triton-windows 3.3.0.post19__cp39-cp39-win_amd64.whl → 3.3.1.post21__cp39-cp39-win_amd64.whl

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.

Potentially problematic release.


This version of triton-windows might be problematic. Click here for more details.

Files changed (116) hide show
  1. triton/_C/libtriton.pyd +0 -0
  2. triton/__init__.py +1 -1
  3. triton/backends/amd/driver.py +6 -1
  4. triton/backends/nvidia/compiler.py +1 -3
  5. triton/backends/nvidia/driver.py +7 -3
  6. triton/runtime/autotuner.py +2 -2
  7. triton/runtime/build.py +5 -5
  8. triton/runtime/tcc/lib/python310.def +1610 -0
  9. triton/runtime/tcc/lib/python311.def +1633 -0
  10. triton/runtime/tcc/lib/python312.def +1703 -0
  11. triton/runtime/tcc/lib/python313.def +1651 -0
  12. triton/runtime/tcc/lib/python313t.def +1656 -0
  13. triton/runtime/tcc/lib/python39.def +1644 -0
  14. triton/runtime/tcc/lib/python3t.def +905 -0
  15. triton/windows_utils.py +11 -4
  16. {triton_windows-3.3.0.post19.dist-info → triton_windows-3.3.1.post21.dist-info}/METADATA +1 -1
  17. {triton_windows-3.3.0.post19.dist-info → triton_windows-3.3.1.post21.dist-info}/RECORD +19 -109
  18. {triton_windows-3.3.0.post19.dist-info → triton_windows-3.3.1.post21.dist-info}/WHEEL +1 -1
  19. triton/backends/amd/include/hip/amd_detail/amd_channel_descriptor.h +0 -358
  20. triton/backends/amd/include/hip/amd_detail/amd_device_functions.h +0 -1010
  21. triton/backends/amd/include/hip/amd_detail/amd_hip_atomic.h +0 -1638
  22. triton/backends/amd/include/hip/amd_detail/amd_hip_bf16.h +0 -1814
  23. triton/backends/amd/include/hip/amd_detail/amd_hip_bfloat16.h +0 -293
  24. triton/backends/amd/include/hip/amd_detail/amd_hip_common.h +0 -32
  25. triton/backends/amd/include/hip/amd_detail/amd_hip_complex.h +0 -174
  26. triton/backends/amd/include/hip/amd_detail/amd_hip_cooperative_groups.h +0 -835
  27. triton/backends/amd/include/hip/amd_detail/amd_hip_fp16.h +0 -1809
  28. triton/backends/amd/include/hip/amd_detail/amd_hip_fp8.h +0 -1391
  29. triton/backends/amd/include/hip/amd_detail/amd_hip_gl_interop.h +0 -108
  30. triton/backends/amd/include/hip/amd_detail/amd_hip_math_constants.h +0 -124
  31. triton/backends/amd/include/hip/amd_detail/amd_hip_runtime.h +0 -405
  32. triton/backends/amd/include/hip/amd_detail/amd_hip_runtime_pt_api.h +0 -196
  33. triton/backends/amd/include/hip/amd_detail/amd_hip_unsafe_atomics.h +0 -565
  34. triton/backends/amd/include/hip/amd_detail/amd_hip_vector_types.h +0 -2226
  35. triton/backends/amd/include/hip/amd_detail/amd_math_functions.h +0 -104
  36. triton/backends/amd/include/hip/amd_detail/amd_surface_functions.h +0 -244
  37. triton/backends/amd/include/hip/amd_detail/amd_warp_functions.h +0 -538
  38. triton/backends/amd/include/hip/amd_detail/amd_warp_sync_functions.h +0 -288
  39. triton/backends/amd/include/hip/amd_detail/concepts.hpp +0 -30
  40. triton/backends/amd/include/hip/amd_detail/device_library_decls.h +0 -133
  41. triton/backends/amd/include/hip/amd_detail/functional_grid_launch.hpp +0 -218
  42. triton/backends/amd/include/hip/amd_detail/grid_launch.h +0 -67
  43. triton/backends/amd/include/hip/amd_detail/grid_launch.hpp +0 -50
  44. triton/backends/amd/include/hip/amd_detail/grid_launch_GGL.hpp +0 -26
  45. triton/backends/amd/include/hip/amd_detail/helpers.hpp +0 -137
  46. triton/backends/amd/include/hip/amd_detail/hip_api_trace.hpp +0 -1446
  47. triton/backends/amd/include/hip/amd_detail/hip_assert.h +0 -101
  48. triton/backends/amd/include/hip/amd_detail/hip_cooperative_groups_helper.h +0 -242
  49. triton/backends/amd/include/hip/amd_detail/hip_fp16_gcc.h +0 -254
  50. triton/backends/amd/include/hip/amd_detail/hip_fp16_math_fwd.h +0 -96
  51. triton/backends/amd/include/hip/amd_detail/hip_ldg.h +0 -100
  52. triton/backends/amd/include/hip/amd_detail/hip_prof_str.h +0 -10570
  53. triton/backends/amd/include/hip/amd_detail/hip_runtime_prof.h +0 -78
  54. triton/backends/amd/include/hip/amd_detail/host_defines.h +0 -184
  55. triton/backends/amd/include/hip/amd_detail/hsa_helpers.hpp +0 -102
