/build/miopen-1weXEh/miopen-7.14+dfsg/include/miopen/miopen.h Source File

/build/miopen-1weXEh/miopen-7.14+dfsg/include/miopen/miopen.h Source File#

MIOpen: /build/miopen-1weXEh/miopen-7.14+dfsg/include/miopen/miopen.h Source File
miopen.h
Go to the documentation of this file.
1// Copyright (c) Advanced Micro Devices, Inc., or its affiliates.
2// SPDX-License-Identifier: MIT
3
4#ifndef MIOPEN_GUARD_MIOPEN_H_
5#define MIOPEN_GUARD_MIOPEN_H_
6
7#ifdef __clang__
8#pragma clang diagnostic push
9#pragma clang diagnostic ignored "-Wextern-c-compat"
10#endif
11
12#include <stddef.h>
13#include <stdbool.h>
14#include <miopen/config.h>
15#include <miopen/export.h>
16
17#if MIOPEN_BACKEND_OPENCL
18#define CL_TARGET_OPENCL_VERSION 120
19#if defined(__APPLE__) || defined(__MACOSX)
20#include <OpenCL/cl.h>
21#else
22#define CL_USE_DEPRECATED_OPENCL_1_2_APIS
23#include <CL/cl.h>
24#endif
25
26#elif MIOPEN_BACKEND_HIP
27#include <hip/hip_runtime_api.h>
28#endif
29
30/*
31 * @defgroup convolutions
32 * @defgroup pooling
33 * @defgroup handle
34 * @defgroup layernorm
35 * @defgroup LRN
36 * @defgroup batchnorm
37 * @defgroup activation
38 * @defgroup tensor
39 * @defgroup softmax
40 * @defgroup RNN
41 * @defgroup fusion
42 * @defgroup LossFunction
43 * @defgroup TensorReduce
44 * @defgroup find2
45 * @defgroup ReduceExtreme
46 * @defgroup groupnorm
47 * @defgroup cat
48 * @defgroup SGD
49 * @defgroup getitem
50 * @defgroup ReduceCalculation
51 * @defgroup RotaryPositionalEmbeddings
52 * @defgroup ReLU
53 *
54 */
55
57#define MIOPEN_DECLARE_OBJECT(name) \
58 struct name \
59 { \
60 }; \
61 typedef struct name* name##_t;
62
63#ifdef __cplusplus
64extern "C" {
65#endif
66
67#if MIOPEN_BACKEND_OPENCL
68typedef cl_command_queue miopenAcceleratorQueue_t;
69#elif MIOPEN_BACKEND_HIP
70typedef hipStream_t miopenAcceleratorQueue_t;
71#endif
72
76MIOPEN_DECLARE_OBJECT(miopenHandle);
77
82
100
101typedef enum
102{
103 // TODO:(LYM) temporary use Pedantic as default until TF32 is fully supported
104 miopenMathDefault = 0,
105 miopenMathPedantic =
106 1,
107} miopenMathType_t;
108
109#ifdef MIOPEN_BETA_API
115#endif
116
124MIOPEN_EXPORT const char* miopenGetErrorString(miopenStatus_t error);
125
134typedef void* (*miopenAllocatorFunction)(void* context, size_t sizeBytes);
135
144typedef void (*miopenDeallocatorFunction)(void* context, void* memory);
145
159MIOPEN_EXPORT miopenStatus_t miopenGetVersion(size_t* major, size_t* minor, size_t* patch);
160
169MIOPEN_EXPORT miopenStatus_t miopenCreate(miopenHandle_t* handle);
170
182MIOPEN_EXPORT miopenStatus_t miopenCreateWithStream(miopenHandle_t* handle,
183 miopenAcceleratorQueue_t stream);
184
191MIOPEN_EXPORT miopenStatus_t miopenDestroy(miopenHandle_t handle);
192
200MIOPEN_EXPORT miopenStatus_t miopenSetStream(miopenHandle_t handle,
201 miopenAcceleratorQueue_t streamID);
202
210MIOPEN_EXPORT miopenStatus_t miopenGetStream(miopenHandle_t handle,
211 miopenAcceleratorQueue_t* streamID);
212
229MIOPEN_EXPORT miopenStatus_t miopenSetAllocator(miopenHandle_t handle,
230 miopenAllocatorFunction allocator,
231 miopenDeallocatorFunction deallocator,
232 void* allocatorContext);
233
245MIOPEN_EXPORT miopenStatus_t miopenGetKernelTime(miopenHandle_t handle, float* time);
246
254MIOPEN_EXPORT miopenStatus_t miopenEnableProfiling(miopenHandle_t handle, bool enable);
256// CLOSEOUT HANDLE DOXYGEN GROUP
257
265MIOPEN_DECLARE_OBJECT(miopenFusionOpDescriptor);
266
274MIOPEN_DECLARE_OBJECT(miopenTensorDescriptor);
275
282MIOPEN_DECLARE_OBJECT(miopenSeqTensorDescriptor);
283
291MIOPEN_DECLARE_OBJECT(miopenConvolutionDescriptor);
292
300MIOPEN_DECLARE_OBJECT(miopenPoolingDescriptor);
301
309MIOPEN_DECLARE_OBJECT(miopenLRNDescriptor);
310
317MIOPEN_DECLARE_OBJECT(miopenActivationDescriptor);
318
322MIOPEN_DECLARE_OBJECT(miopenRNNDescriptor);
323
327MIOPEN_DECLARE_OBJECT(miopenCTCLossDescriptor);
328
332MIOPEN_DECLARE_OBJECT(miopenDropoutDescriptor);
333
337MIOPEN_DECLARE_OBJECT(miopenReduceTensorDescriptor);
338
342MIOPEN_DECLARE_OBJECT(miopenMhaDescriptor);
343
347MIOPEN_DECLARE_OBJECT(miopenSoftmaxDescriptor);
348
353typedef enum
354{
355 miopenHalf = 0,
356 miopenFloat = 1,
357 miopenInt32 = 2,
358 miopenInt8 = 3,
359 // miopenInt8x4 = 4, /*!< Pack of 4x Int8 in NCHW_VECT_C format (Support discontinued) */
360 miopenBFloat16 = 5,
362 miopenDouble = 6,
363 miopenFloat8_fnuz = 7,
364 miopenBFloat8_fnuz = 8,
365 miopenInt64 = 9,
366} miopenDataType_t;
367
385
397
409
421
432
443
455
465#ifdef MIOPEN_BETA_API
484#endif
494
517
528
539
547#define MIOPEN_API_VERSION_REDUCE_TENSOR 1
548
553typedef enum
554{
555 MIOPEN_REDUCE_TENSOR_ADD = 0,
556 MIOPEN_REDUCE_TENSOR_MUL =
557 1,
558 MIOPEN_REDUCE_TENSOR_MIN =
559 2,
560 MIOPEN_REDUCE_TENSOR_MAX =
561 3,
562 MIOPEN_REDUCE_TENSOR_AMAX =
563 4,
564 MIOPEN_REDUCE_TENSOR_AVG =
565 5,
566 MIOPEN_REDUCE_TENSOR_NORM1 =
567 6,
568 MIOPEN_REDUCE_TENSOR_NORM2 = 7,
570 // MIOPEN_REDUCE_TENSOR_MUL_NO_ZEROS =
571 // 8, /*!< the operation is same as MUL, but does not have the zero values considered */
572} miopenReduceTensorOp_t;
573
583
593
605
610typedef enum
611{
612 MIOPEN_CONVOLUTION_ATTRIB_FP16_ALT_IMPL =
613 0,
616 MIOPEN_CONVOLUTION_ATTRIB_DETERMINISTIC =
617 1,
619#ifdef MIOPEN_BETA_API
620 MIOPEN_CONVOLUTION_ATTRIB_FP8_ROUNDING_MODE =
621 2,
625#else
626// miopenReserved1 = 2,
627#endif
628 // TODO:(LYM) temporarily use Pedantic as default until TF32 is fully supported
629 MIOPEN_CONVOLUTION_ATTRIB_MATH_TYPE =
630 3,
631} miopenConvolutionAttrib_t;
632
662
667
674MIOPEN_EXPORT miopenStatus_t miopenCreateTensorDescriptor(miopenTensorDescriptor_t* tensorDesc);
675
688MIOPEN_EXPORT miopenStatus_t miopenSet4dTensorDescriptor(
689 miopenTensorDescriptor_t tensorDesc, miopenDataType_t dataType, int n, int c, int h, int w);
690
701MIOPEN_EXPORT miopenStatus_t
702miopenSetNdTensorDescriptorWithLayout(miopenTensorDescriptor_t tensorDesc,
703 miopenDataType_t dataType,
704 miopenTensorLayout_t tensorLayout,
705 const int* lens,
706 int num_lens);
726MIOPEN_EXPORT miopenStatus_t miopenSet4dTensorDescriptorEx(miopenTensorDescriptor_t tensorDesc,
727 miopenDataType_t dataType,
728 int n,
729 int c,
730 int h,
731 int w,
732 int nStride,
733 int cStride,
734 int hStride,
735 int wStride);
736
753MIOPEN_EXPORT miopenStatus_t miopenGet4dTensorDescriptor(miopenTensorDescriptor_t tensorDesc,
754 miopenDataType_t* dataType,
755 int* n,
756 int* c,
757 int* h,
758 int* w,
759 int* nStride,
760 int* cStride,
761 int* hStride,
762 int* wStride);
763
774MIOPEN_EXPORT miopenStatus_t miopenSetTensorDescriptor(miopenTensorDescriptor_t tensorDesc,
775 miopenDataType_t dataType,
776 int nbDims,
777 const int* dimsA,
778 const int* stridesA);
779
780#ifdef MIOPEN_BETA_API
783MIOPEN_EXPORT miopenStatus_t miopenSetTensorDescriptorV2(miopenTensorDescriptor_t tensorDesc,
784 miopenDataType_t dataType,
785 int nbDims,
786 const size_t* dimsA,
787 const size_t* stridesA);
788#endif
789
790#ifdef MIOPEN_BETA_API
800MIOPEN_EXPORT miopenStatus_t miopenSetTensorCastType(miopenTensorDescriptor_t tensorDesc,
801 miopenDataType_t cast_type);
802#endif
803
812MIOPEN_EXPORT miopenStatus_t miopenGetTensorDescriptorSize(miopenTensorDescriptor_t tensorDesc,
813 int* size);
814
823MIOPEN_EXPORT miopenStatus_t miopenGetTensorDescriptor(miopenTensorDescriptor_t tensorDesc,
824 miopenDataType_t* dataType,
825 int* dimsA,
826 int* stridesA);
827
833MIOPEN_EXPORT miopenStatus_t miopenDestroyTensorDescriptor(miopenTensorDescriptor_t tensorDesc);
834
841MIOPEN_EXPORT miopenStatus_t
842miopenCreateSeqTensorDescriptor(miopenSeqTensorDescriptor_t* tensorDesc);
843
849MIOPEN_EXPORT miopenStatus_t
850miopenDestroySeqTensorDescriptor(miopenSeqTensorDescriptor_t tensorDesc);
851
871MIOPEN_EXPORT miopenStatus_t miopenOpTensor(miopenHandle_t handle,
872 miopenTensorOp_t tensorOp,
873 const void* alpha1,
874 const miopenTensorDescriptor_t aDesc,
875 const void* A,
876 const void* alpha2,
877 const miopenTensorDescriptor_t bDesc,
878 const void* B,
879 const void* beta,
880 const miopenTensorDescriptor_t cDesc,
881 void* C);
882
893MIOPEN_EXPORT miopenStatus_t miopenSetTensor(miopenHandle_t handle,
894 const miopenTensorDescriptor_t yDesc,
895 void* y,
896 const void* alpha);
897
908MIOPEN_EXPORT miopenStatus_t miopenScaleTensor(miopenHandle_t handle,
909 const miopenTensorDescriptor_t yDesc,
910 void* y,
911 const void* alpha);
912
919MIOPEN_EXPORT miopenStatus_t miopenGetTensorNumBytes(miopenTensorDescriptor_t tensorDesc,
920 size_t* numBytes);
921
939MIOPEN_EXPORT miopenStatus_t miopenTransformTensor(miopenHandle_t handle,
940 const void* alpha,
941 const miopenTensorDescriptor_t xDesc,
942 const void* x,
943 const void* beta,
944 const miopenTensorDescriptor_t yDesc,
945 void* y);
946
948// CLOSEOUT TENSOR DOXYGEN GROUP
949
954
960MIOPEN_EXPORT miopenStatus_t
961miopenCreateConvolutionDescriptor(miopenConvolutionDescriptor_t* convDesc);
962
978MIOPEN_EXPORT miopenStatus_t miopenInitConvolutionDescriptor(miopenConvolutionDescriptor_t convDesc,
979 miopenConvolutionMode_t c_mode,
980 int pad_h,
981 int pad_w,
982 int stride_h,
983 int stride_w,
984 int dilation_h,
985 int dilation_w);
986
997MIOPEN_EXPORT miopenStatus_t
998miopenInitConvolutionNdDescriptor(miopenConvolutionDescriptor_t convDesc,
999 int spatialDim,
1000 const int* padA,
1001 const int* strideA,
1002 const int* dilationA,
1003 miopenConvolutionMode_t c_mode);
1004
1011MIOPEN_EXPORT miopenStatus_t miopenGetConvolutionSpatialDim(miopenConvolutionDescriptor_t convDesc,
1012 int* spatialDim);
1013
1029MIOPEN_EXPORT miopenStatus_t miopenGetConvolutionDescriptor(miopenConvolutionDescriptor_t convDesc,
1030 miopenConvolutionMode_t* c_mode,
1031 int* pad_h,
1032 int* pad_w,
1033 int* stride_h,
1034 int* stride_w,
1035 int* dilation_h,
1036 int* dilation_w);
1037
1049MIOPEN_EXPORT miopenStatus_t
1050miopenGetConvolutionNdDescriptor(miopenConvolutionDescriptor_t convDesc,
1051 int requestedSpatialDim,
1052 int* spatialDim,
1053 int* padA,
1054 int* strideA,
1055 int* dilationA,
1056 miopenConvolutionMode_t* c_mode);
1057
1064MIOPEN_EXPORT miopenStatus_t miopenGetConvolutionGroupCount(miopenConvolutionDescriptor_t convDesc,
1065 int* groupCount);
1066
1080MIOPEN_EXPORT miopenStatus_t miopenSetConvolutionGroupCount(miopenConvolutionDescriptor_t convDesc,
1081 int groupCount);
1082
1095MIOPEN_EXPORT miopenStatus_t
1096miopenSetTransposeConvOutputPadding(miopenConvolutionDescriptor_t convDesc, int adj_h, int adj_w);
1097
1110MIOPEN_EXPORT miopenStatus_t miopenSetTransposeConvNdOutputPadding(
1111 miopenConvolutionDescriptor_t convDesc, int spatialDim, const int* adjA);
1112
1130MIOPEN_EXPORT miopenStatus_t
1131miopenGetConvolutionForwardOutputDim(miopenConvolutionDescriptor_t convDesc,
1132 const miopenTensorDescriptor_t inputTensorDesc,
1133 const miopenTensorDescriptor_t filterDesc,
1134 int* n,
1135 int* c,
1136 int* h,
1137 int* w);
1138
1152MIOPEN_EXPORT miopenStatus_t
1153miopenGetConvolutionNdForwardOutputDim(miopenConvolutionDescriptor_t convDesc,
1154 const miopenTensorDescriptor_t inputTensorDesc,
1155 const miopenTensorDescriptor_t filterDesc,
1156 int* nDim,
1157 int* outputTensorDimA);
1158
1164MIOPEN_EXPORT miopenStatus_t
1165miopenDestroyConvolutionDescriptor(miopenConvolutionDescriptor_t convDesc);
1166
1173MIOPEN_EXPORT miopenStatus_t miopenSetConvolutionAttribute(miopenConvolutionDescriptor_t convDesc,
1174 const miopenConvolutionAttrib_t attr,
1175 int value);
1176
1183MIOPEN_EXPORT miopenStatus_t miopenGetConvolutionAttribute(miopenConvolutionDescriptor_t convDesc,
1184 const miopenConvolutionAttrib_t attr,
1185 int* value);
1186
1200MIOPEN_EXPORT miopenStatus_t miopenSetConvolutionFindMode(miopenConvolutionDescriptor_t convDesc,
1201 miopenConvolutionFindMode_t findMode);
1202
1209MIOPEN_EXPORT miopenStatus_t miopenGetConvolutionFindMode(
1210 const miopenConvolutionDescriptor_t convDesc, miopenConvolutionFindMode_t* findMode);
1211
1224
1235
1249
1261
1283
1303
1319MIOPEN_EXPORT miopenStatus_t
1320miopenConvolutionForwardGetSolutionCount(miopenHandle_t handle,
1321 const miopenTensorDescriptor_t wDesc,
1322 const miopenTensorDescriptor_t xDesc,
1323 const miopenConvolutionDescriptor_t convDesc,
1324 const miopenTensorDescriptor_t yDesc,
1325 size_t* solutionCount);
1326
1350MIOPEN_EXPORT miopenStatus_t
1351miopenConvolutionForwardGetSolution(miopenHandle_t handle,
1352 const miopenTensorDescriptor_t wDesc,
1353 const miopenTensorDescriptor_t xDesc,
1354 const miopenConvolutionDescriptor_t convDesc,
1355 const miopenTensorDescriptor_t yDesc,
1356 const size_t maxSolutionCount,
1357 size_t* solutionCount,
1358 miopenConvSolution_t* solutions);
1359
1377MIOPEN_EXPORT miopenStatus_t
1378miopenConvolutionForwardGetSolutionWorkspaceSize(miopenHandle_t handle,
1379 const miopenTensorDescriptor_t wDesc,
1380 const miopenTensorDescriptor_t xDesc,
1381 const miopenConvolutionDescriptor_t convDesc,
1382 const miopenTensorDescriptor_t yDesc,
1383 const uint64_t solution_id,
1384 size_t* workSpaceSize);
1385
1403MIOPEN_EXPORT miopenStatus_t
1404miopenConvolutionForwardCompileSolution(miopenHandle_t handle,
1405 const miopenTensorDescriptor_t wDesc,
1406 const miopenTensorDescriptor_t xDesc,
1407 const miopenConvolutionDescriptor_t convDesc,
1408 const miopenTensorDescriptor_t yDesc,
1409 const uint64_t solution_id);
1410
1428MIOPEN_EXPORT miopenStatus_t
1429miopenConvolutionForwardImmediate(miopenHandle_t handle,
1430 const miopenTensorDescriptor_t wDesc,
1431 const void* w,
1432 const miopenTensorDescriptor_t xDesc,
1433 const void* x,
1434 const miopenConvolutionDescriptor_t convDesc,
1435 const miopenTensorDescriptor_t yDesc,
1436 void* y,
1437 void* workSpace,
1438 size_t workSpaceSize,
1439 const uint64_t solution_id);
1440
1458MIOPEN_EXPORT miopenStatus_t
1459miopenConvolutionBackwardDataGetSolutionCount(miopenHandle_t handle,
1460 const miopenTensorDescriptor_t dyDesc,
1461 const miopenTensorDescriptor_t wDesc,
1462 const miopenConvolutionDescriptor_t convDesc,
1463 const miopenTensorDescriptor_t dxDesc,
1464 size_t* solutionCount);
1465
1490MIOPEN_EXPORT miopenStatus_t
1491miopenConvolutionBackwardDataGetSolution(miopenHandle_t handle,
1492 const miopenTensorDescriptor_t dyDesc,
1493 const miopenTensorDescriptor_t wDesc,
1494 const miopenConvolutionDescriptor_t convDesc,
1495 const miopenTensorDescriptor_t dxDesc,
1496 const size_t maxSolutionCount,
1497 size_t* solutionCount,
1498 miopenConvSolution_t* solutions);
1499
1517MIOPEN_EXPORT miopenStatus_t
1518miopenConvolutionBackwardDataGetSolutionWorkspaceSize(miopenHandle_t handle,
1519 const miopenTensorDescriptor_t dyDesc,
1520 const miopenTensorDescriptor_t wDesc,
1521 const miopenConvolutionDescriptor_t convDesc,
1522 const miopenTensorDescriptor_t dxDesc,
1523 const uint64_t solution_id,
1524 size_t* workSpaceSize);
1525
1544MIOPEN_EXPORT miopenStatus_t
1545miopenConvolutionBackwardDataCompileSolution(miopenHandle_t handle,
1546 const miopenTensorDescriptor_t dyDesc,
1547 const miopenTensorDescriptor_t wDesc,
1548 const miopenConvolutionDescriptor_t convDesc,
1549 const miopenTensorDescriptor_t dxDesc,
1550 const uint64_t solution_id);
1551
1569MIOPEN_EXPORT miopenStatus_t
1570miopenConvolutionBackwardDataImmediate(miopenHandle_t handle,
1571 const miopenTensorDescriptor_t dyDesc,
1572 const void* dy,
1573 const miopenTensorDescriptor_t wDesc,
1574 const void* w,
1575 const miopenConvolutionDescriptor_t convDesc,
1576 const miopenTensorDescriptor_t dxDesc,
1577 void* dx,
1578 void* workSpace,
1579 size_t workSpaceSize,
1580 const uint64_t solution_id);
1581
1599MIOPEN_EXPORT miopenStatus_t
