
+
+typedef struct cudaMemcpy3DBatchAsync_v12080_params_st {
+    size_t numOps;
+    struct cudaMemcpy3DBatchOp *opList;
+    size_t *failIdx;
+    unsigned long long flags;
+    cudaStream_t stream;
+} cudaMemcpy3DBatchAsync_v12080_params;
+
+typedef struct cudaMemset_v3020_params_st {
+    void *devPtr;
+    int value;
+    size_t count;
+} cudaMemset_v3020_params;
+
+typedef struct cudaMemset2D_v3020_params_st {
+    void *devPtr;
+    size_t pitch;
+    int value;
+    size_t width;
+    size_t height;
+} cudaMemset2D_v3020_params;
+
+typedef struct cudaMemset3D_v3020_params_st {
+    struct cudaPitchedPtr pitchedDevPtr;
+    int value;
+    struct cudaExtent extent;
+} cudaMemset3D_v3020_params;
+
+typedef struct cudaMemcpyAsync_v3020_params_st {
+    void *dst;
+    const void *src;
+    size_t count;
+    enum cudaMemcpyKind kind;
+    cudaStream_t stream;
+} cudaMemcpyAsync_v3020_params;
+
+typedef struct cudaMemcpyToSymbolAsync_v3020_params_st {
+    const void *symbol;
+    const void *src;
+    size_t count;
+    size_t offset;
+    enum cudaMemcpyKind kind;
+    cudaStream_t stream;
+} cudaMemcpyToSymbolAsync_v3020_params;
+
+typedef struct cudaMemcpyFromSymbolAsync_v3020_params_st {
+    void *dst;
+    const void *symbol;
+    size_t count;
+    size_t offset;
+    enum cudaMemcpyKind kind;
+    cudaStream_t stream;
+} cudaMemcpyFromSymbolAsync_v3020_params;
+
+typedef struct cudaMemcpy2DAsync_v3020_params_st {
+    void *dst;
+    size_t dpitch;
+    const void *src;
+    size_t spitch;
+    size_t width;
+    size_t height;
+    enum cudaMemcpyKind kind;
+    cudaStream_t stream;
+} cudaMemcpy2DAsync_v3020_params;
+
+typedef struct cudaMemcpyToArrayAsync_v3020_params_st {
+    cudaArray_t dst;
+    size_t wOffset;
+    size_t hOffset;
+    const void *src;
+    size_t count;
+    enum cudaMemcpyKind kind;
+    cudaStream_t stream;
+} cudaMemcpyToArrayAsync_v3020_params;
+
+typedef struct cudaMemcpy2DToArrayAsync_v3020_params_st {
+    cudaArray_t dst;
+    size_t wOffset;
+    size_t hOffset;
+    const void *src;
+    size_t spitch;
+    size_t width;
+    size_t height;
+    enum cudaMemcpyKind kind;
+    cudaStream_t stream;
+} cudaMemcpy2DToArrayAsync_v3020_params;
+
+typedef struct cudaMemcpyFromArrayAsync_v3020_params_st {
+    void *dst;
+    cudaArray_const_t src;
+    size_t wOffset;
+    size_t hOffset;
+    size_t count;
+    enum cudaMemcpyKind kind;
+    cudaStream_t stream;
+} cudaMemcpyFromArrayAsync_v3020_params;
+
+typedef struct cudaMemcpy2DFromArrayAsync_v3020_params_st {
+    void *dst;
+    size_t dpitch;
+    cudaArray_const_t src;
+    size_t wOffset;
+    size_t hOffset;
+    size_t width;
+    size_t height;
+    enum cudaMemcpyKind kind;
+    cudaStream_t stream;
+} cudaMemcpy2DFromArrayAsync_v3020_params;
+
+typedef struct cudaMemcpy3DAsync_v3020_params_st {
+    const struct cudaMemcpy3DParms *p;
+    cudaStream_t stream;
+} cudaMemcpy3DAsync_v3020_params;
+
+typedef struct cudaMemcpy3DPeerAsync_v4000_params_st {
+    const struct cudaMemcpy3DPeerParms *p;
+    cudaStream_t stream;
+} cudaMemcpy3DPeerAsync_v4000_params;
+
+typedef struct cudaMemsetAsync_v3020_params_st {
+    void *devPtr;
+    int value;
+    size_t count;
+    cudaStream_t stream;
+} cudaMemsetAsync_v3020_params;
+
+typedef struct cudaMemset2DAsync_v3020_params_st {
+    void *devPtr;
+    size_t pitch;
+    int value;
+    size_t width;
+    size_t height;
+    cudaStream_t stream;
+} cudaMemset2DAsync_v3020_params;
+
+typedef struct cudaMemset3DAsync_v3020_params_st {
+    struct cudaPitchedPtr pitchedDevPtr;
+    int value;
+    struct cudaExtent extent;
+    cudaStream_t stream;
+} cudaMemset3DAsync_v3020_params;
+
+typedef struct cudaStreamQuery_v3020_params_st {
+    cudaStream_t stream;
+} cudaStreamQuery_v3020_params;
+
+typedef struct cudaStreamGetDevice_v12080_params_st {
+    cudaStream_t hStream;
+    int *device;
+} cudaStreamGetDevice_v12080_params;
+
+typedef struct cudaStreamGetFlags_v5050_params_st {
+    cudaStream_t hStream;
+    unsigned int *flags;
+} cudaStreamGetFlags_v5050_params;
+
+typedef struct cudaStreamGetId_v12000_params_st {
+    cudaStream_t hStream;
+    unsigned long long *streamId;
+} cudaStreamGetId_v12000_params;
+
+typedef struct cudaStreamGetPriority_v5050_params_st {
+    cudaStream_t hStream;
+    int *priority;
+} cudaStreamGetPriority_v5050_params;
+
+typedef struct cudaEventRecord_v3020_params_st {
+    cudaEvent_t event;
+    cudaStream_t stream;
+} cudaEventRecord_v3020_params;
+
+typedef struct cudaEventRecordWithFlags_v11010_params_st {
+    cudaEvent_t event;
+    cudaStream_t stream;
+    unsigned int flags;
+} cudaEventRecordWithFlags_v11010_params;
+
+typedef struct cudaStreamWaitEvent_v3020_params_st {
+    cudaStream_t stream;
+    cudaEvent_t event;
+    unsigned int flags;
+} cudaStreamWaitEvent_v3020_params;
+
+typedef struct cudaStreamAddCallback_v5000_params_st {
+    cudaStream_t stream;
+    cudaStreamCallback_t callback;
+    void *userData;
+    unsigned int flags;
+} cudaStreamAddCallback_v5000_params;
+
+typedef struct cudaStreamAttachMemAsync_v6000_params_st {
+    cudaStream_t stream;
+    void *devPtr;
+    size_t length;
+    unsigned int flags;
+} cudaStreamAttachMemAsync_v6000_params;
+
+typedef struct cudaStreamSynchronize_v3020_params_st {
+    cudaStream_t stream;
+} cudaStreamSynchronize_v3020_params;
+
+typedef struct cudaLaunchKernel_v7000_params_st {
+    const void *func;
+    dim3 gridDim;
+    dim3 blockDim;
+    void **args;
+    size_t sharedMem;
+    cudaStream_t stream;
+} cudaLaunchKernel_v7000_params;
+
+typedef struct cudaLaunchKernelExC_v11060_params_st {
+    const cudaLaunchConfig_t *config;
+    const void *func;
+    void **args;
+} cudaLaunchKernelExC_v11060_params;
+
+typedef struct cudaLaunchCooperativeKernel_v9000_params_st {
+    const void *func;
+    dim3 gridDim;
+    dim3 blockDim;
+    void **args;
+    size_t sharedMem;
+    cudaStream_t stream;
+} cudaLaunchCooperativeKernel_v9000_params;
+
+typedef struct cudaLaunchHostFunc_v10000_params_st {
+    cudaStream_t stream;
+    cudaHostFn_t fn;
+    void *userData;
+} cudaLaunchHostFunc_v10000_params;
+
+typedef struct cudaMemPrefetchAsync_v8000_params_st {
+    const void *devPtr;
+    size_t count;
+    int dstDevice;
+    cudaStream_t stream;
+} cudaMemPrefetchAsync_v8000_params;
+
+typedef struct cudaMemPrefetchAsync_v2_v12020_params_st {
+    const void *devPtr;
+    size_t count;
+    struct cudaMemLocation location;
+    unsigned int flags;
+    cudaStream_t stream;
+} cudaMemPrefetchAsync_v2_v12020_params;
+
+typedef struct cudaSignalExternalSemaphoresAsync_v10000_params_st {
+    const cudaExternalSemaphore_t *extSemArray;
+    const struct cudaExternalSemaphoreSignalParams_v1 *paramsArray;
+    unsigned int numExtSems;
+    cudaStream_t stream;
+} cudaSignalExternalSemaphoresAsync_v10000_params;
+
+typedef struct cudaSignalExternalSemaphoresAsync_ptsz_v10000_params_st {
+    const cudaExternalSemaphore_t *extSemArray;
+    const struct cudaExternalSemaphoreSignalParams_v1 *paramsArray;
+    unsigned int numExtSems;
+    cudaStream_t stream;
+} cudaSignalExternalSemaphoresAsync_ptsz_v10000_params;
+
+typedef struct cudaSignalExternalSemaphoresAsync_v2_v11020_params_st {
+    const cudaExternalSemaphore_t *extSemArray;
+    const struct cudaExternalSemaphoreSignalParams *paramsArray;
+    unsigned int numExtSems;
+    cudaStream_t stream;
+} cudaSignalExternalSemaphoresAsync_v2_v11020_params;
+
+typedef struct cudaWaitExternalSemaphoresAsync_v10000_params_st {
+    const cudaExternalSemaphore_t *extSemArray;
+    const struct cudaExternalSemaphoreWaitParams_v1 *paramsArray;
+    unsigned int numExtSems;
+    cudaStream_t stream;
+} cudaWaitExternalSemaphoresAsync_v10000_params;
+
+typedef struct cudaWaitExternalSemaphoresAsync_ptsz_v10000_params_st {
+    const cudaExternalSemaphore_t *extSemArray;
+    const struct cudaExternalSemaphoreWaitParams_v1 *paramsArray;
+    unsigned int numExtSems;
+    cudaStream_t stream;
+} cudaWaitExternalSemaphoresAsync_ptsz_v10000_params;
+
+typedef struct cudaWaitExternalSemaphoresAsync_v2_v11020_params_st {
+    const cudaExternalSemaphore_t *extSemArray;
+    const struct cudaExternalSemaphoreWaitParams *paramsArray;
+    unsigned int numExtSems;
+    cudaStream_t stream;
+} cudaWaitExternalSemaphoresAsync_v2_v11020_params;
+
+typedef struct cudaGraphInstantiateWithParams_v12000_params_st {
+    cudaGraphExec_t *pGraphExec;
+    cudaGraph_t graph;
+    cudaGraphInstantiateParams *instantiateParams;
+} cudaGraphInstantiateWithParams_v12000_params;
+
+typedef struct cudaGraphUpload_v10000_params_st {
+    cudaGraphExec_t graphExec;
+    cudaStream_t stream;
+} cudaGraphUpload_v10000_params;
+
+typedef struct cudaGraphLaunch_v10000_params_st {
+    cudaGraphExec_t graphExec;
+    cudaStream_t stream;
+} cudaGraphLaunch_v10000_params;
+
+typedef struct cudaStreamBeginCapture_v10000_params_st {
+    cudaStream_t stream;
+    enum cudaStreamCaptureMode mode;
+} cudaStreamBeginCapture_v10000_params;
+
+typedef struct cudaStreamBeginCaptureToGraph_v12030_params_st {
+    cudaStream_t stream;
+    cudaGraph_t graph;
+    const cudaGraphNode_t *dependencies;
+    const cudaGraphEdgeData *dependencyData;
+    size_t numDependencies;
+    enum cudaStreamCaptureMode mode;
+} cudaStreamBeginCaptureToGraph_v12030_params;
+
+typedef struct cudaStreamEndCapture_v10000_params_st {
+    cudaStream_t stream;
+    cudaGraph_t *pGraph;
+} cudaStreamEndCapture_v10000_params;
+
+typedef struct cudaStreamIsCapturing_v10000_params_st {
+    cudaStream_t stream;
+    enum cudaStreamCaptureStatus *pCaptureStatus;
+} cudaStreamIsCapturing_v10000_params;
+
+typedef struct cudaStreamGetCaptureInfo_v10010_params_st {
+    cudaStream_t stream;
+    enum cudaStreamCaptureStatus *captureStatus_out;
+    unsigned long long *id_out;
+} cudaStreamGetCaptureInfo_v10010_params;
+
+typedef struct cudaStreamGetCaptureInfo_ptsz_v10010_params_st {
+    cudaStream_t stream;
+    enum cudaStreamCaptureStatus *captureStatus_out;
+    unsigned long long *id_out;
+} cudaStreamGetCaptureInfo_ptsz_v10010_params;
+
+typedef struct cudaStreamGetCaptureInfo_v2_v11030_params_st {
+    cudaStream_t stream;
+    enum cudaStreamCaptureStatus *captureStatus_out;
+    unsigned long long *id_out;
+    cudaGraph_t *graph_out;
+    const cudaGraphNode_t **dependencies_out;
+    size_t *numDependencies_out;
+} cudaStreamGetCaptureInfo_v2_v11030_params;
+
+typedef struct cudaStreamGetCaptureInfo_v3_v12030_params_st {
+    cudaStream_t stream;
+    enum cudaStreamCaptureStatus *captureStatus_out;
+    unsigned long long *id_out;
+    cudaGraph_t *graph_out;
+    const cudaGraphNode_t **dependencies_out;
+    const cudaGraphEdgeData **edgeData_out;
+    size_t *numDependencies_out;
+} cudaStreamGetCaptureInfo_v3_v12030_params;
+
+typedef struct cudaStreamUpdateCaptureDependencies_v11030_params_st {
+    cudaStream_t stream;
+    cudaGraphNode_t *dependencies;
+    size_t numDependencies;
+    unsigned int flags;
+} cudaStreamUpdateCaptureDependencies_v11030_params;
+
+typedef struct cudaStreamUpdateCaptureDependencies_v2_v12030_params_st {
+    cudaStream_t stream;
+    cudaGraphNode_t *dependencies;
+    const cudaGraphEdgeData *dependencyData;
+    size_t numDependencies;
+    unsigned int flags;
+} cudaStreamUpdateCaptureDependencies_v2_v12030_params;
+
+typedef struct cudaStreamCopyAttributes_v11000_params_st {
+    cudaStream_t dstStream;
+    cudaStream_t srcStream;
+} cudaStreamCopyAttributes_v11000_params;
+
+typedef struct cudaStreamGetAttribute_v11000_params_st {
+    cudaStream_t stream;
+    cudaStreamAttrID attr;
+    cudaStreamAttrValue *value;
+} cudaStreamGetAttribute_v11000_params;
+
+typedef struct cudaStreamSetAttribute_v11000_params_st {
+    cudaStream_t stream;
+    cudaStreamAttrID attr;
+    const cudaStreamAttrValue *param;
+} cudaStreamSetAttribute_v11000_params;
+
+typedef struct cudaMallocAsync_v11020_params_st {
+    void **devPtr;
+    size_t size;
+    cudaStream_t hStream;
+} cudaMallocAsync_v11020_params;
+
+typedef struct cudaFreeAsync_v11020_params_st {
+    void *devPtr;
+    cudaStream_t hStream;
+} cudaFreeAsync_v11020_params;
+
+typedef struct cudaMallocFromPoolAsync_v11020_params_st {
+    void **ptr;
+    size_t size;
+    cudaMemPool_t memPool;
+    cudaStream_t stream;
+} cudaMallocFromPoolAsync_v11020_params;
+
+typedef struct cudaGetDriverEntryPoint_v11030_params_st {
+    const char *symbol;
+    void **funcPtr;
+    unsigned long long flags;
+    enum cudaDriverEntryPointQueryResult *driverStatus;
+} cudaGetDriverEntryPoint_v11030_params;
+
+typedef struct cudaGetDriverEntryPointByVersion_v12050_params_st {
+    const char *symbol;
+    void **funcPtr;
+    unsigned int cudaVersion;
+    unsigned long long flags;
+    enum cudaDriverEntryPointQueryResult *driverStatus;
+} cudaGetDriverEntryPointByVersion_v12050_params;
+
+typedef struct cudaGetDeviceProperties_v3020_params_st {
+    struct cudaDeviceProp *prop;
+    int device;
+} cudaGetDeviceProperties_v3020_params;
+
+// Parameter trace structures for removed functions
+
+
+// End of parameter trace structures

Prism/LLaDA/LLaDA_Prism/.venv/lib/python3.12/site-packages/nvidia/cuda_cupti/include/generated_cuda_vdpau_interop_meta.h

@@ -0,0 +1,38 @@
+// This file is generated.  Any changes you make will be lost during the next clean build.
+
+// CUDA public interface, for type definitions and api function prototypes
+#include "cuda_vdpau_interop.h"
+
+// *************************************************************************
+//      Definitions of structs to hold parameters for each function
+// *************************************************************************
+
+// Currently used parameter trace structures
+typedef struct cudaVDPAUGetDevice_v3020_params_st {
+    int *device;
+    VdpDevice vdpDevice;
+    VdpGetProcAddress *vdpGetProcAddress;
+} cudaVDPAUGetDevice_v3020_params;
+
+typedef struct cudaVDPAUSetVDPAUDevice_v3020_params_st {
+    int device;
+    VdpDevice vdpDevice;
+    VdpGetProcAddress *vdpGetProcAddress;
+} cudaVDPAUSetVDPAUDevice_v3020_params;
+
+typedef struct cudaGraphicsVDPAURegisterVideoSurface_v3020_params_st {
+    struct cudaGraphicsResource **resource;
+    VdpVideoSurface vdpSurface;
+    unsigned int flags;
+} cudaGraphicsVDPAURegisterVideoSurface_v3020_params;
+
+typedef struct cudaGraphicsVDPAURegisterOutputSurface_v3020_params_st {
+    struct cudaGraphicsResource **resource;
+    VdpOutputSurface vdpSurface;
+    unsigned int flags;
+} cudaGraphicsVDPAURegisterOutputSurface_v3020_params;
+
+// Parameter trace structures for removed functions
+
+
+// End of parameter trace structures

Prism/LLaDA/LLaDA_Prism/.venv/lib/python3.12/site-packages/nvidia/cuda_cupti/include/generated_cudart_removed_meta.h

@@ -0,0 +1,162 @@
+// This file is generated.  Any changes you make will be lost during the next clean build.
+
+// CUDA public interface, for type definitions and api function prototypes
+#include "cudart_removed.h"
+
+// *************************************************************************
+//      Definitions of structs to hold parameters for each function
+// *************************************************************************
+
+// Currently used parameter trace structures
+typedef struct cudaStreamDestroy_v3020_params_st {
+    cudaStream_t stream;
+} cudaStreamDestroy_v3020_params;
+
+typedef struct cudaOccupancyMaxActiveBlocksPerMultiprocessor_v6000_params_st {
+    int *numBlocks;
+    const void *func;
+    size_t numDynamicSmemBytes;
+} cudaOccupancyMaxActiveBlocksPerMultiprocessor_v6000_params;
+
+typedef struct cudaConfigureCall_v3020_params_st {
+    dim3 gridDim;
+    dim3 blockDim;
+    size_t sharedMem  __dv;
+    cudaStream_t stream  __dv;
+} cudaConfigureCall_v3020_params;
+
+typedef struct cudaSetupArgument_v3020_params_st {
+    const void *arg;
+    size_t size;
+    size_t offset;
+} cudaSetupArgument_v3020_params;
+
+typedef struct cudaLaunch_v3020_params_st {
+    const void *func;
+} cudaLaunch_v3020_params;
+
+typedef struct cudaLaunch_ptsz_v7000_params_st {
+    const void *func;
+} cudaLaunch_ptsz_v7000_params;
+
+typedef struct cudaStreamSetFlags_v10200_params_st {
+    cudaStream_t hStream;
+    unsigned int flags;
+} cudaStreamSetFlags_v10200_params;
+
+typedef struct cudaStreamSetFlags_ptsz_v10200_params_st {
+    cudaStream_t hStream;
+    unsigned int flags;
+} cudaStreamSetFlags_ptsz_v10200_params;
+
+typedef struct cudaProfilerInitialize_v4000_params_st {
+    const char *configFile;
+    const char *outputFile;
+    cudaOutputMode_t outputMode;
+} cudaProfilerInitialize_v4000_params;
+
+typedef struct cudaThreadSetLimit_v3020_params_st {
+    enum cudaLimit limit;
+    size_t value;
+} cudaThreadSetLimit_v3020_params;
+
+typedef struct cudaThreadGetLimit_v3020_params_st {
+    size_t *pValue;
+    enum cudaLimit limit;
+} cudaThreadGetLimit_v3020_params;
+
+typedef struct cudaThreadGetCacheConfig_v3020_params_st {
+    enum cudaFuncCache *pCacheConfig;
+} cudaThreadGetCacheConfig_v3020_params;
+
+typedef struct cudaThreadSetCacheConfig_v3020_params_st {
+    enum cudaFuncCache cacheConfig;
+} cudaThreadSetCacheConfig_v3020_params;
+
+typedef struct cudaSetDoubleForDevice_v3020_params_st {
+    double *d;
+} cudaSetDoubleForDevice_v3020_params;
+
+typedef struct cudaSetDoubleForHost_v3020_params_st {
+    double *d;
+} cudaSetDoubleForHost_v3020_params;
+
+typedef struct cudaCreateTextureObject_v2_v11080_params_st {
+    cudaTextureObject_t *pTexObject;
+    const struct cudaResourceDesc *pResDesc;
+    const struct cudaTextureDesc *pTexDesc;
+    const struct cudaResourceViewDesc *pResViewDesc;
+} cudaCreateTextureObject_v2_v11080_params;
+
+typedef struct cudaGetTextureObjectTextureDesc_v2_v11080_params_st {
+    struct cudaTextureDesc *pTexDesc;
+    cudaTextureObject_t texObject;
+} cudaGetTextureObjectTextureDesc_v2_v11080_params;
+
+typedef struct cudaBindTexture_v3020_params_st {
+    size_t *offset;
+    const struct textureReference *texref;
+    const void *devPtr;
+    const struct cudaChannelFormatDesc *desc;
+    size_t size  __dv;
+} cudaBindTexture_v3020_params;
+
+typedef struct cudaBindTexture2D_v3020_params_st {
+    size_t *offset;
+    const struct textureReference *texref;
+    const void *devPtr;
+    const struct cudaChannelFormatDesc *desc;
+    size_t width;
+    size_t height;
+    size_t pitch;
+} cudaBindTexture2D_v3020_params;
+
+typedef struct cudaBindTextureToArray_v3020_params_st {
+    const struct textureReference *texref;
+    cudaArray_const_t array;
+    const struct cudaChannelFormatDesc *desc;
+} cudaBindTextureToArray_v3020_params;
+
+typedef struct cudaBindTextureToMipmappedArray_v5000_params_st {
+    const struct textureReference *texref;
+    cudaMipmappedArray_const_t mipmappedArray;
+    const struct cudaChannelFormatDesc *desc;
+} cudaBindTextureToMipmappedArray_v5000_params;
+
+typedef struct cudaUnbindTexture_v3020_params_st {
+    const struct textureReference *texref;
+} cudaUnbindTexture_v3020_params;
+
+typedef struct cudaGetTextureAlignmentOffset_v3020_params_st {
+    size_t *offset;
+    const struct textureReference *texref;
+} cudaGetTextureAlignmentOffset_v3020_params;
+
+typedef struct cudaGetTextureReference_v3020_params_st {
+    const struct textureReference **texref;
+    const void *symbol;
+} cudaGetTextureReference_v3020_params;
+
+typedef struct cudaBindSurfaceToArray_v3020_params_st {
+    const struct surfaceReference *surfref;
+    cudaArray_const_t array;
+    const struct cudaChannelFormatDesc *desc;
+} cudaBindSurfaceToArray_v3020_params;
+
+typedef struct cudaGetSurfaceReference_v3020_params_st {
+    const struct surfaceReference **surfref;
+    const void *symbol;
+} cudaGetSurfaceReference_v3020_params;
+
+typedef struct cudaGraphInstantiate_v10000_params_st {
+    cudaGraphExec_t *pGraphExec;
+    cudaGraph_t graph;
+    cudaGraphNode_t *pErrorNode;
+    char *pLogBuffer;
+    size_t bufferSize;
+} cudaGraphInstantiate_v10000_params;
+
+// Parameter trace structures for removed functions
+
+
+// End of parameter trace structures