  56. triton/backends/amd/include/hip/amd_detail/macro_based_grid_launch.hpp +0 -798
  57. triton/backends/amd/include/hip/amd_detail/math_fwd.h +0 -698
  58. triton/backends/amd/include/hip/amd_detail/ockl_image.h +0 -177
  59. triton/backends/amd/include/hip/amd_detail/program_state.hpp +0 -107
  60. triton/backends/amd/include/hip/amd_detail/texture_fetch_functions.h +0 -491
  61. triton/backends/amd/include/hip/amd_detail/texture_indirect_functions.h +0 -478
  62. triton/backends/amd/include/hip/channel_descriptor.h +0 -39
  63. triton/backends/amd/include/hip/device_functions.h +0 -38
  64. triton/backends/amd/include/hip/driver_types.h +0 -468
  65. triton/backends/amd/include/hip/hip_bf16.h +0 -36
  66. triton/backends/amd/include/hip/hip_bfloat16.h +0 -44
  67. triton/backends/amd/include/hip/hip_common.h +0 -100
  68. triton/backends/amd/include/hip/hip_complex.h +0 -38
  69. triton/backends/amd/include/hip/hip_cooperative_groups.h +0 -46
  70. triton/backends/amd/include/hip/hip_deprecated.h +0 -95
  71. triton/backends/amd/include/hip/hip_ext.h +0 -161
  72. triton/backends/amd/include/hip/hip_fp16.h +0 -36
  73. triton/backends/amd/include/hip/hip_fp8.h +0 -33
  74. triton/backends/amd/include/hip/hip_gl_interop.h +0 -32
  75. triton/backends/amd/include/hip/hip_hcc.h +0 -24
  76. triton/backends/amd/include/hip/hip_math_constants.h +0 -36
  77. triton/backends/amd/include/hip/hip_profile.h +0 -27
  78. triton/backends/amd/include/hip/hip_runtime.h +0 -75
  79. triton/backends/amd/include/hip/hip_runtime_api.h +0 -9261
  80. triton/backends/amd/include/hip/hip_texture_types.h +0 -29
  81. triton/backends/amd/include/hip/hip_vector_types.h +0 -41
  82. triton/backends/amd/include/hip/hip_version.h +0 -17
  83. triton/backends/amd/include/hip/hiprtc.h +0 -421
  84. triton/backends/amd/include/hip/library_types.h +0 -78
  85. triton/backends/amd/include/hip/math_functions.h +0 -42
  86. triton/backends/amd/include/hip/surface_types.h +0 -63
  87. triton/backends/amd/include/hip/texture_types.h +0 -194
  88. triton/backends/amd/include/hsa/Brig.h +0 -1131
  89. triton/backends/amd/include/hsa/amd_hsa_common.h +0 -91
  90. triton/backends/amd/include/hsa/amd_hsa_elf.h +0 -462
  91. triton/backends/amd/include/hsa/amd_hsa_kernel_code.h +0 -269
  92. triton/backends/amd/include/hsa/amd_hsa_queue.h +0 -109
  93. triton/backends/amd/include/hsa/amd_hsa_signal.h +0 -80
  94. triton/backends/amd/include/hsa/hsa.h +0 -5738
  95. triton/backends/amd/include/hsa/hsa_amd_tool.h +0 -91
  96. triton/backends/amd/include/hsa/hsa_api_trace.h +0 -579
  97. triton/backends/amd/include/hsa/hsa_api_trace_version.h +0 -68
  98. triton/backends/amd/include/hsa/hsa_ext_amd.h +0 -3146
  99. triton/backends/amd/include/hsa/hsa_ext_finalize.h +0 -531
  100. triton/backends/amd/include/hsa/hsa_ext_image.h +0 -1454
  101. triton/backends/amd/include/hsa/hsa_ven_amd_aqlprofile.h +0 -488
  102. triton/backends/amd/include/hsa/hsa_ven_amd_loader.h +0 -667
  103. triton/backends/amd/include/hsa/hsa_ven_amd_pc_sampling.h +0 -416
  104. triton/backends/amd/include/roctracer/ext/prof_protocol.h +0 -107
  105. triton/backends/amd/include/roctracer/hip_ostream_ops.h +0 -4515
  106. triton/backends/amd/include/roctracer/hsa_ostream_ops.h +0 -1727
  107. triton/backends/amd/include/roctracer/hsa_prof_str.h +0 -3059
  108. triton/backends/amd/include/roctracer/roctracer.h +0 -779
  109. triton/backends/amd/include/roctracer/roctracer_ext.h +0 -81
  110. triton/backends/amd/include/roctracer/roctracer_hcc.h +0 -24
  111. triton/backends/amd/include/roctracer/roctracer_hip.h +0 -37
  112. triton/backends/amd/include/roctracer/roctracer_hsa.h +0 -112
  113. triton/backends/amd/include/roctracer/roctracer_plugin.h +0 -137
  114. triton/backends/amd/include/roctracer/roctracer_roctx.h +0 -67
  115. triton/backends/amd/include/roctracer/roctx.h +0 -229
  116. {triton_windows-3.3.0.post19.dist-info → triton_windows-3.3.1.post21.dist-info}/top_level.txt +0 -0
@@ -1,46 +0,0 @@
1
- /*
2
- Copyright (c) 2015 - 2021 Advanced Micro Devices, Inc. All rights reserved.