1600miopenConvolutionBackwardWeightsGetSolutionCount(miopenHandle_t handle,
1601 const miopenTensorDescriptor_t dyDesc,
1602 const miopenTensorDescriptor_t xDesc,
1603 const miopenConvolutionDescriptor_t convDesc,
1604 const miopenTensorDescriptor_t dwDesc,
1605 size_t* solutionCount);
1606
1630MIOPEN_EXPORT miopenStatus_t
1631miopenConvolutionBackwardWeightsGetSolution(miopenHandle_t handle,
1632 const miopenTensorDescriptor_t dyDesc,
1633 const miopenTensorDescriptor_t xDesc,
1634 const miopenConvolutionDescriptor_t convDesc,
1635 const miopenTensorDescriptor_t dwDesc,
1636 const size_t maxSolutionCount,
1637 size_t* solutionCount,
1638 miopenConvSolution_t* solutions);
1639
1657MIOPEN_EXPORT miopenStatus_t miopenConvolutionBackwardWeightsGetSolutionWorkspaceSize(
1658 miopenHandle_t handle,
1659 const miopenTensorDescriptor_t dyDesc,
1660 const miopenTensorDescriptor_t xDesc,
1661 const miopenConvolutionDescriptor_t convDesc,
1662 const miopenTensorDescriptor_t dwDesc,
1663 const uint64_t solution_id,
1664 size_t* workSpaceSize);
1665
1683MIOPEN_EXPORT miopenStatus_t
1684miopenConvolutionBackwardWeightsCompileSolution(miopenHandle_t handle,
1685 const miopenTensorDescriptor_t dyDesc,
1686 const miopenTensorDescriptor_t xDesc,
1687 const miopenConvolutionDescriptor_t convDesc,
1688 const miopenTensorDescriptor_t dwDesc,
1689 const uint64_t solution_id);
1690
1709MIOPEN_EXPORT miopenStatus_t
1710miopenConvolutionBackwardWeightsImmediate(miopenHandle_t handle,
1711 const miopenTensorDescriptor_t dyDesc,
1712 const void* dy,
1713 const miopenTensorDescriptor_t xDesc,
1714 const void* x,
1715 const miopenConvolutionDescriptor_t convDesc,
1716 const miopenTensorDescriptor_t dwDesc,
1717 void* dw,
1718 void* workSpace,
1719 size_t workSpaceSize,
1720 const uint64_t solution_id);
1721
1748MIOPEN_EXPORT miopenStatus_t
1749miopenConvolutionForwardGetWorkSpaceSize(miopenHandle_t handle,
1750 const miopenTensorDescriptor_t wDesc,
1751 const miopenTensorDescriptor_t xDesc,
1752 const miopenConvolutionDescriptor_t convDesc,
1753 const miopenTensorDescriptor_t yDesc,
1754 size_t* workSpaceSize);
1755
1799MIOPEN_EXPORT miopenStatus_t
1800miopenFindConvolutionForwardAlgorithm(miopenHandle_t handle,
1801 const miopenTensorDescriptor_t xDesc,
1802 const void* x,
1803 const miopenTensorDescriptor_t wDesc,
1804 const void* w,
1805 const miopenConvolutionDescriptor_t convDesc,
1806 const miopenTensorDescriptor_t yDesc,
1807 void* y,
1808 const int requestAlgoCount,
1809 int* returnedAlgoCount,
1810 miopenConvAlgoPerf_t* perfResults,
1811 void* workSpace,
1812 size_t workSpaceSize,
1813 bool exhaustiveSearch);
1814
1849MIOPEN_EXPORT miopenStatus_t miopenConvolutionForward(miopenHandle_t handle,
1850 const void* alpha,
1851 const miopenTensorDescriptor_t xDesc,
1852 const void* x,
1853 const miopenTensorDescriptor_t wDesc,
1854 const void* w,
1855 const miopenConvolutionDescriptor_t convDesc,
1856 miopenConvFwdAlgorithm_t algo,
1857 const void* beta,
1858 const miopenTensorDescriptor_t yDesc,
1859 void* y,
1860 void* workSpace,
1861 size_t workSpaceSize);
1862
1878MIOPEN_EXPORT miopenStatus_t miopenConvolutionForwardBias(miopenHandle_t handle,
1879 const void* alpha,
1880 const miopenTensorDescriptor_t bDesc,
1881 const void* b,
1882 const void* beta,
1883 const miopenTensorDescriptor_t yDesc,
1884 void* y);
1885
1913MIOPEN_EXPORT miopenStatus_t
1914miopenConvolutionBackwardDataGetWorkSpaceSize(miopenHandle_t handle,
1915 const miopenTensorDescriptor_t dyDesc,
1916 const miopenTensorDescriptor_t wDesc,
1917 const miopenConvolutionDescriptor_t convDesc,
1918 const miopenTensorDescriptor_t dxDesc,
1919 size_t* workSpaceSize);
1920
1964MIOPEN_EXPORT miopenStatus_t
1965miopenFindConvolutionBackwardDataAlgorithm(miopenHandle_t handle,
1966 const miopenTensorDescriptor_t dyDesc,
1967 const void* dy,
1968 const miopenTensorDescriptor_t wDesc,
1969 const void* w,
1970 const miopenConvolutionDescriptor_t convDesc,
1971 const miopenTensorDescriptor_t dxDesc,
1972 void* dx,
1973 const int requestAlgoCount,
1974 int* returnedAlgoCount,
1975 miopenConvAlgoPerf_t* perfResults,
1976 void* workSpace,
1977 size_t workSpaceSize,
1978 bool exhaustiveSearch);
1979
2013MIOPEN_EXPORT miopenStatus_t
2014miopenConvolutionBackwardData(miopenHandle_t handle,
2015 const void* alpha,
2016 const miopenTensorDescriptor_t dyDesc,
2017 const void* dy,
2018 const miopenTensorDescriptor_t wDesc,
2019 const void* w,
2020 const miopenConvolutionDescriptor_t convDesc,
2021 miopenConvBwdDataAlgorithm_t algo,
2022 const void* beta,
2023 const miopenTensorDescriptor_t dxDesc,
2024 void* dx,
2025 void* workSpace,
2026 size_t workSpaceSize);
2027
2055MIOPEN_EXPORT miopenStatus_t
2056miopenConvolutionBackwardWeightsGetWorkSpaceSize(miopenHandle_t handle,
2057 const miopenTensorDescriptor_t dyDesc,
2058 const miopenTensorDescriptor_t xDesc,
2059 const miopenConvolutionDescriptor_t convDesc,
2060 const miopenTensorDescriptor_t dwDesc,
2061 size_t* workSpaceSize);
2062
2106MIOPEN_EXPORT miopenStatus_t
2107miopenFindConvolutionBackwardWeightsAlgorithm(miopenHandle_t handle,
2108 const miopenTensorDescriptor_t dyDesc,
2109 const void* dy,
2110 const miopenTensorDescriptor_t xDesc,
2111 const void* x,
2112 const miopenConvolutionDescriptor_t convDesc,
2113 const miopenTensorDescriptor_t dwDesc,
2114 void* dw,
2115 const int requestAlgoCount,
2116 int* returnedAlgoCount,
2117 miopenConvAlgoPerf_t* perfResults,
2118 void* workSpace,
2119 size_t workSpaceSize,
2120 bool exhaustiveSearch);
2121
2155MIOPEN_EXPORT miopenStatus_t
2156miopenConvolutionBackwardWeights(miopenHandle_t handle,
2157 const void* alpha,
2158 const miopenTensorDescriptor_t dyDesc,
2159 const void* dy,
2160 const miopenTensorDescriptor_t xDesc,
2161 const void* x,
2162 const miopenConvolutionDescriptor_t convDesc,
2163 miopenConvBwdWeightsAlgorithm_t algo,
2164 const void* beta,
2165 const miopenTensorDescriptor_t dwDesc,
2166 void* dw,
2167 void* workSpace,
2168 size_t workSpaceSize);
2169
2189MIOPEN_EXPORT miopenStatus_t miopenConvolutionBackwardBias(miopenHandle_t handle,
2190 const void* alpha,
2191 const miopenTensorDescriptor_t dyDesc,
2192 const void* dy,
2193 const void* beta,
2194 const miopenTensorDescriptor_t dbDesc,
2195 void* db);
2196
2198// CLOSEOUT CONVOLUTIONS DOXYGEN GROUP
2199
2200// Pooling APIs
2205
2211MIOPEN_EXPORT miopenStatus_t miopenCreatePoolingDescriptor(miopenPoolingDescriptor_t* poolDesc);
2212
2221MIOPEN_EXPORT miopenStatus_t miopenSetPoolingIndexType(miopenPoolingDescriptor_t poolDesc,
2222 miopenIndexType_t index_type);
2223
2231MIOPEN_EXPORT miopenStatus_t miopenGetPoolingIndexType(miopenPoolingDescriptor_t poolDesc,
2232 miopenIndexType_t* index_type);
2233
2241MIOPEN_EXPORT miopenStatus_t miopenSetPoolingWorkSpaceIndexMode(
2242 miopenPoolingDescriptor_t poolDesc, miopenPoolingWorkspaceIndexMode_t workspace_index);
2243
2250MIOPEN_EXPORT miopenStatus_t miopenGetPoolingWorkSpaceIndexMode(
2251 miopenPoolingDescriptor_t poolDesc, miopenPoolingWorkspaceIndexMode_t* workspace_index);
2252
2267MIOPEN_EXPORT miopenStatus_t miopenSet2dPoolingDescriptor(miopenPoolingDescriptor_t poolDesc,
2268 miopenPoolingMode_t mode,
2269 int windowHeight,
2270 int windowWidth,
2271 int pad_h,
2272 int pad_w,
2273 int stride_h,
2274 int stride_w);
2275
2290MIOPEN_EXPORT miopenStatus_t miopenGet2dPoolingDescriptor(const miopenPoolingDescriptor_t poolDesc,
2291 miopenPoolingMode_t* mode,
2292 int* windowHeight,
2293 int* windowWidth,
2294 int* pad_h,
2295 int* pad_w,
2296 int* stride_h,
2297 int* stride_w);
2298
2313MIOPEN_EXPORT miopenStatus_t
2314miopenGetPoolingForwardOutputDim(const miopenPoolingDescriptor_t poolDesc,
2315 const miopenTensorDescriptor_t tensorDesc,
2316 int* n,
2317 int* c,
2318 int* h,
2319 int* w);
2320
2336MIOPEN_EXPORT miopenStatus_t miopenSetNdPoolingDescriptor(miopenPoolingDescriptor_t poolDesc,
2337 const miopenPoolingMode_t mode,
2338 int nbDims,
2339 const int* windowDimA,
2340 const int* padA,
2341 const int* stridesA);
2342
2359MIOPEN_EXPORT miopenStatus_t miopenGetNdPoolingDescriptor(const miopenPoolingDescriptor_t poolDesc,
2360 int nbDimsRequested,
2361 miopenPoolingMode_t* mode,
2362 int* nbDims,
2363 int* windowDimA,
2364 int* padA,
2365 int* stridesA);
2366
2379MIOPEN_EXPORT miopenStatus_t
2380miopenGetPoolingNdForwardOutputDim(const miopenPoolingDescriptor_t poolDesc,
2381 const miopenTensorDescriptor_t tensorDesc,
2382 int dims,
2383 int* tensorDimArr);
2384
2397MIOPEN_EXPORT miopenStatus_t miopenPoolingGetWorkSpaceSize(const miopenTensorDescriptor_t yDesc,
2398 size_t* workSpaceSize);
2399
2412MIOPEN_EXPORT miopenStatus_t
2413miopenPoolingGetWorkSpaceSizeV2(const miopenPoolingDescriptor_t poolDesc,
2414 const miopenTensorDescriptor_t yDesc,
2415 size_t* workSpaceSize);
2416
2437MIOPEN_EXPORT miopenStatus_t miopenPoolingForward(miopenHandle_t handle,
2438 const miopenPoolingDescriptor_t poolDesc,
2439 const void* alpha,
2440 const miopenTensorDescriptor_t xDesc,
2441 const void* x,
2442 const void* beta,
2443 const miopenTensorDescriptor_t yDesc,
2444 void* y,
2445 bool do_backward,
2446 void* workSpace,
2447 size_t workSpaceSize);
2448
2469MIOPEN_EXPORT miopenStatus_t miopenPoolingBackward(miopenHandle_t handle,
2470 const miopenPoolingDescriptor_t poolDesc,
2471 const void* alpha,
2472 const miopenTensorDescriptor_t yDesc,
2473 const void* y,
2474 const miopenTensorDescriptor_t dyDesc,
2475 const void* dy,
2476 const miopenTensorDescriptor_t xDesc,
2477 const void* x,
2478 const void* beta,
2479 const miopenTensorDescriptor_t dxDesc,
2480 void* dx,
2481 void* workSpace);
2482
2488MIOPEN_EXPORT miopenStatus_t miopenDestroyPoolingDescriptor(miopenPoolingDescriptor_t poolDesc);
2489
2491// CLOSEOUT POOLING DOXYGEN GROUP
2492
2493// LRN APIs
2503MIOPEN_EXPORT miopenStatus_t miopenCreateLRNDescriptor(miopenLRNDescriptor_t* lrnDesc);
2504
2518MIOPEN_EXPORT miopenStatus_t miopenSetLRNDescriptor(const miopenLRNDescriptor_t lrnDesc,
2519 miopenLRNMode_t mode,
2520 unsigned int lrnN,
2521 double lrnAlpha,
2522 double lrnBeta,
2523 double lrnK);
2524
2537MIOPEN_EXPORT miopenStatus_t miopenGetLRNDescriptor(const miopenLRNDescriptor_t lrnDesc,
2538 miopenLRNMode_t* mode,
2539 unsigned int* lrnN,
2540 double* lrnAlpha,
2541 double* lrnBeta,
2542 double* lrnK);
2543
2553MIOPEN_EXPORT miopenStatus_t miopenLRNGetWorkSpaceSize(const miopenTensorDescriptor_t yDesc,
2554 size_t* workSpaceSize);
2555
2574MIOPEN_EXPORT miopenStatus_t miopenLRNForward(miopenHandle_t handle,
2575 const miopenLRNDescriptor_t lrnDesc,
2576 const void* alpha,
2577 const miopenTensorDescriptor_t xDesc,
2578 const void* x,
2579 const void* beta,
2580 const miopenTensorDescriptor_t yDesc,
2581 void* y,
2582 bool do_backward,
2583 void* workSpace);
2584
2602MIOPEN_EXPORT miopenStatus_t miopenLRNBackward(miopenHandle_t handle,
2603 const miopenLRNDescriptor_t lrnDesc,
2604 const void* alpha,
2605 const miopenTensorDescriptor_t yDesc,
2606 const void* y,
2607 const miopenTensorDescriptor_t dyDesc,
2608 const void* dy,
2609 const miopenTensorDescriptor_t xDesc,
2610 const void* x,
2611 const void* beta,
2612 const miopenTensorDescriptor_t dxDesc,
2613 void* dx,
2614 const void* workSpace);
2615
2621MIOPEN_EXPORT miopenStatus_t miopenDestroyLRNDescriptor(miopenLRNDescriptor_t lrnDesc);
2622
2624// CLOSEOUT LRN DOXYGEN GROUP
2625
2626#ifdef MIOPEN_BETA_API
2627// LayerNorm APIs
2652MIOPEN_EXPORT miopenStatus_t miopenLayerNormForward(miopenHandle_t handle,
2653 miopenNormMode_t mode,
2654 const miopenTensorDescriptor_t xDesc,
2655 const void* x,
2656 const miopenTensorDescriptor_t weightDesc,
2657 const void* weight,
2658 const miopenTensorDescriptor_t biasDesc,
2659 const void* bias,
2660 const float epsilon,
2661 const int32_t normalized_dim,
2662 const miopenTensorDescriptor_t yDesc,
2663 void* y,
2664 const miopenTensorDescriptor_t meanDesc,
2665 void* mean,
2666 const miopenTensorDescriptor_t rstdDesc,
2667 void* rstd);
2668
2686MIOPEN_EXPORT miopenStatus_t
2687miopenGetLayerNormBackwardWorkspaceSize(miopenHandle_t handle,
2688 miopenNormMode_t mode,
2689 const miopenTensorDescriptor_t dyDesc,
2690 const miopenTensorDescriptor_t xDesc,
2691 const miopenTensorDescriptor_t weightDesc,
2692 const miopenTensorDescriptor_t meanDesc,
2693 const miopenTensorDescriptor_t rstdDesc,
2694 const int32_t normalized_dim,
2695 const miopenTensorDescriptor_t dxDesc,
2696 const miopenTensorDescriptor_t dwDesc,
2697 const miopenTensorDescriptor_t dbDesc,
2698 size_t* sizeInBytes);
2699
2725MIOPEN_EXPORT miopenStatus_t miopenLayerNormBackward(miopenHandle_t handle,
2726 miopenNormMode_t mode,
2727 void* workspace,
2728 size_t workspaceSizeInBytes,
2729 const miopenTensorDescriptor_t dyDesc,
2730 const void* dy,
2731 const miopenTensorDescriptor_t xDesc,
2732 const void* x,
2733 const miopenTensorDescriptor_t weightDesc,
2734 const void* weight,
2735 const miopenTensorDescriptor_t meanDesc,
2736 const void* mean,
2737 const miopenTensorDescriptor_t rstdDesc,
2738 const void* rstd,
2739 const int32_t normalized_dim,
2740 const miopenTensorDescriptor_t dxDesc,
2741 void* dx,
2742 const miopenTensorDescriptor_t dwDesc,
2743 void* dw,
2744 const miopenTensorDescriptor_t dbDesc,
2745 void* db);
2746
2748// CLOSEOUT LAYERNORM DOXYGEN GROUP
2749#endif
2750
2751#ifdef MIOPEN_BETA_API
2752// Cat APIs
2768MIOPEN_EXPORT miopenStatus_t miopenCatForward(miopenHandle_t handle,
2769 const int32_t xCount,
2770 const miopenTensorDescriptor_t* xDescs,
2771 const void* const* xs,
2772 const miopenTensorDescriptor_t yDesc,
2773 void* y,
2774 const int32_t dim);
2775
2777// CLOSEOUT CAT DOXYGEN GROUP
2778#endif
2779
2780// Batch-Normalization APIs
2785
2802MIOPEN_EXPORT miopenStatus_t miopenDeriveBNTensorDescriptor(miopenTensorDescriptor_t derivedBnDesc,
2803 const miopenTensorDescriptor_t xDesc,
2804 miopenBatchNormMode_t bn_mode);
2805
2844MIOPEN_EXPORT miopenStatus_t
2845miopenBatchNormalizationForwardTraining(miopenHandle_t handle,
2846 miopenBatchNormMode_t bn_mode,
2847 void* alpha,
2848 void* beta,
2849 const miopenTensorDescriptor_t xDesc,
2850 const void* x,
2851 const miopenTensorDescriptor_t yDesc,
2852 void* y,
2853 const miopenTensorDescriptor_t bnScaleBiasMeanVarDesc,
2854 void* bnScale,
2855 void* bnBias,
2856 double expAvgFactor,
2857 void* resultRunningMean,
2858 void* resultRunningVariance,
2859 double epsilon,
2860 void* resultSaveMean,
2861 void* resultSaveInvVariance);
2902MIOPEN_EXPORT miopenStatus_t
2903miopenBatchNormalizationForwardTraining_V2(miopenHandle_t handle,
2904 miopenBatchNormMode_t bn_mode,
2905 void* alpha,
2906 void* beta,
2907 const miopenTensorDescriptor_t xDesc,
2908 const void* x,
2909 const miopenTensorDescriptor_t yDesc,
2910 void* y,
2911 const miopenTensorDescriptor_t scaleDesc,
2912 const miopenTensorDescriptor_t biasVarDesc,
2913 const miopenTensorDescriptor_t savedMeanDesc,
2914 const miopenTensorDescriptor_t savedVarDesc,
2915 void* bnScale,
2916 void* bnBias,
2917 double expAvgFactor,
2918 void* resultRunningMean,
2919 void* resultRunningVariance,
2920 double epsilon,
2921 void* resultSaveMean,
2922 void* resultSaveInvVariance);
2965MIOPEN_EXPORT miopenStatus_t
2966miopenBatchNormalizationForwardTraining_V3(miopenHandle_t handle,
2967 miopenBatchNormMode_t bn_mode,
2968 void* alpha,
2969 void* beta,
2970 const miopenTensorDescriptor_t xDesc,
2971 const void* x,
2972 const miopenTensorDescriptor_t yDesc,
2973 void* y,
2974 const miopenTensorDescriptor_t scaleDesc,
2975 const miopenTensorDescriptor_t biasVarDesc,
2976 const miopenTensorDescriptor_t savedMeanDesc,
2977 const miopenTensorDescriptor_t savedVarDesc,
2978 void* bnScale,