Prism/LLaDA/LLaDA_Prism/.venv/lib/python3.12/site-packages/nvidia/cuda_cupti/include/generated_nvtx_meta.h

@@ -0,0 +1,247 @@
+/*
+ * Copyright 2013-2018 NVIDIA Corporation.  All rights reserved.
+ *
+ * NOTICE TO LICENSEE:
+ *
+ * This source code and/or documentation ("Licensed Deliverables") are
+ * subject to NVIDIA intellectual property rights under U.S. and
+ * international Copyright laws.
+ *
+ * These Licensed Deliverables contained herein is PROPRIETARY and
+ * CONFIDENTIAL to NVIDIA and is being provided under the terms and
+ * conditions of a form of NVIDIA software license agreement by and
+ * between NVIDIA and Licensee ("License Agreement") or electronically
+ * accepted by Licensee.  Notwithstanding any terms or conditions to
+ * the contrary in the License Agreement, reproduction or disclosure
+ * of the Licensed Deliverables to any third party without the express
+ * written consent of NVIDIA is prohibited.
+ *
+ * NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE
+ * LICENSE AGREEMENT, NVIDIA MAKES NO REPRESENTATION ABOUT THE
+ * SUITABILITY OF THESE LICENSED DELIVERABLES FOR ANY PURPOSE.  IT IS
+ * PROVIDED "AS IS" WITHOUT EXPRESS OR IMPLIED WARRANTY OF ANY KIND.
+ * NVIDIA DISCLAIMS ALL WARRANTIES WITH REGARD TO THESE LICENSED
+ * DELIVERABLES, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY,
+ * NONINFRINGEMENT, AND FITNESS FOR A PARTICULAR PURPOSE.
+ * NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE
+ * LICENSE AGREEMENT, IN NO EVENT SHALL NVIDIA BE LIABLE FOR ANY
+ * SPECIAL, INDIRECT, INCIDENTAL, OR CONSEQUENTIAL DAMAGES, OR ANY
+ * DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS,
+ * WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS
+ * ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE
+ * OF THESE LICENSED DELIVERABLES.
+ *
+ * U.S. Government End Users.  These Licensed Deliverables are a
+ * "commercial item" as that term is defined at 48 C.F.R. 2.101 (OCT
+ * 1995), consisting of "commercial computer software" and "commercial
+ * computer software documentation" as such terms are used in 48
+ * C.F.R. 12.212 (SEPT 1995) and is provided to the U.S. Government
+ * only as a commercial end item.  Consistent with 48 C.F.R.12.212 and
+ * 48 C.F.R. 227.7202-1 through 227.7202-4 (JUNE 1995), all
+ * U.S. Government End Users acquire the Licensed Deliverables with
+ * only those rights set forth herein.
+ *
+ * Any use of the Licensed Deliverables in individual and commercial
+ * software must include, in the user documentation and internal
+ * comments to the code, the above Disclaimer and U.S. Government End
+ * Users Notice.
+ */
+
+#if defined(__GNUC__) && defined(CUPTI_LIB)
+    #pragma GCC visibility push(default)
+#endif
+
+// *************************************************************************
+//      Definitions of structs to hold parameters for each function
+// *************************************************************************
+
+typedef struct nvtxMarkEx_params_st {
+  const nvtxEventAttributes_t* eventAttrib;
+} nvtxMarkEx_params;
+
+typedef struct nvtxMarkA_params_st {
+  const char* message;
+} nvtxMarkA_params;
+
+typedef struct nvtxMarkW_params_st {
+  const wchar_t* message;
+} nvtxMarkW_params;
+
+typedef struct nvtxRangeStartEx_params_st {
+  const nvtxEventAttributes_t* eventAttrib;
+} nvtxRangeStartEx_params;
+
+typedef struct nvtxRangeStartA_params_st {
+  const char* message;
+} nvtxRangeStartA_params;
+
+typedef struct nvtxRangeStartW_params_st {
+  const wchar_t* message;
+} nvtxRangeStartW_params;
+
+typedef struct nvtxRangeEnd_params_st {
+  nvtxRangeId_t id;
+} nvtxRangeEnd_params;
+
+typedef struct nvtxRangePushEx_params_st {
+  const nvtxEventAttributes_t* eventAttrib;
+} nvtxRangePushEx_params;
+
+typedef struct nvtxRangePushA_params_st {
+  const char* message;
+} nvtxRangePushA_params;
+
+typedef struct nvtxRangePushW_params_st {
+  const wchar_t* message;
+} nvtxRangePushW_params;
+
+typedef struct nvtxRangePop_params_st {
+  /* WAR: Windows compiler doesn't allow empty structs */
+  /* This field shouldn't be used */
+  void *dummy;
+} nvtxRangePop_params;
+
+typedef struct nvtxNameCategoryA_params_st {
+  uint32_t category;
+  const char* name;
+} nvtxNameCategoryA_params;
+
+typedef struct nvtxNameCategoryW_params_st {
+  uint32_t category;
+  const wchar_t* name;
+} nvtxNameCategoryW_params;
+
+typedef struct nvtxNameOsThreadA_params_st {
+  uint32_t threadId;
+  const char* name;
+} nvtxNameOsThreadA_params;
+
+typedef struct nvtxNameOsThreadW_params_st {
+  uint32_t threadId;
+  const wchar_t* name;
+} nvtxNameOsThreadW_params;
+
+typedef struct nvtxNameCuDeviceA_params_st {
+  CUdevice device;
+  const char* name;
+} nvtxNameCuDeviceA_params;
+
+typedef struct nvtxNameCuDeviceW_params_st {
+  CUdevice device;
+  const wchar_t* name;
+} nvtxNameCuDeviceW_params;
+
+typedef struct nvtxNameCuContextA_params_st {
+  CUcontext context;
+  const char* name;
+} nvtxNameCuContextA_params;
+
+typedef struct nvtxNameCuContextW_params_st {
+  CUcontext context;
+  const wchar_t* name;
+} nvtxNameCuContextW_params;
+
+typedef struct nvtxNameCuStreamA_params_st {
+  CUstream stream;
+  const char* name;
+} nvtxNameCuStreamA_params;
+
+typedef struct nvtxNameCuStreamW_params_st {
+  CUstream stream;
+  const wchar_t* name;
+} nvtxNameCuStreamW_params;
+
+typedef struct nvtxNameCuEventA_params_st {
+  CUevent event;
+  const char* name;
+} nvtxNameCuEventA_params;
+
+typedef struct nvtxNameCuEventW_params_st {
+  CUevent event;
+  const wchar_t* name;
+} nvtxNameCuEventW_params;
+
+typedef struct nvtxNameCudaDeviceA_params_st {
+  int device;
+  const char* name;
+} nvtxNameCudaDeviceA_params;
+
+typedef struct nvtxNameCudaDeviceW_params_st {
+  int device;
+  const wchar_t* name;
+} nvtxNameCudaDeviceW_params;
+
+typedef struct nvtxNameCudaStreamA_params_st {
+  cudaStream_t stream;
+  const char* name;
+} nvtxNameCudaStreamA_params;
+
+typedef struct nvtxNameCudaStreamW_params_st {
+  cudaStream_t stream;
+  const wchar_t* name;
+} nvtxNameCudaStreamW_params;
+
+typedef struct nvtxNameCudaEventA_params_st {
+  cudaEvent_t event;
+  const char* name;
+} nvtxNameCudaEventA_params;
+
+typedef struct nvtxNameCudaEventW_params_st {
+  cudaEvent_t event;
+  const wchar_t* name;
+} nvtxNameCudaEventW_params;
+
+typedef struct nvtxDomainCreateA_params_st {
+  const char* name;
+} nvtxDomainCreateA_params;
+
+typedef struct nvtxDomainDestroy_params_st {
+  nvtxDomainHandle_t domain;
+} nvtxDomainDestroy_params;
+
+typedef struct nvtxDomainMarkEx_params_st {
+  nvtxDomainHandle_t domain;
+  nvtxMarkEx_params core;
+} nvtxDomainMarkEx_params;
+
+typedef struct nvtxDomainRangeStartEx_params_st {
+  nvtxDomainHandle_t domain;
+  nvtxRangeStartEx_params core;
+} nvtxDomainRangeStartEx_params;
+
+typedef struct nvtxDomainRangeEnd_params_st {
+  nvtxDomainHandle_t domain;
+  nvtxRangeEnd_params core;
+} nvtxDomainRangeEnd_params;
+
+typedef struct nvtxDomainRangePushEx_params_st {
+  nvtxDomainHandle_t domain;
+  nvtxRangePushEx_params core;
+} nvtxDomainRangePushEx_params;
+
+typedef struct nvtxDomainRangePop_params_st {
+  nvtxDomainHandle_t domain;
+} nvtxDomainRangePop_params;
+
+typedef struct nvtxSyncUserCreate_params_st {
+  nvtxDomainHandle_t domain;
+  const nvtxSyncUserAttributes_t* attribs;
+} nvtxSyncUserCreate_params;
+
+typedef struct nvtxSyncUserCommon_params_st {
+  nvtxSyncUser_t handle;
+} nvtxSyncUserCommon_params;
+
+typedef struct nvtxDomainRegisterStringA_params_st {
+    nvtxDomainHandle_t domain;
+    const char* string;
+} nvtxDomainRegisterStringA_params;
+
+typedef struct nvtxDomainRegisterStringW_params_st {
+    nvtxDomainHandle_t domain;
+    const char* string;
+} nvtxDomainRegisterStringW_params;
+
+#if defined(__GNUC__) && defined(CUPTI_LIB)
+    #pragma GCC visibility pop
+#endif

Prism/LLaDA/LLaDA_Prism/.venv/lib/python3.12/site-packages/nvidia/cuda_cupti/include/nvperf_common.h

@@ -0,0 +1,393 @@
+#ifndef NVPERF_COMMON_H
+#define NVPERF_COMMON_H
+
+/*
+ * Copyright 2014-2024 NVIDIA Corporation.  All rights reserved.
+ *
+ * NOTICE TO USER:
+ *
+ * This source code is subject to NVIDIA ownership rights under U.S. and
+ * international Copyright laws.
+ *
+ * This software and the information contained herein is PROPRIETARY and
+ * CONFIDENTIAL to NVIDIA and is being provided under the terms and conditions
+ * of a form of NVIDIA software license agreement.
+ *
+ * NVIDIA MAKES NO REPRESENTATION ABOUT THE SUITABILITY OF THIS SOURCE
+ * CODE FOR ANY PURPOSE.  IT IS PROVIDED "AS IS" WITHOUT EXPRESS OR
+ * IMPLIED WARRANTY OF ANY KIND.  NVIDIA DISCLAIMS ALL WARRANTIES WITH
+ * REGARD TO THIS SOURCE CODE, INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY, NONINFRINGEMENT, AND FITNESS FOR A PARTICULAR PURPOSE.
+ * IN NO EVENT SHALL NVIDIA BE LIABLE FOR ANY SPECIAL, INDIRECT, INCIDENTAL,
+ * OR CONSEQUENTIAL DAMAGES, OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS
+ * OF USE, DATA OR PROFITS,  WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE
+ * OR OTHER TORTIOUS ACTION,  ARISING OUT OF OR IN CONNECTION WITH THE USE
+ * OR PERFORMANCE OF THIS SOURCE CODE.
+ *
+ * U.S. Government End Users.   This source code is a "commercial item" as
+ * that term is defined at  48 C.F.R. 2.101 (OCT 1995), consisting  of
+ * "commercial computer  software"  and "commercial computer software
+ * documentation" as such terms are  used in 48 C.F.R. 12.212 (SEPT 1995)
+ * and is provided to the U.S. Government only as a commercial end item.
+ * Consistent with 48 C.F.R.12.212 and 48 C.F.R. 227.7202-1 through
+ * 227.7202-4 (JUNE 1995), all U.S. Government End Users acquire the
+ * source code with only those rights set forth herein.
+ *
+ * Any use of this source code in individual and commercial software must
+ * include, in the user documentation and internal comments to the code,
+ * the above Disclaimer and U.S. Government End Users Notice.
+ */
+
+#include <stddef.h>
+#include <stdint.h>
+
+#if defined(__GNUC__) && defined(NVPA_SHARED_LIB)
+    #pragma GCC visibility push(default)
+    #if !defined(NVPW_LOCAL)
+        #define NVPW_LOCAL __attribute__ ((visibility ("hidden")))
+    #endif
+#else
+    #if !defined(NVPW_LOCAL)
+        #define NVPW_LOCAL
+    #endif
+#endif
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/**
+ *  @file   nvperf_common.h
+ */
+
+#ifndef NVPERF_NVPA_STATUS_DEFINED
+#define NVPERF_NVPA_STATUS_DEFINED
+
+    /// Error codes.
+    typedef enum NVPA_Status
+    {
+        /// Success
+        NVPA_STATUS_SUCCESS = 0,
+        /// Generic error.
+        NVPA_STATUS_ERROR = 1,
+        /// Internal error.  Please file a bug!
+        NVPA_STATUS_INTERNAL_ERROR = 2,
+        /// NVPW_InitializeTarget() or NVPW_InitializeHost() has not been called yet.
+        NVPA_STATUS_NOT_INITIALIZED = 3,
+        /// The NvPerf DLL/DSO could not be loaded during NVPW_Initialize*(). Please ensure they are placed in the
+        /// appropriate location that can be founder by a dynamic linker. And on Linux systems, confirm that the
+        /// LD_LIBRARY_PATH environment variable is set correctly. Alternatively, you may utilize
+        /// NVPW_SetLibraryLoadPaths() to define additional library search paths.
+        NVPA_STATUS_NOT_LOADED = 4,
+        /// The function was not found in this version of the NvPerf DLL/DSO. Or if you are directly calling
+        /// NVPA_GetProcAddress(), please ensure the function name is spelled correctly.
+        NVPA_STATUS_FUNCTION_NOT_FOUND = 5,
+        /// The request was intentionally not supported.
+        NVPA_STATUS_NOT_SUPPORTED = 6,
+        /// The request was not implemented by this version.
+        NVPA_STATUS_NOT_IMPLEMENTED = 7,
+        /// Invalid argument.
+        NVPA_STATUS_INVALID_ARGUMENT = 8,
+        /// UNUSED
+        NVPA_STATUS_INVALID_METRIC_ID = 9,
+        /// No driver has been loaded via NVPW_*_LoadDriver().
+        NVPA_STATUS_DRIVER_NOT_LOADED = 10,
+        /// Failed memory allocation.
+        NVPA_STATUS_OUT_OF_MEMORY = 11,
+        /// UNUSED
+        NVPA_STATUS_INVALID_THREAD_STATE = 12,
+        /// UNUSED
+        NVPA_STATUS_FAILED_CONTEXT_ALLOC = 13,
+        /// The specified GPU is not supported. It is recommended to call IsGpuSupported() for more information
+        NVPA_STATUS_UNSUPPORTED_GPU = 14,
+        /// The installed NVIDIA driver is too old.
+        NVPA_STATUS_INSUFFICIENT_DRIVER_VERSION = 15,
+        /// UNUSED
+        NVPA_STATUS_OBJECT_NOT_REGISTERED = 16,
+        /// Profiling permission not granted; see https://developer.nvidia.com/nvidia-development-tools-solutions-
+        /// ERR_NVGPUCTRPERM-permission-issue-performance-counters
+        NVPA_STATUS_INSUFFICIENT_PRIVILEGE = 17,
+        /// UNUSED
+        NVPA_STATUS_INVALID_CONTEXT_STATE = 18,
+        /// UNUSED
+        NVPA_STATUS_INVALID_OBJECT_STATE = 19,
+        /// The request could not be fulfilled because a system resource is already in use.
+        NVPA_STATUS_RESOURCE_UNAVAILABLE = 20,
+        /// UNUSED
+        NVPA_STATUS_DRIVER_LOADED_TOO_LATE = 21,
+        /// The provided buffer is not large enough.
+        NVPA_STATUS_INSUFFICIENT_SPACE = 22,
+        /// UNUSED
+        NVPA_STATUS_OBJECT_MISMATCH = 23,
+        /// Virtualized GPU (vGPU) is not supported.
+        NVPA_STATUS_VIRTUALIZED_DEVICE_NOT_SUPPORTED = 24,
+        /// Profiling permission was not granted or the device was disabled.
+        NVPA_STATUS_PROFILING_NOT_ALLOWED = 25,
+        NVPA_STATUS__COUNT
+    } NVPA_Status;
+
+
+    inline void NVPW_NVPAStatusToString(NVPA_Status status, const char** ppStatusStr, const char** ppCommentStr)
+    {
+        switch (status)
+        {
+            case NVPA_STATUS_SUCCESS:
+                *ppStatusStr = "NVPA_STATUS_SUCCESS";
+                *ppCommentStr = "Success";
+                return;
+            case NVPA_STATUS_ERROR:
+                *ppStatusStr = "NVPA_STATUS_ERROR";
+                *ppCommentStr = "Generic error.";
+                return;
+            case NVPA_STATUS_INTERNAL_ERROR:
+                *ppStatusStr = "NVPA_STATUS_INTERNAL_ERROR";
+                *ppCommentStr = "Internal error.  Please file a bug!";
+                return;
+            case NVPA_STATUS_NOT_INITIALIZED:
+                *ppStatusStr = "NVPA_STATUS_NOT_INITIALIZED";
+                *ppCommentStr = "NVPW_InitializeTarget() or NVPW_InitializeHost() has not been called yet.";
+                return;
+            case NVPA_STATUS_NOT_LOADED:
+                *ppStatusStr = "NVPA_STATUS_NOT_LOADED";
+                *ppCommentStr = "The NvPerf DLL/DSO could not be loaded during NVPW_Initialize*(). Please ensure they are placed in the appropriate location that can be founder by a dynamic linker. And on Linux systems, confirm that the LD_LIBRARY_PATH environment variable is set correctly. Alternatively, you may utilize NVPW_SetLibraryLoadPaths() to define additional library search paths.";
+                return;
+            case NVPA_STATUS_FUNCTION_NOT_FOUND:
+                *ppStatusStr = "NVPA_STATUS_FUNCTION_NOT_FOUND";
+                *ppCommentStr = "The function was not found in this version of the NvPerf DLL/DSO. Or if you are directly calling NVPA_GetProcAddress(), please ensure the function name is spelled correctly.";
+                return;
+            case NVPA_STATUS_NOT_SUPPORTED:
+                *ppStatusStr = "NVPA_STATUS_NOT_SUPPORTED";
+                *ppCommentStr = "The request was intentionally not supported.";
+                return;
+            case NVPA_STATUS_NOT_IMPLEMENTED:
+                *ppStatusStr = "NVPA_STATUS_NOT_IMPLEMENTED";
+                *ppCommentStr = "The request was not implemented by this version.";
+                return;
+            case NVPA_STATUS_INVALID_ARGUMENT:
+                *ppStatusStr = "NVPA_STATUS_INVALID_ARGUMENT";
+                *ppCommentStr = "Invalid argument.";
+                return;
+            case NVPA_STATUS_INVALID_METRIC_ID:
+                *ppStatusStr = "NVPA_STATUS_INVALID_METRIC_ID";
+                *ppCommentStr = "UNUSED";
+                return;
+            case NVPA_STATUS_DRIVER_NOT_LOADED:
+                *ppStatusStr = "NVPA_STATUS_DRIVER_NOT_LOADED";
+                *ppCommentStr = "No driver has been loaded via NVPW_*_LoadDriver().";
+                return;
+            case NVPA_STATUS_OUT_OF_MEMORY:
+                *ppStatusStr = "NVPA_STATUS_OUT_OF_MEMORY";
+                *ppCommentStr = "Failed memory allocation.";
+                return;
+            case NVPA_STATUS_INVALID_THREAD_STATE:
+                *ppStatusStr = "NVPA_STATUS_INVALID_THREAD_STATE";
+                *ppCommentStr = "UNUSED";
+                return;
+            case NVPA_STATUS_FAILED_CONTEXT_ALLOC:
+                *ppStatusStr = "NVPA_STATUS_FAILED_CONTEXT_ALLOC";
+                *ppCommentStr = "UNUSED";
+                return;
+            case NVPA_STATUS_UNSUPPORTED_GPU:
+                *ppStatusStr = "NVPA_STATUS_UNSUPPORTED_GPU";
+                *ppCommentStr = "The specified GPU is not supported. It is recommended to call IsGpuSupported() for more information";
+                return;
+            case NVPA_STATUS_INSUFFICIENT_DRIVER_VERSION:
+                *ppStatusStr = "NVPA_STATUS_INSUFFICIENT_DRIVER_VERSION";
+                *ppCommentStr = "The installed NVIDIA driver is too old.";
+                return;
+            case NVPA_STATUS_OBJECT_NOT_REGISTERED:
+                *ppStatusStr = "NVPA_STATUS_OBJECT_NOT_REGISTERED";
+                *ppCommentStr = "UNUSED";
+                return;
+            case NVPA_STATUS_INSUFFICIENT_PRIVILEGE:
+                *ppStatusStr = "NVPA_STATUS_INSUFFICIENT_PRIVILEGE";
+                *ppCommentStr = "Profiling permission not granted; see https://developer.nvidia.com/nvidia-development-tools-solutions-ERR_NVGPUCTRPERM-permission-issue-performance-counters";
+                return;
+            case NVPA_STATUS_INVALID_CONTEXT_STATE:
+                *ppStatusStr = "NVPA_STATUS_INVALID_CONTEXT_STATE";
+                *ppCommentStr = "UNUSED";
+                return;
+            case NVPA_STATUS_INVALID_OBJECT_STATE:
+                *ppStatusStr = "NVPA_STATUS_INVALID_OBJECT_STATE";
+                *ppCommentStr = "UNUSED";
+                return;
+            case NVPA_STATUS_RESOURCE_UNAVAILABLE:
+                *ppStatusStr = "NVPA_STATUS_RESOURCE_UNAVAILABLE";
+                *ppCommentStr = "The request could not be fulfilled because a system resource is already in use.";
+                return;
+            case NVPA_STATUS_DRIVER_LOADED_TOO_LATE:
+                *ppStatusStr = "NVPA_STATUS_DRIVER_LOADED_TOO_LATE";
+                *ppCommentStr = "UNUSED";
+                return;
+            case NVPA_STATUS_INSUFFICIENT_SPACE:
+                *ppStatusStr = "NVPA_STATUS_INSUFFICIENT_SPACE";
+                *ppCommentStr = "The provided buffer is not large enough.";
+                return;
+            case NVPA_STATUS_OBJECT_MISMATCH:
+                *ppStatusStr = "NVPA_STATUS_OBJECT_MISMATCH";
+                *ppCommentStr = "UNUSED";
+                return;
+            case NVPA_STATUS_VIRTUALIZED_DEVICE_NOT_SUPPORTED:
+                *ppStatusStr = "NVPA_STATUS_VIRTUALIZED_DEVICE_NOT_SUPPORTED";
+                *ppCommentStr = "Virtualized GPU (vGPU) is not supported.";
+                return;
+            case NVPA_STATUS_PROFILING_NOT_ALLOWED:
+                *ppStatusStr = "NVPA_STATUS_PROFILING_NOT_ALLOWED";
+                *ppCommentStr = "Profiling permission was not granted or the device was disabled.";
+                return;
+            default:
+                *ppStatusStr = "Unrecognized status";
+                *ppCommentStr = "This status is unrecognized. Is it coming from a newer version of NvPerf library?";
+                return;
+        }
+    }
+
+
+#endif // NVPERF_NVPA_STATUS_DEFINED
+
+
+#ifndef NVPERF_NVPA_ACTIVITY_KIND_DEFINED
+#define NVPERF_NVPA_ACTIVITY_KIND_DEFINED
+
+    /// The configuration's activity-kind dictates which types of data may be collected.
+    typedef enum NVPA_ActivityKind
+    {
+        /// Invalid value.
+        NVPA_ACTIVITY_KIND_INVALID = 0,
+        /// A workload-centric activity for serialized and pipelined collection.
+        /// 
+        /// Profiler is capable of collecting both serialized and pipelined metrics.  The library introduces any
+        /// synchronization required to collect serialized metrics.
+        NVPA_ACTIVITY_KIND_PROFILER,
+        /// A realtime activity for sampling counters from the CPU or GPU.
+        NVPA_ACTIVITY_KIND_REALTIME_SAMPLED,
+        /// A realtime activity for profiling counters from the CPU or GPU without CPU/GPU synchronizations.
+        NVPA_ACTIVITY_KIND_REALTIME_PROFILER,
+        NVPA_ACTIVITY_KIND__COUNT
+    } NVPA_ActivityKind;
+
+
+#endif // NVPERF_NVPA_ACTIVITY_KIND_DEFINED
+
+
+#ifndef NVPERF_NVPA_BOOL_DEFINED
+#define NVPERF_NVPA_BOOL_DEFINED
+    /// The type used for boolean values.
+    typedef uint8_t NVPA_Bool;
+#endif // NVPERF_NVPA_BOOL_DEFINED
+
+#ifndef NVPA_STRUCT_SIZE
+#define NVPA_STRUCT_SIZE(type_, lastfield_)                     (offsetof(type_, lastfield_) + sizeof(((type_*)0)->lastfield_))
+#endif // NVPA_STRUCT_SIZE
+
+#ifndef NVPW_FIELD_EXISTS
+#define NVPW_FIELD_EXISTS(pParams_, name_) \
+    ((pParams_)->structSize >= (size_t)((const uint8_t*)(&(pParams_)->name_) + sizeof(pParams_)->name_ - (const uint8_t*)(pParams_)))
+#endif // NVPW_FIELD_EXISTS
+
+
+#ifndef NVPERF_NVPA_GETPROCADDRESS_DEFINED
+#define NVPERF_NVPA_GETPROCADDRESS_DEFINED
+
+typedef NVPA_Status (*NVPA_GenericFn)(void);
+
+
+    /// 
+    /// Gets the address of an NvPerf API function.
+    /// 
+    /// \return A function pointer to the function, or NULL if the function is not available.
+    /// 
+    /// \param pFunctionName [in] Name of the function to retrieve.
+    NVPA_GenericFn NVPA_GetProcAddress(const char* pFunctionName);
+
+#endif
+
+#ifndef NVPERF_NVPW_SETLIBRARYLOADPATHS_DEFINED
+#define NVPERF_NVPW_SETLIBRARYLOADPATHS_DEFINED
+
+
+    typedef struct NVPW_SetLibraryLoadPaths_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// [in] number of paths in ppPaths
+        size_t numPaths;
+        /// [in] array of null-terminated paths
+        const char** ppPaths;
+    } NVPW_SetLibraryLoadPaths_Params;
+#define NVPW_SetLibraryLoadPaths_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_SetLibraryLoadPaths_Params, ppPaths)
+
+    /// Sets library search path for \ref NVPW_InitializeHost() and \ref NVPW_InitializeTarget().
+    /// \ref NVPW_InitializeHost() and \ref NVPW_InitializeTarget load the NvPerf DLL/DSO.  This function sets
+    /// ordered paths that will be searched with the LoadLibrary() or dlopen() call.
+    /// If load paths are set by this function, the default set of load paths
+    /// will not be attempted.
+    /// Each path must point at a directory (not a file name).
+    /// This function is not thread-safe.
+    /// Example Usage:
+    /// \code
+    ///     const char* paths[] = {
+    ///         "path1", "path2", etc
+    ///     };
+    ///     NVPW_SetLibraryLoadPaths_Params params{NVPW_SetLibraryLoadPaths_Params_STRUCT_SIZE};
+    ///     params.numPaths = sizeof(paths)/sizeof(paths[0]);
+    ///     params.ppPaths = paths;
+    ///     NVPW_SetLibraryLoadPaths(&params);
+    ///     NVPW_InitializeHost();
+    ///     params.numPaths = 0;
+    ///     params.ppPaths = NULL;
+    ///     NVPW_SetLibraryLoadPaths(&params);
+    /// \endcode
+    NVPA_Status NVPW_SetLibraryLoadPaths(NVPW_SetLibraryLoadPaths_Params* pParams);
+
+    typedef struct NVPW_SetLibraryLoadPathsW_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// [in] number of paths in ppwPaths
+        size_t numPaths;
+        /// [in] array of null-terminated paths
+        const wchar_t** ppwPaths;
+    } NVPW_SetLibraryLoadPathsW_Params;
+#define NVPW_SetLibraryLoadPathsW_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_SetLibraryLoadPathsW_Params, ppwPaths)
+
+    /// Sets library search path for \ref NVPW_InitializeHost() and \ref NVPW_InitializeTarget().
+    /// \ref NVPW_InitializeHost() and \ref NVPW_InitializeTarget load the NvPerf DLL/DSO.  This function sets
+    /// ordered paths that will be searched with the LoadLibrary() or dlopen() call.
+    /// If load paths are set by this function, the default set of load paths
+    /// will not be attempted.
+    /// Each path must point at a directory (not a file name).
+    /// This function is not thread-safe.
+    /// Example Usage:
+    /// \code
+    ///     const wchar_t* wpaths[] = {
+    ///         L"path1", L"path2", etc
+    ///     };
+    ///     NVPW_SetLibraryLoadPathsW_Params params{NVPW_SetLibraryLoadPathsW_Params_STRUCT_SIZE};
+    ///     params.numPaths = sizeof(wpaths)/sizeof(wpaths[0]);
+    ///     params.ppwPaths = wpaths;
+    ///     NVPW_SetLibraryLoadPathsW(&params);
+    ///     NVPW_InitializeHost();
+    ///     params.numPaths = 0;
+    ///     params.ppwPaths = NULL;
+    ///     NVPW_SetLibraryLoadPathsW(&params);
+    /// \endcode
+    NVPA_Status NVPW_SetLibraryLoadPathsW(NVPW_SetLibraryLoadPathsW_Params* pParams);
+
+#endif
+
+
+
+#ifdef __cplusplus
+} // extern "C"
+#endif
+
+#if defined(__GNUC__) && defined(NVPA_SHARED_LIB)
+    #pragma GCC visibility pop
+#endif
+
+#endif // NVPERF_COMMON_H