3
-
4
- Permission is hereby granted, free of charge, to any person obtaining a copy
5
- of this software and associated documentation files (the "Software"), to deal
6
- in the Software without restriction, including without limitation the rights
7
- to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
8
- copies of the Software, and to permit persons to whom the Software is
9
- furnished to do so, subject to the following conditions:
10
-
11
- The above copyright notice and this permission notice shall be included in
12
- all copies or substantial portions of the Software.
13
-
14
- THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
15
- IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
16
- FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
17
- AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
18
- LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
19
- OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
20
- THE SOFTWARE.
21
- */
22
-
23
- /**
24
- * @file hip_cooperative_groups.h
25
- *
26
- * @brief Defines new types and device API wrappers for `Cooperative Group`
27
- * feature.
28
- */
29
-
30
- #ifndef HIP_INCLUDE_HIP_HIP_COOPERATIVE_GROUP_H
31
- #define HIP_INCLUDE_HIP_HIP_COOPERATIVE_GROUP_H
32
-
33
- #include <hip/hip_version.h>
34
- #include <hip/hip_common.h>
35
-
36
- #if defined(__HIP_PLATFORM_AMD__) && !defined(__HIP_PLATFORM_NVIDIA__)
37
- #if __cplusplus && defined(__clang__) && defined(__HIP__)
38
- #include <hip/amd_detail/amd_hip_cooperative_groups.h>
39
- #endif
40
- #elif !defined(__HIP_PLATFORM_AMD__) && defined(__HIP_PLATFORM_NVIDIA__)
41
- #include <hip/nvidia_detail/nvidia_hip_cooperative_groups.h>
42
- #else
43
- #error("Must define exactly one of __HIP_PLATFORM_AMD__ or __HIP_PLATFORM_NVIDIA__");
44
- #endif
45
-
46
- #endif // HIP_INCLUDE_HIP_HIP_COOPERATIVE_GROUP_H
@@ -1,95 +0,0 @@
1
- #pragma once
2
-
3
- // This file will add older hip functions used in the versioning system
4
- // Find the deprecated functions and structs in hip_device.cpp
5
-
6
- // This struct is also kept in hip_device.cpp
7
- typedef struct hipDeviceProp_tR0000 {
8
- char name[256]; ///< Device name.
9
- size_t totalGlobalMem; ///< Size of global memory region (in bytes).
10
- size_t sharedMemPerBlock; ///< Size of shared memory region (in bytes).
11
- int regsPerBlock; ///< Registers per block.
12
- int warpSize; ///< Warp size.
13
- int maxThreadsPerBlock; ///< Max work items per work group or workgroup max size.
14
- int maxThreadsDim[3]; ///< Max number of threads in each dimension (XYZ) of a block.
15
- int maxGridSize[3]; ///< Max grid dimensions (XYZ).
16
- int clockRate; ///< Max clock frequency of the multiProcessors in khz.
17
- int memoryClockRate; ///< Max global memory clock frequency in khz.
18
- int memoryBusWidth; ///< Global memory bus width in bits.
19
- size_t totalConstMem; ///< Size of shared memory region (in bytes).
20
- int major; ///< Major compute capability. On HCC, this is an approximation and features may
21
- ///< differ from CUDA CC. See the arch feature flags for portable ways to query
22
- ///< feature caps.
23
- int minor; ///< Minor compute capability. On HCC, this is an approximation and features may
24
- ///< differ from CUDA CC. See the arch feature flags for portable ways to query
25
- ///< feature caps.
26
- int multiProcessorCount; ///< Number of multi-processors (compute units).
27
- int l2CacheSize; ///< L2 cache size.
28
- int maxThreadsPerMultiProcessor; ///< Maximum resident threads per multi-processor.
29
- int computeMode; ///< Compute mode.
30
- int clockInstructionRate; ///< Frequency in khz of the timer used by the device-side "clock*"
31
- ///< instructions. New for HIP.
32
- hipDeviceArch_t arch; ///< Architectural feature flags. New for HIP.