2979 void* bnBias,
2980 double expAvgFactor,
2981 const void* prevResultRunningMean,
2982 const void* prevResultRunningVariance,
2983 void* nextResultRunningMean,
2984 void* nextResultRunningVariance,
2985 double epsilon,
2986 void* resultSaveMean,
2987 void* resultSaveInvVariance);
3029MIOPEN_EXPORT miopenStatus_t
3030miopenBatchNormForwardTrainingActivation(miopenHandle_t handle,
3031 miopenBatchNormMode_t bn_mode,
3032 void* alpha,
3033 void* beta,
3034 const miopenTensorDescriptor_t xDesc,
3035 const void* x,
3036 const miopenTensorDescriptor_t yDesc,
3037 void* y,
3038 const miopenTensorDescriptor_t scaleDesc,
3039 const miopenTensorDescriptor_t biasVarDesc,
3040 const miopenTensorDescriptor_t savedMeanDesc,
3041 const miopenTensorDescriptor_t savedVarDesc,
3042 void* bnScale,
3043 void* bnBias,
3044 double expAvgFactor,
3045 void* resultRunningMean,
3046 void* resultRunningVariance,
3047 double epsilon,
3048 void* resultSaveMean,
3049 void* resultSaveInvVariance,
3050 const miopenActivationDescriptor_t activDesc);
3051
3095MIOPEN_EXPORT miopenStatus_t
3096miopenBatchNormForwardTrainingActivation_V2(miopenHandle_t handle,
3097 miopenBatchNormMode_t bn_mode,
3098 void* alpha,
3099 void* beta,
3100 const miopenTensorDescriptor_t xDesc,
3101 const void* x,
3102 const miopenTensorDescriptor_t yDesc,
3103 void* y,
3104 const miopenTensorDescriptor_t scaleDesc,
3105 const miopenTensorDescriptor_t biasVarDesc,
3106 const miopenTensorDescriptor_t savedMeanDesc,
3107 const miopenTensorDescriptor_t savedVarDesc,
3108 void* bnScale,
3109 void* bnBias,
3110 double expAvgFactor,
3111 const void* prevResultRunningMean,
3112 const void* prevResultRunningVariance,
3113 void* nextResultRunningMean,
3114 void* nextResultRunningVariance,
3115 double epsilon,
3116 void* resultSaveMean,
3117 void* resultSaveInvVariance,
3118 const miopenActivationDescriptor_t activDesc);
3119
3149MIOPEN_EXPORT miopenStatus_t
3150miopenBatchNormalizationForwardInference(miopenHandle_t handle,
3151 miopenBatchNormMode_t bn_mode,
3152 void* alpha,
3153 void* beta,
3154 const miopenTensorDescriptor_t xDesc,
3155 const void* x,
3156 const miopenTensorDescriptor_t yDesc,
3157 void* y,
3158 const miopenTensorDescriptor_t bnScaleBiasMeanVarDesc,
3159 void* bnScale,
3160 void* bnBias,
3161 void* estimatedMean,
3162 void* estimatedVariance,
3163 double epsilon);
3164
3196MIOPEN_EXPORT miopenStatus_t
3197miopenBatchNormalizationForwardInference_V2(miopenHandle_t handle,
3198 miopenBatchNormMode_t bn_mode,
3199 void* alpha,
3200 void* beta,
3201 const miopenTensorDescriptor_t xDesc,
3202 const void* x,
3203 const miopenTensorDescriptor_t yDesc,
3204 void* y,
3205 const miopenTensorDescriptor_t scaleDesc,
3206 const miopenTensorDescriptor_t biasDesc,
3207 const miopenTensorDescriptor_t estMeanDesc,
3208 const miopenTensorDescriptor_t estVarianceDesc,
3209 void* bnScale,
3210 void* bnBias,
3211 void* estimatedMean,
3212 void* estimatedVariance,
3213 double epsilon);
3214
3243MIOPEN_EXPORT miopenStatus_t miopenBatchNormalizationForwardInferenceInvVariance(
3244 miopenHandle_t handle,
3245 miopenBatchNormMode_t bn_mode,
3246 void* alpha,
3247 void* beta,
3248 const miopenTensorDescriptor_t xDesc,
3249 const void* x,
3250 const miopenTensorDescriptor_t yDesc,
3251 void* y,
3252 const miopenTensorDescriptor_t scaleDesc,
3253 const miopenTensorDescriptor_t biasDesc,
3254 const miopenTensorDescriptor_t estMeanDesc,
3255 const miopenTensorDescriptor_t estInvVarianceDesc,
3256 void* bnScale,
3257 void* bnBias,
3258 void* estimatedMean,
3259 void* estimatedInvVariance);
3260
3291MIOPEN_EXPORT miopenStatus_t miopenBatchNormForwardInferenceActivationInvVariance(
3292 miopenHandle_t handle,
3293 miopenBatchNormMode_t bn_mode,
3294 void* alpha,
3295 void* beta,
3296 const miopenTensorDescriptor_t xDesc,
3297 const void* x,
3298 const miopenTensorDescriptor_t yDesc,
3299 void* y,
3300 const miopenTensorDescriptor_t scaleDesc,
3301 const miopenTensorDescriptor_t biasDesc,
3302 const miopenTensorDescriptor_t estMeanDesc,
3303 const miopenTensorDescriptor_t estInvVarianceDesc,
3304 void* bnScale,
3305 void* bnBias,
3306 void* estimatedMean,
3307 void* estimatedInvVariance,
3308 const miopenActivationDescriptor_t activDesc);
3309
3342MIOPEN_EXPORT miopenStatus_t
3343miopenBatchNormForwardInferenceActivation(miopenHandle_t handle,
3344 miopenBatchNormMode_t bn_mode,
3345 void* alpha,
3346 void* beta,
3347 const miopenTensorDescriptor_t xDesc,
3348 const void* x,
3349 const miopenTensorDescriptor_t yDesc,
3350 void* y,
3351 const miopenTensorDescriptor_t scaleDesc,
3352 const miopenTensorDescriptor_t biasDesc,
3353 const miopenTensorDescriptor_t estMeanDesc,
3354 const miopenTensorDescriptor_t estVarianceDesc,
3355 void* bnScale,
3356 void* bnBias,
3357 void* estimatedMean,
3358 void* estimatedVariance,
3359 double epsilon,
3360 const miopenActivationDescriptor_t activDesc);
3361
3396MIOPEN_EXPORT miopenStatus_t
3397miopenBatchNormalizationBackward(miopenHandle_t handle,
3398 miopenBatchNormMode_t bn_mode,
3399 const void* alphaDataDiff,
3400 const void* betaDataDiff,
3401 const void* alphaParamDiff,
3402 const void* betaParamDiff,
3403 const miopenTensorDescriptor_t xDesc,
3404 const void* x,
3405 const miopenTensorDescriptor_t dyDesc,
3406 const void* dy,
3407 const miopenTensorDescriptor_t dxDesc,
3408 void* dx,
3409 const miopenTensorDescriptor_t bnScaleBiasDiffDesc,
3410 const void* bnScale,
3411 void* resultBnScaleDiff,
3412 void* resultBnBiasDiff,
3413 double epsilon,
3414 const void* savedMean,
3415 const void* savedInvVariance);
3416
3455MIOPEN_EXPORT miopenStatus_t
3456miopenBatchNormalizationBackward_V2(miopenHandle_t handle,
3457 miopenBatchNormMode_t bn_mode,
3458 const void* alphaDataDiff,
3459 const void* betaDataDiff,
3460 const void* alphaParamDiff,
3461 const void* betaParamDiff,
3462 const miopenTensorDescriptor_t xDesc,
3463 const void* x,
3464 const miopenTensorDescriptor_t dyDesc,
3465 const void* dy,
3466 const miopenTensorDescriptor_t dxDesc,
3467 void* dx,
3468 const miopenTensorDescriptor_t scaleDesc,
3469 const miopenTensorDescriptor_t biasDesc,
3470 const miopenTensorDescriptor_t savedMeanDesc,
3471 const miopenTensorDescriptor_t savedVarDesc,
3472 const void* bnScale,
3473 void* resultBnScaleDiff,
3474 void* resultBnBiasDiff,
3475 double epsilon,
3476 const void* savedMean,
3477 const void* savedInvVariance);
3478
3519MIOPEN_EXPORT miopenStatus_t
3520miopenBatchNormBackwardActivation(miopenHandle_t handle,
3521 miopenBatchNormMode_t bn_mode,
3522 const void* alphaDataDiff,
3523 const void* betaDataDiff,
3524 const void* alphaParamDiff,
3525 const void* betaParamDiff,
3526 const miopenTensorDescriptor_t xDesc,
3527 const void* x,
3528 const miopenTensorDescriptor_t dyDesc,
3529 const void* dy,
3530 const miopenTensorDescriptor_t dxDesc,
3531 void* dx,
3532 const miopenTensorDescriptor_t scaleDesc,
3533 const miopenTensorDescriptor_t biasDesc,
3534 const miopenTensorDescriptor_t savedMeanDesc,
3535 const miopenTensorDescriptor_t savedVarianceDesc,
3536 const void* bnScale,
3537 const void* bnBias,
3538 void* resultBnScaleDiff,
3539 void* resultBnBiasDiff,
3540 double epsilon,
3541 const void* savedMean,
3542 const void* savedInvVariance,
3543 const miopenActivationDescriptor_t activDesc);
3545// CLOSEOUT BATCHNORM DOXYGEN GROUP
3546
3547// Activation APIs
3557MIOPEN_EXPORT miopenStatus_t
3558miopenCreateActivationDescriptor(miopenActivationDescriptor_t* activDesc);
3559
3571MIOPEN_EXPORT miopenStatus_t
3572miopenSetActivationDescriptor(const miopenActivationDescriptor_t activDesc,
3573 miopenActivationMode_t mode,
3574 double activAlpha,
3575 double activBeta,
3576 double activGamma);
3577
3589MIOPEN_EXPORT miopenStatus_t
3590miopenGetActivationDescriptor(const miopenActivationDescriptor_t activDesc,
3591 miopenActivationMode_t* mode,
3592 double* activAlpha,
3593 double* activBeta,
3594 double* activGamma);
3595
3608MIOPEN_EXPORT miopenStatus_t miopenActivationForward(miopenHandle_t handle,
3609 const miopenActivationDescriptor_t activDesc,
3610 const void* alpha,
3611 const miopenTensorDescriptor_t xDesc,
3612 const void* x,
3613 const void* beta,
3614 const miopenTensorDescriptor_t yDesc,
3615 void* y);
3616
3633MIOPEN_EXPORT miopenStatus_t miopenActivationBackward(miopenHandle_t handle,
3634 const miopenActivationDescriptor_t activDesc,
3635 const void* alpha,
3636 const miopenTensorDescriptor_t yDesc,
3637 const void* y,
3638 const miopenTensorDescriptor_t dyDesc,
3639 const void* dy,
3640 const miopenTensorDescriptor_t xDesc,
3641 const void* x,
3642 const void* beta,
3643 const miopenTensorDescriptor_t dxDesc,
3644 void* dx);
3645
3651MIOPEN_EXPORT miopenStatus_t
3652miopenDestroyActivationDescriptor(miopenActivationDescriptor_t activDesc);
3653
3655// CLOSEOUT ACTIVATION DOXYGEN GROUP
3656
3657#ifdef MIOPEN_BETA_API
3662
3673MIOPEN_EXPORT miopenStatus_t miopenGLUForward(miopenHandle_t handle,
3674 const miopenTensorDescriptor_t inputDesc,
3675 const void* input,
3676 const miopenTensorDescriptor_t outputDesc,
3677 void* output,
3678 const uint32_t dim);
3679
3692MIOPEN_EXPORT miopenStatus_t miopenGLUBackward(miopenHandle_t handle,
3693 const miopenTensorDescriptor_t inputDesc,
3694 const void* input,
3695 const miopenTensorDescriptor_t outputGradDesc,
3696 const void* outputGrad,
3697 const miopenTensorDescriptor_t inputGradDesc,
3698 void* inputGrad,
3699 const uint32_t dim);
3700
3702// CLOSEOUT ACTIVATION DOXYGEN GROUP
3703#endif // MIOPEN_BETA_API
3704
3705// Softmax APIs
3723MIOPEN_EXPORT miopenStatus_t miopenSoftmaxForward(miopenHandle_t handle,
3724 const void* alpha,
3725 const miopenTensorDescriptor_t xDesc,
3726 const void* x,
3727 const void* beta,
3728 const miopenTensorDescriptor_t yDesc,
3729 void* y);
3730
3746MIOPEN_EXPORT miopenStatus_t miopenSoftmaxBackward(miopenHandle_t handle,
3747 const void* alpha,
3748 const miopenTensorDescriptor_t yDesc,
3749 const void* y,
3750 const miopenTensorDescriptor_t dyDesc,
3751 const void* dy,
3752 const void* beta,
3753 const miopenTensorDescriptor_t dxDesc,
3754 void* dx);
3755
3769MIOPEN_EXPORT miopenStatus_t miopenSoftmaxForward_V2(miopenHandle_t handle,
3770 const void* alpha,
3771 const miopenTensorDescriptor_t xDesc,
3772 const void* x,
3773 const void* beta,
3774 const miopenTensorDescriptor_t yDesc,
3775 void* y,
3776 miopenSoftmaxAlgorithm_t algorithm,
3777 miopenSoftmaxMode_t mode);
3778
3794MIOPEN_EXPORT miopenStatus_t miopenSoftmaxBackward_V2(miopenHandle_t handle,
3795 const void* alpha,
3796 const miopenTensorDescriptor_t yDesc,
3797 const void* y,
3798 const miopenTensorDescriptor_t dyDesc,
3799 const void* dy,
3800 const void* beta,
3801 const miopenTensorDescriptor_t dxDesc,
3802 void* dx,
3803 miopenSoftmaxAlgorithm_t algorithm,
3804 miopenSoftmaxMode_t mode);
3805
3807// CLOSEOUT SOFTMAX DOXYGEN GROUP
3808
3812MIOPEN_DECLARE_OBJECT(miopenFusionPlanDescriptor);
3813MIOPEN_DECLARE_OBJECT(miopenOperatorDescriptor);
3814MIOPEN_DECLARE_OBJECT(miopenOperatorArgs);
3815
3820
3829
3837MIOPEN_EXPORT miopenStatus_t miopenCreateFusionPlan(miopenFusionPlanDescriptor_t* fusePlanDesc,
3838 const miopenFusionDirection_t fuseDirection,
3839 const miopenTensorDescriptor_t inputDesc);
3840
3846MIOPEN_EXPORT miopenStatus_t miopenDestroyFusionPlan(miopenFusionPlanDescriptor_t fusePlanDesc);
3847
3854MIOPEN_EXPORT miopenStatus_t miopenCompileFusionPlan(miopenHandle_t handle,
3855 miopenFusionPlanDescriptor_t fusePlanDesc);
3856
3867MIOPEN_EXPORT miopenStatus_t miopenFusionPlanGetOp(miopenFusionPlanDescriptor_t fusePlanDesc,
3868 const int op_idx,
3869 miopenFusionOpDescriptor_t* op);
3870
3878MIOPEN_EXPORT miopenStatus_t
3879miopenFusionPlanGetWorkSpaceSize(miopenHandle_t handle,
3880 miopenFusionPlanDescriptor_t fusePlanDesc,
3881 size_t* workSpaceSize,
3882 miopenConvFwdAlgorithm_t algo);
3883
3901MIOPEN_EXPORT miopenStatus_t
3902miopenFusionPlanConvolutionGetAlgo(miopenFusionPlanDescriptor_t fusePlanDesc,
3903 const int requestAlgoCount,
3904 int* returnedAlgoCount,
3905 miopenConvFwdAlgorithm_t* returnedAlgos);
3906
3916MIOPEN_EXPORT miopenStatus_t miopenFusionPlanConvolutionSetAlgo(
3917 miopenFusionPlanDescriptor_t fusePlanDesc, miopenConvFwdAlgorithm_t algo);
3918
3927MIOPEN_EXPORT miopenStatus_t miopenCreateOpConvForward(miopenFusionPlanDescriptor_t fusePlanDesc,
3928 miopenFusionOpDescriptor_t* convOp,
3929 miopenConvolutionDescriptor_t convDesc,
3930 const miopenTensorDescriptor_t wDesc);
3931
3932//---
3933
3934// Activation forward create ops ---
3942MIOPEN_EXPORT miopenStatus_t
3943miopenCreateOpActivationForward(miopenFusionPlanDescriptor_t fusePlanDesc,
3944 miopenFusionOpDescriptor_t* activFwdOp,
3945 miopenActivationMode_t mode);
3946
3947// Activation backward create ops ---
3955MIOPEN_EXPORT miopenStatus_t
3956miopenCreateOpActivationBackward(miopenFusionPlanDescriptor_t fusePlanDesc,
3957 miopenFusionOpDescriptor_t* activBwdOp,
3958 miopenActivationMode_t mode);
3959
3960// Bias create ops ---
3968MIOPEN_EXPORT miopenStatus_t miopenCreateOpBiasForward(miopenFusionPlanDescriptor_t fusePlanDesc,
3969 miopenFusionOpDescriptor_t* biasOp,
3970 const miopenTensorDescriptor_t bDesc);
3971
3972// Batch normalization create ops ---
3981MIOPEN_EXPORT miopenStatus_t
3982miopenCreateOpBatchNormInference(miopenFusionPlanDescriptor_t fusePlanDesc,
3983 miopenFusionOpDescriptor_t* bnOp,
3984 const miopenBatchNormMode_t bn_mode,
3985 const miopenTensorDescriptor_t bnScaleBiasMeanVarDesc);
3986
3996MIOPEN_EXPORT miopenStatus_t
3997miopenCreateOpBatchNormForward(miopenFusionPlanDescriptor_t fusePlanDesc,
3998 miopenFusionOpDescriptor_t* bnFwdOp,
3999 const miopenBatchNormMode_t bn_mode,
4000 bool runningMeanVariance);
4001
4009MIOPEN_EXPORT miopenStatus_t
4010miopenCreateOpBatchNormBackward(miopenFusionPlanDescriptor_t fusePlanDesc,
4011 miopenFusionOpDescriptor_t* bnBwdOp,
4012 const miopenBatchNormMode_t bn_mode);
4013
4014//---
4020MIOPEN_EXPORT miopenStatus_t miopenCreateOperatorArgs(miopenOperatorArgs_t* args);
4021
4027MIOPEN_EXPORT miopenStatus_t miopenDestroyOperatorArgs(miopenOperatorArgs_t args);
4028
4029// Convolution set arguments ---
4039MIOPEN_EXPORT miopenStatus_t miopenSetOpArgsConvForward(miopenOperatorArgs_t args,
4040 const miopenFusionOpDescriptor_t convOp,
4041 const void* alpha,
4042 const void* beta,
4043 const void* w);
4044// Activation set arguments ---
4056MIOPEN_EXPORT miopenStatus_t
4057miopenSetOpArgsActivForward(miopenOperatorArgs_t args,
4058 const miopenFusionOpDescriptor_t activFwdOp,
4059 const void* alpha,
4060 const void* beta,
4061 double activAlpha,
4062 double activBeta,
4063 double activGamma);
4064
4078MIOPEN_EXPORT miopenStatus_t
4079miopenSetOpArgsActivBackward(miopenOperatorArgs_t args,
4080 const miopenFusionOpDescriptor_t activBwdOp,
4081 const void* alpha,
4082 const void* beta,
4083 const void* y,
4084 const void* reserved,
4085 double activAlpha,
4086 double activBeta,
4087 double activGamma);
4088
4089// Batch Normalization set arguments ---
4103MIOPEN_EXPORT miopenStatus_t
4104miopenSetOpArgsBatchNormInference(miopenOperatorArgs_t args,
4105 const miopenFusionOpDescriptor_t bnOp,
4106 const void* alpha,
4107 const void* beta,
4108 const void* bnScale,
4109 const void* bnBias,
4110 const void* estimatedMean,
4111 const void* estimatedVariance,
4112 double epsilon);
4113
4130MIOPEN_EXPORT miopenStatus_t miopenSetOpArgsBatchNormForward(miopenOperatorArgs_t args,
4131 const miopenFusionOpDescriptor_t bnOp,
4132 const void* alpha,
4133 const void* beta,
4134 const void* bnScale,
4135 const void* bnBias,
4136 void* savedMean,
4137 void* savedInvVariance,