Prism/LLaDA/LLaDA_Prism/.venv/lib/python3.12/site-packages/nvidia/cuda_cupti/include/nvperf_cuda_host.h

@@ -0,0 +1,179 @@
+#ifndef NVPERF_CUDA_HOST_H
+#define NVPERF_CUDA_HOST_H
+
+/*
+ * Copyright 2014-2024 NVIDIA Corporation.  All rights reserved.
+ *
+ * NOTICE TO USER:
+ *
+ * This source code is subject to NVIDIA ownership rights under U.S. and
+ * international Copyright laws.
+ *
+ * This software and the information contained herein is PROPRIETARY and
+ * CONFIDENTIAL to NVIDIA and is being provided under the terms and conditions
+ * of a form of NVIDIA software license agreement.
+ *
+ * NVIDIA MAKES NO REPRESENTATION ABOUT THE SUITABILITY OF THIS SOURCE
+ * CODE FOR ANY PURPOSE.  IT IS PROVIDED "AS IS" WITHOUT EXPRESS OR
+ * IMPLIED WARRANTY OF ANY KIND.  NVIDIA DISCLAIMS ALL WARRANTIES WITH
+ * REGARD TO THIS SOURCE CODE, INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY, NONINFRINGEMENT, AND FITNESS FOR A PARTICULAR PURPOSE.
+ * IN NO EVENT SHALL NVIDIA BE LIABLE FOR ANY SPECIAL, INDIRECT, INCIDENTAL,
+ * OR CONSEQUENTIAL DAMAGES, OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS
+ * OF USE, DATA OR PROFITS,  WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE
+ * OR OTHER TORTIOUS ACTION,  ARISING OUT OF OR IN CONNECTION WITH THE USE
+ * OR PERFORMANCE OF THIS SOURCE CODE.
+ *
+ * U.S. Government End Users.   This source code is a "commercial item" as
+ * that term is defined at  48 C.F.R. 2.101 (OCT 1995), consisting  of
+ * "commercial computer  software"  and "commercial computer software
+ * documentation" as such terms are  used in 48 C.F.R. 12.212 (SEPT 1995)
+ * and is provided to the U.S. Government only as a commercial end item.
+ * Consistent with 48 C.F.R.12.212 and 48 C.F.R. 227.7202-1 through
+ * 227.7202-4 (JUNE 1995), all U.S. Government End Users acquire the
+ * source code with only those rights set forth herein.
+ *
+ * Any use of this source code in individual and commercial software must
+ * include, in the user documentation and internal comments to the code,
+ * the above Disclaimer and U.S. Government End Users Notice.
+ */
+
+#include <stddef.h>
+#include <stdint.h>
+#include "nvperf_common.h"
+#include "nvperf_host.h"
+
+#if defined(__GNUC__) && defined(NVPA_SHARED_LIB)
+    #pragma GCC visibility push(default)
+    #if !defined(NVPW_LOCAL)
+        #define NVPW_LOCAL __attribute__ ((visibility ("hidden")))
+    #endif
+#else
+    #if !defined(NVPW_LOCAL)
+        #define NVPW_LOCAL
+    #endif
+#endif
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/**
+ *  @file   nvperf_cuda_host.h
+ */
+
+    typedef struct NVPW_CUDA_RawMetricsConfig_Create_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// [in]
+        NVPA_ActivityKind activityKind;
+        /// [in]
+        const char* pChipName;
+        /// [out] new NVPA_RawMetricsConfig object
+        struct NVPA_RawMetricsConfig* pRawMetricsConfig;
+    } NVPW_CUDA_RawMetricsConfig_Create_Params;
+#define NVPW_CUDA_RawMetricsConfig_Create_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_CUDA_RawMetricsConfig_Create_Params, pRawMetricsConfig)
+
+    NVPA_Status NVPW_CUDA_RawMetricsConfig_Create(NVPW_CUDA_RawMetricsConfig_Create_Params* pParams);
+
+    typedef struct NVPW_CUDA_RawMetricsConfig_Create_V2_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// [in]
+        NVPA_ActivityKind activityKind;
+        /// [in] accepted for chips supported at the time-of-release.
+        const char* pChipName;
+        /// [in] buffer with counter availability image - required for future chip support
+        const uint8_t* pCounterAvailabilityImage;
+        /// [out] new NVPA_RawMetricsConfig object
+        struct NVPA_RawMetricsConfig* pRawMetricsConfig;
+    } NVPW_CUDA_RawMetricsConfig_Create_V2_Params;
+#define NVPW_CUDA_RawMetricsConfig_Create_V2_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_CUDA_RawMetricsConfig_Create_V2_Params, pRawMetricsConfig)
+
+    /// Use either 'pChipName' or 'pCounterAvailabilityImage'.
+    NVPA_Status NVPW_CUDA_RawMetricsConfig_Create_V2(NVPW_CUDA_RawMetricsConfig_Create_V2_Params* pParams);
+
+    typedef struct NVPW_CUDA_CounterDataBuilder_Create_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// [in] accepted for chips supported at the time-of-release.
+        const char* pChipName;
+        /// [in] buffer with counter availability image - required for future chip support
+        const uint8_t* pCounterAvailabilityImage;
+        /// [out] new NVPA_CounterDataBuilder object
+        struct NVPA_CounterDataBuilder* pCounterDataBuilder;
+    } NVPW_CUDA_CounterDataBuilder_Create_Params;
+#define NVPW_CUDA_CounterDataBuilder_Create_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_CUDA_CounterDataBuilder_Create_Params, pCounterDataBuilder)
+
+    /// Use either 'pChipName' or 'pCounterAvailabilityImage'.
+    NVPA_Status NVPW_CUDA_CounterDataBuilder_Create(NVPW_CUDA_CounterDataBuilder_Create_Params* pParams);
+
+    typedef struct NVPW_MetricsEvaluator NVPW_MetricsEvaluator;
+
+    typedef struct NVPW_CUDA_MetricsEvaluator_CalculateScratchBufferSize_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// [in] accepted for chips supported at the time-of-release.
+        const char* pChipName;
+        /// [in] buffer with counter availability image - required for future chip support
+        const uint8_t* pCounterAvailabilityImage;
+        /// [out]
+        size_t scratchBufferSize;
+    } NVPW_CUDA_MetricsEvaluator_CalculateScratchBufferSize_Params;
+#define NVPW_CUDA_MetricsEvaluator_CalculateScratchBufferSize_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_CUDA_MetricsEvaluator_CalculateScratchBufferSize_Params, scratchBufferSize)
+
+    /// Use either 'pChipName' or 'pCounterAvailabilityImage'.
+    NVPA_Status NVPW_CUDA_MetricsEvaluator_CalculateScratchBufferSize(NVPW_CUDA_MetricsEvaluator_CalculateScratchBufferSize_Params* pParams);
+
+    typedef struct NVPW_CUDA_MetricsEvaluator_Initialize_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// [in]
+        uint8_t* pScratchBuffer;
+        /// [in] the size of the 'pScratchBuffer' array, should be at least the size of the 'scratchBufferSize' returned
+        /// by 'NVPW_CUDA_MetricsEvaluator_CalculateScratchBufferSize'
+        size_t scratchBufferSize;
+        /// [in] accepted for chips supported at the time-of-release.
+        const char* pChipName;
+        /// [in] buffer with counter availability image - required for future chip support
+        const uint8_t* pCounterAvailabilityImage;
+        /// [in]
+        const uint8_t* pCounterDataImage;
+        /// [in] must be provided if 'pCounterDataImage' is not NULL
+        size_t counterDataImageSize;
+        /// [out]
+        struct NVPW_MetricsEvaluator* pMetricsEvaluator;
+    } NVPW_CUDA_MetricsEvaluator_Initialize_Params;
+#define NVPW_CUDA_MetricsEvaluator_Initialize_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_CUDA_MetricsEvaluator_Initialize_Params, pMetricsEvaluator)
+
+    /// Use one of 'pChipName', 'pCounterAvailabilityImage', or 'pCounterDataImage'. 'pChipName' or
+    /// 'pCounterAvailabilityImage' will create a metrics evaluator based on a virtual device while 'pCounterDataImage'
+    /// will create a metrics evaluator based on the actual device.
+    NVPA_Status NVPW_CUDA_MetricsEvaluator_Initialize(NVPW_CUDA_MetricsEvaluator_Initialize_Params* pParams);
+
+
+
+#ifdef __cplusplus
+} // extern "C"
+#endif
+
+#if defined(__GNUC__) && defined(NVPA_SHARED_LIB)
+    #pragma GCC visibility pop
+#endif
+
+#endif // NVPERF_CUDA_HOST_H

Prism/LLaDA/LLaDA_Prism/.venv/lib/python3.12/site-packages/nvidia/cuda_cupti/include/nvperf_host.h