33
- int concurrentKernels; ///< Device can possibly execute multiple kernels concurrently.
34
- int pciDomainID; ///< PCI Domain ID
35
- int pciBusID; ///< PCI Bus ID.
36
- int pciDeviceID; ///< PCI Device ID.
37
- size_t maxSharedMemoryPerMultiProcessor; ///< Maximum Shared Memory Per Multiprocessor.
38
- int isMultiGpuBoard; ///< 1 if device is on a multi-GPU board, 0 if not.
39
- int canMapHostMemory; ///< Check whether HIP can map host memory
40
- int gcnArch; ///< DEPRECATED: use gcnArchName instead
41
- char gcnArchName[256]; ///< AMD GCN Arch Name.
42
- int integrated; ///< APU vs dGPU
43
- int cooperativeLaunch; ///< HIP device supports cooperative launch
44
- int cooperativeMultiDeviceLaunch; ///< HIP device supports cooperative launch on multiple
45
- ///< devices
46
- int maxTexture1DLinear; ///< Maximum size for 1D textures bound to linear memory
47
- int maxTexture1D; ///< Maximum number of elements in 1D images
48
- int maxTexture2D[2]; ///< Maximum dimensions (width, height) of 2D images, in image elements
49
- int maxTexture3D[3]; ///< Maximum dimensions (width, height, depth) of 3D images, in image
50
- ///< elements
51
- unsigned int* hdpMemFlushCntl; ///< Addres of HDP_MEM_COHERENCY_FLUSH_CNTL register
52
- unsigned int* hdpRegFlushCntl; ///< Addres of HDP_REG_COHERENCY_FLUSH_CNTL register
53
- size_t memPitch; ///< Maximum pitch in bytes allowed by memory copies
54
- size_t textureAlignment; ///< Alignment requirement for textures
55
- size_t texturePitchAlignment; ///< Pitch alignment requirement for texture references bound to
56
- ///< pitched memory
57
- int kernelExecTimeoutEnabled; ///< Run time limit for kernels executed on the device
58
- int ECCEnabled; ///< Device has ECC support enabled
59
- int tccDriver; ///< 1:If device is Tesla device using TCC driver, else 0
60
- int cooperativeMultiDeviceUnmatchedFunc; ///< HIP device supports cooperative launch on
61
- ///< multiple
62
- /// devices with unmatched functions
63
- int cooperativeMultiDeviceUnmatchedGridDim; ///< HIP device supports cooperative launch on
64
- ///< multiple
65
- /// devices with unmatched grid dimensions
66
- int cooperativeMultiDeviceUnmatchedBlockDim; ///< HIP device supports cooperative launch on
67
- ///< multiple
68
- /// devices with unmatched block dimensions
69
- int cooperativeMultiDeviceUnmatchedSharedMem; ///< HIP device supports cooperative launch on
70
- ///< multiple
71
- /// devices with unmatched shared memories
72
- int isLargeBar; ///< 1: if it is a large PCI bar device, else 0
73
- int asicRevision; ///< Revision of the GPU in this device
74
- int managedMemory; ///< Device supports allocating managed memory on this system
75
- int directManagedMemAccessFromHost; ///< Host can directly access managed memory on the device
76
- ///< without migration
77
- int concurrentManagedAccess; ///< Device can coherently access managed memory concurrently with
78
- ///< the CPU
79
- int pageableMemoryAccess; ///< Device supports coherently accessing pageable memory
80
- ///< without calling hipHostRegister on it
81
- int pageableMemoryAccessUsesHostPageTables; ///< Device accesses pageable memory via the host's
82
- ///< page tables
83
- } hipDeviceProp_tR0000;
84
-
85
-
86
- #ifdef __cplusplus
87
- extern "C" {
88
- #endif
89
-
90
- hipError_t hipGetDevicePropertiesR0000(hipDeviceProp_tR0000* prop, int device);
91
- hipError_t hipChooseDeviceR0000(int* device, const hipDeviceProp_tR0000* prop);
92
-
93
- #ifdef __cplusplus
94
- }
95
- #endif
@@ -1,161 +0,0 @@
1
- /*
2
- Copyright (c) 2015 - 2021 Advanced Micro Devices, Inc. All rights reserved.
3
-
4
- Permission is hereby granted, free of charge, to any person obtaining a copy
5
- of this software and associated documentation files (the "Software"), to deal
6
- in the Software without restriction, including without limitation the rights
7
- to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
8
- copies of the Software, and to permit persons to whom the Software is
9
- furnished to do so, subject to the following conditions:
10
-
11
- The above copyright notice and this permission notice shall be included in
12
- all copies or substantial portions of the Software.
13
-
14
- THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
15
- IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
16
- FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
17
- AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
18
- LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
19
- OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
20
- THE SOFTWARE.