4138 void* runningMean,
4139 void* runningVariance,
4140 double expAvgFactor,
4141 double epsilon);
4142
4158MIOPEN_EXPORT miopenStatus_t miopenSetOpArgsBatchNormBackward(miopenOperatorArgs_t args,
4159 const miopenFusionOpDescriptor_t bnOp,
4160 const void* alpha,
4161 const void* beta,
4162 const void* x,
4163 const void* bnScale,
4164 const void* bnBias,
4165 void* resultBnScaleDiff,
4166 void* resultBnBiasDiff,
4167 const void* savedMean,
4168 const void* savedInvVariance);
4169
4170// Bias forward set arguments ---
4180MIOPEN_EXPORT miopenStatus_t miopenSetOpArgsBiasForward(miopenOperatorArgs_t args,
4181 const miopenFusionOpDescriptor_t biasOp,
4182 const void* alpha,
4183 const void* beta,
4184 const void* bias);
4185
4200MIOPEN_EXPORT miopenStatus_t
4201miopenExecuteFusionPlan(const miopenHandle_t handle,
4202 const miopenFusionPlanDescriptor_t fusePlanDesc,
4203 const miopenTensorDescriptor_t inputDesc,
4204 const void* input,
4205 const miopenTensorDescriptor_t outputDesc,
4206 void* output,
4207 miopenOperatorArgs_t args);
4208
4223MIOPEN_EXPORT miopenStatus_t
4224miopenExecuteFusionPlan_v2(const miopenHandle_t handle,
4225 const miopenFusionPlanDescriptor_t fusePlanDesc,
4226 const miopenTensorDescriptor_t inputDesc,
4227 const void* input,
4228 const miopenTensorDescriptor_t outputDesc,
4229 void* output,
4230 miopenOperatorArgs_t args,
4231 void* workspace,
4232 size_t workspaceSize);
4233
4256
4257MIOPEN_EXPORT miopenStatus_t
4258miopenConvolutionBiasActivationForward(miopenHandle_t handle,
4259 const void* alpha1,
4260 const miopenTensorDescriptor_t xDesc,
4261 const void* x,
4262 const miopenTensorDescriptor_t wDesc,
4263 const void* w,
4264 const miopenConvolutionDescriptor_t convDesc,
4265 miopenConvFwdAlgorithm_t algo,
4266 void* workspace,
4267 size_t workspaceSizeInBytes,
4268 const void* alpha2,
4269 const miopenTensorDescriptor_t zDesc,
4270 const void* z,
4271 const miopenTensorDescriptor_t biasDesc,
4272 const void* bias,
4273 const miopenActivationDescriptor_t activationDesc,
4274 const miopenTensorDescriptor_t yDesc,
4275 void* y);
4277// CLOSEOUT FUSION DOXYGEN GROUP
4278
4283
4294
4303
4316
4325
4334
4342
4351
4360
4371
4378MIOPEN_EXPORT miopenStatus_t miopenCreateRNNDescriptor(miopenRNNDescriptor_t* rnnDesc);
4379
4392MIOPEN_EXPORT miopenStatus_t miopenGetRNNDescriptor(miopenRNNDescriptor_t rnnDesc,
4393 miopenRNNMode_t* rnnMode,
4394 miopenRNNAlgo_t* algoMode,
4395 miopenRNNInputMode_t* inputMode,
4396 miopenRNNDirectionMode_t* dirMode,
4397 miopenRNNBiasMode_t* biasMode,
4398 int* hiddenSize,
4399 int* layer);
4400
4417MIOPEN_EXPORT miopenStatus_t miopenGetRNNDescriptor_V2(miopenRNNDescriptor_t rnnDesc,
4418 int* hiddenSize,
4419 int* layer,
4420 miopenDropoutDescriptor_t* dropoutDesc,
4421 miopenRNNInputMode_t* inputMode,
4422 miopenRNNDirectionMode_t* dirMode,
4423 miopenRNNMode_t* rnnMode,
4424 miopenRNNBiasMode_t* biasMode,
4425 miopenRNNAlgo_t* algoMode,
4426 miopenDataType_t* dataType);
4427
4433MIOPEN_EXPORT miopenStatus_t miopenDestroyRNNDescriptor(miopenRNNDescriptor_t rnnDesc);
4434
4450MIOPEN_EXPORT miopenStatus_t miopenSetRNNDescriptor(miopenRNNDescriptor_t rnnDesc,
4451 const int hsize,
4452 const int nlayers,
4453 miopenRNNInputMode_t inMode,
4454 miopenRNNDirectionMode_t direction,
4455 miopenRNNMode_t rnnMode,
4456 miopenRNNBiasMode_t biasMode,
4457 miopenRNNAlgo_t algo,
4458 miopenDataType_t dataType);
4459
4478MIOPEN_EXPORT miopenStatus_t miopenSetRNNDescriptor_V2(miopenRNNDescriptor_t rnnDesc,
4479 const int hsize,
4480 const int nlayers,
4481 miopenDropoutDescriptor_t dropoutDesc,
4482 miopenRNNInputMode_t inMode,
4483 miopenRNNDirectionMode_t direction,
4484 miopenRNNMode_t rnnMode,
4485 miopenRNNBiasMode_t biasMode,
4486 miopenRNNAlgo_t algo,
4487 miopenDataType_t dataType);
4488
4503MIOPEN_EXPORT miopenStatus_t
4504miopenSetRNNDataSeqTensorDescriptor(miopenSeqTensorDescriptor_t seqTensorDesc,
4505 miopenDataType_t dataType,
4506 miopenRNNBaseLayout_t layout,
4507 int maxSequenceLen,
4508 int batchSize,
4509 int vectorSize,
4510 const int* sequenceLenArray,
4511 void* paddingMarker);
4512
4530
4531MIOPEN_EXPORT miopenStatus_t
4532miopenGetRNNDataSeqTensorDescriptor(miopenSeqTensorDescriptor_t seqTensorDesc,
4533 miopenDataType_t* dataType,
4534 miopenRNNBaseLayout_t* layout,
4535 int* maxSequenceLen,
4536 int* batchSize,
4537 int* vectorSize,
4538 int sequenceLenArrayLimit,
4539 int* sequenceLenArray,
4540 void* paddingMarker);
4541
4558MIOPEN_EXPORT miopenStatus_t miopenGetRNNWorkspaceSize(miopenHandle_t handle,
4559 const miopenRNNDescriptor_t rnnDesc,
4560 const int sequenceLen,
4561 const miopenTensorDescriptor_t* xDesc,
4562 size_t* numBytes);
4563
4580MIOPEN_EXPORT miopenStatus_t miopenGetRNNTrainingReserveSize(miopenHandle_t handle,
4581 miopenRNNDescriptor_t rnnDesc,
4582 const int sequenceLen,
4583 const miopenTensorDescriptor_t* xDesc,
4584 size_t* numBytes);
4585
4602MIOPEN_EXPORT miopenStatus_t miopenGetRNNTempSpaceSizes(miopenHandle_t handle,
4603 miopenRNNDescriptor_t rnnDesc,
4604 miopenSeqTensorDescriptor_t xDesc,
4605 miopenRNNFWDMode_t fwdMode,
4606 size_t* workSpaceSize,
4607 size_t* reserveSpaceSize);
4608
4621MIOPEN_EXPORT miopenStatus_t miopenGetRNNParamsSize(miopenHandle_t handle,
4622 miopenRNNDescriptor_t rnnDesc,
4623 miopenTensorDescriptor_t xDesc,
4624 size_t* numBytes,
4625 miopenDataType_t dtype);
4626
4639MIOPEN_EXPORT miopenStatus_t miopenGetRNNParamsDescriptor(miopenHandle_t handle,
4640 miopenRNNDescriptor_t rnnDesc,
4641 miopenTensorDescriptor_t xDesc,
4642 miopenTensorDescriptor_t wDesc,
4643 miopenDataType_t dtype);
4644
4662MIOPEN_EXPORT miopenStatus_t miopenGetRNNInputTensorSize(miopenHandle_t handle,
4663 miopenRNNDescriptor_t rnnDesc,
4664 const int seqLen,
4665 miopenTensorDescriptor_t* xDesc,
4666 size_t* numBytes);
4667
4680MIOPEN_EXPORT miopenStatus_t miopenGetRNNHiddenTensorSize(miopenHandle_t handle,
4681 miopenRNNDescriptor_t rnnDesc,
4682 const int seqLen,
4683 miopenTensorDescriptor_t* xDesc,
4684 size_t* numBytes);
4685
4726MIOPEN_EXPORT miopenStatus_t miopenGetRNNLayerParamSize(miopenHandle_t handle,
4727 miopenRNNDescriptor_t rnnDesc,
4728 const int layer,
4729 miopenTensorDescriptor_t xDesc,
4730 const int paramID,
4731 size_t* numBytes);
4732
4770MIOPEN_EXPORT miopenStatus_t miopenGetRNNLayerBiasSize(miopenHandle_t handle,
4771 miopenRNNDescriptor_t rnnDesc,
4772 const int layer,
4773 const int biasID,
4774 size_t* numBytes);
4775
4834MIOPEN_EXPORT miopenStatus_t miopenGetRNNLayerParam(miopenHandle_t handle,
4835 miopenRNNDescriptor_t rnnDesc,
4836 const int layer,
4837 miopenTensorDescriptor_t xDesc,
4838 miopenTensorDescriptor_t wDesc,
4839 const void* w,
4840 const int paramID,
4841 miopenTensorDescriptor_t paramDesc,
4842 void* layerParam);
4843
4901MIOPEN_EXPORT miopenStatus_t miopenGetRNNLayerBias(miopenHandle_t handle,
4902 miopenRNNDescriptor_t rnnDesc,
4903 const int layer,
4904 miopenTensorDescriptor_t xDesc,
4905 miopenTensorDescriptor_t wDesc,
4906 const void* w,
4907 const int biasID,
4908 miopenTensorDescriptor_t biasDesc,
4909 void* layerBias);
4910
4965MIOPEN_EXPORT miopenStatus_t miopenGetRNNLayerParamOffset(miopenRNNDescriptor_t rnnDesc,
4966 const int layer,
4967 miopenTensorDescriptor_t xDesc,
4968 const int paramID,
4969 miopenTensorDescriptor_t paramDesc,
4970 size_t* layerParamOffset);
4971
5022MIOPEN_EXPORT miopenStatus_t miopenGetRNNLayerBiasOffset(miopenRNNDescriptor_t rnnDesc,
5023 const int layer,
5024 miopenTensorDescriptor_t xDesc,
5025 const int biasID,
5026 miopenTensorDescriptor_t biasDesc,
5027 size_t* layerBiasOffset);
5028
5081MIOPEN_EXPORT miopenStatus_t miopenSetRNNLayerParam(miopenHandle_t handle,
5082 miopenRNNDescriptor_t rnnDesc,
5083 const int layer,
5084 miopenTensorDescriptor_t xDesc,
5085 miopenTensorDescriptor_t wDesc,
5086 void* w,
5087 const int paramID,
5088 miopenTensorDescriptor_t paramDesc,
5089 const void* layerParam);
5090
5141MIOPEN_EXPORT miopenStatus_t miopenSetRNNLayerBias(miopenHandle_t handle,
5142 miopenRNNDescriptor_t rnnDesc,
5143 const int layer,
5144 miopenTensorDescriptor_t xDesc,
5145 miopenTensorDescriptor_t wDesc,
5146 void* w,
5147 const int biasID,
5148 miopenTensorDescriptor_t biasDesc,
5149 const void* layerBias);
5150
5162MIOPEN_EXPORT miopenStatus_t miopenSetRNNPaddingMode(miopenRNNDescriptor_t rnnDesc,
5163 miopenRNNPaddingMode_t paddingMode);
5164
5171
5172MIOPEN_EXPORT miopenStatus_t miopenGetRNNPaddingMode(miopenRNNDescriptor_t rnnDesc,
5173 miopenRNNPaddingMode_t* paddingMode);
5174
5225MIOPEN_EXPORT miopenStatus_t miopenRNNForward(miopenHandle_t handle,
5226 const miopenRNNDescriptor_t rnnDesc,
5227 miopenRNNFWDMode_t fwdMode,
5228 const miopenSeqTensorDescriptor_t xDesc,
5229 const void* x,
5230 const miopenTensorDescriptor_t hDesc,
5231 const void* hx,
5232 void* hy,
5233 const miopenTensorDescriptor_t cDesc,
5234 const void* cx,
5235 void* cy,
5236 const miopenSeqTensorDescriptor_t yDesc,
5237 void* y,
5238 const void* w,
5239 size_t weightSpaceSize,
5240 void* workSpace,
5241 size_t workSpaceNumBytes,
5242 void* reserveSpace,
5243 size_t reserveSpaceNumBytes);
5244
5294MIOPEN_EXPORT miopenStatus_t miopenRNNBackwardSeqData(miopenHandle_t handle,
5295 const miopenRNNDescriptor_t rnnDesc,
5296 const miopenSeqTensorDescriptor_t yDesc,
5297 const void* y,
5298 const void* dy,
5299 const miopenTensorDescriptor_t hDesc,
5300 const void* hx,
5301 const void* dhy,
5302 void* dhx,
5303 const miopenTensorDescriptor_t cDesc,
5304 const void* cx,
5305 const void* dcy,
5306 void* dcx,
5307 const miopenSeqTensorDescriptor_t xDesc,
5308 void* dx,
5309 const void* w,
5310 size_t weightSpaceSize,
5311 void* workSpace,
5312 size_t workSpaceNumBytes,
5313 void* reserveSpace,
5314 size_t reserveSpaceNumBytes);
5315
5349MIOPEN_EXPORT miopenStatus_t
5350miopenRNNBackwardWeightsSeqTensor(miopenHandle_t handle,
5351 const miopenRNNDescriptor_t rnnDesc,
5352 const miopenSeqTensorDescriptor_t xDesc,
5353 const void* x,
5354 const miopenTensorDescriptor_t hDesc,
5355 const void* hx,
5356 const miopenSeqTensorDescriptor_t yDesc,
5357 const void* y,
5358 void* dw,
5359 size_t weightSpaceSize,
5360 void* workSpace,
5361 size_t workSpaceNumBytes,
5362 const void* reserveSpace,
5363 size_t reserveSpaceNumBytes);
5364
5422MIOPEN_EXPORT miopenStatus_t miopenRNNForwardTraining(miopenHandle_t handle,
5423 const miopenRNNDescriptor_t rnnDesc,
5424 const int sequenceLen,
5425 const miopenTensorDescriptor_t* xDesc,
5426 const void* x,
5427 const miopenTensorDescriptor_t hxDesc,
5428 const void* hx,
5429 const miopenTensorDescriptor_t cxDesc,
5430 const void* cx,
5431 const miopenTensorDescriptor_t wDesc,
5432 const void* w,
5433 const miopenTensorDescriptor_t* yDesc,
5434 void* y,
5435 const miopenTensorDescriptor_t hyDesc,
5436 void* hy,
5437 const miopenTensorDescriptor_t cyDesc,
5438 void* cy,
5439 void* workSpace,
5440 size_t workSpaceNumBytes,
5441 void* reserveSpace,
5442 size_t reserveSpaceNumBytes);
5443
5516MIOPEN_EXPORT miopenStatus_t miopenRNNBackwardData(miopenHandle_t handle,
5517 const miopenRNNDescriptor_t rnnDesc,
5518 const int sequenceLen,
5519 const miopenTensorDescriptor_t* yDesc,
5520 const void* y,
5521 const miopenTensorDescriptor_t* dyDesc,
5522 const void* dy,
5523 const miopenTensorDescriptor_t dhyDesc,
5524 const void* dhy,
5525 const miopenTensorDescriptor_t dcyDesc,
5526 const void* dcy,
5527 const miopenTensorDescriptor_t wDesc,
5528 const void* w,
5529 const miopenTensorDescriptor_t hxDesc,
5530 const void* hx,
5531 const miopenTensorDescriptor_t cxDesc,
5532 const void* cx,
5533 const miopenTensorDescriptor_t* dxDesc,
5534 void* dx,
5535 const miopenTensorDescriptor_t dhxDesc,
5536 void* dhx,
5537 const miopenTensorDescriptor_t dcxDesc,
5538 void* dcx,
5539 void* workSpace,
5540 size_t workSpaceNumBytes,
5541 void* reserveSpace,
5542 size_t reserveSpaceNumBytes);
5543
5580MIOPEN_EXPORT miopenStatus_t miopenRNNBackwardWeights(miopenHandle_t handle,
5581 const miopenRNNDescriptor_t rnnDesc,
5582 const int sequenceLen,
5583 const miopenTensorDescriptor_t* xDesc,
5584 const void* x,
5585 const miopenTensorDescriptor_t hxDesc,
5586 const void* hx,
5587 const miopenTensorDescriptor_t* yDesc,
5588 const void* y,
5589 const miopenTensorDescriptor_t dwDesc,
5590 void* dw,
5591 void* workSpace,
5592 size_t workSpaceNumBytes,
5593 const void* reserveSpace,
5594 size_t reserveSpaceNumBytes);
5595
5651MIOPEN_EXPORT miopenStatus_t miopenRNNForwardInference(miopenHandle_t handle,
5652 miopenRNNDescriptor_t rnnDesc,
5653 const int sequenceLen,
5654 const miopenTensorDescriptor_t* xDesc,
5655 const void* x,
5656 const miopenTensorDescriptor_t hxDesc,
5657 const void* hx,
5658 const miopenTensorDescriptor_t cxDesc,
5659 const void* cx,
5660 const miopenTensorDescriptor_t wDesc,
5661 const void* w,
5662 const miopenTensorDescriptor_t* yDesc,
5663 void* y,
5664 const miopenTensorDescriptor_t hyDesc,
5665 void* hy,
5666 const miopenTensorDescriptor_t cyDesc,
5667 void* cy,
5668 void* workSpace,
5669 size_t workSpaceNumBytes);
5670
5672// CLOSEOUT RNN DOXYGEN GROUP
5673
5678
5686
5693MIOPEN_EXPORT miopenStatus_t miopenCreateCTCLossDescriptor(miopenCTCLossDescriptor_t* ctcLossDesc);
5694
5704MIOPEN_EXPORT miopenStatus_t miopenGetCTCLossDescriptor(miopenCTCLossDescriptor_t ctcLossDesc,
5705 miopenDataType_t* dataType,
5706 int* blank_label_id,
5707 bool* apply_softmax_layer);
5708
5714MIOPEN_EXPORT miopenStatus_t miopenDestroyCTCLossDescriptor(miopenCTCLossDescriptor_t ctcLossDesc);
5715
5725MIOPEN_EXPORT miopenStatus_t miopenSetCTCLossDescriptor(miopenCTCLossDescriptor_t ctcLossDesc,
5726 miopenDataType_t dataType,
5727 const int blank_label_id,
5728 bool apply_softmax_layer);
5729
5746MIOPEN_EXPORT miopenStatus_t
5747miopenGetCTCLossWorkspaceSize(miopenHandle_t handle,
5748 const miopenTensorDescriptor_t probsDesc,
5749 const miopenTensorDescriptor_t gradientsDesc,
5750 const int* labels,
5751 const int* labelLengths,
5752 const int* inputLengths,
5753 miopenCTCLossAlgo_t algo,
5754 const miopenCTCLossDescriptor_t ctcLossDesc,
5755 size_t* workSpaceSize);
5756
5776MIOPEN_EXPORT miopenStatus_t miopenCTCLoss(miopenHandle_t handle,
5777 const miopenTensorDescriptor_t probsDesc,
5778 const void* probs,
5779 const int* labels,
5780 const int* labelLengths,
5781 const int* inputLengths,
5782 void* losses,
5783 const miopenTensorDescriptor_t gradientsDesc,
5784 void* gradients,
5785 miopenCTCLossAlgo_t algo,
5786 const miopenCTCLossDescriptor_t ctcLossDesc,
5787 void* workSpace,
5788 size_t workSpaceSize);
5789
5791// CLOSEOUT LossFunction DOXYGEN GROUP
5792
5793// Dropout APIs
5798
5802typedef enum
5803{
5804 MIOPEN_RNG_PSEUDO_XORWOW = 0,
5805} miopenRNGType_t;
5806
5812MIOPEN_EXPORT miopenStatus_t miopenCreateDropoutDescriptor(miopenDropoutDescriptor_t* dropoutDesc);
5813
5819MIOPEN_EXPORT miopenStatus_t miopenDestroyDropoutDescriptor(miopenDropoutDescriptor_t dropoutDesc);
5820
5829MIOPEN_EXPORT miopenStatus_t miopenDropoutGetReserveSpaceSize(const miopenTensorDescriptor_t xDesc,
5830 size_t* reserveSpaceSizeInBytes);
5831
5840MIOPEN_EXPORT miopenStatus_t miopenDropoutGetStatesSize(miopenHandle_t handle,
5841 size_t* stateSizeInBytes);
5842
5859MIOPEN_EXPORT miopenStatus_t miopenGetDropoutDescriptor(miopenDropoutDescriptor_t dropoutDesc,
5860 miopenHandle_t handle,
5861 float* dropout,
5862 void** states,
5863 unsigned long long* seed,
5864 bool* use_mask,
5865 bool* state_evo,
5866 miopenRNGType_t* rng_mode);
5867
5890MIOPEN_EXPORT miopenStatus_t miopenRestoreDropoutDescriptor(miopenDropoutDescriptor_t dropoutDesc,
5891 miopenHandle_t handle,