@@ -0,0 +1,1178 @@
+#ifndef NVPERF_HOST_H
+#define NVPERF_HOST_H
+
+/*
+ * Copyright 2014-2024 NVIDIA Corporation.  All rights reserved.
+ *
+ * NOTICE TO USER:
+ *
+ * This source code is subject to NVIDIA ownership rights under U.S. and
+ * international Copyright laws.
+ *
+ * This software and the information contained herein is PROPRIETARY and
+ * CONFIDENTIAL to NVIDIA and is being provided under the terms and conditions
+ * of a form of NVIDIA software license agreement.
+ *
+ * NVIDIA MAKES NO REPRESENTATION ABOUT THE SUITABILITY OF THIS SOURCE
+ * CODE FOR ANY PURPOSE.  IT IS PROVIDED "AS IS" WITHOUT EXPRESS OR
+ * IMPLIED WARRANTY OF ANY KIND.  NVIDIA DISCLAIMS ALL WARRANTIES WITH
+ * REGARD TO THIS SOURCE CODE, INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY, NONINFRINGEMENT, AND FITNESS FOR A PARTICULAR PURPOSE.
+ * IN NO EVENT SHALL NVIDIA BE LIABLE FOR ANY SPECIAL, INDIRECT, INCIDENTAL,
+ * OR CONSEQUENTIAL DAMAGES, OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS
+ * OF USE, DATA OR PROFITS,  WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE
+ * OR OTHER TORTIOUS ACTION,  ARISING OUT OF OR IN CONNECTION WITH THE USE
+ * OR PERFORMANCE OF THIS SOURCE CODE.
+ *
+ * U.S. Government End Users.   This source code is a "commercial item" as
+ * that term is defined at  48 C.F.R. 2.101 (OCT 1995), consisting  of
+ * "commercial computer  software"  and "commercial computer software
+ * documentation" as such terms are  used in 48 C.F.R. 12.212 (SEPT 1995)
+ * and is provided to the U.S. Government only as a commercial end item.
+ * Consistent with 48 C.F.R.12.212 and 48 C.F.R. 227.7202-1 through
+ * 227.7202-4 (JUNE 1995), all U.S. Government End Users acquire the
+ * source code with only those rights set forth herein.
+ *
+ * Any use of this source code in individual and commercial software must
+ * include, in the user documentation and internal comments to the code,
+ * the above Disclaimer and U.S. Government End Users Notice.
+ */
+
+#include <stddef.h>
+#include <stdint.h>
+#include "nvperf_common.h"
+
+#if defined(__GNUC__) && defined(NVPA_SHARED_LIB)
+    #pragma GCC visibility push(default)
+    #if !defined(NVPW_LOCAL)
+        #define NVPW_LOCAL __attribute__ ((visibility ("hidden")))
+    #endif
+#else
+    #if !defined(NVPW_LOCAL)
+        #define NVPW_LOCAL
+    #endif
+#endif
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/**
+ *  @file   nvperf_host.h
+ */
+
+
+// Guard against multiple definition of NvPerf host types
+#ifndef NVPERF_HOST_API_DEFINED
+#define NVPERF_HOST_API_DEFINED
+
+
+/***************************************************************************//**
+ *  @name   Host Configuration
+ *  @{
+ */
+
+    typedef struct NVPW_InitializeHost_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+    } NVPW_InitializeHost_Params;
+#define NVPW_InitializeHost_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_InitializeHost_Params, pPriv)
+
+    /// Load the host library.
+    NVPA_Status NVPW_InitializeHost(NVPW_InitializeHost_Params* pParams);
+
+    typedef struct NVPW_CounterData_CalculateCounterDataImageCopySize_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// The CounterDataPrefix generated from e.g.    nvperf2 initdata   or
+        /// NVPW_CounterDataBuilder_GetCounterDataPrefix().  Must be align(8).
+        const uint8_t* pCounterDataPrefix;
+        size_t counterDataPrefixSize;
+        /// max number of ranges that can be profiled
+        uint32_t maxNumRanges;
+        /// max number of RangeTree nodes; must be >= maxNumRanges
+        uint32_t maxNumRangeTreeNodes;
+        /// max string length of each RangeName, including the trailing NUL character
+        uint32_t maxRangeNameLength;
+        const uint8_t* pCounterDataSrc;
+        /// [out] required size of the copy buffer
+        size_t copyDataImageCounterSize;
+    } NVPW_CounterData_CalculateCounterDataImageCopySize_Params;
+#define NVPW_CounterData_CalculateCounterDataImageCopySize_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_CounterData_CalculateCounterDataImageCopySize_Params, copyDataImageCounterSize)
+
+    NVPA_Status NVPW_CounterData_CalculateCounterDataImageCopySize(NVPW_CounterData_CalculateCounterDataImageCopySize_Params* pParams);
+
+    typedef struct NVPW_CounterData_InitializeCounterDataImageCopy_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// The CounterDataPrefix generated from e.g.    nvperf2 initdata   or
+        /// NVPW_CounterDataBuilder_GetCounterDataPrefix().  Must be align(8).
+        const uint8_t* pCounterDataPrefix;
+        size_t counterDataPrefixSize;
+        /// max number of ranges that can be profiled
+        uint32_t maxNumRanges;
+        /// max number of RangeTree nodes; must be >= maxNumRanges
+        uint32_t maxNumRangeTreeNodes;
+        /// max string length of each RangeName, including the trailing NUL character
+        uint32_t maxRangeNameLength;
+        const uint8_t* pCounterDataSrc;
+        uint8_t* pCounterDataDst;
+    } NVPW_CounterData_InitializeCounterDataImageCopy_Params;
+#define NVPW_CounterData_InitializeCounterDataImageCopy_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_CounterData_InitializeCounterDataImageCopy_Params, pCounterDataDst)
+
+    NVPA_Status NVPW_CounterData_InitializeCounterDataImageCopy(NVPW_CounterData_InitializeCounterDataImageCopy_Params* pParams);
+
+    typedef struct NVPW_CounterData_ExtractCounterDataPrefix_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// The source buffer to extract the prefix from.
+        const uint8_t* pCounterDataSrc;
+        size_t counterDataSrcSize;
+        /// [in] If not NULL, the prefix will be copied into this buffer.
+        uint8_t* pCounterDataPrefix;
+        /// [inout] if 'pCounterDataPrefix' is NULL, size of counter data prefix will be returned; otherwise it should
+        /// be set to the size of buffer allocated for 'pCounterDataPrefix'.
+        size_t counterDataPrefixSize;
+    } NVPW_CounterData_ExtractCounterDataPrefix_Params;
+#define NVPW_CounterData_ExtractCounterDataPrefix_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_CounterData_ExtractCounterDataPrefix_Params, counterDataPrefixSize)
+
+    NVPA_Status NVPW_CounterData_ExtractCounterDataPrefix(NVPW_CounterData_ExtractCounterDataPrefix_Params* pParams);
+
+    typedef struct NVPA_CounterDataCombiner NVPA_CounterDataCombiner;
+
+    typedef struct NVPW_CounterDataCombiner_Create_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// The destination counter data into which the source datas will be combined
+        uint8_t* pCounterDataDst;
+        /// [out] The created counter data combiner
+        NVPA_CounterDataCombiner* pCounterDataCombiner;
+    } NVPW_CounterDataCombiner_Create_Params;
+#define NVPW_CounterDataCombiner_Create_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_CounterDataCombiner_Create_Params, pCounterDataCombiner)
+
+    NVPA_Status NVPW_CounterDataCombiner_Create(NVPW_CounterDataCombiner_Create_Params* pParams);
+
+    typedef struct NVPW_CounterDataCombiner_Destroy_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        NVPA_CounterDataCombiner* pCounterDataCombiner;
+    } NVPW_CounterDataCombiner_Destroy_Params;
+#define NVPW_CounterDataCombiner_Destroy_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_CounterDataCombiner_Destroy_Params, pCounterDataCombiner)
+
+    NVPA_Status NVPW_CounterDataCombiner_Destroy(NVPW_CounterDataCombiner_Destroy_Params* pParams);
+
+    typedef struct NVPW_CounterDataCombiner_CreateRange_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        NVPA_CounterDataCombiner* pCounterDataCombiner;
+        size_t numDescriptions;
+        const char* const* ppDescriptions;
+        /// [out]
+        size_t rangeIndexDst;
+    } NVPW_CounterDataCombiner_CreateRange_Params;
+#define NVPW_CounterDataCombiner_CreateRange_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_CounterDataCombiner_CreateRange_Params, rangeIndexDst)
+
+    NVPA_Status NVPW_CounterDataCombiner_CreateRange(NVPW_CounterDataCombiner_CreateRange_Params* pParams);
+
+    typedef struct NVPW_CounterDataCombiner_CopyIntoRange_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// [in]
+        NVPA_CounterDataCombiner* pCounterDataCombiner;
+        /// [in]
+        size_t rangeIndexDst;
+        /// [in]
+        const uint8_t* pCounterDataSrc;
+        /// [in]
+        size_t rangeIndexSrc;
+    } NVPW_CounterDataCombiner_CopyIntoRange_Params;
+#define NVPW_CounterDataCombiner_CopyIntoRange_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_CounterDataCombiner_CopyIntoRange_Params, rangeIndexSrc)
+
+    /// In order to use this API, the source counter data and the destination counter data must have identical counters
+    NVPA_Status NVPW_CounterDataCombiner_CopyIntoRange(NVPW_CounterDataCombiner_CopyIntoRange_Params* pParams);
+
+    typedef struct NVPW_CounterDataCombiner_AccumulateIntoRange_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        NVPA_CounterDataCombiner* pCounterDataCombiner;
+        size_t rangeIndexDst;
+        uint32_t dstMultiplier;
+        const uint8_t* pCounterDataSrc;
+        size_t rangeIndexSrc;
+        uint32_t srcMultiplier;
+    } NVPW_CounterDataCombiner_AccumulateIntoRange_Params;
+#define NVPW_CounterDataCombiner_AccumulateIntoRange_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_CounterDataCombiner_AccumulateIntoRange_Params, srcMultiplier)
+
+    NVPA_Status NVPW_CounterDataCombiner_AccumulateIntoRange(NVPW_CounterDataCombiner_AccumulateIntoRange_Params* pParams);
+
+    typedef struct NVPW_CounterDataCombiner_SumIntoRange_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        NVPA_CounterDataCombiner* pCounterDataCombiner;
+        size_t rangeIndexDst;
+        const uint8_t* pCounterDataSrc;
+        size_t rangeIndexSrc;
+    } NVPW_CounterDataCombiner_SumIntoRange_Params;
+#define NVPW_CounterDataCombiner_SumIntoRange_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_CounterDataCombiner_SumIntoRange_Params, rangeIndexSrc)
+
+    NVPA_Status NVPW_CounterDataCombiner_SumIntoRange(NVPW_CounterDataCombiner_SumIntoRange_Params* pParams);
+
+    typedef struct NVPW_CounterDataCombiner_WeightedSumIntoRange_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        NVPA_CounterDataCombiner* pCounterDataCombiner;
+        size_t rangeIndexDst;
+        double dstMultiplier;
+        const uint8_t* pCounterDataSrc;
+        size_t rangeIndexSrc;
+        double srcMultiplier;
+    } NVPW_CounterDataCombiner_WeightedSumIntoRange_Params;
+#define NVPW_CounterDataCombiner_WeightedSumIntoRange_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_CounterDataCombiner_WeightedSumIntoRange_Params, srcMultiplier)
+
+    NVPA_Status NVPW_CounterDataCombiner_WeightedSumIntoRange(NVPW_CounterDataCombiner_WeightedSumIntoRange_Params* pParams);
+
+/**
+ *  @}
+ ******************************************************************************/
+ 
+/***************************************************************************//**
+ *  @name   Metrics Configuration
+ *  @{
+ */
+
+    typedef struct NVPA_RawMetricsConfig NVPA_RawMetricsConfig;
+
+    typedef struct NVPA_RawMetricRequest
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// in
+        const char* pMetricName;
+        /// in
+        NVPA_Bool isolated;
+        /// in; ignored by AddMetric but observed by CounterData initialization
+        NVPA_Bool keepInstances;
+    } NVPA_RawMetricRequest;
+#define NVPA_RAW_METRIC_REQUEST_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPA_RawMetricRequest, keepInstances)
+
+    typedef struct NVPW_GetSupportedChipNames_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// [out]
+        const char* const* ppChipNames;
+        /// [out]
+        size_t numChipNames;
+    } NVPW_GetSupportedChipNames_Params;
+#define NVPW_GetSupportedChipNames_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_GetSupportedChipNames_Params, numChipNames)
+
+    NVPA_Status NVPW_GetSupportedChipNames(NVPW_GetSupportedChipNames_Params* pParams);
+
+    typedef struct NVPW_RawMetricsConfig_Destroy_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        NVPA_RawMetricsConfig* pRawMetricsConfig;
+    } NVPW_RawMetricsConfig_Destroy_Params;
+#define NVPW_RawMetricsConfig_Destroy_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_RawMetricsConfig_Destroy_Params, pRawMetricsConfig)
+
+    NVPA_Status NVPW_RawMetricsConfig_Destroy(NVPW_RawMetricsConfig_Destroy_Params* pParams);
+
+    typedef struct NVPW_RawMetricsConfig_SetCounterAvailability_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        NVPA_RawMetricsConfig* pRawMetricsConfig;
+        /// [in] buffer with counter availability image
+        const uint8_t* pCounterAvailabilityImage;
+    } NVPW_RawMetricsConfig_SetCounterAvailability_Params;
+#define NVPW_RawMetricsConfig_SetCounterAvailability_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_RawMetricsConfig_SetCounterAvailability_Params, pCounterAvailabilityImage)
+
+    NVPA_Status NVPW_RawMetricsConfig_SetCounterAvailability(NVPW_RawMetricsConfig_SetCounterAvailability_Params* pParams);
+
+    typedef struct NVPW_RawMetricsConfig_BeginPassGroup_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        NVPA_RawMetricsConfig* pRawMetricsConfig;
+        size_t maxPassCount;
+    } NVPW_RawMetricsConfig_BeginPassGroup_Params;
+#define NVPW_RawMetricsConfig_BeginPassGroup_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_RawMetricsConfig_BeginPassGroup_Params, maxPassCount)
+
+    NVPA_Status NVPW_RawMetricsConfig_BeginPassGroup(NVPW_RawMetricsConfig_BeginPassGroup_Params* pParams);
+
+    typedef struct NVPW_RawMetricsConfig_EndPassGroup_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        NVPA_RawMetricsConfig* pRawMetricsConfig;
+    } NVPW_RawMetricsConfig_EndPassGroup_Params;
+#define NVPW_RawMetricsConfig_EndPassGroup_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_RawMetricsConfig_EndPassGroup_Params, pRawMetricsConfig)
+
+    NVPA_Status NVPW_RawMetricsConfig_EndPassGroup(NVPW_RawMetricsConfig_EndPassGroup_Params* pParams);
+
+    typedef struct NVPW_RawMetricsConfig_GetNumMetrics_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        const NVPA_RawMetricsConfig* pRawMetricsConfig;
+        /// [out]
+        size_t numMetrics;
+    } NVPW_RawMetricsConfig_GetNumMetrics_Params;
+#define NVPW_RawMetricsConfig_GetNumMetrics_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_RawMetricsConfig_GetNumMetrics_Params, numMetrics)
+
+    NVPA_Status NVPW_RawMetricsConfig_GetNumMetrics(NVPW_RawMetricsConfig_GetNumMetrics_Params* pParams);
+
+    typedef struct NVPW_RawMetricsConfig_GetMetricProperties_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        const NVPA_RawMetricsConfig* pRawMetricsConfig;
+        size_t metricIndex;
+        /// [out]
+        const char* pMetricName;
+        /// [out]
+        NVPA_Bool supportsPipelined;
+        /// [out]
+        NVPA_Bool supportsIsolated;
+    } NVPW_RawMetricsConfig_GetMetricProperties_Params;
+#define NVPW_RawMetricsConfig_GetMetricProperties_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_RawMetricsConfig_GetMetricProperties_Params, supportsIsolated)
+
+    NVPA_Status NVPW_RawMetricsConfig_GetMetricProperties(NVPW_RawMetricsConfig_GetMetricProperties_Params* pParams);
+
+    typedef struct NVPW_RawMetricsConfig_GetMetricProperties_V2_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        const NVPA_RawMetricsConfig* pRawMetricsConfig;
+        size_t metricIndex;
+        /// [out]
+        const char* pMetricName;
+    } NVPW_RawMetricsConfig_GetMetricProperties_V2_Params;
+#define NVPW_RawMetricsConfig_GetMetricProperties_V2_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_RawMetricsConfig_GetMetricProperties_V2_Params, pMetricName)
+
+    NVPA_Status NVPW_RawMetricsConfig_GetMetricProperties_V2(NVPW_RawMetricsConfig_GetMetricProperties_V2_Params* pParams);
+
+    typedef struct NVPW_RawMetricsConfig_AddMetrics_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        NVPA_RawMetricsConfig* pRawMetricsConfig;
+        const NVPA_RawMetricRequest* pRawMetricRequests;
+        size_t numMetricRequests;
+    } NVPW_RawMetricsConfig_AddMetrics_Params;
+#define NVPW_RawMetricsConfig_AddMetrics_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_RawMetricsConfig_AddMetrics_Params, numMetricRequests)
+
+    NVPA_Status NVPW_RawMetricsConfig_AddMetrics(NVPW_RawMetricsConfig_AddMetrics_Params* pParams);
+
+    typedef struct NVPW_RawMetricsConfig_IsAddMetricsPossible_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        const NVPA_RawMetricsConfig* pRawMetricsConfig;
+        const NVPA_RawMetricRequest* pRawMetricRequests;
+        size_t numMetricRequests;
+        /// [out]
+        NVPA_Bool isPossible;
+    } NVPW_RawMetricsConfig_IsAddMetricsPossible_Params;
+#define NVPW_RawMetricsConfig_IsAddMetricsPossible_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_RawMetricsConfig_IsAddMetricsPossible_Params, isPossible)
+
+    NVPA_Status NVPW_RawMetricsConfig_IsAddMetricsPossible(NVPW_RawMetricsConfig_IsAddMetricsPossible_Params* pParams);
+
+    typedef struct NVPW_RawMetricsConfig_GenerateConfigImage_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        NVPA_RawMetricsConfig* pRawMetricsConfig;
+        /// [in] If true, all existing pass groups may be merged to reduce number of passes.
+        /// If merge was successful, distribution of counters in passes may be updated as a side-effect. The effects
+        /// will be persistent in pRawMetricsConfig.
+        NVPA_Bool mergeAllPassGroups;
+    } NVPW_RawMetricsConfig_GenerateConfigImage_Params;
+#define NVPW_RawMetricsConfig_GenerateConfigImage_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_RawMetricsConfig_GenerateConfigImage_Params, mergeAllPassGroups)
+
+    /// This API may fail if called inside a pass group with `mergeAllPassGroups` = true.
+    NVPA_Status NVPW_RawMetricsConfig_GenerateConfigImage(NVPW_RawMetricsConfig_GenerateConfigImage_Params* pParams);
+
+    typedef struct NVPW_RawMetricsConfig_GetConfigImage_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        const NVPA_RawMetricsConfig* pRawMetricsConfig;
+        /// [in] Number of bytes allocated for pBuffer
+        size_t bytesAllocated;
+        /// [out] [optional] Buffer receiving the config image
+        uint8_t* pBuffer;
+        /// [out] Count of bytes that would be copied into pBuffer
+        size_t bytesCopied;
+    } NVPW_RawMetricsConfig_GetConfigImage_Params;
+#define NVPW_RawMetricsConfig_GetConfigImage_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_RawMetricsConfig_GetConfigImage_Params, bytesCopied)
+
+    NVPA_Status NVPW_RawMetricsConfig_GetConfigImage(NVPW_RawMetricsConfig_GetConfigImage_Params* pParams);
+
+    typedef struct NVPW_RawMetricsConfig_GetNumPasses_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        const NVPA_RawMetricsConfig* pRawMetricsConfig;
+        /// [out]
+        size_t numPipelinedPasses;
+        /// [out]
+        size_t numIsolatedPasses;
+    } NVPW_RawMetricsConfig_GetNumPasses_Params;
+#define NVPW_RawMetricsConfig_GetNumPasses_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_RawMetricsConfig_GetNumPasses_Params, numIsolatedPasses)
+
+    /// Total num passes = numPipelinedPasses + numIsolatedPasses * numNestingLevels
+    NVPA_Status NVPW_RawMetricsConfig_GetNumPasses(NVPW_RawMetricsConfig_GetNumPasses_Params* pParams);
+
+    typedef struct NVPW_RawMetricsConfig_GetNumPasses_V2_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// [in]
+        const NVPA_RawMetricsConfig* pRawMetricsConfig;
+        /// [out]
+        size_t numPasses;
+    } NVPW_RawMetricsConfig_GetNumPasses_V2_Params;
+#define NVPW_RawMetricsConfig_GetNumPasses_V2_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_RawMetricsConfig_GetNumPasses_V2_Params, numPasses)
+
+    /// Total num passes = numPasses * numNestingLevels
+    NVPA_Status NVPW_RawMetricsConfig_GetNumPasses_V2(NVPW_RawMetricsConfig_GetNumPasses_V2_Params* pParams);
+
+    typedef struct NVPW_PeriodicSampler_Config_GetSocEstimatedSampleSize_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// [in] Typically created by e.g. NVPW_RawMetricsConfig_GetConfigImage(), must be align(8).
+        const uint8_t* pConfig;
+        /// [in]
+        size_t configSize;
+        /// [out]
+        size_t sampleSize;
+    } NVPW_PeriodicSampler_Config_GetSocEstimatedSampleSize_Params;
+#define NVPW_PeriodicSampler_Config_GetSocEstimatedSampleSize_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_PeriodicSampler_Config_GetSocEstimatedSampleSize_Params, sampleSize)
+
+    /// Estimate per sample records size based on a virtual device
+    NVPA_Status NVPW_PeriodicSampler_Config_GetSocEstimatedSampleSize(NVPW_PeriodicSampler_Config_GetSocEstimatedSampleSize_Params* pParams);
+
+    typedef struct NVPW_PeriodicSampler_Config_GetGpuEstimatedSampleSize_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// [in] Typically created by e.g. NVPW_RawMetricsConfig_GetConfigImage(), must be align(8).
+        const uint8_t* pConfig;
+        /// [in]
+        size_t configSize;
+        /// [out]
+        size_t sampleSize;
+    } NVPW_PeriodicSampler_Config_GetGpuEstimatedSampleSize_Params;
+#define NVPW_PeriodicSampler_Config_GetGpuEstimatedSampleSize_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_PeriodicSampler_Config_GetGpuEstimatedSampleSize_Params, sampleSize)
+
+    /// Estimate per sample records size based on a virtual device
+    NVPA_Status NVPW_PeriodicSampler_Config_GetGpuEstimatedSampleSize(NVPW_PeriodicSampler_Config_GetGpuEstimatedSampleSize_Params* pParams);
+
+/**
+ *  @}
+ ******************************************************************************/
+ 
+    typedef struct NVPW_Config_GetRawCounterInfo_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// [in]
+        const uint8_t* pConfig;
+        /// [in]
+        size_t configSize;
+        /// [in]
+        const char* pRawCounterName;
+        /// [inout] array containing indices of passes the counter resides in. 'pPassIndices' is in, '*pPassIndices' is
+        /// out.
+        size_t* pPassIndices;
+        /// [inout] if 'pPassIndices' is NULL, the count of passes this counter resides in will be returned; otherwise
+        /// it should be set to the capacity of 'pPassIndices' array, and on return, it will be overwritten to reflect
+        /// the actual count filled into 'pPassIndices'
+        size_t numPassIndices;
+    } NVPW_Config_GetRawCounterInfo_Params;
+#define NVPW_Config_GetRawCounterInfo_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_Config_GetRawCounterInfo_Params, numPassIndices)
+
+    NVPA_Status NVPW_Config_GetRawCounterInfo(NVPW_Config_GetRawCounterInfo_Params* pParams);
+
+    typedef struct NVPW_Config_GetRawCounters_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// [in]
+        const uint8_t* pConfig;
+        /// [in]
+        size_t configSize;
+        /// [in]
+        size_t passIndex;
+        /// [inout] array containing raw counter names. 'ppRawCounterNames' is in, '*ppRawCounterNames' is out.
+        const char** ppRawCounterNames;
+        /// [inout] if 'ppRawCounterNames' is NULL, the count of raw counters will be returned; otherwise it should be
+        /// set to the capacity of 'ppRawCounterNames' array, and on return, it will be overwritten to reflect the
+        /// actual count filled into 'ppRawCounterNames'
+        size_t numRawCounters;
+    } NVPW_Config_GetRawCounters_Params;
+#define NVPW_Config_GetRawCounters_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_Config_GetRawCounters_Params, numRawCounters)
+