21
- */
22
-
23
- #ifndef HIP_INCLUDE_HIP_HIP_EXT_H
24
- #define HIP_INCLUDE_HIP_HIP_EXT_H
25
- #include "hip/hip_runtime.h"
26
- #if defined(__cplusplus)
27
- #include <tuple>
28
- #include <type_traits>
29
- #endif
30
- /** @addtogroup Module Module Management
31
- * @{
32
- */
33
-
34
- /**
35
- * @brief Launches kernel with parameters and shared memory on stream with arguments passed
36
- * to kernel params or extra arguments.
37
- *
38
- * @param [in] f Kernel to launch.
39
- * @param [in] globalWorkSizeX X grid dimension specified in work-items.
40
- * @param [in] globalWorkSizeY Y grid dimension specified in work-items.
41
- * @param [in] globalWorkSizeZ Z grid dimension specified in work-items.
42
- * @param [in] localWorkSizeX X block dimension specified in work-items.
43
- * @param [in] localWorkSizeY Y block dimension specified in work-items.
44
- * @param [in] localWorkSizeZ Z block dimension specified in work-items.
45
- * @param [in] sharedMemBytes Amount of dynamic shared memory to allocate for this kernel.
46
- * HIP-Clang compiler provides support for extern shared declarations.
47
- * @param [in] hStream Stream where the kernel should be dispatched.
48
- * May be 0, in which case the default stream is used with associated synchronization rules.
49
- * @param [in] kernelParams pointer to kernel parameters.
50
- * @param [in] extra Pointer to kernel arguments. These are passed directly to the kernel and
51
- * must be in the memory layout and alignment expected by the kernel.
52
- * All passed arguments must be naturally aligned according to their type. The memory address of each
53
- * argument should be a multiple of its size in bytes. Please refer to hip_porting_driver_api.md
54
- * for sample usage.
55
- * @param [in] startEvent If non-null, specified event will be updated to track the start time of
56
- * the kernel launch. The event must be created before calling this API.
57
- * @param [in] stopEvent If non-null, specified event will be updated to track the stop time of
58
- * the kernel launch. The event must be created before calling this API.
59
- * @param [in] flags The value of hipExtAnyOrderLaunch, signifies if kernel can be
60
- * launched in any order.
61
- * @returns #hipSuccess, #hipInvalidDeviceId, #hipErrorNotInitialized, #hipErrorInvalidValue.
62
- *
63
- * HIP/ROCm actually updates the start event when the associated kernel completes.
64
- * Currently, timing between startEvent and stopEvent does not include the time it takes to perform
65
- * a system scope release/cache flush - only the time it takes to issues writes to cache.
66
- *
67
- * @note For this HIP API, the flag 'hipExtAnyOrderLaunch' is not supported on AMD GFX9xx boards.
68
- *
69
- */
70
- HIP_PUBLIC_API
71
- extern "C" hipError_t hipExtModuleLaunchKernel(hipFunction_t f, uint32_t globalWorkSizeX,
72
- uint32_t globalWorkSizeY, uint32_t globalWorkSizeZ,
73
- uint32_t localWorkSizeX, uint32_t localWorkSizeY,
74
- uint32_t localWorkSizeZ, size_t sharedMemBytes,
75
- hipStream_t hStream, void** kernelParams, void** extra,
76
- hipEvent_t startEvent __dparm(NULL),
77
- hipEvent_t stopEvent __dparm(NULL),
78
- uint32_t flags __dparm(0));
79
- /**
80
- * @brief This HIP API is deprecated, please use hipExtModuleLaunchKernel() instead.
81
- *
82
- */
83
- DEPRECATED("use hipExtModuleLaunchKernel instead")
84
- HIP_PUBLIC_API
85
- extern "C" hipError_t hipHccModuleLaunchKernel(hipFunction_t f, uint32_t globalWorkSizeX,
86
- uint32_t globalWorkSizeY, uint32_t globalWorkSizeZ,
87
- uint32_t localWorkSizeX, uint32_t localWorkSizeY,
88
- uint32_t localWorkSizeZ, size_t sharedMemBytes,
89
- hipStream_t hStream, void** kernelParams, void** extra,
90
- hipEvent_t startEvent __dparm(NULL),
91
- hipEvent_t stopEvent __dparm(NULL));
92
-
93
- #if defined(__cplusplus)
94
-
95
- /**
96
- * @brief Launches kernel from the pointer address, with arguments and shared memory on stream.
97
- *
98
- * @param [in] function_address pointer to the Kernel to launch.
99
- * @param [in] numBlocks number of blocks.
100
- * @param [in] dimBlocks dimension of a block.
101
- * @param [in] args pointer to kernel arguments.
102
- * @param [in] sharedMemBytes Amount of dynamic shared memory to allocate for this kernel.
103
- * HIP-Clang compiler provides support for extern shared declarations.
104
- * @param [in] stream Stream where the kernel should be dispatched.
105
- * May be 0, in which case the default stream is used with associated synchronization rules.