5892 float dropout,
5893 void* states,
5894 size_t stateSizeInBytes,
5895 unsigned long long seed,
5896 bool use_mask,
5897 bool state_evo,
5898 miopenRNGType_t rng_mode);
5899
5919MIOPEN_EXPORT miopenStatus_t miopenSetDropoutDescriptor(miopenDropoutDescriptor_t dropoutDesc,
5920 miopenHandle_t handle,
5921 float dropout,
5922 void* states,
5923 size_t stateSizeInBytes,
5924 unsigned long long seed,
5925 bool use_mask,
5926 bool state_evo,
5927 miopenRNGType_t rng_mode);
5928
5946MIOPEN_EXPORT miopenStatus_t miopenDropoutForward(miopenHandle_t handle,
5947 const miopenDropoutDescriptor_t dropoutDesc,
5948 const miopenTensorDescriptor_t noise_shape,
5949 const miopenTensorDescriptor_t xDesc,
5950 const void* x,
5951 const miopenTensorDescriptor_t yDesc,
5952 void* y,
5953 void* reserveSpace,
5954 size_t reserveSpaceSizeInBytes);
5955
5973MIOPEN_EXPORT miopenStatus_t miopenDropoutBackward(miopenHandle_t handle,
5974 const miopenDropoutDescriptor_t dropoutDesc,
5975 const miopenTensorDescriptor_t noise_shape,
5976 const miopenTensorDescriptor_t dyDesc,
5977 const void* dy,
5978 const miopenTensorDescriptor_t dxDesc,
5979 void* dx,
5980 void* reserveSpace,
5981 size_t reserveSpaceSizeInBytes);
5982
5984// CLOSEOUT DROPOUT DOXYGEN GROUP
5985
5986// TensorReduce APIs
5991
5997MIOPEN_EXPORT miopenStatus_t
5998miopenCreateReduceTensorDescriptor(miopenReduceTensorDescriptor_t* reduceTensorDesc);
5999
6005MIOPEN_EXPORT miopenStatus_t
6006miopenDestroyReduceTensorDescriptor(miopenReduceTensorDescriptor_t reduceTensorDesc);
6007
6020MIOPEN_EXPORT miopenStatus_t
6021miopenSetReduceTensorDescriptor(miopenReduceTensorDescriptor_t reduceTensorDesc,
6022 miopenReduceTensorOp_t reduceTensorOp,
6023 miopenDataType_t reduceTensorCompType,
6024 miopenNanPropagation_t reduceTensorNanOpt,
6025 miopenReduceTensorIndices_t reduceTensorIndices,
6026 miopenIndicesType_t reduceTensorIndicesType);
6027
6043MIOPEN_EXPORT miopenStatus_t
6044miopenGetReduceTensorDescriptor(const miopenReduceTensorDescriptor_t reduceTensorDesc,
6045 miopenReduceTensorOp_t* reduceTensorOp,
6046 miopenDataType_t* reduceTensorCompType,
6047 miopenNanPropagation_t* reduceTensorNanOpt,
6048 miopenReduceTensorIndices_t* reduceTensorIndices,
6049 miopenIndicesType_t* reduceTensorIndicesType);
6050
6060MIOPEN_EXPORT miopenStatus_t
6061miopenGetReductionIndicesSize(miopenHandle_t handle,
6062 const miopenReduceTensorDescriptor_t reduceTensorDesc,
6063 const miopenTensorDescriptor_t aDesc,
6064 const miopenTensorDescriptor_t cDesc,
6065 size_t* sizeInBytes);
6066
6076MIOPEN_EXPORT miopenStatus_t
6077miopenGetReductionWorkspaceSize(miopenHandle_t handle,
6078 const miopenReduceTensorDescriptor_t reduceTensorDesc,
6079 const miopenTensorDescriptor_t aDesc,
6080 const miopenTensorDescriptor_t cDesc,
6081 size_t* sizeInBytes);
6082
6106MIOPEN_EXPORT miopenStatus_t
6107miopenReduceTensor(miopenHandle_t handle,
6108 const miopenReduceTensorDescriptor_t reduceTensorDesc,
6109 void* indices,
6110 size_t indicesSizeInBytes,
6111 void* workspace,
6112 size_t workspaceSizeInBytes,
6113 const void* alpha,
6114 const miopenTensorDescriptor_t aDesc,
6115 const void* A,
6116 const void* beta,
6117 const miopenTensorDescriptor_t cDesc,
6118 void* C);
6119
6121// CLOSEOUT TensorReduce DOXYGEN GROUP
6122
6123// Find 2.0 API
6128
6134MIOPEN_DECLARE_OBJECT(miopenProblem);
6135
6148
6152typedef enum
6153{
6154 miopenTensorArgumentIdInvalid = 0,
6155 miopenTensorConvolutionX = 1,
6156 miopenTensorConvolutionW = 2,
6157 miopenTensorConvolutionY = 3,
6158
6159 miopenTensorMhaK = 4,
6160 miopenTensorMhaQ = 5,
6161 miopenTensorMhaV = 6,
6162 miopenTensorMhaDescaleK = 7,
6163 miopenTensorMhaDescaleQ = 8,
6164 miopenTensorMhaDescaleV = 9,
6165 miopenTensorMhaDescaleS = 10,
6166 miopenTensorMhaScaleS = 11,
6167 miopenTensorMhaScaleO = 12,
6168 miopenTensorMhaDropoutProbability = 13,
6169 miopenTensorMhaDropoutSeed = 14,
6170 miopenTensorMhaDropoutOffset = 15,
6171 miopenTensorMhaO = 16,
6172 miopenTensorMhaAmaxO = 17,
6173 miopenTensorMhaAmaxS = 18,
6174 miopenTensorMhaM = 19,
6175 miopenTensorMhaZInv = 20,
6176 miopenTensorMhaDO = 21,
6177 miopenTensorMhaDescaleO = 22,
6178 miopenTensorMhaDescaleDO = 23,
6179 miopenTensorMhaDescaleDS = 24,
6180 miopenTensorMhaScaleDS = 25,
6181 miopenTensorMhaScaleDQ = 26,
6182 miopenTensorMhaScaleDK = 27,
6183 miopenTensorMhaScaleDV = 28,
6184 miopenTensorMhaDQ = 29,
6185 miopenTensorMhaDK = 30,
6186 miopenTensorMhaDV = 31,
6187 miopenTensorMhaAmaxDQ = 32,
6188 miopenTensorMhaAmaxDK = 33,
6189 miopenTensorMhaAmaxDV = 34,
6190 miopenTensorMhaAmaxDS = 35,
6191 miopenTensorMhaBias = 36,
6192
6193#ifdef MIOPEN_BETA_API
6194 miopenTensorActivationX = 37,
6195 miopenTensorActivationY = 38,
6196 miopenTensorActivationDX = 39,
6197 miopenTensorActivationDY = 40,
6198 miopenTensorBiasX = 41,
6199 miopenTensorBiasY = 42,
6200 miopenTensorBias = 43,
6201 miopenTensorSoftmaxX = 44,
6202 miopenTensorSoftmaxY = 45,
6203 miopenTensorSoftmaxDX = 46,
6204 miopenTensorSoftmaxDY = 47,
6205 miopenTensorBatchnormX = 48,
6206 miopenTensorBatchnormY = 49,
6207 miopenTensorBatchnormRunningMean = 50,
6208 miopenTensorBatchnormRunningVariance = 51,
6209 miopenTensorBatchnormSavedMean = 52,
6210 miopenTensorBatchnormSavedVariance = 53,
6211 miopenTensorBatchnormScale = 54,
6212 miopenTensorBatchnormScaleDiff = 55,
6213 miopenTensorBatchnormEstimatedMean = 56,
6214 miopenTensorBatchnormEstimatedVariance = 57,
6215 miopenTensorBatchnormBias = 58,
6216 miopenTensorBatchnormBiasDiff = 59,
6217 miopenTensorBatchnormDX = 60,
6218 miopenTensorBatchnormDY = 61,
6219#endif
6220
6221 miopenTensorArgumentIsScalar = 1U << 31,
6222
6223 miopenTensorMhaMask = miopenTensorArgumentIsScalar | 1,
6224#ifdef MIOPEN_BETA_API
6225 miopenScalarBatchnormExpAvgFactor = miopenTensorArgumentIsScalar | 2,
6226 miopenScalarBatchnormEpsilon = miopenTensorArgumentIsScalar | 3,
6227#endif
6228} miopenTensorArgumentId_t;
6229
6238
6246MIOPEN_EXPORT miopenStatus_t miopenCreateConvProblem(miopenProblem_t* problem,
6247 miopenConvolutionDescriptor_t operatorDesc,
6248 miopenProblemDirection_t direction);
6249
6257
6266
6267MIOPEN_EXPORT miopenStatus_t miopenCreateMhaProblem(miopenProblem_t* problem,
6268 miopenMhaDescriptor_t operatorDesc,
6269 miopenProblemDirection_t direction);
6270
6276
6277MIOPEN_EXPORT miopenStatus_t miopenCreateMhaDescriptor(miopenMhaDescriptor_t* mhaDesc);
6278
6287
6288MIOPEN_EXPORT miopenStatus_t miopenSetMhaDescriptor(miopenMhaDescriptor_t mhaDesc, float scale);
6289
6298
6299MIOPEN_EXPORT miopenStatus_t miopenGetMhaDescriptor(miopenMhaDescriptor_t mhaDesc, float* scale);
6300
6306
6307MIOPEN_EXPORT miopenStatus_t miopenCreateSoftmaxDescriptor(miopenSoftmaxDescriptor_t* softmaxDesc);
6308
6320MIOPEN_EXPORT miopenStatus_t miopenSetSoftmaxDescriptor(miopenSoftmaxDescriptor_t softmaxDesc,
6321 float alpha,
6322 float beta,
6323 miopenSoftmaxAlgorithm_t algorithm,
6324 miopenSoftmaxMode_t mode);
6325
6337MIOPEN_EXPORT miopenStatus_t miopenGetSoftmaxDescriptor(const miopenSoftmaxDescriptor_t softmaxDesc,
6338 float* alpha,
6339 float* beta,
6340 miopenSoftmaxAlgorithm_t* algorithm,
6341 miopenSoftmaxMode_t* mode);
6342
6348MIOPEN_EXPORT miopenStatus_t miopenDestroyProblem(miopenProblem_t problem);
6349
6357MIOPEN_EXPORT miopenStatus_t
6358miopenSetProblemTensorDescriptor(miopenProblem_t problem,
6359 miopenTensorArgumentId_t id,
6360 const miopenTensorDescriptor_t descriptor);
6361
6364MIOPEN_DECLARE_OBJECT(miopenFindOptions);
6365
6371MIOPEN_EXPORT miopenStatus_t miopenCreateFindOptions(miopenFindOptions_t* options);
6372
6378MIOPEN_EXPORT miopenStatus_t miopenDestroyFindOptions(miopenFindOptions_t options);
6379
6386MIOPEN_EXPORT miopenStatus_t miopenSetFindOptionTuning(miopenFindOptions_t options, int value);
6387
6394MIOPEN_EXPORT miopenStatus_t miopenSetFindOptionResultsOrder(miopenFindOptions_t options,
6395 miopenFindResultsOrder_t value);
6396
6404MIOPEN_EXPORT miopenStatus_t miopenSetFindOptionWorkspaceLimit(miopenFindOptions_t options,
6405 size_t value);
6406
6414MIOPEN_EXPORT miopenStatus_t miopenSetFindOptionPreallocatedWorkspace(miopenFindOptions_t options,
6415 void* buffer,
6416 size_t size);
6417
6426MIOPEN_EXPORT miopenStatus_t miopenSetFindOptionPreallocatedTensor(miopenFindOptions_t options,
6427 miopenTensorArgumentId_t id,
6428 void* buffer);
6429
6437MIOPEN_EXPORT miopenStatus_t miopenSetFindOptionAttachBinaries(miopenFindOptions_t options,
6438 unsigned attach);
6439
6442MIOPEN_DECLARE_OBJECT(miopenSolution);
6443
6455MIOPEN_EXPORT miopenStatus_t miopenFindSolutions(miopenHandle_t handle,
6456 miopenProblem_t problem,
6457 miopenFindOptions_t options,
6458 miopenSolution_t* solutions,
6459 size_t* numSolutions,
6460 size_t maxSolutions);
6461
6464struct miopenTensorArgument_t
6465{
6466 /* @brief Identifier of the tensor argument.
6467 */
6468 miopenTensorArgumentId_t id;
6469 /* @brief Tensor descriptor to override the value stored in the solution.
6470 *
6471 * Some solvers may support overriding input and output tensor descriptors, but right now there
6472 * is no way to tell from the API. Intended for the future use.
6473 */
6474 miopenTensorDescriptor_t* descriptor;
6475 /* @brief Pointer to the device memory buffer to use for the operation or to the host memory if
6476 * the value is scalar.
6477 */
6478 void* buffer;
6479};
6480
6492MIOPEN_EXPORT miopenStatus_t miopenRunSolution(miopenHandle_t handle,
6493 miopenSolution_t solution,
6494 size_t nInputs,
6495 const miopenTensorArgument_t* tensors,
6496 void* workspace,
6497 size_t workspaceSize);
6498
6504MIOPEN_EXPORT miopenStatus_t miopenDestroySolution(miopenSolution_t solution);
6505
6513MIOPEN_EXPORT miopenStatus_t miopenLoadSolution(miopenSolution_t* solution,
6514 const char* data,
6515 size_t size);
6516
6523MIOPEN_EXPORT miopenStatus_t miopenSaveSolution(miopenSolution_t solution, char* data);
6524
6531MIOPEN_EXPORT miopenStatus_t miopenGetSolutionSize(miopenSolution_t solution, size_t* size);
6532
6539MIOPEN_EXPORT miopenStatus_t miopenGetSolutionWorkspaceSize(miopenSolution_t solution,
6540 size_t* workspaceSize);
6541
6548MIOPEN_EXPORT miopenStatus_t miopenGetSolutionTime(miopenSolution_t solution, float* time);
6549
6556MIOPEN_EXPORT miopenStatus_t miopenGetSolutionSolverId(miopenSolution_t solution,
6557 uint64_t* solverId);
6558
6565MIOPEN_EXPORT miopenStatus_t miopenGetSolverIdConvAlgorithm(uint64_t solverId,
6566 miopenConvAlgorithm_t* result);
6567
6568#ifdef MIOPEN_BETA_API
6569
6578MIOPEN_EXPORT miopenStatus_t
6579miopenCreateActivationProblem(miopenProblem_t* problem,
6580 miopenActivationDescriptor_t operatorDesc,
6581 miopenProblemDirection_t direction);
6582
6591MIOPEN_EXPORT miopenStatus_t miopenCreateBatchnormProblem(miopenProblem_t* problem,
6592 miopenBatchNormMode_t mode,
6593 bool runningMeanVariance,
6594 miopenProblemDirection_t direction);
6595
6615MIOPEN_EXPORT miopenStatus_t miopenFuseProblems(miopenProblem_t problem1, miopenProblem_t problem2);
6616
6624MIOPEN_EXPORT miopenStatus_t miopenCreateBiasProblem(miopenProblem_t* problem,
6625 miopenProblemDirection_t direction);
6626
6634
6635MIOPEN_EXPORT miopenStatus_t miopenCreateSoftmaxProblem(miopenProblem_t* problem,
6636 miopenSoftmaxDescriptor_t operatorDesc,
6637 miopenProblemDirection_t direction);
6638
6639#endif
6640
6642// CLOSEOUT find2 DOXYGEN GROUP
6643
6644#ifdef MIOPEN_BETA_API
6645
6656
6657// ReduceCalculation APIs
6662
6673
6683MIOPEN_EXPORT miopenStatus_t
6684miopenGetReduceCalculationWorkspaceSize(miopenHandle_t handle,
6685 const miopenTensorDescriptor_t xDesc,
6686 const int32_t dim,
6687 const miopenReduceCalculationOp_t reduceCalculationOp,
6688 const miopenTensorDescriptor_t reduceDesc,
6689 size_t* sizeInBytes);
6690
6704MIOPEN_EXPORT miopenStatus_t
6705miopenReduceCalculationForward(miopenHandle_t handle,
6706 miopenReduceCalculationNanPropagation_t nanPropagation,
6707 void* workspace,
6708 size_t workspaceSizeInBytes,
6709 const miopenTensorDescriptor_t xDesc,
6710 const void* x,
6711 const int32_t dim,
6712 const miopenReduceCalculationOp_t reduceCalculationOp,
6713 const miopenTensorDescriptor_t reduceDesc,
6714 void* y);
6715
6717// CLOSEOUT REDUCE CALCULATION DOXYGEN GROUP
6718#endif // MIOPEN_BETA_API
6719
6720#ifdef MIOPEN_BETA_API
6721
6737
6738// ReduceExtreme APIs
6743
6758MIOPEN_EXPORT miopenStatus_t
6759miopenReduceExtremeForward(miopenHandle_t handle,
6760 const miopenTensorDescriptor_t xDesc,
6761 const void* x,
6762 const int32_t dim,
6763 const miopenReduceExtremeOp_t reduceExtremeOp,
6764 const miopenTensorDescriptor_t yDesc,
6765 void* y,
6766 const miopenTensorDescriptor_t indiceDesc,
6767 void* indice);
6768
6770// CLOSEOUT REDUCEEXTREME DOXYGEN GROUP
6771#endif // MIOPEN_BETA_API
6772
6773#ifdef MIOPEN_BETA_API
6774// GroupNorm APIs
6799MIOPEN_EXPORT miopenStatus_t miopenGroupNormForward(miopenHandle_t handle,
6800 miopenNormMode_t mode,
6801 const miopenTensorDescriptor_t xDesc,
6802 const void* x,
6803 const miopenTensorDescriptor_t weightDesc,
6804 const void* weight,
6805 const miopenTensorDescriptor_t biasDesc,
6806 const void* bias,
6807 const uint64_t num_groups,
6808 const float epsilon,
6809 const miopenTensorDescriptor_t yDesc,
6810 void* y,
6811 const miopenTensorDescriptor_t meanDesc,
6812 void* mean,
6813 const miopenTensorDescriptor_t rstdDesc,
6814 void* rstd);
6815
6817// CLOSEOUT groupnorm DOXYGEN GROUP
6818#endif // MIOPEN_BETA_API
6819
6820#ifdef MIOPEN_BETA_API
6821// LayerNorm APIs
6848MIOPEN_EXPORT miopenStatus_t miopenAddLayerNormForward(miopenHandle_t handle,
6849 miopenNormMode_t mode,
6850 const miopenTensorDescriptor_t xDesc,
6851 const void* x,
6852 const miopenTensorDescriptor_t x2Desc,
6853 const void* x2,
6854 const miopenTensorDescriptor_t weightDesc,
6855 const void* weight,
6856 const miopenTensorDescriptor_t biasDesc,
6857 const void* bias,
6858 const float epsilon,
6859 const int32_t normalized_dim,
6860 const miopenTensorDescriptor_t yDesc,
6861 void* y,
6862 const miopenTensorDescriptor_t meanDesc,
6863 void* mean,
6864 const miopenTensorDescriptor_t rstdDesc,
6865 void* rstd);
6866
6868// CLOSEOUT LAYERNORM DOXYGEN GROUP
6869#endif // MIOPEN_BETA_API
6870
6871#ifdef MIOPEN_BETA_API
6872// LayerNorm APIs
6892MIOPEN_EXPORT miopenStatus_t miopenT5LayerNormForward(miopenHandle_t handle,
6893 miopenNormMode_t mode,
6894 const miopenTensorDescriptor_t xDesc,
6895 const void* x,
6896 const miopenTensorDescriptor_t weightDesc,
6897 const void* weight,
6898 const float epsilon,
6899 const miopenTensorDescriptor_t yDesc,
6900 void* y,
6901 const miopenTensorDescriptor_t rstdDesc,
6902 void* rstd);
6903
6918MIOPEN_EXPORT miopenStatus_t
6919miopenGetT5LayerNormBackwardWorkspaceSize(miopenHandle_t handle,
6920 miopenNormMode_t mode,
6921 const miopenTensorDescriptor_t dyDesc,
6922 const miopenTensorDescriptor_t xDesc,
6923 const miopenTensorDescriptor_t weightDesc,
6924 const miopenTensorDescriptor_t rstdDesc,
6925 const miopenTensorDescriptor_t dxDesc,
6926 const miopenTensorDescriptor_t dwDesc,
6927 size_t* sizeInBytes);
6928
6949MIOPEN_EXPORT miopenStatus_t miopenT5LayerNormBackward(miopenHandle_t handle,
6950 miopenNormMode_t mode,
6951 void* workspace,
6952 size_t workspaceSizeInBytes,
6953 const miopenTensorDescriptor_t dyDesc,
6954 const void* dy,
6955 const miopenTensorDescriptor_t xDesc,
6956 const void* x,
6957 const miopenTensorDescriptor_t weightDesc,
6958 const void* weight,
6959 const miopenTensorDescriptor_t rstdDesc,
6960 const void* rstd,
6961 const miopenTensorDescriptor_t dxDesc,
6962 void* dx,
6963 const miopenTensorDescriptor_t dwDesc,
6964 void* dw);