+    NVPA_Status NVPW_Config_GetRawCounters(NVPW_Config_GetRawCounters_Params* pParams);
+
+/***************************************************************************//**
+ *  @name   CounterData Creation
+ *  @{
+ */
+
+    typedef struct NVPA_CounterDataBuilder NVPA_CounterDataBuilder;
+
+    typedef struct NVPW_CounterDataBuilder_Create_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// [out]
+        NVPA_CounterDataBuilder* pCounterDataBuilder;
+        const char* pChipName;
+    } NVPW_CounterDataBuilder_Create_Params;
+#define NVPW_CounterDataBuilder_Create_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_CounterDataBuilder_Create_Params, pChipName)
+
+    NVPA_Status NVPW_CounterDataBuilder_Create(NVPW_CounterDataBuilder_Create_Params* pParams);
+
+    typedef struct NVPW_CounterDataBuilder_Destroy_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        NVPA_CounterDataBuilder* pCounterDataBuilder;
+    } NVPW_CounterDataBuilder_Destroy_Params;
+#define NVPW_CounterDataBuilder_Destroy_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_CounterDataBuilder_Destroy_Params, pCounterDataBuilder)
+
+    NVPA_Status NVPW_CounterDataBuilder_Destroy(NVPW_CounterDataBuilder_Destroy_Params* pParams);
+
+    typedef struct NVPW_CounterDataBuilder_AddMetrics_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        NVPA_CounterDataBuilder* pCounterDataBuilder;
+        const NVPA_RawMetricRequest* pRawMetricRequests;
+        size_t numMetricRequests;
+    } NVPW_CounterDataBuilder_AddMetrics_Params;
+#define NVPW_CounterDataBuilder_AddMetrics_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_CounterDataBuilder_AddMetrics_Params, numMetricRequests)
+
+    NVPA_Status NVPW_CounterDataBuilder_AddMetrics(NVPW_CounterDataBuilder_AddMetrics_Params* pParams);
+
+    typedef struct NVPW_CounterDataBuilder_GetCounterDataPrefix_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        NVPA_CounterDataBuilder* pCounterDataBuilder;
+        /// [in] Number of bytes allocated for pBuffer
+        size_t bytesAllocated;
+        /// [out] [optional] Buffer receiving the counter data prefix
+        uint8_t* pBuffer;
+        /// [out] Count of bytes that would be copied to pBuffer
+        size_t bytesCopied;
+    } NVPW_CounterDataBuilder_GetCounterDataPrefix_Params;
+#define NVPW_CounterDataBuilder_GetCounterDataPrefix_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_CounterDataBuilder_GetCounterDataPrefix_Params, bytesCopied)
+
+    NVPA_Status NVPW_CounterDataBuilder_GetCounterDataPrefix(NVPW_CounterDataBuilder_GetCounterDataPrefix_Params* pParams);
+
+/**
+ *  @}
+ ******************************************************************************/
+ 
+/***************************************************************************//**
+ *  @name   Metrics Evaluator
+ *  @{
+ */
+
+    typedef struct NVPW_MetricsEvaluator NVPW_MetricsEvaluator;
+
+#ifndef NVPW_DIM_UNIT_DEFINED
+#define NVPW_DIM_UNIT_DEFINED
+    typedef enum NVPW_DimUnitName
+    {
+        NVPW_DIM_UNIT_INVALID = 3518299157,
+        NVPW_DIM_UNIT_UNITLESS = 2126137902,
+        NVPW_DIM_UNIT_ATTRIBUTES = 3776338729,
+        NVPW_DIM_UNIT_BYTES = 3797850191,
+        NVPW_DIM_UNIT_CTAS = 1960564139,
+        NVPW_DIM_UNIT_CTC_CYCLES = 2224883873,
+        NVPW_DIM_UNIT_DRAM_CYCLES = 2650981327,
+        NVPW_DIM_UNIT_FBP_CYCLES = 1785238957,
+        NVPW_DIM_UNIT_FE_OPS = 2919159083,
+        NVPW_DIM_UNIT_GPC_CYCLES = 1222631184,
+        NVPW_DIM_UNIT_IDC_REQUESTS = 2012649669,
+        NVPW_DIM_UNIT_INSTRUCTIONS = 1418625543,
+        NVPW_DIM_UNIT_KILOBYTES = 1335980302,
+        NVPW_DIM_UNIT_L1DATA_BANK_ACCESSES = 1479493682,
+        NVPW_DIM_UNIT_L1DATA_BANK_CONFLICTS = 3433170787,
+        NVPW_DIM_UNIT_L1TEX_REQUESTS = 1306473767,
+        NVPW_DIM_UNIT_L1TEX_TAGS = 26573010,
+        NVPW_DIM_UNIT_L1TEX_WAVEFRONTS = 129373765,
+        NVPW_DIM_UNIT_L2_REQUESTS = 1143695106,
+        NVPW_DIM_UNIT_L2_SECTORS = 3424101564,
+        NVPW_DIM_UNIT_L2_TAGS = 3755612781,
+        NVPW_DIM_UNIT_LRC_REQUESTS = 2280914327,
+        NVPW_DIM_UNIT_LRC_SECTORS = 7212034,
+        NVPW_DIM_UNIT_MCC_CYCLES = 1826685787,
+        NVPW_DIM_UNIT_NANOSECONDS = 3047500672,
+        NVPW_DIM_UNIT_NVDLA_CYCLES = 3374059789,
+        NVPW_DIM_UNIT_NVENC_CYCLES = 2267185244,
+        NVPW_DIM_UNIT_NVLRX_CYCLES = 4059934930,
+        NVPW_DIM_UNIT_NVLTX_CYCLES = 1814350488,
+        NVPW_DIM_UNIT_OFA_CYCLES = 4290210307,
+        NVPW_DIM_UNIT_PCIE_CYCLES = 1230450943,
+        NVPW_DIM_UNIT_PERCENT = 1284354694,
+        NVPW_DIM_UNIT_PIXELS = 4227616663,
+        NVPW_DIM_UNIT_PIXEL_SHADER_BARRIERS = 3705502518,
+        NVPW_DIM_UNIT_PRIMITIVES = 2373084002,
+        NVPW_DIM_UNIT_PVAVPU_CYCLES = 2238259366,
+        NVPW_DIM_UNIT_PVA_CYCLES = 202044173,
+        NVPW_DIM_UNIT_QUADS = 1539753497,
+        NVPW_DIM_UNIT_REGISTERS = 2837260947,
+        NVPW_DIM_UNIT_SAMPLES = 746046551,
+        NVPW_DIM_UNIT_SECONDS = 1164825258,
+        NVPW_DIM_UNIT_SYSL2_REQUESTS = 2165109286,
+        NVPW_DIM_UNIT_SYSL2_SECTORS = 2268734175,
+        NVPW_DIM_UNIT_SYSL2_TAGS = 3308651352,
+        NVPW_DIM_UNIT_SYSLRC_REQUESTS = 3328245480,
+        NVPW_DIM_UNIT_SYSLRC_SECTORS = 1190477493,
+        NVPW_DIM_UNIT_SYS_CYCLES = 3310821688,
+        NVPW_DIM_UNIT_TEXELS = 1293214069,
+        NVPW_DIM_UNIT_THREADS = 164261907,
+        NVPW_DIM_UNIT_TMEM_ACCESSES = 3742902067,
+        NVPW_DIM_UNIT_VERTICES = 1873662209,
+        NVPW_DIM_UNIT_VIC_CYCLES = 103143588,
+        NVPW_DIM_UNIT_WARPS = 97951949,
+        NVPW_DIM_UNIT_WORKIDS = 1971113483,
+        NVPW_DIM_UNIT_WORKLOADS = 1728142656
+    } NVPW_DimUnitName;
+#endif //NVPW_DIM_UNIT_DEFINED
+
+#ifndef NVPW_HW_UNIT_DEFINED
+#define NVPW_HW_UNIT_DEFINED
+    typedef enum NVPW_HwUnit
+    {
+        NVPW_HW_UNIT_INVALID = 3498035701,
+        NVPW_HW_UNIT_CROP = 2872137846,
+        NVPW_HW_UNIT_CTC = 4123164475,
+        NVPW_HW_UNIT_DRAM = 1662616918,
+        NVPW_HW_UNIT_DRAMC = 1401232876,
+        NVPW_HW_UNIT_FBP = 2947194306,
+        NVPW_HW_UNIT_FBPA = 690045803,
+        NVPW_HW_UNIT_FE = 2204924321,
+        NVPW_HW_UNIT_GPC = 1911735839,
+        NVPW_HW_UNIT_GPU = 1014363534,
+        NVPW_HW_UNIT_GR = 2933618517,
+        NVPW_HW_UNIT_IDC = 842765289,
+        NVPW_HW_UNIT_L1TEX = 893940957,
+        NVPW_HW_UNIT_LRC = 4004756136,
+        NVPW_HW_UNIT_LTS = 2333266697,
+        NVPW_HW_UNIT_MCC = 3980130194,
+        NVPW_HW_UNIT_NVDLA = 4201167892,
+        NVPW_HW_UNIT_NVENC = 207708260,
+        NVPW_HW_UNIT_NVLRX = 3091684901,
+        NVPW_HW_UNIT_NVLTX = 869679659,
+        NVPW_HW_UNIT_OFA = 70307371,
+        NVPW_HW_UNIT_PCIE = 3433264174,
+        NVPW_HW_UNIT_PDA = 345193251,
+        NVPW_HW_UNIT_PES = 804128425,
+        NVPW_HW_UNIT_PROP = 3339255507,
+        NVPW_HW_UNIT_PVA = 2565499490,
+        NVPW_HW_UNIT_PVAVPU = 1656645655,
+        NVPW_HW_UNIT_RASTER = 187932504,
+        NVPW_HW_UNIT_SM = 724224710,
+        NVPW_HW_UNIT_SMSP = 2837616917,
+        NVPW_HW_UNIT_SYS = 768990063,
+        NVPW_HW_UNIT_SYSLRC = 3247626950,
+        NVPW_HW_UNIT_SYSLTS = 4137740217,
+        NVPW_HW_UNIT_TPC = 1889024613,
+        NVPW_HW_UNIT_VAF = 753670509,
+        NVPW_HW_UNIT_VIC = 322439594,
+        NVPW_HW_UNIT_VPC = 275561583,
+        NVPW_HW_UNIT_ZCULL = 2401248356,
+        NVPW_HW_UNIT_ZROP = 979500456
+    } NVPW_HwUnit;
+#endif //NVPW_HW_UNIT_DEFINED
+
+    typedef enum NVPW_RollupOp
+    {
+        NVPW_ROLLUP_OP_AVG = 0,
+        NVPW_ROLLUP_OP_MAX,
+        NVPW_ROLLUP_OP_MIN,
+        NVPW_ROLLUP_OP_SUM,
+        NVPW_ROLLUP_OP__COUNT
+    } NVPW_RollupOp;
+
+    typedef enum NVPW_MetricType
+    {
+        NVPW_METRIC_TYPE_COUNTER = 0,
+        NVPW_METRIC_TYPE_RATIO,
+        NVPW_METRIC_TYPE_THROUGHPUT,
+        NVPW_METRIC_TYPE__COUNT
+    } NVPW_MetricType;
+
+    typedef enum NVPW_Submetric
+    {
+        NVPW_SUBMETRIC_NONE = 0,
+        NVPW_SUBMETRIC_PEAK_SUSTAINED = 1,
+        NVPW_SUBMETRIC_PEAK_SUSTAINED_ACTIVE = 2,
+        NVPW_SUBMETRIC_PEAK_SUSTAINED_ACTIVE_PER_SECOND = 3,
+        NVPW_SUBMETRIC_PEAK_SUSTAINED_ELAPSED = 4,
+        NVPW_SUBMETRIC_PEAK_SUSTAINED_ELAPSED_PER_SECOND = 5,
+        NVPW_SUBMETRIC_PEAK_SUSTAINED_FRAME = 6,
+        NVPW_SUBMETRIC_PEAK_SUSTAINED_FRAME_PER_SECOND = 7,
+        NVPW_SUBMETRIC_PEAK_SUSTAINED_REGION = 8,
+        NVPW_SUBMETRIC_PEAK_SUSTAINED_REGION_PER_SECOND = 9,
+        NVPW_SUBMETRIC_PER_CYCLE_ACTIVE = 10,
+        NVPW_SUBMETRIC_PER_CYCLE_ELAPSED = 11,
+        NVPW_SUBMETRIC_PER_CYCLE_IN_FRAME = 12,
+        NVPW_SUBMETRIC_PER_CYCLE_IN_REGION = 13,
+        NVPW_SUBMETRIC_PER_SECOND = 14,
+        NVPW_SUBMETRIC_PCT_OF_PEAK_SUSTAINED_ACTIVE = 15,
+        NVPW_SUBMETRIC_PCT_OF_PEAK_SUSTAINED_ELAPSED = 16,
+        NVPW_SUBMETRIC_PCT_OF_PEAK_SUSTAINED_FRAME = 17,
+        NVPW_SUBMETRIC_PCT_OF_PEAK_SUSTAINED_REGION = 18,
+        NVPW_SUBMETRIC_MAX_RATE = 19,
+        NVPW_SUBMETRIC_PCT = 20,
+        NVPW_SUBMETRIC_RATIO = 21,
+        NVPW_SUBMETRIC__COUNT
+    } NVPW_Submetric;
+
+    typedef struct NVPW_MetricEvalRequest
+    {
+        /// the metric index as in 'NVPW_MetricsEvaluator_GetMetricNames'
+        size_t metricIndex;
+        /// one of 'NVPW_MetricType'
+        uint8_t metricType;
+        /// one of 'NVPW_RollupOp', required for Counter and Throughput, doesn't apply to Ratio
+        uint8_t rollupOp;
+        /// one of 'NVPW_Submetric', required for Ratio and Throughput, optional for Counter
+        uint16_t submetric;
+    } NVPW_MetricEvalRequest;
+#define NVPW_MetricEvalRequest_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_MetricEvalRequest, submetric)
+
+    typedef struct NVPW_DimUnitFactor
+    {
+        /// one of 'NVPW_DimUnitName'
+        uint32_t dimUnit;
+        int8_t exponent;
+    } NVPW_DimUnitFactor;
+#define NVPW_DimUnitFactor_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_DimUnitFactor, exponent)
+
+    typedef struct NVPW_MetricsEvaluator_Destroy_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// [in]
+        struct NVPW_MetricsEvaluator* pMetricsEvaluator;
+    } NVPW_MetricsEvaluator_Destroy_Params;
+#define NVPW_MetricsEvaluator_Destroy_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_MetricsEvaluator_Destroy_Params, pMetricsEvaluator)
+
+    NVPA_Status NVPW_MetricsEvaluator_Destroy(NVPW_MetricsEvaluator_Destroy_Params* pParams);
+
+    typedef struct NVPW_MetricsEvaluator_GetMetricNames_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// [in]
+        struct NVPW_MetricsEvaluator* pMetricsEvaluator;
+        /// [in] one of 'NVPW_MetricType'
+        uint8_t metricType;
+        /// [out]
+        const char* pMetricNames;
+        /// [out]
+        const size_t* pMetricNameBeginIndices;
+        /// [out]
+        size_t numMetrics;
+    } NVPW_MetricsEvaluator_GetMetricNames_Params;
+#define NVPW_MetricsEvaluator_GetMetricNames_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_MetricsEvaluator_GetMetricNames_Params, numMetrics)
+
+    NVPA_Status NVPW_MetricsEvaluator_GetMetricNames(NVPW_MetricsEvaluator_GetMetricNames_Params* pParams);
+
+    typedef struct NVPW_MetricsEvaluator_GetMetricTypeAndIndex_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// [in]
+        struct NVPW_MetricsEvaluator* pMetricsEvaluator;
+        /// [in] can be either a base metric or a metric
+        const char* pMetricName;
+        /// [out] one of 'NVPW_MetricType'
+        uint8_t metricType;
+        /// [out] the metric index as in 'NVPW_MetricsEvaluator_GetMetricNames'
+        size_t metricIndex;
+    } NVPW_MetricsEvaluator_GetMetricTypeAndIndex_Params;
+#define NVPW_MetricsEvaluator_GetMetricTypeAndIndex_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_MetricsEvaluator_GetMetricTypeAndIndex_Params, metricIndex)
+
+    NVPA_Status NVPW_MetricsEvaluator_GetMetricTypeAndIndex(NVPW_MetricsEvaluator_GetMetricTypeAndIndex_Params* pParams);
+
+    typedef struct NVPW_MetricsEvaluator_ConvertMetricNameToMetricEvalRequest_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// [in]
+        struct NVPW_MetricsEvaluator* pMetricsEvaluator;
+        /// [in]
+        const char* pMetricName;
+        /// [inout] 'pMetricEvalRequest' is in, '*pMetricEvalRequest' is out
+        struct NVPW_MetricEvalRequest* pMetricEvalRequest;
+        /// [in] set to 'NVPW_MetricEvalRequest_STRUCT_SIZE'
+        size_t metricEvalRequestStructSize;
+    } NVPW_MetricsEvaluator_ConvertMetricNameToMetricEvalRequest_Params;
+#define NVPW_MetricsEvaluator_ConvertMetricNameToMetricEvalRequest_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_MetricsEvaluator_ConvertMetricNameToMetricEvalRequest_Params, metricEvalRequestStructSize)
+
+    NVPA_Status NVPW_MetricsEvaluator_ConvertMetricNameToMetricEvalRequest(NVPW_MetricsEvaluator_ConvertMetricNameToMetricEvalRequest_Params* pParams);
+
+    typedef struct NVPW_MetricsEvaluator_HwUnitToString_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// [in]
+        struct NVPW_MetricsEvaluator* pMetricsEvaluator;
+        /// [in] one of 'NVPW_HwUnit'
+        uint32_t hwUnit;
+        /// [out]
+        const char* pHwUnitName;
+    } NVPW_MetricsEvaluator_HwUnitToString_Params;
+#define NVPW_MetricsEvaluator_HwUnitToString_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_MetricsEvaluator_HwUnitToString_Params, pHwUnitName)
+
+    NVPA_Status NVPW_MetricsEvaluator_HwUnitToString(NVPW_MetricsEvaluator_HwUnitToString_Params* pParams);
+
+    typedef struct NVPW_MetricsEvaluator_GetCounterProperties_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// [in]
+        struct NVPW_MetricsEvaluator* pMetricsEvaluator;
+        /// [in] the metric index as in 'NVPW_MetricsEvaluator_GetMetricNames'
+        size_t counterIndex;
+        /// [out]
+        const char* pDescription;
+        /// [out] one of 'NVPW_HwUnit'
+        uint32_t hwUnit;
+    } NVPW_MetricsEvaluator_GetCounterProperties_Params;
+#define NVPW_MetricsEvaluator_GetCounterProperties_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_MetricsEvaluator_GetCounterProperties_Params, hwUnit)
+
+    NVPA_Status NVPW_MetricsEvaluator_GetCounterProperties(NVPW_MetricsEvaluator_GetCounterProperties_Params* pParams);
+
+    typedef struct NVPW_MetricsEvaluator_GetRatioMetricProperties_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// [in]
+        struct NVPW_MetricsEvaluator* pMetricsEvaluator;
+        /// [in] the metric index as in 'NVPW_MetricsEvaluator_GetMetricNames'
+        size_t ratioMetricIndex;
+        /// [out]
+        const char* pDescription;
+        /// [out]
+        uint64_t hwUnit;
+    } NVPW_MetricsEvaluator_GetRatioMetricProperties_Params;
+#define NVPW_MetricsEvaluator_GetRatioMetricProperties_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_MetricsEvaluator_GetRatioMetricProperties_Params, hwUnit)
+
+    NVPA_Status NVPW_MetricsEvaluator_GetRatioMetricProperties(NVPW_MetricsEvaluator_GetRatioMetricProperties_Params* pParams);
+
+    typedef struct NVPW_MetricsEvaluator_GetThroughputMetricProperties_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// [in]
+        struct NVPW_MetricsEvaluator* pMetricsEvaluator;
+        /// [in] the metric index as in 'NVPW_MetricsEvaluator_GetMetricNames'
+        size_t throughputMetricIndex;
+        /// [out]
+        const char* pDescription;
+        /// [out]
+        uint32_t hwUnit;
+        /// [out] number of constituent counters for the throughput metric
+        size_t numCounters;
+        /// [out] metric indices as in 'NVPW_MetricsEvaluator_GetMetricNames', valid if 'numCounters' > 0, otherwise
+        /// returned as nullptr
+        const size_t* pCounterIndices;
+        /// [out] number of constituent sub-throughputs for the throughput metric
+        size_t numSubThroughputs;
+        /// [out] metric indices as in 'NVPW_MetricsEvaluator_GetMetricNames', valid if 'numSubThroughputs' > 0,
+        /// otherwise returned as nullptr
+        const size_t* pSubThroughputIndices;
+    } NVPW_MetricsEvaluator_GetThroughputMetricProperties_Params;
+#define NVPW_MetricsEvaluator_GetThroughputMetricProperties_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_MetricsEvaluator_GetThroughputMetricProperties_Params, pSubThroughputIndices)
+
+    NVPA_Status NVPW_MetricsEvaluator_GetThroughputMetricProperties(NVPW_MetricsEvaluator_GetThroughputMetricProperties_Params* pParams);
+
+    typedef struct NVPW_MetricsEvaluator_GetSupportedSubmetrics_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// [in]
+        struct NVPW_MetricsEvaluator* pMetricsEvaluator;
+        /// [in] one of 'NVPW_MetricType'
+        uint8_t metricType;
+        /// [out] an array of 'NVPW_Submetric'
+        const uint16_t* pSupportedSubmetrics;
+        /// [out]
+        size_t numSupportedSubmetrics;
+    } NVPW_MetricsEvaluator_GetSupportedSubmetrics_Params;
+#define NVPW_MetricsEvaluator_GetSupportedSubmetrics_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_MetricsEvaluator_GetSupportedSubmetrics_Params, numSupportedSubmetrics)
+
+    NVPA_Status NVPW_MetricsEvaluator_GetSupportedSubmetrics(NVPW_MetricsEvaluator_GetSupportedSubmetrics_Params* pParams);
+
+    typedef struct NVPW_MetricsEvaluator_GetMetricRawDependencies_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// [in]
+        struct NVPW_MetricsEvaluator* pMetricsEvaluator;
+        /// [in]
+        const struct NVPW_MetricEvalRequest* pMetricEvalRequests;
+        /// [in]
+        size_t numMetricEvalRequests;
+        /// [in] set to 'NVPW_MetricEvalRequest_STRUCT_SIZE'
+        size_t metricEvalRequestStructSize;
+        /// [in] set to sizeof('NVPW_MetricEvalRequest')
+        size_t metricEvalRequestStrideSize;
+        /// [inout] 'ppRawDependencies' is in, '*ppRawDependencies' is out
+        const char** ppRawDependencies;
+        /// [inout] if 'ppRawDependencies' is NULL, number of raw dependencies available will be returned; otherwise it
+        /// should be set to the number of elements allocated for 'ppRawDependencies', and on return, it will be
+        /// overwritten by number of elements copied to 'ppRawDependencies'
+        size_t numRawDependencies;
+        /// [inout] 'ppOptionalRawDependencies' is in, '*ppOptionalRawDependencies' is out
+        const char** ppOptionalRawDependencies;
+        /// [inout] if 'ppOptionalRawDependencies' is NULL, number of optional raw dependencies available will be
+        /// returned; otherwise it should be set to the number of elements allocated for 'ppOptionalRawDependencies',
+        /// and on return, it will be overwritten by number of elements copied to 'ppOptionalRawDependencies'
+        size_t numOptionalRawDependencies;
+    } NVPW_MetricsEvaluator_GetMetricRawDependencies_Params;
+#define NVPW_MetricsEvaluator_GetMetricRawDependencies_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_MetricsEvaluator_GetMetricRawDependencies_Params, numOptionalRawDependencies)
+
+    NVPA_Status NVPW_MetricsEvaluator_GetMetricRawDependencies(NVPW_MetricsEvaluator_GetMetricRawDependencies_Params* pParams);
+
+    typedef struct NVPW_MetricsEvaluator_DimUnitToString_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// [in]
+        struct NVPW_MetricsEvaluator* pMetricsEvaluator;
+        /// [in] one of 'NVPW_DimUnitName'
+        uint32_t dimUnit;
+        /// [out]
+        const char* pSingularName;
+        /// [out]
+        const char* pPluralName;
+    } NVPW_MetricsEvaluator_DimUnitToString_Params;
+#define NVPW_MetricsEvaluator_DimUnitToString_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_MetricsEvaluator_DimUnitToString_Params, pPluralName)
+
+    NVPA_Status NVPW_MetricsEvaluator_DimUnitToString(NVPW_MetricsEvaluator_DimUnitToString_Params* pParams);
+
+    typedef struct NVPW_MetricsEvaluator_GetMetricDimUnits_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// [in]
+        struct NVPW_MetricsEvaluator* pMetricsEvaluator;
+        /// [in]
+        const struct NVPW_MetricEvalRequest* pMetricEvalRequest;
+        /// [in] set to 'NVPW_MetricEvalRequest_STRUCT_SIZE'
+        size_t metricEvalRequestStructSize;
+        /// [inout] 'pDimUnits' is in, '*pDimUnits' is out
+        NVPW_DimUnitFactor* pDimUnits;
+        /// [inout] if 'pDimUnits' is NULL, number of dim-units available will be returned; otherwise it should be set
+        /// to the number of elements allocated for 'pDimUnits', and on return, it will be overwritten by number of
+        /// elements copied to 'pDimUnits'
+        size_t numDimUnits;
+        /// [in] set to 'NVPW_DimUnitFactor_STRUCT_SIZE'
+        size_t dimUnitFactorStructSize;
+    } NVPW_MetricsEvaluator_GetMetricDimUnits_Params;
+#define NVPW_MetricsEvaluator_GetMetricDimUnits_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_MetricsEvaluator_GetMetricDimUnits_Params, dimUnitFactorStructSize)
+
+    NVPA_Status NVPW_MetricsEvaluator_GetMetricDimUnits(NVPW_MetricsEvaluator_GetMetricDimUnits_Params* pParams);
+
+    typedef struct NVPW_MetricsEvaluator_SetUserData_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// [in]
+        struct NVPW_MetricsEvaluator* pMetricsEvaluator;
+        /// [in] duration in ns of user defined frame
+        double frameDuration;
+        /// [in] duration in ns of user defined region
+        double regionDuration;
+        /// [in]
+        NVPA_Bool isolated;
+    } NVPW_MetricsEvaluator_SetUserData_Params;
+#define NVPW_MetricsEvaluator_SetUserData_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_MetricsEvaluator_SetUserData_Params, isolated)
+
+    NVPA_Status NVPW_MetricsEvaluator_SetUserData(NVPW_MetricsEvaluator_SetUserData_Params* pParams);
+
+    typedef struct NVPW_MetricsEvaluator_EvaluateToGpuValues_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// [in]
+        struct NVPW_MetricsEvaluator* pMetricsEvaluator;
+        /// [in]
+        const struct NVPW_MetricEvalRequest* pMetricEvalRequests;
+        /// [in]
+        size_t numMetricEvalRequests;
+        /// [in] set to 'NVPW_MetricEvalRequest_STRUCT_SIZE'
+        size_t metricEvalRequestStructSize;
+        /// [in] set to sizeof('NVPW_MetricEvalRequest')
+        size_t metricEvalRequestStrideSize;
+        /// [in]
+        const uint8_t* pCounterDataImage;
+        /// [in]
+        size_t counterDataImageSize;
+        /// [in]
+        size_t rangeIndex;
+        /// [in]
+        NVPA_Bool isolated;
+        /// [inout] 'pMetricValues' is in, '*pMetricValues' is out
+        double* pMetricValues;
+    } NVPW_MetricsEvaluator_EvaluateToGpuValues_Params;
+#define NVPW_MetricsEvaluator_EvaluateToGpuValues_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_MetricsEvaluator_EvaluateToGpuValues_Params, pMetricValues)
+
+    NVPA_Status NVPW_MetricsEvaluator_EvaluateToGpuValues(NVPW_MetricsEvaluator_EvaluateToGpuValues_Params* pParams);
+
+    typedef struct NVPW_MetricsEvaluator_SetDeviceAttributes_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// [in]
+        struct NVPW_MetricsEvaluator* pMetricsEvaluator;
+        /// [in]
+        const uint8_t* pCounterDataImage;
+        /// [in]
+        size_t counterDataImageSize;
+    } NVPW_MetricsEvaluator_SetDeviceAttributes_Params;
+#define NVPW_MetricsEvaluator_SetDeviceAttributes_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_MetricsEvaluator_SetDeviceAttributes_Params, counterDataImageSize)
+
+    NVPA_Status NVPW_MetricsEvaluator_SetDeviceAttributes(NVPW_MetricsEvaluator_SetDeviceAttributes_Params* pParams);
+
+/**
+ *  @}
+ ******************************************************************************/
+ 
+
+#endif // NVPERF_HOST_API_DEFINED
+
+
+
+
+#ifdef __cplusplus
+} // extern "C"
+#endif
+
+#if defined(__GNUC__) && defined(NVPA_SHARED_LIB)
+    #pragma GCC visibility pop
+#endif
+
+#endif // NVPERF_HOST_H