106
- * @param [in] startEvent If non-null, specified event will be updated to track the start time of
107
- * the kernel launch. The event must be created before calling this API.
108
- * @param [in] stopEvent If non-null, specified event will be updated to track the stop time of
109
- * the kernel launch. The event must be created before calling this API.
110
- * @param [in] flags The value of hipExtAnyOrderLaunch, signifies if kernel can be
111
- * launched in any order.
112
- * @returns #hipSuccess, #hipInvalidDeviceId, #hipErrorNotInitialized, #hipErrorInvalidValue.
113
- *
114
- */
115
- extern "C" hipError_t hipExtLaunchKernel(const void* function_address, dim3 numBlocks,
116
- dim3 dimBlocks, void** args, size_t sharedMemBytes,
117
- hipStream_t stream, hipEvent_t startEvent,
118
- hipEvent_t stopEvent, int flags);
119
-
120
- /**
121
- * @brief Launches kernel with dimention parameters and shared memory on stream with templated kernel and arguments.
122
- *
123
- * @param [in] kernel Kernel to launch.
124
- * @param [in] numBlocks const number of blocks.
125
- * @param [in] dimBlocks const dimension of a block.
126
- * @param [in] sharedMemBytes Amount of dynamic shared memory to allocate for this kernel.
127
- * HIP-Clang compiler provides support for extern shared declarations.
128
- * @param [in] stream Stream where the kernel should be dispatched.
129
- * May be 0, in which case the default stream is used with associated synchronization rules.
130
- * @param [in] startEvent If non-null, specified event will be updated to track the start time of
131
- * the kernel launch. The event must be created before calling this API.
132
- * @param [in] stopEvent If non-null, specified event will be updated to track the stop time of
133
- * the kernel launch. The event must be created before calling this API.
134
- * @param [in] flags The value of hipExtAnyOrderLaunch, signifies if kernel can be
135
- * launched in any order.
136
- * @param [in] args templated kernel arguments.
137
- *
138
- */
139
- template <typename... Args, typename F = void (*)(Args...)>
140
- inline void hipExtLaunchKernelGGL(F kernel, const dim3& numBlocks, const dim3& dimBlocks,
141
- std::uint32_t sharedMemBytes, hipStream_t stream,
142
- hipEvent_t startEvent, hipEvent_t stopEvent, std::uint32_t flags,
143
- Args... args) {
144
- constexpr size_t count = sizeof...(Args);
145
- auto tup_ = std::tuple<Args...>{args...};
146
- auto tup = validateArgsCountType(kernel, tup_);
147
- void* _Args[count];
148
- pArgs<0>(tup, _Args);
149
-
150
- auto k = reinterpret_cast<void*>(kernel);
151
- hipExtLaunchKernel(k, numBlocks, dimBlocks, _Args, sharedMemBytes, stream, startEvent,
152
- stopEvent, (int)flags);
153
- }
154
-
155
- #endif // defined(__cplusplus)
156
-
157
- // doxygen end AMD-specific features
158
- /**
159
- * @}
160
- */
161
- #endif // #iidef HIP_INCLUDE_HIP_HIP_EXT_H
@@ -1,36 +0,0 @@
1
- /*
2
- Copyright (c) 2015 - 2021 Advanced Micro Devices, Inc. All rights reserved.
3
-
4
- Permission is hereby granted, free of charge, to any person obtaining a copy
5
- of this software and associated documentation files (the "Software"), to deal
6
- in the Software without restriction, including without limitation the rights
7
- to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
8
- copies of the Software, and to permit persons to whom the Software is
9
- furnished to do so, subject to the following conditions:
10
-
11
- The above copyright notice and this permission notice shall be included in
12
- all copies or substantial portions of the Software.
13
-
14
- THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
15
- IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
16
- FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
17
- AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
18
- LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
19
- OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
20
- THE SOFTWARE.
21
- */
22
-
23
- #ifndef HIP_INCLUDE_HIP_HIP_FP16_H
24
- #define HIP_INCLUDE_HIP_HIP_FP16_H
25
-
26
- #include <hip/hip_common.h>
27
-
28
- #if defined(__HIP_PLATFORM_AMD__) && !defined(__HIP_PLATFORM_NVIDIA__)
29
- #include <hip/amd_detail/amd_hip_fp16.h>
30
- #elif !defined(__HIP_PLATFORM_AMD__) && defined(__HIP_PLATFORM_NVIDIA__)
31
- #include "cuda_fp16.h"
32
- #else
33
- #error("Must define exactly one of __HIP_PLATFORM_AMD__ or __HIP_PLATFORM_NVIDIA__");
34
- #endif
35
-
36
- #endif
@@ -1,33 +0,0 @@
1
- /*
2
- Copyright (c) 2024 Advanced Micro Devices, Inc. All rights reserved.