6966// CLOSEOUT LAYERNORM DOXYGEN GROUP
6967#endif // MIOPEN_BETA_API
6968
6973typedef enum
6974{
6975 /* IDENTITY alpha = 1.0 and beta = 0.0 */
6976 /* SCALE alpha = 4.2 and beta = 0.0 */
6977 /* BILINEAR alpha = 3.2 and beta = 1.1 */
6978 /* ERROR_STATE alpha = 0.0 and beta = 3.1 */
6979
6980 DEFAULT = 0, /* alpha = 1.0 and beta = 0.0.*/
6981 SCALE = 1, /* alpha with some value and beta 0.0*/
6982 BILINEAR = 2, /* both alpha and beta with some value*/
6983 ERROR_STATE = 3 /* alpha 0.0 and beta with some value, this should not occur.
6984 But used to check for errors.*/
6985} miopenAlphaBetaCase_t;
6986
6987#ifdef MIOPEN_BETA_API
6988// FusedAdam APIs
7131MIOPEN_EXPORT miopenStatus_t miopenFusedAdam(miopenHandle_t handle,
7132 const miopenTensorDescriptor_t paramDesc,
7133 void* param,
7134 const miopenTensorDescriptor_t gradDesc,
7135 const void* grad,
7136 const miopenTensorDescriptor_t expAvgDesc,
7137 void* expAvg,
7138 const miopenTensorDescriptor_t expAvgSqDesc,
7139 void* expAvgSq,
7140 const miopenTensorDescriptor_t maxExpAvgSqDesc,
7141 void* maxExpAvgSq,
7142 const miopenTensorDescriptor_t stateStepDesc,
7143 void* stateStep,
7144 const unsigned int state_step,
7145 const float lr,
7146 const float beta1,
7147 const float beta2,
7148 const float weight_decay,
7149 const float eps,
7150 const bool amsgrad,
7151 const bool maximize,
7152 const bool adamw,
7153 const miopenTensorDescriptor_t gradScaleDesc,
7154 const void* gradScale,
7155 const miopenTensorDescriptor_t foundInfDesc,
7156 const void* foundInf);
7157
7298MIOPEN_EXPORT miopenStatus_t
7299miopenFusedAdamWithOutput(miopenHandle_t handle,
7300 const miopenTensorDescriptor_t paramInDesc,
7301 void* paramIn,
7302 const miopenTensorDescriptor_t paramOutDesc,
7303 void* paramOut,
7304 const miopenTensorDescriptor_t paramOutFloat16Desc,
7305 void* paramOutFloat16,
7306 const miopenTensorDescriptor_t gradInDesc,
7307 const void* gradIn,
7308 const miopenTensorDescriptor_t expAvgInDesc,
7309 void* expAvgIn,
7310 const miopenTensorDescriptor_t expAvgOutDesc,
7311 void* expAvgOut,
7312 const miopenTensorDescriptor_t expAvgSqInDesc,
7313 void* expAvgSqIn,
7314 const miopenTensorDescriptor_t expAvgSqOutDesc,
7315 void* expAvgSqOut,
7316 const miopenTensorDescriptor_t maxExpAvgSqInDesc,
7317 void* maxExpAvgSqIn,
7318 const miopenTensorDescriptor_t maxExpAvgSqOutDesc,
7319 void* maxExpAvgSqOut,
7320 const miopenTensorDescriptor_t stateStepInDesc,
7321 void* stateStepIn,
7322 const miopenTensorDescriptor_t stateStepOutDesc,
7323 void* stateStepOut,
7324 const unsigned int state_step,
7325 const float lr,
7326 const float beta1,
7327 const float beta2,
7328 const float weight_decay,
7329 const float eps,
7330 const bool amsgrad,
7331 const bool maximize,
7332 const bool adamw,
7333 const miopenTensorDescriptor_t gradScaleDesc,
7334 const void* gradScale,
7335 const miopenTensorDescriptor_t foundInfDesc,
7336 const void* foundInf);
7337
7339// CLOSEOUT SGD DOXYGEN GROUP
7340#endif // MIOPEN_BETA_API
7341
7342#ifdef MIOPEN_BETA_API
7343// TransformersAdamW APIs
7434MIOPEN_EXPORT miopenStatus_t miopenTransformersAdamW(miopenHandle_t handle,
7435 const miopenTensorDescriptor_t paramDesc,
7436 void* param,
7437 const miopenTensorDescriptor_t gradDesc,
7438 const void* grad,
7439 const miopenTensorDescriptor_t expAvgDesc,
7440 void* expAvg,
7441 const miopenTensorDescriptor_t expAvgSqDesc,
7442 void* expAvgSq,
7443 const miopenTensorDescriptor_t stateStepDesc,
7444 void* stateStep,
7445 const unsigned int state_step,
7446 const float lr,
7447 const float beta1,
7448 const float beta2,
7449 const float weight_decay,
7450 const float eps,
7451 const bool correct_bias,
7452 const miopenTensorDescriptor_t gradScaleDesc,
7453 const void* gradScale,
7454 const miopenTensorDescriptor_t foundInfDesc,
7455 const void* foundInf);
7456
7583MIOPEN_EXPORT miopenStatus_t
7584miopenTransformersAdamWWithOutput(miopenHandle_t handle,
7585 const miopenTensorDescriptor_t paramInDesc,
7586 void* paramIn,
7587 const miopenTensorDescriptor_t paramOutDesc,
7588 void* paramOut,
7589 const miopenTensorDescriptor_t paramOutFloat16Desc,
7590 void* paramOutFloat16,
7591 const miopenTensorDescriptor_t gradInDesc,
7592 const void* gradIn,
7593 const miopenTensorDescriptor_t expAvgInDesc,
7594 void* expAvgIn,
7595 const miopenTensorDescriptor_t expAvgOutDesc,
7596 void* expAvgOut,
7597 const miopenTensorDescriptor_t expAvgSqInDesc,
7598 void* expAvgSqIn,
7599 const miopenTensorDescriptor_t expAvgSqOutDesc,
7600 void* expAvgSqOut,
7601 const miopenTensorDescriptor_t stateStepInDesc,
7602 void* stateStepIn,
7603 const miopenTensorDescriptor_t stateStepOutDesc,
7604 void* stateStepOut,
7605 const unsigned int state_step,
7606 const float lr,
7607 const float beta1,
7608 const float beta2,
7609 const float weight_decay,
7610 const float eps,
7611 const float step_size,
7612 const bool correct_bias,
7613 const miopenTensorDescriptor_t gradScaleDesc,
7614 const void* gradScale,
7615 const miopenTensorDescriptor_t foundInfDesc,
7616 const void* foundInf);
7617
7619// CLOSEOUT SGD DOXYGEN GROUP
7620#endif // MIOPEN_BETA_API
7621
7622#ifdef MIOPEN_BETA_API
7623// GetItem APIs
7636MIOPEN_EXPORT miopenStatus_t
7637miopenGetGetitemWorkspaceSize(miopenHandle_t handle,
7638 uint32_t indexCount,
7639 const miopenTensorDescriptor_t* indexDescs,
7640 size_t* sizeInBytes);
7641
7666MIOPEN_EXPORT miopenStatus_t miopenGetitemBackward(miopenHandle_t handle,
7667 void* workspace,
7668 size_t workspaceSizeInBytes,
7669 const miopenTensorDescriptor_t dyDesc,
7670 const void* dy,
7671 uint32_t indexCount,
7672 const miopenTensorDescriptor_t* indexDescs,
7673 const void* const* indexs,
7674 const miopenTensorDescriptor_t dxDesc,
7675 void* dx,
7676 const miopenTensorDescriptor_t errorDesc,
7677 void* error,
7678 uint32_t dimCount,
7679 const int32_t* dims,
7680 uint32_t sliceCount,
7681 const int32_t* slices,
7682 uint32_t offset);
7683
7685// CLOSEOUT GETITEM DOXYGEN GROUP
7686#endif // MIOPEN_BETA_API
7687
7688#ifdef MIOPEN_BETA_API
7689// RotaryPositionalEmbeddings APIs
7707MIOPEN_EXPORT miopenStatus_t miopenRoPEForward(miopenHandle_t handle,
7708 const miopenTensorDescriptor_t xDesc,
7709 const void* x,
7710 const miopenTensorDescriptor_t cosDesc,
7711 const void* cos,
7712 const miopenTensorDescriptor_t sinDesc,
7713 const void* sin,
7714 const miopenTensorDescriptor_t yDesc,
7715 void* y);
7716
7730MIOPEN_EXPORT miopenStatus_t miopenRoPEBackward(miopenHandle_t handle,
7731 const miopenTensorDescriptor_t dyDesc,
7732 const void* dy,
7733 const miopenTensorDescriptor_t cosDesc,
7734 const void* cos,
7735 const miopenTensorDescriptor_t sinDesc,
7736 const void* sin,
7737 const miopenTensorDescriptor_t dxDesc,
7738 void* dx);
7740// CLOSEOUT ROPE DOXYGEN GROUP
7741// kthvalue APIs
7746
7762MIOPEN_EXPORT miopenStatus_t miopenKthvalueForward(miopenHandle_t handle,
7763 miopenTensorDescriptor_t inputDesc,
7764 const void* input,
7765 miopenTensorDescriptor_t outputDesc,
7766 void* output,
7767 miopenTensorDescriptor_t indicesDesc,
7768 size_t* indices,
7769 size_t k,
7770 int32_t dim = -1,
7771 bool keepDim = false);
7772
7774// CLOSEOUT kthvalue DOXYGEN GROUP
7775#endif // MIOPEN_BETA_API
7776
7777#ifdef MIOPEN_BETA_API
7782
7791MIOPEN_EXPORT miopenStatus_t
7792miopenGetPReLUBackwardWorkspaceSize(miopenHandle_t handle,
7793 miopenTensorDescriptor_t inputDesc,
7794 miopenTensorDescriptor_t weightDesc,
7795 size_t* sizeInBytes);
7796
7813MIOPEN_EXPORT miopenStatus_t miopenPReLUBackward(miopenHandle_t handle,
7814 void* workspace,
7815 size_t workspaceSizeInBytes,
7816 miopenTensorDescriptor_t inputDesc,
7817 const void* input,
7818 miopenTensorDescriptor_t weightDesc,
7819 const void* weight,
7820 miopenTensorDescriptor_t doutputDesc,
7821 const void* doutput,
7822 miopenTensorDescriptor_t dinputDesc,
7823 void* dinput,
7824 miopenTensorDescriptor_t dweightDesc,
7825 void* dweight);
7826
7828// CLOSEOUT RELU DOXYGEN GROUP
7829#endif // MIOPEN_BETA_API
7830
7831#ifdef MIOPEN_BETA_API
7832
7844
7845// SoftMarginLoss APIs
7850
7863MIOPEN_EXPORT miopenStatus_t
7864miopenGetSoftMarginLossForwardWorkspaceSize(miopenHandle_t handle,
7865 miopenTensorDescriptor_t inputDesc,
7866 miopenTensorDescriptor_t targetDesc,
7867 miopenTensorDescriptor_t outputDesc,
7868 miopenLossReductionMode_t reduction,
7869 size_t* sizeInBytes);
7870
7887MIOPEN_EXPORT miopenStatus_t miopenSoftMarginLossForward(miopenHandle_t handle,
7888 miopenTensorDescriptor_t inputDesc,
7889 const void* input,
7890 miopenTensorDescriptor_t targetDesc,
7891 const void* target,
7892 miopenTensorDescriptor_t outputDesc,
7893 void* output,
7894 miopenLossReductionMode_t reduction,
7895 void* workspace = nullptr,
7896 size_t workspaceSizeInBytes = 0);
7897
7912MIOPEN_EXPORT miopenStatus_t miopenSoftMarginLossBackward(miopenHandle_t handle,
7913 miopenTensorDescriptor_t inputDesc,
7914 const void* input,
7915 miopenTensorDescriptor_t targetDesc,
7916 const void* target,
7917 miopenTensorDescriptor_t doutputDesc,
7918 const void* doutput,
7919 miopenTensorDescriptor_t dinputDesc,
7920 void* dinput,
7921 miopenLossReductionMode_t reduction);
7922
7924// CLOSEOUT LossFunction DOXYGEN GROUP
7925#endif
7926
7927#ifdef MIOPEN_BETA_API
7928// MultiMarginLoss APIs
7933
7952MIOPEN_EXPORT miopenStatus_t
7953miopenGetMultiMarginLossForwardWorkspaceSize(miopenHandle_t handle,
7954 miopenTensorDescriptor_t inputDesc,
7955 miopenTensorDescriptor_t targetDesc,
7956 miopenTensorDescriptor_t weightDesc,
7957 miopenTensorDescriptor_t outputDesc,
7958 long p,
7959 float margin,
7960 miopenLossReductionMode_t reduction,
7961 size_t* sizeInBytes);
7962
7990MIOPEN_EXPORT miopenStatus_t miopenMultiMarginLossForward(miopenHandle_t handle,
7991 miopenTensorDescriptor_t inputDesc,
7992 const void* input,
7993 miopenTensorDescriptor_t targetDesc,
7994 const void* target,
7995 miopenTensorDescriptor_t weightDesc,
7996 const void* weight,
7997 miopenTensorDescriptor_t outputDesc,
7998 void* output,
7999 long p,
8000 float margin,
8001 miopenLossReductionMode_t reduction,
8002 void* workspace,
8003 size_t workspaceSizeInBytes);
8004
8006// CLOSEOUT LossFunction DOXYGEN GROUP
8007#endif // MIOPEN_BETA_API
8008
8029
8038MIOPEN_EXPORT miopenStatus_t miopenSetTuningPolicy(miopenHandle_t handle,
8039 miopenTuningPolicy_t newValue);
8040
8048MIOPEN_EXPORT miopenStatus_t miopenGetTuningPolicy(miopenHandle_t handle,
8049 miopenTuningPolicy_t* value);
8050
8051#ifdef __cplusplus
8052}
8053#endif
8054
8055#ifdef __clang__
8056#pragma clang diagnostic pop
8057#endif
8058
8059#endif // MIOPEN_GUARD_MIOPEN_H_
miopenCreateOpActivationBackward
miopenStatus_t miopenCreateOpActivationBackward(miopenFusionPlanDescriptor_t fusePlanDesc, miopenFusionOpDescriptor_t *activBwdOp, miopenActivationMode_t mode)
Creates a backward activation operator.
miopenCreateOpBatchNormForward
miopenStatus_t miopenCreateOpBatchNormForward(miopenFusionPlanDescriptor_t fusePlanDesc, miopenFusionOpDescriptor_t *bnFwdOp, const miopenBatchNormMode_t bn_mode, bool runningMeanVariance)
Creates a forward training batch normalization operator.
miopenFusionPlanConvolutionGetAlgo
miopenStatus_t miopenFusionPlanConvolutionGetAlgo(miopenFusionPlanDescriptor_t fusePlanDesc, const int requestAlgoCount, int *returnedAlgoCount, miopenConvFwdAlgorithm_t *returnedAlgos)
Returns the supported algorithms for the convolution operator in the Fusion Plan.
miopenFusionPlanGetWorkSpaceSize
miopenStatus_t miopenFusionPlanGetWorkSpaceSize(miopenHandle_t handle, miopenFusionPlanDescriptor_t fusePlanDesc, size_t *workSpaceSize, miopenConvFwdAlgorithm_t algo)
Query the workspace size required for the fusion plan.
miopenFusionPlanConvolutionSetAlgo
miopenStatus_t miopenFusionPlanConvolutionSetAlgo(miopenFusionPlanDescriptor_t fusePlanDesc, miopenConvFwdAlgorithm_t algo)
Requests the fusion runtime to choose a particular algorithm for the added convolution operation.
miopenCreateOpBatchNormInference
miopenStatus_t miopenCreateOpBatchNormInference(miopenFusionPlanDescriptor_t fusePlanDesc, miopenFusionOpDescriptor_t *bnOp, const miopenBatchNormMode_t bn_mode, const miopenTensorDescriptor_t bnScaleBiasMeanVarDesc)
Creates a forward inference batch normalization operator.
miopenSetOpArgsBiasForward
miopenStatus_t miopenSetOpArgsBiasForward(miopenOperatorArgs_t args, const miopenFusionOpDescriptor_t biasOp, const void *alpha, const void *beta, const void *bias)
Sets the arguments for forward bias op.
miopenCreateOpConvForward
miopenStatus_t miopenCreateOpConvForward(miopenFusionPlanDescriptor_t fusePlanDesc, miopenFusionOpDescriptor_t *convOp, miopenConvolutionDescriptor_t convDesc, const miopenTensorDescriptor_t wDesc)
Creates forward convolution operator.
miopenSetOpArgsBatchNormInference
miopenStatus_t miopenSetOpArgsBatchNormInference(miopenOperatorArgs_t args, const miopenFusionOpDescriptor_t bnOp, const void *alpha, const void *beta, const void *bnScale, const void *bnBias, const void *estimatedMean, const void *estimatedVariance, double epsilon)
Sets the arguments for inference batch normalization op.
miopenSetOpArgsBatchNormForward
miopenStatus_t miopenSetOpArgsBatchNormForward(miopenOperatorArgs_t args, const miopenFusionOpDescriptor_t bnOp, const void *alpha, const void *beta, const void *bnScale, const void *bnBias, void *savedMean, void *savedInvVariance, void *runningMean, void *runningVariance, double expAvgFactor, double epsilon)
Sets the arguments for forward batch normalization op.
miopenExecuteFusionPlan
miopenStatus_t miopenExecuteFusionPlan(const miopenHandle_t handle, const miopenFusionPlanDescriptor_t fusePlanDesc, const miopenTensorDescriptor_t inputDesc, const void *input, const miopenTensorDescriptor_t outputDesc, void *output, miopenOperatorArgs_t args)
Executes the fusion plan. Only compatible with NHWC/NDHWC tensor layouts.
miopenFusionPlanGetOp
miopenStatus_t miopenFusionPlanGetOp(miopenFusionPlanDescriptor_t fusePlanDesc, const int op_idx, miopenFusionOpDescriptor_t *op)
Allows access to the operators in a fusion plan.
miopenDestroyFusionPlan
miopenStatus_t miopenDestroyFusionPlan(miopenFusionPlanDescriptor_t fusePlanDesc)
Destroy the fusion plan descriptor object.
miopenCreateOpActivationForward
miopenStatus_t miopenCreateOpActivationForward(miopenFusionPlanDescriptor_t fusePlanDesc, miopenFusionOpDescriptor_t *activFwdOp, miopenActivationMode_t mode)
Creates a forward activation operator.
miopenExecuteFusionPlan_v2
miopenStatus_t miopenExecuteFusionPlan_v2(const miopenHandle_t handle, const miopenFusionPlanDescriptor_t fusePlanDesc, const miopenTensorDescriptor_t inputDesc, const void *input, const miopenTensorDescriptor_t outputDesc, void *output, miopenOperatorArgs_t args, void *workspace, size_t workspaceSize)
Executes the fusion plan with a workspace buffer for layout transformations.
miopenFusionDirection_t
miopenFusionDirection_t
Kernel fusion direction in the network.
Definition miopen.h:3825
miopenSetOpArgsActivBackward
miopenStatus_t miopenSetOpArgsActivBackward(miopenOperatorArgs_t args, const miopenFusionOpDescriptor_t activBwdOp, const void *alpha, const void *beta, const void *y, const void *reserved, double activAlpha, double activBeta, double activGamma)
Sets the arguments for backward activation op.
miopenCompileFusionPlan
miopenStatus_t miopenCompileFusionPlan(miopenHandle_t handle, miopenFusionPlanDescriptor_t fusePlanDesc)
Compiles the fusion plan.
miopenSetOpArgsBatchNormBackward
miopenStatus_t miopenSetOpArgsBatchNormBackward(miopenOperatorArgs_t args, const miopenFusionOpDescriptor_t bnOp, const void *alpha, const void *beta, const void *x, const void *bnScale, const void *bnBias, void *resultBnScaleDiff, void *resultBnBiasDiff, const void *savedMean, const void *savedInvVariance)
Sets the arguments for backward batch normalization op.
miopenSetOpArgsActivForward
miopenStatus_t miopenSetOpArgsActivForward(miopenOperatorArgs_t args, const miopenFusionOpDescriptor_t activFwdOp, const void *alpha, const void *beta, double activAlpha, double activBeta, double activGamma)
Sets the arguments for forward activation op.
miopenDestroyOperatorArgs
miopenStatus_t miopenDestroyOperatorArgs(miopenOperatorArgs_t args)
Destroys an operator argument object.
miopenCreateOpBatchNormBackward
miopenStatus_t miopenCreateOpBatchNormBackward(miopenFusionPlanDescriptor_t fusePlanDesc, miopenFusionOpDescriptor_t *bnBwdOp, const miopenBatchNormMode_t bn_mode)
Creates a back propagation batch normalization operator.
miopenCreateOperatorArgs
miopenStatus_t miopenCreateOperatorArgs(miopenOperatorArgs_t *args)
Creates an operator argument object.
miopenSetOpArgsConvForward
miopenStatus_t miopenSetOpArgsConvForward(miopenOperatorArgs_t args, const miopenFusionOpDescriptor_t convOp, const void *alpha, const void *beta, const void *w)
Sets the arguments for forward convolution op.
miopenConvolutionBiasActivationForward
miopenStatus_t miopenConvolutionBiasActivationForward(miopenHandle_t handle, const void *alpha1, const miopenTensorDescriptor_t xDesc, const void *x, const miopenTensorDescriptor_t wDesc, const void *w, const miopenConvolutionDescriptor_t convDesc, miopenConvFwdAlgorithm_t algo, void *workspace, size_t workspaceSizeInBytes, const void *alpha2, const miopenTensorDescriptor_t zDesc, const void *z, const miopenTensorDescriptor_t biasDesc, const void *bias, const miopenActivationDescriptor_t activationDesc, const miopenTensorDescriptor_t yDesc, void *y)
Prepares and executes the Convlution+Bias+Activation Fusion.
miopenCreateFusionPlan
miopenStatus_t miopenCreateFusionPlan(miopenFusionPlanDescriptor_t *fusePlanDesc, const miopenFusionDirection_t fuseDirection, const miopenTensorDescriptor_t inputDesc)
Creates the kenrel fusion plan descriptor object.
miopenCreateOpBiasForward
miopenStatus_t miopenCreateOpBiasForward(miopenFusionPlanDescriptor_t fusePlanDesc, miopenFusionOpDescriptor_t *biasOp, const miopenTensorDescriptor_t bDesc)
Creates a forward bias operator.
miopenHorizontalFusion
@ miopenHorizontalFusion
Definition miopen.h:3827
miopenVerticalFusion
@ miopenVerticalFusion
Definition miopen.h:3826
miopenSetLRNDescriptor
miopenStatus_t miopenSetLRNDescriptor(const miopenLRNDescriptor_t lrnDesc, miopenLRNMode_t mode, unsigned int lrnN, double lrnAlpha, double lrnBeta, double lrnK)
Sets a LRN layer descriptor details.
miopenCreateLRNDescriptor
miopenStatus_t miopenCreateLRNDescriptor(miopenLRNDescriptor_t *lrnDesc)
Creates a local response normalization (LRN) layer descriptor.
miopenGetLRNDescriptor
miopenStatus_t miopenGetLRNDescriptor(const miopenLRNDescriptor_t lrnDesc, miopenLRNMode_t *mode, unsigned int *lrnN, double *lrnAlpha, double *lrnBeta, double *lrnK)
Gets a LRN layer descriptor details.
miopenLRNBackward
miopenStatus_t miopenLRNBackward(miopenHandle_t handle, const miopenLRNDescriptor_t lrnDesc, const void *alpha, const miopenTensorDescriptor_t yDesc, const void *y, const miopenTensorDescriptor_t dyDesc, const void *dy, const miopenTensorDescriptor_t xDesc, const void *x, const void *beta, const miopenTensorDescriptor_t dxDesc, void *dx, const void *workSpace)
Execute a LRN backward layer.
miopenLRNGetWorkSpaceSize
miopenStatus_t miopenLRNGetWorkSpaceSize(const miopenTensorDescriptor_t yDesc, size_t *workSpaceSize)
Determine the workspace requirements.
miopenLRNMode_t
miopenLRNMode_t
Definition miopen.h:461
miopenLRNForward
miopenStatus_t miopenLRNForward(miopenHandle_t handle, const miopenLRNDescriptor_t lrnDesc, const void *alpha, const miopenTensorDescriptor_t xDesc, const void *x, const void *beta, const miopenTensorDescriptor_t yDesc, void *y, bool do_backward, void *workSpace)
Execute a LRN forward layer.
miopenDestroyLRNDescriptor
miopenStatus_t miopenDestroyLRNDescriptor(miopenLRNDescriptor_t lrnDesc)
Destroys the LRN descriptor object.
miopenGetCTCLossWorkspaceSize
miopenStatus_t miopenGetCTCLossWorkspaceSize(miopenHandle_t handle, const miopenTensorDescriptor_t probsDesc, const miopenTensorDescriptor_t gradientsDesc, const int *labels, const int *labelLengths, const int *inputLengths, miopenCTCLossAlgo_t algo, const miopenCTCLossDescriptor_t ctcLossDesc, size_t *workSpaceSize)
Query the amount of memory required to execute miopenCTCLoss.
miopenGetSoftMarginLossForwardWorkspaceSize
miopenStatus_t miopenGetSoftMarginLossForwardWorkspaceSize(miopenHandle_t handle, miopenTensorDescriptor_t inputDesc, miopenTensorDescriptor_t targetDesc, miopenTensorDescriptor_t outputDesc, miopenLossReductionMode_t reduction, size_t *sizeInBytes)
Helper function to query the minimum workspace size required by the SoftMarginLossForward call.
miopenLossReductionMode_t
miopenLossReductionMode_t
Definition miopen.h:7838
miopenGetMultiMarginLossForwardWorkspaceSize
miopenStatus_t miopenGetMultiMarginLossForwardWorkspaceSize(miopenHandle_t handle, miopenTensorDescriptor_t inputDesc, miopenTensorDescriptor_t targetDesc, miopenTensorDescriptor_t weightDesc, miopenTensorDescriptor_t outputDesc, long p, float margin, miopenLossReductionMode_t reduction, size_t *sizeInBytes)
Helper function to query the minimum workspace size required by the MultiMarginLossForward call.
miopenCTCLossAlgo_t
miopenCTCLossAlgo_t
Definition miopen.h:5683
miopenSetCTCLossDescriptor
miopenStatus_t miopenSetCTCLossDescriptor(miopenCTCLossDescriptor_t ctcLossDesc, miopenDataType_t dataType, const int blank_label_id, bool apply_softmax_layer)
Set the details of a CTC loss function descriptor.
miopenCTCLoss
miopenStatus_t miopenCTCLoss(miopenHandle_t handle, const miopenTensorDescriptor_t probsDesc, const void *probs, const int *labels, const int *labelLengths, const int *inputLengths, void *losses, const miopenTensorDescriptor_t gradientsDesc, void *gradients, miopenCTCLossAlgo_t algo, const miopenCTCLossDescriptor_t ctcLossDesc, void *workSpace, size_t workSpaceSize)
Execute forward inference for CTCLoss layer.
miopenGetCTCLossDescriptor
miopenStatus_t miopenGetCTCLossDescriptor(miopenCTCLossDescriptor_t ctcLossDesc, miopenDataType_t *dataType, int *blank_label_id, bool *apply_softmax_layer)
Retrieves a CTC loss function descriptor's details.
miopenCreateCTCLossDescriptor
miopenStatus_t miopenCreateCTCLossDescriptor(miopenCTCLossDescriptor_t *ctcLossDesc)
Create a CTC loss function Descriptor.
miopenMultiMarginLossForward
miopenStatus_t miopenMultiMarginLossForward(miopenHandle_t handle, miopenTensorDescriptor_t inputDesc, const void *input, miopenTensorDescriptor_t targetDesc, const void *target, miopenTensorDescriptor_t weightDesc, const void *weight, miopenTensorDescriptor_t outputDesc, void *output, long p, float margin, miopenLossReductionMode_t reduction, void *workspace, size_t workspaceSizeInBytes)
Execute a MultiMarginLoss forward layer.
miopenSoftMarginLossBackward
miopenStatus_t miopenSoftMarginLossBackward(miopenHandle_t handle, miopenTensorDescriptor_t inputDesc, const void *input, miopenTensorDescriptor_t targetDesc, const void *target, miopenTensorDescriptor_t doutputDesc, const void *doutput, miopenTensorDescriptor_t dinputDesc, void *dinput, miopenLossReductionMode_t reduction)
Execute a SoftMarginLoss backward layer.
miopenSoftMarginLossForward
miopenStatus_t miopenSoftMarginLossForward(miopenHandle_t handle, miopenTensorDescriptor_t inputDesc, const void *input, miopenTensorDescriptor_t targetDesc, const void *target, miopenTensorDescriptor_t outputDesc, void *output, miopenLossReductionMode_t reduction, void *workspace=nullptr, size_t workspaceSizeInBytes=0)
Execute a SoftMarginLoss forward layer.
miopenDestroyCTCLossDescriptor
miopenStatus_t miopenDestroyCTCLossDescriptor(miopenCTCLossDescriptor_t ctcLossDesc)
Destroys a CTC loss function descriptor object.