Prism/LLaDA/LLaDA_Prism/.venv/lib/python3.12/site-packages/nvidia/cuda_cupti/include/nvperf_target.h

@@ -0,0 +1,626 @@
+#ifndef NVPERF_TARGET_H
+#define NVPERF_TARGET_H
+
+/*
+ * Copyright 2014-2024 NVIDIA Corporation.  All rights reserved.
+ *
+ * NOTICE TO USER:
+ *
+ * This source code is subject to NVIDIA ownership rights under U.S. and
+ * international Copyright laws.
+ *
+ * This software and the information contained herein is PROPRIETARY and
+ * CONFIDENTIAL to NVIDIA and is being provided under the terms and conditions
+ * of a form of NVIDIA software license agreement.
+ *
+ * NVIDIA MAKES NO REPRESENTATION ABOUT THE SUITABILITY OF THIS SOURCE
+ * CODE FOR ANY PURPOSE.  IT IS PROVIDED "AS IS" WITHOUT EXPRESS OR
+ * IMPLIED WARRANTY OF ANY KIND.  NVIDIA DISCLAIMS ALL WARRANTIES WITH
+ * REGARD TO THIS SOURCE CODE, INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY, NONINFRINGEMENT, AND FITNESS FOR A PARTICULAR PURPOSE.
+ * IN NO EVENT SHALL NVIDIA BE LIABLE FOR ANY SPECIAL, INDIRECT, INCIDENTAL,
+ * OR CONSEQUENTIAL DAMAGES, OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS
+ * OF USE, DATA OR PROFITS,  WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE
+ * OR OTHER TORTIOUS ACTION,  ARISING OUT OF OR IN CONNECTION WITH THE USE
+ * OR PERFORMANCE OF THIS SOURCE CODE.
+ *
+ * U.S. Government End Users.   This source code is a "commercial item" as
+ * that term is defined at  48 C.F.R. 2.101 (OCT 1995), consisting  of
+ * "commercial computer  software"  and "commercial computer software
+ * documentation" as such terms are  used in 48 C.F.R. 12.212 (SEPT 1995)
+ * and is provided to the U.S. Government only as a commercial end item.
+ * Consistent with 48 C.F.R.12.212 and 48 C.F.R. 227.7202-1 through
+ * 227.7202-4 (JUNE 1995), all U.S. Government End Users acquire the
+ * source code with only those rights set forth herein.
+ *
+ * Any use of this source code in individual and commercial software must
+ * include, in the user documentation and internal comments to the code,
+ * the above Disclaimer and U.S. Government End Users Notice.
+ */
+
+#include <stddef.h>
+#include <stdint.h>
+#include "nvperf_common.h"
+
+#if defined(__GNUC__) && defined(NVPA_SHARED_LIB)
+    #pragma GCC visibility push(default)
+    #if !defined(NVPW_LOCAL)
+        #define NVPW_LOCAL __attribute__ ((visibility ("hidden")))
+    #endif
+#else
+    #if !defined(NVPW_LOCAL)
+        #define NVPW_LOCAL
+    #endif
+#endif
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/**
+ *  @file   nvperf_target.h
+ */
+
+#ifndef NVPW_GPU_ARCHITECTURE_SUPPORT_LEVEL_DEFINED
+#define NVPW_GPU_ARCHITECTURE_SUPPORT_LEVEL_DEFINED
+    /// GPU architecture support level
+    typedef enum NVPW_GpuArchitectureSupportLevel
+    {
+        NVPW_GPU_ARCHITECTURE_SUPPORT_LEVEL_UNKNOWN = 0,
+        NVPW_GPU_ARCHITECTURE_SUPPORT_LEVEL_UNSUPPORTED,
+        NVPW_GPU_ARCHITECTURE_SUPPORT_LEVEL_SUPPORTED
+    } NVPW_GpuArchitectureSupportLevel;
+#endif //NVPW_GPU_ARCHITECTURE_SUPPORT_LEVEL_DEFINED
+
+#ifndef NVPW_SLI_SUPPORT_LEVEL_DEFINED
+#define NVPW_SLI_SUPPORT_LEVEL_DEFINED
+    /// SLI configuration support level
+    typedef enum NVPW_SliSupportLevel
+    {
+        NVPW_SLI_SUPPORT_LEVEL_UNKNOWN = 0,
+        NVPW_SLI_SUPPORT_LEVEL_UNSUPPORTED,
+        /// Only Non-SLI configurations are supported.
+        NVPW_SLI_SUPPORT_LEVEL_SUPPORTED_NON_SLI_CONFIGURATION
+    } NVPW_SliSupportLevel;
+#endif //NVPW_SLI_SUPPORT_LEVEL_DEFINED
+
+#ifndef NVPW_VGPU_SUPPORT_LEVEL_DEFINED
+#define NVPW_VGPU_SUPPORT_LEVEL_DEFINED
+    /// Virtualized GPU configuration support level
+    typedef enum NVPW_VGpuSupportLevel
+    {
+        NVPW_VGPU_SUPPORT_LEVEL_UNKNOWN = 0,
+        NVPW_VGPU_SUPPORT_LEVEL_UNSUPPORTED,
+        /// Supported but not allowed by system admin.
+        NVPW_VGPU_SUPPORT_LEVEL_SUPPORTED_DISALLOWED,
+        NVPW_VGPU_SUPPORT_LEVEL_SUPPORTED_ALLOWED,
+        NVPW_VGPU_SUPPORT_LEVEL_SUPPORTED_NON_VGPU_CONFIGURATION
+    } NVPW_VGpuSupportLevel;
+#endif //NVPW_VGPU_SUPPORT_LEVEL_DEFINED
+
+#ifndef NVPW_CONF_COMPUTE_SUPPORT_LEVEL_DEFINED
+#define NVPW_CONF_COMPUTE_SUPPORT_LEVEL_DEFINED
+    /// Confidential Compute mode support level
+    typedef enum NVPW_ConfidentialComputeSupportLevel
+    {
+        NVPW_CONF_COMPUTE_SUPPORT_LEVEL_UNKNOWN = 0,
+        NVPW_CONF_COMPUTE_SUPPORT_LEVEL_UNSUPPORTED,
+        NVPW_CONF_COMPUTE_SUPPORT_LEVEL_SUPPORTED_NON_CONF_COMPUTE_CONFIGURATION,
+        NVPW_CONF_COMPUTE_SUPPORT_LEVEL_SUPPORTED_CONF_COMPUTE_DEVTOOLS_MODE
+    } NVPW_ConfidentialComputeSupportLevel;
+#endif //NVPW_CONF_COMPUTE_SUPPORT_LEVEL_DEFINED
+
+#ifndef NVPW_CMP_SUPPORT_LEVEL_DEFINED
+#define NVPW_CMP_SUPPORT_LEVEL_DEFINED
+    /// CMP support level
+    typedef enum NVPW_CmpSupportLevel
+    {
+        NVPW_CMP_SUPPORT_LEVEL_UNKNOWN = 0,
+        NVPW_CMP_SUPPORT_LEVEL_UNSUPPORTED,
+        NVPW_CMP_SUPPORT_LEVEL_SUPPORTED_NON_CMP_CONFIGURATON
+    } NVPW_CmpSupportLevel;
+#endif //NVPW_CMP_SUPPORT_LEVEL_DEFINED
+
+#ifndef NVPW_WSL_SUPPORT_LEVEL_DEFINED
+#define NVPW_WSL_SUPPORT_LEVEL_DEFINED
+    /// WSL support level
+    typedef enum NVPW_WslSupportLevel
+    {
+        NVPW_WSL_SUPPORT_LEVEL_UNKNOWN = 0,
+        NVPW_WSL_SUPPORT_LEVEL_UNSUPPORTED_INSUFFICIENT_DRIVER_VERSION,
+        NVPW_WSL_SUPPORT_LEVEL_SUPPORTED,
+        NVPW_WSL_SUPPORT_LEVEL_SUPPORTED_NON_WSL_CONFIGURATION
+    } NVPW_WslSupportLevel;
+#endif //NVPW_WSL_SUPPORT_LEVEL_DEFINED
+
+#ifndef NVPW_MIG_SUPPORT_LEVEL_DEFINED
+#define NVPW_MIG_SUPPORT_LEVEL_DEFINED
+    /// MIG support level
+    typedef enum NVPW_MigSupportLevel
+    {
+        NVPW_MIG_SUPPORT_LEVEL_UNKNOWN = 0,
+        NVPW_MIG_SUPPORT_LEVEL_UNSUPPORTED,
+        NVPW_MIG_SUPPORT_LEVEL_SUPPORTED,
+        NVPW_MIG_SUPPORT_LEVEL_SUPPORTED_NON_MIG_CONFIGURATION
+    } NVPW_MigSupportLevel;
+#endif //NVPW_MIG_SUPPORT_LEVEL_DEFINED
+
+    typedef struct NVPW_InitializeTarget_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+    } NVPW_InitializeTarget_Params;
+#define NVPW_InitializeTarget_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_InitializeTarget_Params, pPriv)
+
+    /// Load the target library.
+    NVPA_Status NVPW_InitializeTarget(NVPW_InitializeTarget_Params* pParams);
+
+    typedef struct NVPW_GetDeviceCount_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        size_t numDevices;
+    } NVPW_GetDeviceCount_Params;
+#define NVPW_GetDeviceCount_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_GetDeviceCount_Params, numDevices)
+
+    NVPA_Status NVPW_GetDeviceCount(NVPW_GetDeviceCount_Params* pParams);
+
+    typedef struct NVPW_Device_GetNames_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        size_t deviceIndex;
+        const char* pDeviceName;
+        const char* pChipName;
+    } NVPW_Device_GetNames_Params;
+#define NVPW_Device_GetNames_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_Device_GetNames_Params, pChipName)
+
+    NVPA_Status NVPW_Device_GetNames(NVPW_Device_GetNames_Params* pParams);
+
+    typedef struct NVPW_PciBusId
+    {
+        /// The PCI domain on which the device bus resides.
+        uint32_t domain;
+        ///  The bus on which the device resides.
+        uint16_t bus;
+        /// device ID.
+        uint16_t device;
+    } NVPW_PciBusId;
+#define NVPW_PciBusId_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_PciBusId, device)
+
+    typedef struct NVPW_Device_GetPciBusIds_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// [in] caller-allocated array of NVPW_PciBusId, indexed by NVPW deviceIndex
+        NVPW_PciBusId* pBusIds;
+        /// [in] size of the pBusIDs array; use result from NVPW_GetDeviceCount
+        size_t numDevices;
+    } NVPW_Device_GetPciBusIds_Params;
+#define NVPW_Device_GetPciBusIds_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_Device_GetPciBusIds_Params, numDevices)
+
+    NVPA_Status NVPW_Device_GetPciBusIds(NVPW_Device_GetPciBusIds_Params* pParams);
+
+
+#define NVPW_DEVICE_MIG_GPU_INSTANCE_ID_INVALID     0xFFFFFFFFu
+#define NVPW_DEVICE_MIG_GPU_INSTANCE_ID_FULLCHIP    0xFFFFFFFEu
+
+
+    typedef struct NVPW_Device_GetMigAttributes_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// [in]
+        size_t deviceIndex;
+        /// [out]
+        NVPA_Bool isMigPartition;
+        /// [out]
+        uint32_t gpuInstanceId;
+        /// [out]
+        uint32_t computeInstanceId;
+    } NVPW_Device_GetMigAttributes_Params;
+#define NVPW_Device_GetMigAttributes_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_Device_GetMigAttributes_Params, computeInstanceId)
+
+    NVPA_Status NVPW_Device_GetMigAttributes(NVPW_Device_GetMigAttributes_Params* pParams);
+
+    typedef struct NVPW_Adapter_GetDeviceIndex_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// [in]
+        struct IDXGIAdapter* pAdapter;
+        /// [in]
+        size_t sliIndex;
+        /// [out]
+        size_t deviceIndex;
+    } NVPW_Adapter_GetDeviceIndex_Params;
+#define NVPW_Adapter_GetDeviceIndex_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_Adapter_GetDeviceIndex_Params, deviceIndex)
+
+    NVPA_Status NVPW_Adapter_GetDeviceIndex(NVPW_Adapter_GetDeviceIndex_Params* pParams);
+
+    typedef struct NVPW_CounterData_GetNumRanges_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        const uint8_t* pCounterDataImage;
+        size_t numRanges;
+    } NVPW_CounterData_GetNumRanges_Params;
+#define NVPW_CounterData_GetNumRanges_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_CounterData_GetNumRanges_Params, numRanges)
+
+    NVPA_Status NVPW_CounterData_GetNumRanges(NVPW_CounterData_GetNumRanges_Params* pParams);
+
+    typedef struct NVPW_CounterData_GetChipName_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// [in]
+        const uint8_t* pCounterDataImage;
+        /// [in]
+        size_t counterDataImageSize;
+        /// [out]
+        const char* pChipName;
+    } NVPW_CounterData_GetChipName_Params;
+#define NVPW_CounterData_GetChipName_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_CounterData_GetChipName_Params, pChipName)
+
+    NVPA_Status NVPW_CounterData_GetChipName(NVPW_CounterData_GetChipName_Params* pParams);
+
+    typedef struct NVPW_Config_GetNumPasses_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// [in]
+        const uint8_t* pConfig;
+        /// [out]
+        size_t numPipelinedPasses;
+        /// [out]
+        size_t numIsolatedPasses;
+    } NVPW_Config_GetNumPasses_Params;
+#define NVPW_Config_GetNumPasses_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_Config_GetNumPasses_Params, numIsolatedPasses)
+
+    /// Total num passes = numPipelinedPasses + numIsolatedPasses * numNestingLevels
+    NVPA_Status NVPW_Config_GetNumPasses(NVPW_Config_GetNumPasses_Params* pParams);
+
+    typedef struct NVPW_Config_GetNumPasses_V2_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// [in]
+        const uint8_t* pConfig;
+        /// [out]
+        size_t numPasses;
+    } NVPW_Config_GetNumPasses_V2_Params;
+#define NVPW_Config_GetNumPasses_V2_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_Config_GetNumPasses_V2_Params, numPasses)
+
+    /// Total num passes = numPasses * numNestingLevels
+    NVPA_Status NVPW_Config_GetNumPasses_V2(NVPW_Config_GetNumPasses_V2_Params* pParams);
+
+#define NVPW_API_SET_CUDA_PROFILER             0x18209d0775b2f89dULL
+
+#define NVPW_API_SET_D3D11_PROFILER            0xca55c6738445db2bULL
+
+#define NVPW_API_SET_D3D12_PROFILER            0xc0c2d46dd7c7ad78ULL
+
+#define NVPW_API_SET_EGL_PROFILER              0x3c3747dae1f9565cULL
+
+#define NVPW_API_SET_GPU_PERIODICSAMPLER       0x9f4c2571fc0b2e8aULL
+
+#define NVPW_API_SET_METRICSEVALUATOR          0x0368a8768d811af9ULL
+
+#define NVPW_API_SET_METRICS_AD10X_COMP        0xbe57278e12cb5288ULL
+
+#define NVPW_API_SET_METRICS_AD10X_GRFX        0x5cbf0774f81bf491ULL
+
+#define NVPW_API_SET_METRICS_GA100_COMP        0x16b7d8c20d8b4915ULL
+
+#define NVPW_API_SET_METRICS_GA100_GRFX        0xc94eaabec04a94faULL
+
+#define NVPW_API_SET_METRICS_GA10X_COMP        0xb5d6391c2e299ab5ULL
+
+#define NVPW_API_SET_METRICS_GA10X_GRFX        0x6ebc121178b5ce0bULL
+
+#define NVPW_API_SET_METRICS_GV100_COMP        0x863705cc57919f72ULL
+
+#define NVPW_API_SET_METRICS_GV100_GRFX        0x9900da75d164fecfULL
+
+#define NVPW_API_SET_METRICS_GV11B_COMP        0xd3f79a859235848fULL
+
+#define NVPW_API_SET_METRICS_GV11B_GRFX        0xeb8e26220106e227ULL
+
+#define NVPW_API_SET_METRICS_TU10X_COMP        0x70f40be0afd35da8ULL
+
+#define NVPW_API_SET_METRICS_TU10X_GRFX        0xdf219cb838db6968ULL
+
+#define NVPW_API_SET_METRICS_TU11X_COMP        0xeb0069d7d0956678ULL
+
+#define NVPW_API_SET_METRICS_TU11X_GRFX        0x0977d9342bd62743ULL
+
+#define NVPW_API_SET_OPENGL_PROFILER           0xe4cd9ea40f2ee777ULL
+
+#define NVPW_API_SET_VULKAN_PROFILER           0x8c56b6a03d779689ULL
+
+#define NVPW_SDK_VERSION               0x1e128b6f001423fcULL
+
+    typedef struct NVPW_QueryVersionNumber_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// [in]
+        uint64_t apiSet;
+        /// [out]
+        uint32_t major;
+        /// [out]
+        uint32_t minor;
+        /// [out]
+        uint32_t patch;
+        /// [out]
+        uint32_t relMajor;
+        /// [out]
+        uint32_t relMinor;
+        /// [out]
+        uint32_t relPatch;
+    } NVPW_QueryVersionNumber_Params;
+#define NVPW_QueryVersionNumber_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_QueryVersionNumber_Params, relPatch)
+
+    /// Query version number of an API set
+    NVPA_Status NVPW_QueryVersionNumber(NVPW_QueryVersionNumber_Params* pParams);
+
+    typedef enum NVPW_Device_ClockStatus
+    {
+        /// clock status is unknown
+        NVPW_DEVICE_CLOCK_STATUS_UNKNOWN,
+        /// clocks are locked to rated tdp values - Deprecated, use NVPW_DEVICE_CLOCK_STATUS_LOCKED instead
+        NVPW_DEVICE_CLOCK_STATUS_LOCKED_TO_RATED_TDP,
+        /// clocks are not locked and can boost above rated tdp
+        NVPW_DEVICE_CLOCK_STATUS_BOOST_ENABLED,
+        /// clocks are not locked and will not go above rated tdp
+        NVPW_DEVICE_CLOCK_STATUS_BOOST_DISABLED,
+        /// clocks are locked
+        NVPW_DEVICE_CLOCK_STATUS_LOCKED,
+        /// clocks are not locked
+        NVPW_DEVICE_CLOCK_STATUS_UNLOCKED,
+        NVPW_DEVICE_CLOCK_STATUS__COUNT
+    } NVPW_Device_ClockStatus;
+
+    typedef enum NVPW_Device_ClockLevel
+    {
+        /// clock level is invalid
+        NVPW_DEVICE_CLOCK_LEVEL_INVALID,
+        /// clock level is at rated tdp
+        NVPW_DEVICE_CLOCK_LEVEL_RATED_TDP,
+        /// clock level is at turbo boost
+        NVPW_DEVICE_CLOCK_LEVEL_TURBO_BOOST,
+        NVPW_DEVICE_CLOCK_LEVEL__COUNT
+    } NVPW_Device_ClockLevel;
+
+    typedef struct NVPW_Device_GetClockStatus_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        size_t deviceIndex;
+        /// [in]
+        NVPW_Device_ClockStatus clockStatus;
+        /// [in]
+        NVPW_Device_ClockLevel clockLevel;
+    } NVPW_Device_GetClockStatus_Params;
+#define NVPW_Device_GetClockStatus_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_Device_GetClockStatus_Params, clockLevel)
+
+    NVPA_Status NVPW_Device_GetClockStatus(NVPW_Device_GetClockStatus_Params* pParams);
+
+    typedef enum NVPW_Device_ClockSetting
+    {
+        /// invalid op, specify valid clocks operation during profiling
+        NVPW_DEVICE_CLOCK_SETTING_INVALID,
+        /// default to driver/application config (normally unlocked and not boosted, but could be unlocked boosted, or
+        /// locked to rated TDP)
+        NVPW_DEVICE_CLOCK_SETTING_DEFAULT,
+        /// lock clocks at rated tdp base values
+        NVPW_DEVICE_CLOCK_SETTING_LOCK_TO_RATED_TDP,
+        /// lock clocks at turbo boost values
+        NVPW_DEVICE_CLOCK_SETTING_LOCK_TO_TURBO_BOOST,
+        NVPW_DEVICE_CLOCK_SETTING__COUNT
+    } NVPW_Device_ClockSetting;
+
+    typedef struct NVPW_Device_SetClockSetting_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        size_t deviceIndex;
+        /// [in]
+        NVPW_Device_ClockSetting clockSetting;
+    } NVPW_Device_SetClockSetting_Params;
+#define NVPW_Device_SetClockSetting_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_Device_SetClockSetting_Params, clockSetting)
+
+    NVPA_Status NVPW_Device_SetClockSetting(NVPW_Device_SetClockSetting_Params* pParams);
+
+    typedef struct NVPW_CounterData_GetRangeDescriptions_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        const uint8_t* pCounterDataImage;
+        size_t rangeIndex;
+        /// [inout] Number of descriptions allocated in ppDescriptions
+        size_t numDescriptions;
+        const char** ppDescriptions;
+    } NVPW_CounterData_GetRangeDescriptions_Params;
+#define NVPW_CounterData_GetRangeDescriptions_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_CounterData_GetRangeDescriptions_Params, ppDescriptions)
+
+    NVPA_Status NVPW_CounterData_GetRangeDescriptions(NVPW_CounterData_GetRangeDescriptions_Params* pParams);
+
+    typedef struct NVPW_Profiler_CounterData_GetRangeDescriptions_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        const uint8_t* pCounterDataImage;
+        size_t rangeIndex;
+        /// [inout] Number of descriptions allocated in ppDescriptions
+        size_t numDescriptions;
+        const char** ppDescriptions;
+    } NVPW_Profiler_CounterData_GetRangeDescriptions_Params;
+#define NVPW_Profiler_CounterData_GetRangeDescriptions_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_Profiler_CounterData_GetRangeDescriptions_Params, ppDescriptions)
+
+    NVPA_Status NVPW_Profiler_CounterData_GetRangeDescriptions(NVPW_Profiler_CounterData_GetRangeDescriptions_Params* pParams);
+
+#ifndef NVPW_PERIODIC_SAMPLER_COUNTER_DATA_APPEND_MODE_DEFINED
+#define NVPW_PERIODIC_SAMPLER_COUNTER_DATA_APPEND_MODE_DEFINED
+    typedef enum NVPW_PeriodicSampler_CounterData_AppendMode
+    {
+        NVPW_PERIODIC_SAMPLER_COUNTER_DATA_APPEND_MODE_LINEAR = 0,
+        NVPW_PERIODIC_SAMPLER_COUNTER_DATA_APPEND_MODE_CIRCULAR = 1,
+        NVPW_PERIODIC_SAMPLER_COUNTER_DATA_APPEND_MODE__COUNT
+    } NVPW_PeriodicSampler_CounterData_AppendMode;
+#endif //NVPW_PERIODIC_SAMPLER_COUNTER_DATA_APPEND_MODE_DEFINED
+
+    typedef struct NVPW_PeriodicSampler_CounterData_GetSampleTime_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// [in]
+        const uint8_t* pCounterDataImage;
+        /// [in]
+        size_t rangeIndex;
+        /// [out]
+        uint64_t timestampStart;
+        /// [out]
+        uint64_t timestampEnd;
+    } NVPW_PeriodicSampler_CounterData_GetSampleTime_Params;
+#define NVPW_PeriodicSampler_CounterData_GetSampleTime_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_PeriodicSampler_CounterData_GetSampleTime_Params, timestampEnd)
+
+    NVPA_Status NVPW_PeriodicSampler_CounterData_GetSampleTime(NVPW_PeriodicSampler_CounterData_GetSampleTime_Params* pParams);
+
+    typedef struct NVPW_PeriodicSampler_CounterData_TrimInPlace_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// [in]
+        uint8_t* pCounterDataImage;
+        /// [in]
+        size_t counterDataImageSize;
+        /// [out]
+        size_t counterDataImageTrimmedSize;
+    } NVPW_PeriodicSampler_CounterData_TrimInPlace_Params;
+#define NVPW_PeriodicSampler_CounterData_TrimInPlace_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_PeriodicSampler_CounterData_TrimInPlace_Params, counterDataImageTrimmedSize)
+
+    NVPA_Status NVPW_PeriodicSampler_CounterData_TrimInPlace(NVPW_PeriodicSampler_CounterData_TrimInPlace_Params* pParams);
+
+    typedef struct NVPW_PeriodicSampler_CounterData_GetInfo_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// [in]
+        const uint8_t* pCounterDataImage;
+        /// [in]
+        size_t counterDataImageSize;
+        /// [out] total number of ranges in the counter data
+        size_t numTotalRanges;
+        /// [out] if in "linear" mode, this API returns the number of "populated" ranges; if it's in "circular" mode,
+        /// then it returns the last "populated" range index + 1, when there is no such range, it returns 0.
+        size_t numPopulatedRanges;
+        /// [out] if in "linear" mode, this API returns the number of "completed" ranges; if it's in "circular" mode,
+        /// then it returns the last "completed" range index + 1, when there is no such range, it returns 0.
+        size_t numCompletedRanges;
+    } NVPW_PeriodicSampler_CounterData_GetInfo_Params;
+#define NVPW_PeriodicSampler_CounterData_GetInfo_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_PeriodicSampler_CounterData_GetInfo_Params, numCompletedRanges)
+
+    /// In periodic sampler, a range in counter data stores exactly one sample's data. For better performance, periodic
+    /// sampler may operate in an out-of-order fashion when populating sample data, i.e. it may not fully populate all
+    /// counters of a sample/range before starting to populate the next sample/range. As a result, we have two concepts
+    /// here, "populated" & "completed": a range is considered "populated" even if only partial counters have been
+    /// written; on the other hand, a range is only considered "completed" if all the collecting counters have been
+    /// written.
+    NVPA_Status NVPW_PeriodicSampler_CounterData_GetInfo(NVPW_PeriodicSampler_CounterData_GetInfo_Params* pParams);
+
+    typedef struct NVPW_PeriodicSampler_CounterData_GetTriggerCount_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// [in]
+        const uint8_t* pCounterDataImage;
+        /// [in]
+        size_t counterDataImageSize;
+        /// [in]
+        size_t rangeIndex;
+        /// [out]
+        uint32_t triggerCount;
+    } NVPW_PeriodicSampler_CounterData_GetTriggerCount_Params;
+#define NVPW_PeriodicSampler_CounterData_GetTriggerCount_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_PeriodicSampler_CounterData_GetTriggerCount_Params, triggerCount)
+
+    NVPA_Status NVPW_PeriodicSampler_CounterData_GetTriggerCount(NVPW_PeriodicSampler_CounterData_GetTriggerCount_Params* pParams);
+
+    typedef struct NVPW_PeriodicSampler_CounterData_IsDataComplete_Params
+    {
+        /// [in]
+        size_t structSize;
+        /// [in] assign to NULL
+        void* pPriv;
+        /// [in]
+        const uint8_t* pCounterDataImage;
+        /// [in]
+        size_t counterDataImageSize;
+        /// [in]
+        size_t rangeIndex;
+        /// [out]
+        NVPA_Bool isComplete;
+    } NVPW_PeriodicSampler_CounterData_IsDataComplete_Params;
+#define NVPW_PeriodicSampler_CounterData_IsDataComplete_Params_STRUCT_SIZE NVPA_STRUCT_SIZE(NVPW_PeriodicSampler_CounterData_IsDataComplete_Params, isComplete)
+
+    /// Checks whether a given sample's data is complete. See also 'NVPW_PeriodicSampler_CounterData_GetInfo'
+    NVPA_Status NVPW_PeriodicSampler_CounterData_IsDataComplete(NVPW_PeriodicSampler_CounterData_IsDataComplete_Params* pParams);
+
+
+    typedef struct NVPW_TimestampReport
+    {
+        uint32_t payload;
+        uint8_t reserved0004[4];
+        uint64_t timestamp;
+    } NVPW_TimestampReport;
+
+
+
+
+#ifdef __cplusplus
+} // extern "C"
+#endif
+
+#if defined(__GNUC__) && defined(NVPA_SHARED_LIB)
+    #pragma GCC visibility pop
+#endif
+
+#endif // NVPERF_TARGET_H