3
-
4
- Permission is hereby granted, free of charge, to any person obtaining a copy
5
- of this software and associated documentation files (the "Software"), to deal
6
- in the Software without restriction, including without limitation the rights
7
- to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
8
- copies of the Software, and to permit persons to whom the Software is
9
- furnished to do so, subject to the following conditions:
10
-
11
- The above copyright notice and this permission notice shall be included in
12
- all copies or substantial portions of the Software.
13
-
14
- THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
15
- IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
16
- FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
17
- AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
18
- LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
19
- OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
20
- THE SOFTWARE.
21
- */
22
-
23
- #ifndef HIP_INCLUDE_HIP_HIP_FP8_H
24
- #define HIP_INCLUDE_HIP_HIP_FP8_H
25
-
26
- #include <hip/hip_common.h>
27
-
28
- #if defined(__HIP_PLATFORM_AMD__) && !defined(__HIP_PLATFORM_NVIDIA__)
29
- // We only have fnuz defs for now, which are not supported by other platforms
30
- #include <hip/amd_detail/amd_hip_fp8.h>
31
- #endif
32
-
33
- #endif // HIP_INCLUDE_HIP_HIP_FP8_H
@@ -1,32 +0,0 @@
1
- /*
2
- Copyright (c) 2023 Advanced Micro Devices, Inc. All rights reserved.
3
-
4
- Permission is hereby granted, free of charge, to any person obtaining a copy
5
- of this software and associated documentation files (the "Software"), to deal
6
- in the Software without restriction, including without limitation the rights
7
- to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
8
- copies of the Software, and to permit persons to whom the Software is
9
- furnished to do so, subject to the following conditions:
10
-
11
- The above copyright notice and this permission notice shall be included in
12
- all copies or substantial portions of the Software.
13
-
14
- THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
15
- IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
16
- FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
17
- AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
18
- LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
19
- OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
20
- THE SOFTWARE.
21
- */
22
- #ifndef HIP_GL_INTEROP_H
23
- #define HIP_GL_INTEROP_H
24
-
25
- #include <hip/hip_common.h>
26
-
27
- #if defined(__HIP_PLATFORM_AMD__) && !defined(__HIP_PLATFORM_NVIDIA__)
28
- #include "hip/amd_detail/amd_hip_gl_interop.h"
29
- #elif !defined(__HIP_PLATFORM_AMD__) && defined(__HIP_PLATFORM_NVIDIA__)
30
- #include "hip/nvidia_detail/nvidia_hip_gl_interop.h"
31
- #endif
32
- #endif
@@ -1,24 +0,0 @@
1
- /*
2
- Copyright (c) 2015 - 2021 Advanced Micro Devices, Inc. All rights reserved.
3
- Permission is hereby granted, free of charge, to any person obtaining a copy
4
- of this software and associated documentation files (the "Software"), to deal
5
- in the Software without restriction, including without limitation the rights
6
- to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
7
- copies of the Software, and to permit persons to whom the Software is
8
- furnished to do so, subject to the following conditions:
9
- The above copyright notice and this permission notice shall be included in
10
- all copies or substantial portions of the Software.
11
- THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
12
- IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
13
- FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
14
- AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
15
- LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
16
- OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
17
- THE SOFTWARE.
18
- */
19
-
20
- #ifndef HIP_INCLUDE_HIP_HIP_HCC_H
21
- #define HIP_INCLUDE_HIP_HIP_HCC_H
22
- #warning "hip/hip_hcc.h is deprecated, please use hip/hip_ext.h"
23
- #include "hip/hip_ext.h"
24
- #endif // #ifdef HIP_INCLUDE_HIP_HIP_HCC_H
@@ -1,36 +0,0 @@
1
- /*
2
- Copyright (c) 2015 - 2022 Advanced Micro Devices, Inc. All rights reserved.
3
-
4
- Permission is hereby granted, free of charge, to any person obtaining a copy
5
- of this software and associated documentation files (the "Software"), to deal
6
- in the Software without restriction, including without limitation the rights
7
- to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
8
- copies of the Software, and to permit persons to whom the Software is
9
- furnished to do so, subject to the following conditions:
10
-
11
- The above copyright notice and this permission notice shall be included in
12
- all copies or substantial portions of the Software.
13
-
14
- THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
15
- IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
16
- FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
17
- AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
18
- LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
19
- OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
20
- THE SOFTWARE.
21
- */
22
- #ifndef HIP_MATH_CONSTANTS_H
23
- #define HIP_MATH_CONSTANTS_H
24
-
25
- #if !defined(__HIPCC_RTC__)
26
- #include <hip/hip_common.h>
27
- #endif
28
-
29
- #if defined(__HIP_PLATFORM_AMD__) && !defined(__HIP_PLATFORM_NVIDIA__)
30
- #include "hip/amd_detail/amd_hip_math_constants.h"
31
- #elif !defined(__HIP_PLATFORM_AMD__) && defined(__HIP_PLATFORM_NVIDIA__)
32
- #include "hip/nvidia_detail/nvidia_hip_math_constants.h"
33
- #else
34
- #error("Must define exactly one of __HIP_PLATFORM_AMD__ or __HIP_PLATFORM_NVIDIA__");
35
- #endif
36
- #endif
@@ -1,27 +0,0 @@
1
- /*
2
- Copyright (c) 2015 - 2021 Advanced Micro Devices, Inc. All rights reserved.