MIOPEN_CTC_LOSS_ALGO_DETERMINISTIC
@ MIOPEN_CTC_LOSS_ALGO_DETERMINISTIC
Definition miopen.h:5684
miopenRNNMode_t
miopenRNNMode_t
Definition miopen.h:4288
miopenGetRNNDescriptor_V2
miopenStatus_t miopenGetRNNDescriptor_V2(miopenRNNDescriptor_t rnnDesc, int *hiddenSize, int *layer, miopenDropoutDescriptor_t *dropoutDesc, miopenRNNInputMode_t *inputMode, miopenRNNDirectionMode_t *dirMode, miopenRNNMode_t *rnnMode, miopenRNNBiasMode_t *biasMode, miopenRNNAlgo_t *algoMode, miopenDataType_t *dataType)
Retrieves a RNN layer descriptor's details version 2. This version enables retrieving information of ...
miopenRNNInputMode_t
miopenRNNInputMode_t
Definition miopen.h:4299
miopenSetRNNLayerBias
miopenStatus_t miopenSetRNNLayerBias(miopenHandle_t handle, miopenRNNDescriptor_t rnnDesc, const int layer, miopenTensorDescriptor_t xDesc, miopenTensorDescriptor_t wDesc, void *w, const int biasID, miopenTensorDescriptor_t biasDesc, const void *layerBias)
Sets a bias for a specific layer in an RNN stack.
miopenRNNBackwardWeightsSeqTensor
miopenStatus_t miopenRNNBackwardWeightsSeqTensor(miopenHandle_t handle, const miopenRNNDescriptor_t rnnDesc, const miopenSeqTensorDescriptor_t xDesc, const void *x, const miopenTensorDescriptor_t hDesc, const void *hx, const miopenSeqTensorDescriptor_t yDesc, const void *y, void *dw, size_t weightSpaceSize, void *workSpace, size_t workSpaceNumBytes, const void *reserveSpace, size_t reserveSpaceNumBytes)
Execute backward weights for recurrent layer.
miopenGetRNNParamsSize
miopenStatus_t miopenGetRNNParamsSize(miopenHandle_t handle, miopenRNNDescriptor_t rnnDesc, miopenTensorDescriptor_t xDesc, size_t *numBytes, miopenDataType_t dtype)
Query the amount of parameter memory required for RNN training.
miopenGetRNNLayerBiasOffset
miopenStatus_t miopenGetRNNLayerBiasOffset(miopenRNNDescriptor_t rnnDesc, const int layer, miopenTensorDescriptor_t xDesc, const int biasID, miopenTensorDescriptor_t biasDesc, size_t *layerBiasOffset)
Gets a bias index offset for a specific layer in an RNN stack.
miopenGetRNNLayerParamOffset
miopenStatus_t miopenGetRNNLayerParamOffset(miopenRNNDescriptor_t rnnDesc, const int layer, miopenTensorDescriptor_t xDesc, const int paramID, miopenTensorDescriptor_t paramDesc, size_t *layerParamOffset)
Gets an index offset for a specific weight matrix for a layer in the RNN stack.
miopenRNNDirectionMode_t
miopenRNNDirectionMode_t
Definition miopen.h:4321
miopenRNNBackwardSeqData
miopenStatus_t miopenRNNBackwardSeqData(miopenHandle_t handle, const miopenRNNDescriptor_t rnnDesc, const miopenSeqTensorDescriptor_t yDesc, const void *y, const void *dy, const miopenTensorDescriptor_t hDesc, const void *hx, const void *dhy, void *dhx, const miopenTensorDescriptor_t cDesc, const void *cx, const void *dcy, void *dcx, const miopenSeqTensorDescriptor_t xDesc, void *dx, const void *w, size_t weightSpaceSize, void *workSpace, size_t workSpaceNumBytes, void *reserveSpace, size_t reserveSpaceNumBytes)
Execute backward data for recurrent layer.
miopenRNNBiasMode_t
miopenRNNBiasMode_t
Definition miopen.h:4330
miopenGetRNNLayerBias
miopenStatus_t miopenGetRNNLayerBias(miopenHandle_t handle, miopenRNNDescriptor_t rnnDesc, const int layer, miopenTensorDescriptor_t xDesc, miopenTensorDescriptor_t wDesc, const void *w, const int biasID, miopenTensorDescriptor_t biasDesc, void *layerBias)
Gets a bias for a specific layer in an RNN stack.
miopenRNNBackwardWeights
miopenStatus_t miopenRNNBackwardWeights(miopenHandle_t handle, const miopenRNNDescriptor_t rnnDesc, const int sequenceLen, const miopenTensorDescriptor_t *xDesc, const void *x, const miopenTensorDescriptor_t hxDesc, const void *hx, const miopenTensorDescriptor_t *yDesc, const void *y, const miopenTensorDescriptor_t dwDesc, void *dw, void *workSpace, size_t workSpaceNumBytes, const void *reserveSpace, size_t reserveSpaceNumBytes)
Execute backward weights for recurrent layer.
miopenGetRNNTrainingReserveSize
miopenStatus_t miopenGetRNNTrainingReserveSize(miopenHandle_t handle, miopenRNNDescriptor_t rnnDesc, const int sequenceLen, const miopenTensorDescriptor_t *xDesc, size_t *numBytes)
Query the amount of memory required for RNN training.
miopenGetRNNLayerBiasSize
miopenStatus_t miopenGetRNNLayerBiasSize(miopenHandle_t handle, miopenRNNDescriptor_t rnnDesc, const int layer, const int biasID, size_t *numBytes)
Gets the number of bytes of a bias.
miopenSetRNNDescriptor
miopenStatus_t miopenSetRNNDescriptor(miopenRNNDescriptor_t rnnDesc, const int hsize, const int nlayers, miopenRNNInputMode_t inMode, miopenRNNDirectionMode_t direction, miopenRNNMode_t rnnMode, miopenRNNBiasMode_t biasMode, miopenRNNAlgo_t algo, miopenDataType_t dataType)
Set the details of the RNN descriptor.
miopenRNNBaseLayout_t
miopenRNNBaseLayout_t
Definition miopen.h:4365
miopenRNNBackwardData
miopenStatus_t miopenRNNBackwardData(miopenHandle_t handle, const miopenRNNDescriptor_t rnnDesc, const int sequenceLen, const miopenTensorDescriptor_t *yDesc, const void *y, const miopenTensorDescriptor_t *dyDesc, const void *dy, const miopenTensorDescriptor_t dhyDesc, const void *dhy, const miopenTensorDescriptor_t dcyDesc, const void *dcy, const miopenTensorDescriptor_t wDesc, const void *w, const miopenTensorDescriptor_t hxDesc, const void *hx, const miopenTensorDescriptor_t cxDesc, const void *cx, const miopenTensorDescriptor_t *dxDesc, void *dx, const miopenTensorDescriptor_t dhxDesc, void *dhx, const miopenTensorDescriptor_t dcxDesc, void *dcx, void *workSpace, size_t workSpaceNumBytes, void *reserveSpace, size_t reserveSpaceNumBytes)
Execute backward data for recurrent layer.
miopenGetRNNDescriptor
miopenStatus_t miopenGetRNNDescriptor(miopenRNNDescriptor_t rnnDesc, miopenRNNMode_t *rnnMode, miopenRNNAlgo_t *algoMode, miopenRNNInputMode_t *inputMode, miopenRNNDirectionMode_t *dirMode, miopenRNNBiasMode_t *biasMode, int *hiddenSize, int *layer)
Retrieves a RNN layer descriptor's details.
miopenRNNAlgo_t
miopenRNNAlgo_t
Definition miopen.h:4308
miopenGetRNNTempSpaceSizes
miopenStatus_t miopenGetRNNTempSpaceSizes(miopenHandle_t handle, miopenRNNDescriptor_t rnnDesc, miopenSeqTensorDescriptor_t xDesc, miopenRNNFWDMode_t fwdMode, size_t *workSpaceSize, size_t *reserveSpaceSize)
Query the amount of additional memory required for this RNN layer execution.
miopenGetRNNPaddingMode
miopenStatus_t miopenGetRNNPaddingMode(miopenRNNDescriptor_t rnnDesc, miopenRNNPaddingMode_t *paddingMode)
This function retrieves the RNN padding mode from the RNN descriptor.
miopenRNNForward
miopenStatus_t miopenRNNForward(miopenHandle_t handle, const miopenRNNDescriptor_t rnnDesc, miopenRNNFWDMode_t fwdMode, const miopenSeqTensorDescriptor_t xDesc, const void *x, const miopenTensorDescriptor_t hDesc, const void *hx, void *hy, const miopenTensorDescriptor_t cDesc, const void *cx, void *cy, const miopenSeqTensorDescriptor_t yDesc, void *y, const void *w, size_t weightSpaceSize, void *workSpace, size_t workSpaceNumBytes, void *reserveSpace, size_t reserveSpaceNumBytes)
Execute forward training for recurrent layer.
miopenGetRNNDataSeqTensorDescriptor
miopenStatus_t miopenGetRNNDataSeqTensorDescriptor(miopenSeqTensorDescriptor_t seqTensorDesc, miopenDataType_t *dataType, miopenRNNBaseLayout_t *layout, int *maxSequenceLen, int *batchSize, int *vectorSize, int sequenceLenArrayLimit, int *sequenceLenArray, void *paddingMarker)
Get shape of RNN seqData tensor.
miopenGetRNNInputTensorSize
miopenStatus_t miopenGetRNNInputTensorSize(miopenHandle_t handle, miopenRNNDescriptor_t rnnDesc, const int seqLen, miopenTensorDescriptor_t *xDesc, size_t *numBytes)
Obtain the size in bytes of the RNN input tensor.
miopenRNNFWDMode_t
miopenRNNFWDMode_t
Definition miopen.h:4356
miopenSetRNNLayerParam
miopenStatus_t miopenSetRNNLayerParam(miopenHandle_t handle, miopenRNNDescriptor_t rnnDesc, const int layer, miopenTensorDescriptor_t xDesc, miopenTensorDescriptor_t wDesc, void *w, const int paramID, miopenTensorDescriptor_t paramDesc, const void *layerParam)
Sets a weight matrix for a specific layer in an RNN stack.
miopenGetRNNLayerParamSize
miopenStatus_t miopenGetRNNLayerParamSize(miopenHandle_t handle, miopenRNNDescriptor_t rnnDesc, const int layer, miopenTensorDescriptor_t xDesc, const int paramID, size_t *numBytes)
Gets the number of bytes of a parameter matrix.
miopenCreateRNNDescriptor
miopenStatus_t miopenCreateRNNDescriptor(miopenRNNDescriptor_t *rnnDesc)
Create a RNN layer Descriptor.
miopenRNNGEMMalgoMode_t
miopenRNNGEMMalgoMode_t
Definition miopen.h:4339
miopenGetRNNLayerParam
miopenStatus_t miopenGetRNNLayerParam(miopenHandle_t handle, miopenRNNDescriptor_t rnnDesc, const int layer, miopenTensorDescriptor_t xDesc, miopenTensorDescriptor_t wDesc, const void *w, const int paramID, miopenTensorDescriptor_t paramDesc, void *layerParam)
Gets a weight matrix for a specific layer in an RNN stack.
miopenGetRNNWorkspaceSize
miopenStatus_t miopenGetRNNWorkspaceSize(miopenHandle_t handle, const miopenRNNDescriptor_t rnnDesc, const int sequenceLen, const miopenTensorDescriptor_t *xDesc, size_t *numBytes)
Query the amount of memory required to execute the RNN layer.
miopenSetRNNDataSeqTensorDescriptor
miopenStatus_t miopenSetRNNDataSeqTensorDescriptor(miopenSeqTensorDescriptor_t seqTensorDesc, miopenDataType_t dataType, miopenRNNBaseLayout_t layout, int maxSequenceLen, int batchSize, int vectorSize, const int *sequenceLenArray, void *paddingMarker)
Set shape of RNN seqData tensor.
miopenGetRNNParamsDescriptor
miopenStatus_t miopenGetRNNParamsDescriptor(miopenHandle_t handle, miopenRNNDescriptor_t rnnDesc, miopenTensorDescriptor_t xDesc, miopenTensorDescriptor_t wDesc, miopenDataType_t dtype)
Obtain a weight tensor descriptor for RNNs.
miopenRNNForwardTraining
miopenStatus_t miopenRNNForwardTraining(miopenHandle_t handle, const miopenRNNDescriptor_t rnnDesc, const int sequenceLen, const miopenTensorDescriptor_t *xDesc, const void *x, const miopenTensorDescriptor_t hxDesc, const void *hx, const miopenTensorDescriptor_t cxDesc, const void *cx, const miopenTensorDescriptor_t wDesc, const void *w, const miopenTensorDescriptor_t *yDesc, void *y, const miopenTensorDescriptor_t hyDesc, void *hy, const miopenTensorDescriptor_t cyDesc, void *cy, void *workSpace, size_t workSpaceNumBytes, void *reserveSpace, size_t reserveSpaceNumBytes)
Execute forward training for recurrent layer.
miopenRNNPaddingMode_t
miopenRNNPaddingMode_t
Definition miopen.h:4347
miopenSetRNNDescriptor_V2
miopenStatus_t miopenSetRNNDescriptor_V2(miopenRNNDescriptor_t rnnDesc, const int hsize, const int nlayers, miopenDropoutDescriptor_t dropoutDesc, miopenRNNInputMode_t inMode, miopenRNNDirectionMode_t direction, miopenRNNMode_t rnnMode, miopenRNNBiasMode_t biasMode, miopenRNNAlgo_t algo, miopenDataType_t dataType)
Set the details of the RNN descriptor version 2. This version enables the use of dropout in rnn.
miopenDestroyRNNDescriptor
miopenStatus_t miopenDestroyRNNDescriptor(miopenRNNDescriptor_t rnnDesc)
Destroys the tensor descriptor object.
miopenGetRNNHiddenTensorSize
miopenStatus_t miopenGetRNNHiddenTensorSize(miopenHandle_t handle, miopenRNNDescriptor_t rnnDesc, const int seqLen, miopenTensorDescriptor_t *xDesc, size_t *numBytes)
Obtain the size in bytes of the RNN hidden tensor.
miopenSetRNNPaddingMode
miopenStatus_t miopenSetRNNPaddingMode(miopenRNNDescriptor_t rnnDesc, miopenRNNPaddingMode_t paddingMode)
Sets a bias for a specific layer in an RNN stack.
miopenRNNForwardInference
miopenStatus_t miopenRNNForwardInference(miopenHandle_t handle, miopenRNNDescriptor_t rnnDesc, const int sequenceLen, const miopenTensorDescriptor_t *xDesc, const void *x, const miopenTensorDescriptor_t hxDesc, const void *hx, const miopenTensorDescriptor_t cxDesc, const void *cx, const miopenTensorDescriptor_t wDesc, const void *w, const miopenTensorDescriptor_t *yDesc, void *y, const miopenTensorDescriptor_t hyDesc, void *hy, const miopenTensorDescriptor_t cyDesc, void *cy, void *workSpace, size_t workSpaceNumBytes)
Execute forward inference for RNN layer.
miopenRNNTANH
@ miopenRNNTANH
Definition miopen.h:4290
miopenLSTM
@ miopenLSTM
Definition miopen.h:4291
miopenGRU
@ miopenGRU
Definition miopen.h:4292
miopenRNNRELU
@ miopenRNNRELU
Definition miopen.h:4289
miopenRNNlinear
@ miopenRNNlinear
Definition miopen.h:4300
miopenRNNskip
@ miopenRNNskip
Definition miopen.h:4301
miopenRNNbidirection
@ miopenRNNbidirection
Definition miopen.h:4323
miopenRNNunidirection
@ miopenRNNunidirection
Definition miopen.h:4322
miopenRNNwithBias
@ miopenRNNwithBias
Definition miopen.h:4332
miopenRNNNoBias
@ miopenRNNNoBias
Definition miopen.h:4331
miopenRNNDataSeqMajorPadded
@ miopenRNNDataSeqMajorPadded
Definition miopen.h:4368
miopenRNNDataSeqMajorNotPadded
@ miopenRNNDataSeqMajorNotPadded
Definition miopen.h:4367
miopenRNNDataBatchMajorPadded
@ miopenRNNDataBatchMajorPadded
Definition miopen.h:4369
miopenRNNDataUnknownLayout
@ miopenRNNDataUnknownLayout
Definition miopen.h:4366
miopenRNNroundedDynamic
@ miopenRNNroundedDynamic
Definition miopen.h:4313
miopenRNNfundamental
@ miopenRNNfundamental
Definition miopen.h:4311
miopenRNNdefault
@ miopenRNNdefault
Definition miopen.h:4309
miopenRNNTraining
@ miopenRNNTraining
Definition miopen.h:4357
miopenRNNInference
@ miopenRNNInference
Definition miopen.h:4358
miopenRNNAlgoGEMM
@ miopenRNNAlgoGEMM
Definition miopen.h:4340
miopenRNNIOWithPadding
@ miopenRNNIOWithPadding
Definition miopen.h:4349
miopenRNNIONotPadded
@ miopenRNNIONotPadded
Definition miopen.h:4348
miopenGetPReLUBackwardWorkspaceSize
miopenStatus_t miopenGetPReLUBackwardWorkspaceSize(miopenHandle_t handle, miopenTensorDescriptor_t inputDesc, miopenTensorDescriptor_t weightDesc, size_t *sizeInBytes)
Helper function to query the minimum workspace size required by the PReLU backward call.
miopenPReLUBackward
miopenStatus_t miopenPReLUBackward(miopenHandle_t handle, void *workspace, size_t workspaceSizeInBytes, miopenTensorDescriptor_t inputDesc, const void *input, miopenTensorDescriptor_t weightDesc, const void *weight, miopenTensorDescriptor_t doutputDesc, const void *doutput, miopenTensorDescriptor_t dinputDesc, void *dinput, miopenTensorDescriptor_t dweightDesc, void *dweight)
Execute a PReLU backward layer.
miopenReduceExtremeForward
miopenStatus_t miopenReduceExtremeForward(miopenHandle_t handle, const miopenTensorDescriptor_t xDesc, const void *x, const int32_t dim, const miopenReduceExtremeOp_t reduceExtremeOp, const miopenTensorDescriptor_t yDesc, void *y, const miopenTensorDescriptor_t indiceDesc, void *indice)
Find the the extreme (minimum, maximum) value and index of a tensor across Dimension.
miopenReduceExtremeOp_t
miopenReduceExtremeOp_t
Definition miopen.h:6727
miopenRoPEBackward
miopenStatus_t miopenRoPEBackward(miopenHandle_t handle, const miopenTensorDescriptor_t dyDesc, const void *dy, const miopenTensorDescriptor_t cosDesc, const void *cos, const miopenTensorDescriptor_t sinDesc, const void *sin, const miopenTensorDescriptor_t dxDesc, void *dx)
Execute a rope backward layer.