Prism/LLaDA/LLaDA_Prism/.venv/lib/python3.12/site-packages/nvidia/cufile/include/cufile.h

@@ -0,0 +1,740 @@
+/*
+ * Copyright 1993-2023 NVIDIA Corporation.  All rights reserved.
+ *
+ * NOTICE TO LICENSEE:
+ *
+ * This source code and/or documentation ("Licensed Deliverables") are
+ * subject to NVIDIA intellectual property rights under U.S. and
+ * international Copyright laws.
+ *
+ * These Licensed Deliverables contained herein is PROPRIETARY and
+ * CONFIDENTIAL to NVIDIA and is being provided under the terms and
+ * conditions of a form of NVIDIA software license agreement by and
+ * between NVIDIA and Licensee ("License Agreement") or electronically
+ * accepted by Licensee.  Notwithstanding any terms or conditions to
+ * the contrary in the License Agreement, reproduction or disclosure
+ * of the Licensed Deliverables to any third party without the express
+ * written consent of NVIDIA is prohibited.
+ *
+ * NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE
+ * LICENSE AGREEMENT, NVIDIA MAKES NO REPRESENTATION ABOUT THE
+ * SUITABILITY OF THESE LICENSED DELIVERABLES FOR ANY PURPOSE.  IT IS
+ * PROVIDED "AS IS" WITHOUT EXPRESS OR IMPLIED WARRANTY OF ANY KIND.
+ * NVIDIA DISCLAIMS ALL WARRANTIES WITH REGARD TO THESE LICENSED
+ * DELIVERABLES, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY,
+ * NONINFRINGEMENT, AND FITNESS FOR A PARTICULAR PURPOSE.
+ * NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE
+ * LICENSE AGREEMENT, IN NO EVENT SHALL NVIDIA BE LIABLE FOR ANY
+ * SPECIAL, INDIRECT, INCIDENTAL, OR CONSEQUENTIAL DAMAGES, OR ANY
+ * DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS,
+ * WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS
+ * ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE
+ * OF THESE LICENSED DELIVERABLES.
+ *
+ * U.S. Government End Users.  These Licensed Deliverables are a
+ * "commercial item" as that term is defined at 48 C.F.R. 2.101 (OCT
+ * 1995), consisting of "commercial computer software" and "commercial
+ * computer software documentation" as such terms are used in 48
+ * C.F.R. 12.212 (SEPT 1995) and is provided to the U.S. Government
+ * only as a commercial end item.  Consistent with 48 C.F.R.12.212 and
+ * 48 C.F.R. 227.7202-1 through 227.7202-4 (JUNE 1995), all
+ * U.S. Government End Users acquire the Licensed Deliverables with
+ * only those rights set forth herein.
+ *
+ * Any use of the Licensed Deliverables in individual and commercial
+ * software must include, in the user documentation and internal
+ * comments to the code, the above Disclaimer and U.S. Government End
+ * Users Notice.
+ */
+
+/**
+ * @file cufile.h
+ * @brief  cuFile C APIs
+ *
+ * This file contains all the C APIs to perform GPUDirect Storage supported IO operations
+ */
+
+#ifdef __cplusplus
+extern "C"
+{
+#endif
+
+/// @cond DOXYGEN_SKIP_MACRO
+#ifndef __CUFILE_H_
+#define __CUFILE_H_
+
+#include <stdlib.h>
+#include <stdbool.h>
+
+#include <cuda.h>
+#include <arpa/inet.h>
+#include <sys/socket.h>
+
+#define CUFILEOP_BASE_ERR 5000
+
+//Note :Data path errors are captured via standard error codes
+#define CUFILEOP_STATUS_ENTRIES \
+	CUFILE_OP(0,                      CU_FILE_SUCCESS, cufile success) \
+	CUFILE_OP(CUFILEOP_BASE_ERR + 1,  CU_FILE_DRIVER_NOT_INITIALIZED, nvidia-fs driver is not loaded. Set allow_compat_mode to true in cufile.json file to enable compatible mode) \
+	CUFILE_OP(CUFILEOP_BASE_ERR + 2,  CU_FILE_DRIVER_INVALID_PROPS, invalid property) \
+	CUFILE_OP(CUFILEOP_BASE_ERR + 3,  CU_FILE_DRIVER_UNSUPPORTED_LIMIT, property range error) \
+	CUFILE_OP(CUFILEOP_BASE_ERR + 4,  CU_FILE_DRIVER_VERSION_MISMATCH, nvidia-fs driver version mismatch) \
+	CUFILE_OP(CUFILEOP_BASE_ERR + 5,  CU_FILE_DRIVER_VERSION_READ_ERROR, nvidia-fs driver version read error) \
+	CUFILE_OP(CUFILEOP_BASE_ERR + 6,  CU_FILE_DRIVER_CLOSING, driver shutdown in progress) \
+	CUFILE_OP(CUFILEOP_BASE_ERR + 7,  CU_FILE_PLATFORM_NOT_SUPPORTED, GPUDirect Storage not supported on current platform) \
+	CUFILE_OP(CUFILEOP_BASE_ERR + 8,  CU_FILE_IO_NOT_SUPPORTED, GPUDirect Storage not supported on current file) \
+	CUFILE_OP(CUFILEOP_BASE_ERR + 9,  CU_FILE_DEVICE_NOT_SUPPORTED, GPUDirect Storage not supported on current GPU) \
+	CUFILE_OP(CUFILEOP_BASE_ERR + 10, CU_FILE_NVFS_DRIVER_ERROR, nvidia-fs driver ioctl error) \
+	CUFILE_OP(CUFILEOP_BASE_ERR + 11, CU_FILE_CUDA_DRIVER_ERROR, CUDA Driver API error) \
+	CUFILE_OP(CUFILEOP_BASE_ERR + 12, CU_FILE_CUDA_POINTER_INVALID, invalid device pointer) \
+	CUFILE_OP(CUFILEOP_BASE_ERR + 13, CU_FILE_CUDA_MEMORY_TYPE_INVALID, invalid pointer memory type) \
+	CUFILE_OP(CUFILEOP_BASE_ERR + 14, CU_FILE_CUDA_POINTER_RANGE_ERROR, pointer range exceeds allocated address range) \
+	CUFILE_OP(CUFILEOP_BASE_ERR + 15, CU_FILE_CUDA_CONTEXT_MISMATCH, cuda context mismatch) \
+	CUFILE_OP(CUFILEOP_BASE_ERR + 16, CU_FILE_INVALID_MAPPING_SIZE, access beyond maximum pinned size) \
+	CUFILE_OP(CUFILEOP_BASE_ERR + 17, CU_FILE_INVALID_MAPPING_RANGE, access beyond mapped size) \
+	CUFILE_OP(CUFILEOP_BASE_ERR + 18, CU_FILE_INVALID_FILE_TYPE, unsupported file type) \
+	CUFILE_OP(CUFILEOP_BASE_ERR + 19, CU_FILE_INVALID_FILE_OPEN_FLAG, unsupported file open flags) \
+	CUFILE_OP(CUFILEOP_BASE_ERR + 20, CU_FILE_DIO_NOT_SET, fd direct IO not set) \
+	CUFILE_OP(CUFILEOP_BASE_ERR + 22, CU_FILE_INVALID_VALUE, invalid arguments) \
+	CUFILE_OP(CUFILEOP_BASE_ERR + 23, CU_FILE_MEMORY_ALREADY_REGISTERED, device pointer already registered) \
+	CUFILE_OP(CUFILEOP_BASE_ERR + 24, CU_FILE_MEMORY_NOT_REGISTERED, device pointer lookup failure) \
+	CUFILE_OP(CUFILEOP_BASE_ERR + 25, CU_FILE_PERMISSION_DENIED, driver or file access error) \
+	CUFILE_OP(CUFILEOP_BASE_ERR + 26, CU_FILE_DRIVER_ALREADY_OPEN, driver is already open) \
+	CUFILE_OP(CUFILEOP_BASE_ERR + 27, CU_FILE_HANDLE_NOT_REGISTERED, file descriptor is not registered) \
+	CUFILE_OP(CUFILEOP_BASE_ERR + 28, CU_FILE_HANDLE_ALREADY_REGISTERED, file descriptor is already registered) \
+	CUFILE_OP(CUFILEOP_BASE_ERR + 29, CU_FILE_DEVICE_NOT_FOUND, GPU device not found) \
+	CUFILE_OP(CUFILEOP_BASE_ERR + 30, CU_FILE_INTERNAL_ERROR, internal error) \
+	CUFILE_OP(CUFILEOP_BASE_ERR + 31, CU_FILE_GETNEWFD_FAILED, failed to obtain new file descriptor) \
+	CUFILE_OP(CUFILEOP_BASE_ERR + 33, CU_FILE_NVFS_SETUP_ERROR, NVFS driver initialization error) \
+	CUFILE_OP(CUFILEOP_BASE_ERR + 34, CU_FILE_IO_DISABLED, GPUDirect Storage disabled by config on current file)\
+	CUFILE_OP(CUFILEOP_BASE_ERR + 35, CU_FILE_BATCH_SUBMIT_FAILED, failed to submit batch operation)\
+	CUFILE_OP(CUFILEOP_BASE_ERR + 36, CU_FILE_GPU_MEMORY_PINNING_FAILED, failed to allocate pinned GPU Memory) \
+	CUFILE_OP(CUFILEOP_BASE_ERR + 37, CU_FILE_BATCH_FULL, queue full for batch operation) \
+	CUFILE_OP(CUFILEOP_BASE_ERR + 38, CU_FILE_ASYNC_NOT_SUPPORTED, cuFile stream operation not supported) \
+	CUFILE_OP(CUFILEOP_BASE_ERR + 39, CU_FILE_IO_MAX_ERROR, GPUDirect Storage Max Error)
+
+
+/**
+ * @brief cufileop status enum
+ *
+ * @note on success the error code is set to  @ref CU_FILE_SUCCESS.
+ * @note The error code can be inspected using @ref IS_CUFILE_ERR and @ref CUFILE_ERRSTR.
+ * @note The error code if set to @ref CU_FILE_CUDA_DRIVER_ERROR, then cuda error can be inspected using @ref IS_CUDA_ERR and @ref CU_FILE_CUDA_ERR.
+ * @note Data path errors are captured via standard error codes
+ */
+typedef enum CUfileOpError {
+        /// @cond DOXYGEN_SKIP_MACRO
+	#define CUFILE_OP(code, name, string) name = code,
+	CUFILEOP_STATUS_ENTRIES
+	#undef CUFILE_OP
+        ///@endcond
+} CUfileOpError;
+
+/// @endcond
+
+/**
+ * @brief cufileop status string
+ */
+static inline const char *cufileop_status_error(CUfileOpError status)
+{
+	switch (status) {
+	/// @cond DOXYGEN_SKIP_MACRO
+	#define CUFILE_OP(code, name, string) \
+	case name: return #string;
+	CUFILEOP_STATUS_ENTRIES
+	#undef CUFILE_OP
+	///@endcond
+	default:return "unknown cufile error";
+	}
+}
+
+/**
+ * @brief cufileop status string
+ */
+typedef struct CUfileError {
+
+	CUfileOpError err; // cufile error
+
+	CUresult cu_err; // cuda driver error
+
+}CUfileError_t;
+
+/**
+ * @brief  error macros to inspect error status of type @ref CUfileOpError
+ */
+
+#define IS_CUFILE_ERR(err) \
+	(abs((err)) > CUFILEOP_BASE_ERR)
+
+#define CUFILE_ERRSTR(err) \
+	cufileop_status_error((CUfileOpError)abs((err)))
+
+#define IS_CUDA_ERR(status) \
+	((status).err == CU_FILE_CUDA_DRIVER_ERROR)
+
+#define CU_FILE_CUDA_ERR(status) ((status).cu_err)
+
+/* driver properties */
+typedef enum CUfileDriverStatusFlags {
+        CU_FILE_LUSTRE_SUPPORTED = 0, /*!< Support for DDN LUSTRE */
+
+        CU_FILE_WEKAFS_SUPPORTED = 1, /*!< Support for WEKAFS */
+
+        CU_FILE_NFS_SUPPORTED = 2, /*!< Support for NFS */
+
+        CU_FILE_GPFS_SUPPORTED = 3, /*! < Support for GPFS */
+
+        CU_FILE_NVME_SUPPORTED = 4, /*!< Support for NVMe */
+
+        CU_FILE_NVMEOF_SUPPORTED = 5, /*!< Support for NVMeOF */
+
+        CU_FILE_SCSI_SUPPORTED = 6, /*!< Support for SCSI */
+
+	CU_FILE_SCALEFLUX_CSD_SUPPORTED = 7, /*!< Support for Scaleflux CSD*/
+
+	CU_FILE_NVMESH_SUPPORTED = 8, /*!< Support for NVMesh Block Dev*/
+	CU_FILE_BEEGFS_SUPPORTED = 9, /*!< Support for BeeGFS */
+    //10 is reserved for YRCloudFile
+        CU_FILE_NVME_P2P_SUPPORTED = 11,   /*!< Support for NVMe using PCI P2PDMA */
+
+}CUfileDriverStatusFlags_t;
+
+typedef enum CUfileDriverControlFlags {
+        CU_FILE_USE_POLL_MODE = 0 , /*!< use POLL mode. properties.use_poll_mode*/
+
+        CU_FILE_ALLOW_COMPAT_MODE = 1/*!< allow COMPATIBILITY mode. properties.allow_compat_mode*/
+
+}CUfileDriverControlFlags_t;
+
+typedef enum CUfileFeatureFlags {
+        CU_FILE_DYN_ROUTING_SUPPORTED = 0, /*!< Support for Dynamic routing to handle devices across the PCIe bridges */
+
+        CU_FILE_BATCH_IO_SUPPORTED = 1, /*!<  Unsupported */
+
+        CU_FILE_STREAMS_SUPPORTED = 2, /*!<  Unsupported */
+
+        CU_FILE_PARALLEL_IO_SUPPORTED = 3 /*!<  Unsupported */
+}CUfileFeatureFlags_t;
+
+typedef struct CUfileDrvProps {
+        struct {
+                unsigned int major_version;
+
+                unsigned int minor_version;
+
+                size_t poll_thresh_size;
+
+                size_t max_direct_io_size;
+
+                unsigned int dstatusflags;
+
+                unsigned int dcontrolflags;
+
+        } nvfs;
+
+        unsigned int fflags;
+
+        unsigned int max_device_cache_size;
+
+	unsigned int per_buffer_cache_size;
+
+        unsigned int max_device_pinned_mem_size;
+
+        unsigned int max_batch_io_size;
+        unsigned int max_batch_io_timeout_msecs;
+}CUfileDrvProps_t;
+
+typedef struct sockaddr sockaddr_t;
+
+typedef struct cufileRDMAInfo
+{
+        int version;
+        int desc_len;
+        const char *desc_str;
+}cufileRDMAInfo_t;
+
+#define CU_FILE_RDMA_REGISTER 1
+#define CU_FILE_RDMA_RELAXED_ORDERING (1<<1)
+
+
+
+typedef struct CUfileFSOps {
+      /* NULL means discover using fstat */
+      const char* (*fs_type) (void *handle);
+
+      /* list of host addresses to use,  NULL means no restriction */
+      int (*getRDMADeviceList)(void *handle, sockaddr_t **hostaddrs);
+
+      /* -1 no pref */
+      int (*getRDMADevicePriority)(void *handle, char*, size_t,
+                                loff_t, sockaddr_t* hostaddr);
+
+      /* NULL means try VFS */
+      ssize_t (*read) (void *handle, char*, size_t, loff_t, cufileRDMAInfo_t*);
+      ssize_t (*write) (void *handle, const char *, size_t, loff_t , cufileRDMAInfo_t*);
+}CUfileFSOps_t;
+
+/* File Handle */
+enum CUfileFileHandleType {
+	CU_FILE_HANDLE_TYPE_OPAQUE_FD = 1,   /*!< Linux based fd */
+
+	CU_FILE_HANDLE_TYPE_OPAQUE_WIN32 = 2, /*!< Windows based handle (unsupported) */
+
+        CU_FILE_HANDLE_TYPE_USERSPACE_FS = 3,  /* Userspace based FS */
+};
+
+typedef struct CUfileDescr_t {
+	enum CUfileFileHandleType type; /* type of file being registered */
+	union {
+		int fd; 		/* Linux   */
+		void *handle; 		/* Windows */
+	} handle;
+        const CUfileFSOps_t *fs_ops;     /* file system operation table */
+}CUfileDescr_t;
+
+/**
+ * @brief File handle type
+ *
+ */
+typedef void* CUfileHandle_t;
+
+
+#pragma GCC visibility push(default)
+
+/**
+ * @brief cuFileHandleRegister is required, and performs extra checking that is memoized to provide increased performance on later cuFile operations.
+ *
+ * @param fh @ref CUfileHandle_t opaque file handle for IO operations
+ * @param descr @ref CUfileDescr_t  file descriptor (OS agnostic)
+ *
+ * @return      CU_FILE_SUCCESS on successful completion. fh will be updated for use in @ref cuFileRead, @ref cuFileWrite, @ref cuFileHandleDeregister
+ * @return      CU_FILE_DRIVER_NOT_INITIALIZED on failure to load driver
+ * @return      CU_FILE_IO_NOT_SUPPORTED -  if filesystem is not supported
+ * @return      CU_FILE_INVALID_VALUE if null or bad api arguments
+ * @return      CU_FILE_INVALID_FILE_OPEN_FLAG if file is opened with unsupported modes like no O_DIRECT
+ * @return      CU_FILE_INVALID_FILE_TYPE if filepath is not valid or is not a regular file
+ * @return      CU_FILE_HANDLE_ALREADY_REGISTERED if file handle/descriptor is already registered
+ *
+ * <b>Description</b>
+ * cuFileHandleRegister registers the open file descriptor for use with cuFile IO operations.
+ *
+ * This API will ensure that the file’s descriptor is checked for GPUDirect Storage support and returns a valid file handle on CU_FILE_SUCCESS.
+ *
+ * @note the file needs to be opened in O_DIRECT mode to support GPUDirect Storage.
+ *
+ * @see cuFileRead
+ * @see cuFileWrite
+ * @see cuFileHandleDeregister
+ *
+ */
+CUfileError_t cuFileHandleRegister(CUfileHandle_t *fh, CUfileDescr_t *descr);
+
+/**
+ * @brief releases a registered filehandle from cuFile
+ *
+ * @param fh @ref CUfileHandle_t file handle
+ *
+ * @return void
+ *
+ * @see cuFileHandleRegister
+ */
+void cuFileHandleDeregister(CUfileHandle_t fh);
+
+/**
+ * @brief register an existing cudaMalloced memory with cuFile to pin for GPUDirect Storage access or 
+ * register host allocated memory with cuFile.
+ *
+ * @param bufPtr_base buffer pointer allocated
+ * @param length  size of memory region from the above specified bufPtr
+ * @param flags   CU_FILE_RDMA_REGISTER
+ *
+ * @return  CU_FILE_SUCCESS on success
+ * @return  CU_FILE_NVFS_DRIVER_ERROR
+ * @return  CU_FILE_INVALID_VALUE
+ * @return  CU_FILE_CUDA_ERROR for unsuported memory type
+ * @return  CU_FILE_MEMORY_ALREADY_REGISTERED on error
+ * @return  CU_FILE_GPU_MEMORY_PINNING_FAILED if not enough pinned memory is available
+ * @note This memory will be use to perform GPU direct DMA from the supported storage.
+ * @warning This API is intended for usecases where the memory is used as streaming buffer that is reused across multiple cuFile IO operations before calling @ref cuFileBufDeregister
+ *
+ * @see cuFileBufDeregister
+ * @see cuFileRead
+ * @see cuFileWrite
+ */
+CUfileError_t cuFileBufRegister(const void *bufPtr_base, size_t length, int flags);
+
+/**
+ * @brief  deregister an already registered device or host memory from cuFile
+ *
+ * @param bufPtr_base  buffer pointer to deregister
+ *