3
- Permission is hereby granted, free of charge, to any person obtaining a copy
4
- of this software and associated documentation files (the "Software"), to deal
5
- in the Software without restriction, including without limitation the rights
6
- to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
7
- copies of the Software, and to permit persons to whom the Software is
8
- furnished to do so, subject to the following conditions:
9
- The above copyright notice and this permission notice shall be included in
10
- all copies or substantial portions of the Software.
11
- THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
12
- IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
13
- FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
14
- AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
15
- LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
16
- OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
17
- THE SOFTWARE.
18
- */
19
-
20
- #ifndef HIP_INCLUDE_HIP_HIP_PROFILE_H
21
- #define HIP_INCLUDE_HIP_HIP_PROFILE_H
22
-
23
- #define HIP_SCOPED_MARKER(markerName, group)
24
- #define HIP_BEGIN_MARKER(markerName, group)
25
- #define HIP_END_MARKER()
26
-
27
- #endif
@@ -1,75 +0,0 @@
1
- /*
2
- Copyright (c) 2015 - 2021 Advanced Micro Devices, Inc. All rights reserved.
3
-
4
- Permission is hereby granted, free of charge, to any person obtaining a copy
5
- of this software and associated documentation files (the "Software"), to deal
6
- in the Software without restriction, including without limitation the rights
7
- to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
8
- copies of the Software, and to permit persons to whom the Software is
9
- furnished to do so, subject to the following conditions:
10
-
11
- The above copyright notice and this permission notice shall be included in
12
- all copies or substantial portions of the Software.
13
-
14
- THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
15
- IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
16
- FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
17
- AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
18
- LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
19
- OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
20
- THE SOFTWARE.
21
- */
22
-
23
- //! HIP = Heterogeneous-compute Interface for Portability
24
- //!
25
- //! Define a extremely thin runtime layer that allows source code to be compiled unmodified
26
- //! through either AMD CLANG or NVCC. Key features tend to be in the spirit
27
- //! and terminology of CUDA, but with a portable path to other accelerators as well:
28
- //
29
- //! Both paths support rich C++ features including classes, templates, lambdas, etc.
30
- //! Runtime API is C
31
- //! Memory management is based on pure pointers and resembles malloc/free/copy.
32
- //
33
- //! hip_runtime.h : includes everything in hip_api.h, plus math builtins and kernel launch
34
- //! macros. hip_runtime_api.h : Defines HIP API. This is a C header file and does not use any C++
35
- //! features.
36
-
37
- #ifndef HIP_INCLUDE_HIP_HIP_RUNTIME_H
38
- #define HIP_INCLUDE_HIP_HIP_RUNTIME_H
39
-
40
- #if __HIP_DEVICE_COMPILE__ && !__GFX7__ && !__GFX8__ && !__GFX9__ && __AMDGCN_WAVEFRONT_SIZE == 64
41
- #error HIP is not supported on the specified GPU ARCH with wavefront size 64
42
- #endif
43
-
44
- #if !defined(__HIPCC_RTC__)
45
- // Some standard header files, these are included by hc.hpp and so want to make them avail on both
46
- // paths to provide a consistent include env and avoid "missing symbol" errors that only appears
47
- // on NVCC path:
48
- #include <stdint.h>
49
- #include <stdio.h>
50
- #include <stdlib.h>
51
- #include <assert.h>
52
-
53
- #if __cplusplus > 199711L
54
- #include <thread>
55
- #endif
56
- #endif // !defined(__HIPCC_RTC__)
57
-
58
- #include <hip/hip_version.h>
59
- #include <hip/hip_common.h>
60
-
61
- #if defined(__HIP_PLATFORM_AMD__) && !defined(__HIP_PLATFORM_NVIDIA__)
62
- #include <hip/amd_detail/amd_hip_runtime.h>
63
- #elif !defined(__HIP_PLATFORM_AMD__) && defined(__HIP_PLATFORM_NVIDIA__)
64
- #include <hip/nvidia_detail/nvidia_hip_runtime.h>
65
- #else
66
- #error("Must define exactly one of __HIP_PLATFORM_AMD__ or __HIP_PLATFORM_NVIDIA__");
67
- #endif
68
-
69
- #if !defined(__HIPCC_RTC__)
70
- #include <hip/hip_runtime_api.h>
71
- #include <hip/library_types.h>
72
- #endif // !defined(__HIPCC_RTC__)
73
- #include <hip/hip_vector_types.h>
74
-
75
- #endif