+ * @return  CU_FILE_SUCCESS on success
+ * @return  CU_FILE_INVALID_VALUE on invalid memory pointer or unregistered memory pointer
+ *
+ * @see cuFileBufRegister
+ * @see cuFileRead
+ * @see cuFileWrite
+ */
+
+CUfileError_t cuFileBufDeregister(const void *bufPtr_base);
+
+/**
+ * @brief read data from a registered file handle to a specified device or host memory
+ *
+ * @param fh @ref CUfileHandle_t opaque file handle
+ * @param bufPtr_base  base address of buffer in device or host memory
+ * @param size    size bytes to read
+ * @param file_offset  file-offset from begining of the file
+ * @param bufPtr_offset  offset relative to the bufPtr_base pointer to read into.
+ *
+ * @return  size of bytes successfully read
+ * @return  -1 on error, in which case errno is set to indicate filesystem errors.
+ * @return  all other errors will return a negative integer value of @ref CUfileOpError enum value.
+ *
+ * @note  If the bufPtr is not registered with @ref cuFileBufRegister, the data will be buffered through preallocated pinned buffers if needed.
+ * @note  This is useful for applications that need to perform IO to unaligned file offsets and/or size. This is also recommended
+ *        for cases where the BAR1 memory size is smaller than the size of the allocated memory.
+ *
+ * @see cuFileBufRegister
+ * @see cuFileHandleRegister
+ * @see cuFileWrite
+ */
+
+ssize_t cuFileRead(CUfileHandle_t fh, void *bufPtr_base, size_t size, off_t file_offset, off_t bufPtr_offset);
+
+/**
+ * @brief  write data from a specified device or host memory to a registered file handle
+ *
+ * @param fh @ref CUfileHandle_t opaque file handle
+ * @param bufPtr_base  base address of buffer in device or host memory
+ * @param size    size bytes to write
+ * @param file_offset  file-offset from begining of the file
+ * @param bufPtr_offset  offset relative to the bufPtr_base pointer to write from.
+ *
+ * @return  size of bytes successfully written
+ * @return  -1 on error, in which case errno is set to indicate filesystem errors.
+ * @return  all other errors will return a negative integer value of @ref CUfileOpError enum value.
+ *
+ * @note  If the bufPtr is not registered with @ref cuFileBufRegister, the data will be buffered through preallocated pinned buffers if needed.
+ * @note  This is useful for applications that need to perform IO to unaligned file offsets and/or size. This is also recommended
+ *        for cases where the BAR1 memory size is smaller than the size of the allocated memory.
+ *
+ * @see cuFileBufRegister
+ * @see cuFileHandleRegister
+ * @see cuFileRead
+ */
+
+ssize_t cuFileWrite(CUfileHandle_t fh, const void *bufPtr_base, size_t size, off_t file_offset, off_t bufPtr_offset);
+
+// CUFile Driver APIs
+
+/**
+ * @brief
+ * Initialize the cuFile library and open the nvidia-fs driver
+ *
+ * @return CU_FILE_SUCCESS on success
+ * @return CU_FILE_DRIVER_NOT_INITIALIZED
+ * @return CU_FILE_DRIVER_VERSION_MISMATCH on driver version mismatch error
+ *
+ * @see cuFileDriverClose
+ */
+CUfileError_t cuFileDriverOpen(void);
+
+CUfileError_t cuFileDriverClose(void);
+#define cuFileDriverClose cuFileDriverClose_v2
+/**
+ * @brief
+ * reset the cuFile library and release the nvidia-fs driver
+ *
+ * @return CU_FILE_SUCCESS on success
+ * @return CU_FILE_DRIVER_CLOSING if there are any active IO operations using @ref cuFileRead or @ref cuFileWrite
+ *
+ * @see cuFileDriverOpen
+ */
+CUfileError_t cuFileDriverClose(void);
+
+/**
+ * @brief
+ * returns use count of cufile drivers at that moment by the process.
+ */
+long cuFileUseCount(void);
+
+/**
+ * @brief
+ * Gets the Driver session properties
+ *
+ * @return CU_FILE_SUCCESS on success
+ *
+ * @see cuFileDriverSetPollMode
+ * @see cuFileDriverSetMaxDirectIOSize
+ * @see cuFileDriverSetMaxCacheSize
+ * @see cuFileDriverSetMaxPinnedMemSize
+ */
+CUfileError_t cuFileDriverGetProperties(CUfileDrvProps_t *props);
+
+/**
+ * @brief
+ * Sets whether the Read/Write APIs use polling to do IO operations
+ *
+ * @param  poll boolean to indicate whether to use poll mode or not
+ * @param  poll_threshold_size max IO size to use for POLLING mode in KB
+ *
+ * @return CU_FILE_SUCCESS on success
+ * @return CU_FILE_DRIVER_NOT_INITIALIZED if the driver is not initialized
+ * @return CU_FILE_DRIVER_VERSION_MISMATCH, CU_FILE_DRIVER_UNSUPPORTED_LIMIT on error
+ *
+ * @warning This is an advanced command and should be tuned based on available system memory
+ *
+ * @see cuFileDriverGetProperties
+ */
+CUfileError_t cuFileDriverSetPollMode(bool poll, size_t poll_threshold_size);
+
+/**
+ * @brief
+ * Control parameter to set max IO size(KB) used by the library to talk to nvidia-fs driver
+ *
+ * @param  max_direct_io_size maximum allowed direct io size in KB
+ *
+ * @return CU_FILE_SUCCESS on success
+ * @return CU_FILE_DRIVER_NOT_INITIALIZED if the driver is not initialized
+ * @return CU_FILE_DRIVER_VERSION_MISMATCH, CU_FILE_DRIVER_UNSUPPORTED_LIMIT on error
+ *
+ * @warning This is an advanced command and should be tuned based on available system memory
+ *
+ * @see cuFileDriverGetProperties
+ *
+ */
+CUfileError_t cuFileDriverSetMaxDirectIOSize(size_t max_direct_io_size);
+
+/**
+ * @brief
+ * Control parameter to set maximum GPU memory reserved per device by the library for internal buffering
+ *
+ * @param  max_cache_size The maximum GPU buffer space per device used for internal use in KB
+ *
+ * @return CU_FILE_SUCCESS on success
+ * @return CU_FILE_DRIVER_NOT_INITIALIZED if the driver is not initialized
+ * @return CU_FILE_DRIVER_VERSION_MISMATCH, CU_FILE_DRIVER_UNSUPPORTED_LIMIT on error
+ *
+ * @warning This is an advanced command and should be tuned based on supported GPU memory
+ *
+ * @see cuFileDriverGetProperties
+ */
+CUfileError_t cuFileDriverSetMaxCacheSize(size_t max_cache_size);
+
+/**
+ * @brief
+ * Sets maximum buffer space that is pinned in KB for use by @ref cuFileBufRegister
+ *
+ * @param max_pinned_size maximum buffer space that is pinned in KB
+ *
+ * @return CU_FILE_SUCCESS on success
+ * @return CU_FILE_DRIVER_NOT_INITIALIZED if the driver is not initialized
+ * @return CU_FILE_DRIVER_VERSION_MISMATCH, CU_FILE_DRIVER_UNSUPPORTED_LIMIT on error
+ *
+ * @warning This is an advanced command and should be tuned based on supported GPU memory
+ *
+ * @see cuFileDriverGetProperties
+ *
+ */
+CUfileError_t cuFileDriverSetMaxPinnedMemSize(size_t max_pinned_size);
+
+//Experimental Batch API's
+
+
+typedef enum CUfileOpcode {
+		CUFILE_READ = 0,
+		CUFILE_WRITE
+}CUfileOpcode_t;
+
+typedef enum CUFILEStatus_enum {
+	CUFILE_WAITING = 0x000001,  /* required value prior to submission */
+	CUFILE_PENDING = 0x000002,  /* once enqueued */
+	CUFILE_INVALID = 0x000004,  /* request was ill-formed or could not be enqueued */
+	CUFILE_CANCELED = 0x000008, /* request successfully canceled */
+	CUFILE_COMPLETE = 0x0000010, /* request successfully completed */
+	CUFILE_TIMEOUT = 0x0000020,  /* request timed out */
+	CUFILE_FAILED  = 0x0000040  /* unable to complete */
+}CUfileStatus_t;
+typedef enum cufileBatchMode {
+	CUFILE_BATCH = 1,
+} CUfileBatchMode_t;
+typedef struct CUfileIOParams {
+	CUfileBatchMode_t mode; // Must be the very first field.
+	union {
+		struct  {
+			void *devPtr_base; //This can be a device memory or a host memory pointer.
+			off_t file_offset;
+			off_t devPtr_offset; 
+			size_t size;
+		}batch;
+	}u;
+	CUfileHandle_t fh;
+	CUfileOpcode_t opcode;
+	void *cookie;
+}CUfileIOParams_t;
+typedef struct CUfileIOEvents {
+	void *cookie;
+	CUfileStatus_t   status;      /* status of the operation */
+	size_t ret; /* -ve error or amount of I/O done. */
+}CUfileIOEvents_t;
+
+typedef void* CUfileBatchHandle_t;
+
+CUfileError_t cuFileBatchIOSetUp(CUfileBatchHandle_t *batch_idp, unsigned nr);
+CUfileError_t cuFileBatchIOSubmit(CUfileBatchHandle_t batch_idp, unsigned nr, CUfileIOParams_t *iocbp, unsigned int flags);
+CUfileError_t cuFileBatchIOGetStatus(CUfileBatchHandle_t batch_idp, unsigned min_nr, unsigned* nr,
+				CUfileIOEvents_t *iocbp, struct timespec* timeout);
+CUfileError_t cuFileBatchIOCancel(CUfileBatchHandle_t batch_idp);
+void cuFileBatchIODestroy(CUfileBatchHandle_t batch_idp);
+
+//Async API's with cuda streams
+
+// cuFile stream API registration flags
+// buffer pointer offset is set at submission time
+#define CU_FILE_STREAM_FIXED_BUF_OFFSET         1
+// file offset is set at submission time
+#define CU_FILE_STREAM_FIXED_FILE_OFFSET        2
+// file size is set at submission time
+#define CU_FILE_STREAM_FIXED_FILE_SIZE          4
+// size, offset and buffer offset are 4k aligned
+#define CU_FILE_STREAM_PAGE_ALIGNED_INPUTS      8
+
+/**
+ *@brief
+
+ * @param fh The cuFile handle for the file.
+ * @param bufPtr_base  base address of buffer in device or host memory
+ * @param size_p  pointer to size bytes to read
+ * @note  *size_p if the size is not known at the time of submission, then must provide the max possible size for I/O request.
+ * @param file_offset_p  pointer to file-offset from begining of the file
+ * @param bufPtr_offset_p  pointer to offset relative to the bufPtr_base pointer to read into.
+ * @param bytes_read_p  pointer to the number of bytes that were successfully read.
+ * @param CUstream stream cuda stream for the operation.
+ *
+ * @return  size of bytes successfully read in *bytes_read_p
+ * @return  -1 on error, in which case errno is set to indicate filesystem errors.
+ * @return  all other errors will return a negative integer value of @ref CUfileOpError enum value.
+ *
+ * @note  If the bufPtr_base is not registered with @ref cuFileBufRegister, the data will be buffered through preallocated pinned buffers.
+ * @note  This is useful for applications that need to perform IO to unaligned file offsets and/or size. This is also recommended
+ *        for cases where the BAR1 memory size is smaller than the size of the allocated memory.
+ * @note  If the stream is registered with cuFileStreamRegister, the IO setup and teardown overhead will be reduced.
+ * @note  on cuda stream errors, the user must call cuFileStreamDeregister to release any outstanding cuFile resources for the stream.
+ *
+ *
+ * @see cuFileBufRegister
+ * @see cuFileHandleRegister
+ * @see cuFileRead
+ * @see cuFileStreamRegister
+ * @see cuFileStreamDeregister
+ */
+
+CUfileError_t cuFileReadAsync(CUfileHandle_t fh, void *bufPtr_base,
+                        size_t *size_p, off_t *file_offset_p, off_t *bufPtr_offset_p, ssize_t *bytes_read_p, CUstream stream);
+
+/**
+ *@brief
+
+* @param fh The cuFile handle for the file.
+ * @param bufPtr_base  base address of buffer in device or host memory
+ * @param size_p    pointer to size bytes to write.
+ * @note  *size_p if the size is not known at the time of submission, then must provide the max possible size for I/O request.
+ * @param file_offset_p  pointer to file-offset from begining of the file
+ * @param bufPtr_offset_p  pointer to offset relative to the bufPtr_base pointer to write from.
+ * @param bytes_written_p pointer to the number of bytes that were successfully written.
+ * @param CUstream cuda stream for the operation.
+ *
+ * @return  size of bytes successfully written in *bytes_written_p
+ * @return  -1 on error, in which case errno is set to indicate filesystem errors.
+ * @return  all other errors will return a negative integer value of @ref CUfileOpError enum value.
+ *
+ * @note  If the bufPtr_base is not registered with @ref cuFileBufRegister, the data will be buffered through preallocated pinned buffers.
+ * @note  This is useful for applications that need to perform IO to unaligned file offsets and/or size. This is also recommended
+ *        for cases where the BAR1 memory size is smaller than the size of the allocated memory.
+ * @note  If the stream is registered with cuFileStreamRegister prior to this call, the IO setup and teardown overhead will be reduced.
+ * @note  on cuda stream errors, the user must call cuFileStreamDeregister to release any outstanding cuFile resources for the stream.
+ *
+ * @see cuFileBufRegister
+ * @see cuFileHandleRegister
+ * @see cuFileWrite
+ * @see cuFileStreamRegister
+ * @see cuFileStreamDeregister
+ */
+
+CUfileError_t cuFileWriteAsync(CUfileHandle_t fh, void *bufPtr_base,
+                        size_t *size_p, off_t *file_offset_p, off_t *bufPtr_offset_p, ssize_t *bytes_written_p, CUstream stream);
+
+/**
+ *@brief
+
+ * @param CUstream cuda stream for the operation.
+ * @param flags for the stream to improve the stream execution of IO based on input parameters.
+ * @note  supported FLAGS are
+ * @note CU_FILE_STREAM_FIXED_BUF_OFFSET - buffer pointer offset is set at submission time
+ * @note CU_FILE_STREAM_FIXED_FILE_OFFSET - file offset is set at submission time
+ * @note CU_FILE_STREAM_FIXED_FILE_SIZE  - file size is set at submission time
+ * @note CU_FILE_STREAM_PAGE_ALIGNED_INPUTS - size, offset and buffer offset are 4k aligned
+ *
+ * @note  allocates resources needed to support cuFile operations asynchronously for the cuda stream
+ * @note  This is useful for applications that need to perform IO to unaligned file offsets and/or size. This is also recommended
+ *        for cases where the BAR1 memory size is smaller than the size of the allocated memory.
+ *
+ * @return CU_FILE_SUCCESS on success
+ * @return CU_FILE_DRIVER_NOT_INITIALIZED if the driver is not initialized
+ * @return CU_FILE_INVALID_VALUE if the stream is invalid
+ *
+ * @see cuFileReadAsync
+ * @see cuFileWriteAsync
+ * @see cuFileStreamDeregister
+ */
+
+CUfileError_t cuFileStreamRegister(CUstream stream, unsigned flags);
+
+/**
+ *@brief
+
+ * @param CUstream cuda stream for the operation.
+ *
+ * @note  deallocates resources used by previous cuFile asynchronous operations for the cuda stream
+ * @note  highly recommend to call after cuda stream errors to release any outstanding cuFile resources for this stream
+ * @note  must be called before cuStreamDestroy call for the specified stream.
+ * @note  This is useful for applications that need to perform IO to unaligned file offsets and/or size. This is also recommended
+ *        for cases where the BAR1 memory size is smaller than the size of the allocated memory.
+ *
+ * @return CU_FILE_SUCCESS on success
+ * @return CU_FILE_DRIVER_NOT_INITIALIZED if the driver is not initialized
+ * @return CU_FILE_INVALID_VALUE if the stream is invalid
+ *
+ * @see cuFileReadAsync
+ * @see cuFileWriteAsync
+ * @see cuFileStreamRegister
+ */
+
+CUfileError_t cuFileStreamDeregister(CUstream stream);
+
+/**
+ *@brief
+
+ * @returns cufile library version.
+ * 
+ * @The version is returned as (1000 major + 10 minor).
+ * @For example, CUFILE 1.7.0 would be represented by 1070.
+ * @note  This is useful for applications that need to inquire the library.
+ *
+ * @return CU_FILE_SUCCESS on success
+ * @return CU_FILE_INVALID_VALUE if the input parameter is null.
+ * @return CU_FILE_DRIVER_VERSION_READ_ERROR if the version is not available.
+ *
+ */
+
+CUfileError_t cuFileGetVersion(int *version);
+
+#pragma GCC visibility pop
+
+/// @cond DOXYGEN_SKIP_MACRO
+#endif // CUFILE_H
+/// @endcond
+#ifdef __cplusplus
+}
+#endif