miopenRoPEForward
miopenStatus_t miopenRoPEForward(miopenHandle_t handle, const miopenTensorDescriptor_t xDesc, const void *x, const miopenTensorDescriptor_t cosDesc, const void *cos, const miopenTensorDescriptor_t sinDesc, const void *sin, const miopenTensorDescriptor_t yDesc, void *y)
Execute a rope forward layer.
miopenFusedAdam
miopenStatus_t miopenFusedAdam(miopenHandle_t handle, const miopenTensorDescriptor_t paramDesc, void *param, const miopenTensorDescriptor_t gradDesc, const void *grad, const miopenTensorDescriptor_t expAvgDesc, void *expAvg, const miopenTensorDescriptor_t expAvgSqDesc, void *expAvgSq, const miopenTensorDescriptor_t maxExpAvgSqDesc, void *maxExpAvgSq, const miopenTensorDescriptor_t stateStepDesc, void *stateStep, const unsigned int state_step, const float lr, const float beta1, const float beta2, const float weight_decay, const float eps, const bool amsgrad, const bool maximize, const bool adamw, const miopenTensorDescriptor_t gradScaleDesc, const void *gradScale, const miopenTensorDescriptor_t foundInfDesc, const void *foundInf)
Perform Fused Adam optimization for a single tensor (Adaptive Moment Estimation).
miopenTransformersAdamW
miopenStatus_t miopenTransformersAdamW(miopenHandle_t handle, const miopenTensorDescriptor_t paramDesc, void *param, const miopenTensorDescriptor_t gradDesc, const void *grad, const miopenTensorDescriptor_t expAvgDesc, void *expAvg, const miopenTensorDescriptor_t expAvgSqDesc, void *expAvgSq, const miopenTensorDescriptor_t stateStepDesc, void *stateStep, const unsigned int state_step, const float lr, const float beta1, const float beta2, const float weight_decay, const float eps, const bool correct_bias, const miopenTensorDescriptor_t gradScaleDesc, const void *gradScale, const miopenTensorDescriptor_t foundInfDesc, const void *foundInf)
Implements Adam algorithm with weight decay fix as introduced in Decoupled Weight Decay Regularizatio...
miopenFusedAdamWithOutput
miopenStatus_t miopenFusedAdamWithOutput(miopenHandle_t handle, const miopenTensorDescriptor_t paramInDesc, void *paramIn, const miopenTensorDescriptor_t paramOutDesc, void *paramOut, const miopenTensorDescriptor_t paramOutFloat16Desc, void *paramOutFloat16, const miopenTensorDescriptor_t gradInDesc, const void *gradIn, const miopenTensorDescriptor_t expAvgInDesc, void *expAvgIn, const miopenTensorDescriptor_t expAvgOutDesc, void *expAvgOut, const miopenTensorDescriptor_t expAvgSqInDesc, void *expAvgSqIn, const miopenTensorDescriptor_t expAvgSqOutDesc, void *expAvgSqOut, const miopenTensorDescriptor_t maxExpAvgSqInDesc, void *maxExpAvgSqIn, const miopenTensorDescriptor_t maxExpAvgSqOutDesc, void *maxExpAvgSqOut, const miopenTensorDescriptor_t stateStepInDesc, void *stateStepIn, const miopenTensorDescriptor_t stateStepOutDesc, void *stateStepOut, const unsigned int state_step, const float lr, const float beta1, const float beta2, const float weight_decay, const float eps, const bool amsgrad, const bool maximize, const bool adamw, const miopenTensorDescriptor_t gradScaleDesc, const void *gradScale, const miopenTensorDescriptor_t foundInfDesc, const void *foundInf)
Execute single tensor Adam optimization and receive the result in a separate output tensor.
miopenTransformersAdamWWithOutput
miopenStatus_t miopenTransformersAdamWWithOutput(miopenHandle_t handle, const miopenTensorDescriptor_t paramInDesc, void *paramIn, const miopenTensorDescriptor_t paramOutDesc, void *paramOut, const miopenTensorDescriptor_t paramOutFloat16Desc, void *paramOutFloat16, const miopenTensorDescriptor_t gradInDesc, const void *gradIn, const miopenTensorDescriptor_t expAvgInDesc, void *expAvgIn, const miopenTensorDescriptor_t expAvgOutDesc, void *expAvgOut, const miopenTensorDescriptor_t expAvgSqInDesc, void *expAvgSqIn, const miopenTensorDescriptor_t expAvgSqOutDesc, void *expAvgSqOut, const miopenTensorDescriptor_t stateStepInDesc, void *stateStepIn, const miopenTensorDescriptor_t stateStepOutDesc, void *stateStepOut, const unsigned int state_step, const float lr, const float beta1, const float beta2, const float weight_decay, const float eps, const float step_size, const bool correct_bias, const miopenTensorDescriptor_t gradScaleDesc, const void *gradScale, const miopenTensorDescriptor_t foundInfDesc, const void *foundInf)
Execute single tensor Adam optimization and receive the result in a separate output tensor.
miopenReduceTensorIndices_t
miopenReduceTensorIndices_t
Definition miopen.h:589
miopenSetReduceTensorDescriptor
miopenStatus_t miopenSetReduceTensorDescriptor(miopenReduceTensorDescriptor_t reduceTensorDesc, miopenReduceTensorOp_t reduceTensorOp, miopenDataType_t reduceTensorCompType, miopenNanPropagation_t reduceTensorNanOpt, miopenReduceTensorIndices_t reduceTensorIndices, miopenIndicesType_t reduceTensorIndicesType)
Initialize a ReduceTensor descriptor object.
miopenCreateReduceTensorDescriptor
miopenStatus_t miopenCreateReduceTensorDescriptor(miopenReduceTensorDescriptor_t *reduceTensorDesc)
Creates the ReduceTensor descriptor object.
miopenReduceTensorOp_t
miopenReduceTensorOp_t
Definition miopen.h:554
miopenIndicesType_t
miopenIndicesType_t
Definition miopen.h:599
miopenReduceTensor
miopenStatus_t miopenReduceTensor(miopenHandle_t handle, const miopenReduceTensorDescriptor_t reduceTensorDesc, void *indices, size_t indicesSizeInBytes, void *workspace, size_t workspaceSizeInBytes, const void *alpha, const miopenTensorDescriptor_t aDesc, const void *A, const void *beta, const miopenTensorDescriptor_t cDesc, void *C)
TensorReduce function doing reduction on tensor A by implementing C = alpha * reduceOp(A).
miopenGetReductionIndicesSize
miopenStatus_t miopenGetReductionIndicesSize(miopenHandle_t handle, const miopenReduceTensorDescriptor_t reduceTensorDesc, const miopenTensorDescriptor_t aDesc, const miopenTensorDescriptor_t cDesc, size_t *sizeInBytes)
Helper function to query the minimum index space size required by the ReduceTensor call.
miopenDestroyReduceTensorDescriptor
miopenStatus_t miopenDestroyReduceTensorDescriptor(miopenReduceTensorDescriptor_t reduceTensorDesc)
Destroy the ReduceTensor descriptor object.
miopenGetReductionWorkspaceSize
miopenStatus_t miopenGetReductionWorkspaceSize(miopenHandle_t handle, const miopenReduceTensorDescriptor_t reduceTensorDesc, const miopenTensorDescriptor_t aDesc, const miopenTensorDescriptor_t cDesc, size_t *sizeInBytes)
Helper function to query the minimum workspace size required by the ReduceTensor call.
miopenGetReduceTensorDescriptor
miopenStatus_t miopenGetReduceTensorDescriptor(const miopenReduceTensorDescriptor_t reduceTensorDesc, miopenReduceTensorOp_t *reduceTensorOp, miopenDataType_t *reduceTensorCompType, miopenNanPropagation_t *reduceTensorNanOpt, miopenReduceTensorIndices_t *reduceTensorIndices, miopenIndicesType_t *reduceTensorIndicesType)
Query a ReduceTensor descriptor object.
miopenActivationForward
miopenStatus_t miopenActivationForward(miopenHandle_t handle, const miopenActivationDescriptor_t activDesc, const void *alpha, const miopenTensorDescriptor_t xDesc, const void *x, const void *beta, const miopenTensorDescriptor_t yDesc, void *y)
Execute an activation forward layer.
miopenGLUBackward
miopenStatus_t miopenGLUBackward(miopenHandle_t handle, const miopenTensorDescriptor_t inputDesc, const void *input, const miopenTensorDescriptor_t outputGradDesc, const void *outputGrad, const miopenTensorDescriptor_t inputGradDesc, void *inputGrad, const uint32_t dim)
Execute a GLU backward layer.
miopenActivationBackward
miopenStatus_t miopenActivationBackward(miopenHandle_t handle, const miopenActivationDescriptor_t activDesc, const void *alpha, const miopenTensorDescriptor_t yDesc, const void *y, const miopenTensorDescriptor_t dyDesc, const void *dy, const miopenTensorDescriptor_t xDesc, const void *x, const void *beta, const miopenTensorDescriptor_t dxDesc, void *dx)
Execute a activation backwards layer.
miopenActivationMode_t
miopenActivationMode_t
Definition miopen.h:500
miopenSetActivationDescriptor
miopenStatus_t miopenSetActivationDescriptor(const miopenActivationDescriptor_t activDesc, miopenActivationMode_t mode, double activAlpha, double activBeta, double activGamma)
Sets the activation layer descriptor details.
miopenCreateActivationDescriptor
miopenStatus_t miopenCreateActivationDescriptor(miopenActivationDescriptor_t *activDesc)
Creates the Activation descriptor object.
miopenGetActivationDescriptor
miopenStatus_t miopenGetActivationDescriptor(const miopenActivationDescriptor_t activDesc, miopenActivationMode_t *mode, double *activAlpha, double *activBeta, double *activGamma)
Gets the activation layer descriptor details.
miopenGLUForward
miopenStatus_t miopenGLUForward(miopenHandle_t handle, const miopenTensorDescriptor_t inputDesc, const void *input, const miopenTensorDescriptor_t outputDesc, void *output, const uint32_t dim)
Execute a GLU forward layer.
miopenDestroyActivationDescriptor
miopenStatus_t miopenDestroyActivationDescriptor(miopenActivationDescriptor_t activDesc)
Destroys the activation descriptor object.
miopenBatchNormalizationForwardInference
miopenStatus_t miopenBatchNormalizationForwardInference(miopenHandle_t handle, miopenBatchNormMode_t bn_mode, void *alpha, void *beta, const miopenTensorDescriptor_t xDesc, const void *x, const miopenTensorDescriptor_t yDesc, void *y, const miopenTensorDescriptor_t bnScaleBiasMeanVarDesc, void *bnScale, void *bnBias, void *estimatedMean, void *estimatedVariance, double epsilon)
Execute forward inference layer for batch normalization.
miopenBatchNormalizationForwardInference_V2
miopenStatus_t miopenBatchNormalizationForwardInference_V2(miopenHandle_t handle, miopenBatchNormMode_t bn_mode, void *alpha, void *beta, const miopenTensorDescriptor_t xDesc, const void *x, const miopenTensorDescriptor_t yDesc, void *y, const miopenTensorDescriptor_t scaleDesc, const miopenTensorDescriptor_t biasDesc, const miopenTensorDescriptor_t estMeanDesc, const miopenTensorDescriptor_t estVarianceDesc, void *bnScale, void *bnBias, void *estimatedMean, void *estimatedVariance, double epsilon)
Execute forward inference layer for batch normalization.
miopenBatchNormForwardTrainingActivation
miopenStatus_t miopenBatchNormForwardTrainingActivation(miopenHandle_t handle, miopenBatchNormMode_t bn_mode, void *alpha, void *beta, const miopenTensorDescriptor_t xDesc, const void *x, const miopenTensorDescriptor_t yDesc, void *y, const miopenTensorDescriptor_t scaleDesc, const miopenTensorDescriptor_t biasVarDesc, const miopenTensorDescriptor_t savedMeanDesc, const miopenTensorDescriptor_t savedVarDesc, void *bnScale, void *bnBias, double expAvgFactor, void *resultRunningMean, void *resultRunningVariance, double epsilon, void *resultSaveMean, void *resultSaveInvVariance, const miopenActivationDescriptor_t activDesc)
Execute forward training layer for batch normalization with fused activation.
miopenBatchNormBackwardActivation
miopenStatus_t miopenBatchNormBackwardActivation(miopenHandle_t handle, miopenBatchNormMode_t bn_mode, const void *alphaDataDiff, const void *betaDataDiff, const void *alphaParamDiff, const void *betaParamDiff, const miopenTensorDescriptor_t xDesc, const void *x, const miopenTensorDescriptor_t dyDesc, const void *dy, const miopenTensorDescriptor_t dxDesc, void *dx, const miopenTensorDescriptor_t scaleDesc, const miopenTensorDescriptor_t biasDesc, const miopenTensorDescriptor_t savedMeanDesc, const miopenTensorDescriptor_t savedVarianceDesc, const void *bnScale, const void *bnBias, void *resultBnScaleDiff, void *resultBnBiasDiff, double epsilon, const void *savedMean, const void *savedInvVariance, const miopenActivationDescriptor_t activDesc)
Execute backwards propagation layer for batch normalization with fused activation.
miopenBatchNormalizationForwardTraining_V3
miopenStatus_t miopenBatchNormalizationForwardTraining_V3(miopenHandle_t handle, miopenBatchNormMode_t bn_mode, void *alpha, void *beta, const miopenTensorDescriptor_t xDesc, const void *x, const miopenTensorDescriptor_t yDesc, void *y, const miopenTensorDescriptor_t scaleDesc, const miopenTensorDescriptor_t biasVarDesc, const miopenTensorDescriptor_t savedMeanDesc, const miopenTensorDescriptor_t savedVarDesc, void *bnScale, void *bnBias, double expAvgFactor, const void *prevResultRunningMean, const void *prevResultRunningVariance, void *nextResultRunningMean, void *nextResultRunningVariance, double epsilon, void *resultSaveMean, void *resultSaveInvVariance)
Execute forward training layer for batch normalization.
miopenBatchNormalizationForwardTraining
miopenStatus_t miopenBatchNormalizationForwardTraining(miopenHandle_t handle, miopenBatchNormMode_t bn_mode, void *alpha, void *beta, const miopenTensorDescriptor_t xDesc, const void *x, const miopenTensorDescriptor_t yDesc, void *y, const miopenTensorDescriptor_t bnScaleBiasMeanVarDesc, void *bnScale, void *bnBias, double expAvgFactor, void *resultRunningMean, void *resultRunningVariance, double epsilon, void *resultSaveMean, void *resultSaveInvVariance)
Execute forward training layer for batch normalization.
miopenBatchNormalizationForwardTraining_V2
miopenStatus_t miopenBatchNormalizationForwardTraining_V2(miopenHandle_t handle, miopenBatchNormMode_t bn_mode, void *alpha, void *beta, const miopenTensorDescriptor_t xDesc, const void *x, const miopenTensorDescriptor_t yDesc, void *y, const miopenTensorDescriptor_t scaleDesc, const miopenTensorDescriptor_t biasVarDesc, const miopenTensorDescriptor_t savedMeanDesc, const miopenTensorDescriptor_t savedVarDesc, void *bnScale, void *bnBias, double expAvgFactor, void *resultRunningMean, void *resultRunningVariance, double epsilon, void *resultSaveMean, void *resultSaveInvVariance)
Execute forward training layer for batch normalization.
miopenBatchNormalizationBackward_V2
miopenStatus_t miopenBatchNormalizationBackward_V2(miopenHandle_t handle, miopenBatchNormMode_t bn_mode, const void *alphaDataDiff, const void *betaDataDiff, const void *alphaParamDiff, const void *betaParamDiff, const miopenTensorDescriptor_t xDesc, const void *x, const miopenTensorDescriptor_t dyDesc, const void *dy, const miopenTensorDescriptor_t dxDesc, void *dx, const miopenTensorDescriptor_t scaleDesc, const miopenTensorDescriptor_t biasDesc, const miopenTensorDescriptor_t savedMeanDesc, const miopenTensorDescriptor_t savedVarDesc, const void *bnScale, void *resultBnScaleDiff, void *resultBnBiasDiff, double epsilon, const void *savedMean, const void *savedInvVariance)
Execute backwards propagation layer for batch normalization.
miopenBatchNormForwardTrainingActivation_V2
miopenStatus_t miopenBatchNormForwardTrainingActivation_V2(miopenHandle_t handle, miopenBatchNormMode_t bn_mode, void *alpha, void *beta, const miopenTensorDescriptor_t xDesc, const void *x, const miopenTensorDescriptor_t yDesc, void *y, const miopenTensorDescriptor_t scaleDesc, const miopenTensorDescriptor_t biasVarDesc, const miopenTensorDescriptor_t savedMeanDesc, const miopenTensorDescriptor_t savedVarDesc, void *bnScale, void *bnBias, double expAvgFactor, const void *prevResultRunningMean, const void *prevResultRunningVariance, void *nextResultRunningMean, void *nextResultRunningVariance, double epsilon, void *resultSaveMean, void *resultSaveInvVariance, const miopenActivationDescriptor_t activDesc)
Execute forward training layer for batch normalization with fused activation.
miopenBatchNormalizationForwardInferenceInvVariance
miopenStatus_t miopenBatchNormalizationForwardInferenceInvVariance(miopenHandle_t handle, miopenBatchNormMode_t bn_mode, void *alpha, void *beta, const miopenTensorDescriptor_t xDesc, const void *x, const miopenTensorDescriptor_t yDesc, void *y, const miopenTensorDescriptor_t scaleDesc, const miopenTensorDescriptor_t biasDesc, const miopenTensorDescriptor_t estMeanDesc, const miopenTensorDescriptor_t estInvVarianceDesc, void *bnScale, void *bnBias, void *estimatedMean, void *estimatedInvVariance)
Execute forward inference layer for batch normalization using inverse variance.
miopenBatchNormalizationBackward
miopenStatus_t miopenBatchNormalizationBackward(miopenHandle_t handle, miopenBatchNormMode_t bn_mode, const void *alphaDataDiff, const void *betaDataDiff, const void *alphaParamDiff, const void *betaParamDiff, const miopenTensorDescriptor_t xDesc, const void *x, const miopenTensorDescriptor_t dyDesc, const void *dy, const miopenTensorDescriptor_t dxDesc, void *dx, const miopenTensorDescriptor_t bnScaleBiasDiffDesc, const void *bnScale, void *resultBnScaleDiff, void *resultBnBiasDiff, double epsilon, const void *savedMean, const void *savedInvVariance)
Execute backwards propagation layer for batch normalization.
miopenBatchNormMode_t
miopenBatchNormMode_t
Definition miopen.h:490
miopenBatchNormForwardInferenceActivation
miopenStatus_t miopenBatchNormForwardInferenceActivation(miopenHandle_t handle, miopenBatchNormMode_t bn_mode, void *alpha, void *beta, const miopenTensorDescriptor_t xDesc, const void *x, const miopenTensorDescriptor_t yDesc, void *y, const miopenTensorDescriptor_t scaleDesc, const miopenTensorDescriptor_t biasDesc, const miopenTensorDescriptor_t estMeanDesc, const miopenTensorDescriptor_t estVarianceDesc, void *bnScale, void *bnBias, void *estimatedMean, void *estimatedVariance, double epsilon, const miopenActivationDescriptor_t activDesc)
Execute forward inference layer for batch normalization with fused activation.
miopenBatchNormForwardInferenceActivationInvVariance
miopenStatus_t miopenBatchNormForwardInferenceActivationInvVariance(miopenHandle_t handle, miopenBatchNormMode_t bn_mode, void *alpha, void *beta, const miopenTensorDescriptor_t xDesc, const void *x, const miopenTensorDescriptor_t yDesc, void *y, const miopenTensorDescriptor_t scaleDesc, const miopenTensorDescriptor_t biasDesc, const miopenTensorDescriptor_t estMeanDesc, const miopenTensorDescriptor_t estInvVarianceDesc, void *bnScale, void *bnBias, void *estimatedMean, void *estimatedInvVariance, const miopenActivationDescriptor_t activDesc)
Execute forward inference layer for batch normalization with fused activation using inverse variance.
miopenDeriveBNTensorDescriptor
miopenStatus_t miopenDeriveBNTensorDescriptor(miopenTensorDescriptor_t derivedBnDesc, const miopenTensorDescriptor_t xDesc, miopenBatchNormMode_t bn_mode)
Derive tensor for gamma and beta from input tensor descriptor.
miopenCatForward
miopenStatus_t miopenCatForward(miopenHandle_t handle, const int32_t xCount, const miopenTensorDescriptor_t *xDescs, const void *const *xs, const miopenTensorDescriptor_t yDesc, void *y, const int32_t dim)
Execute a cat forward layer.
miopenInitConvolutionNdDescriptor
miopenStatus_t miopenInitConvolutionNdDescriptor(miopenConvolutionDescriptor_t convDesc, int spatialDim, const int *padA, const int *strideA, const int *dilationA, miopenConvolutionMode_t c_mode)
Creates a N-dimensional convolution layer descriptor.
miopenConvolutionForwardImmediate
miopenStatus_t miopenConvolutionForwardImmediate(miopenHandle_t handle, const miopenTensorDescriptor_t wDesc, const void *w, const miopenTensorDescriptor_t xDesc, const void *x, const miopenConvolutionDescriptor_t convDesc, const miopenTensorDescriptor_t yDesc, void *y, void *workSpace, size_t workSpaceSize, const uint64_t solution_id)
Executes the Forward convolution operation based on the provided solution ID.
miopenConvolutionBackwardDataCompileSolution
miopenStatus_t miopenConvolutionBackwardDataCompileSolution(miopenHandle_t handle, const miopenTensorDescriptor_t dyDesc, const miopenTensorDescriptor_t wDesc, const miopenConvolutionDescriptor_t convDesc, const miopenTensorDescriptor_t dxDesc, const uint64_t solution_id)
Compiles the solution provided by the user, this solution may be acquired by the miopenConvolutionBac...
miopenConvolutionForwardGetSolutionWorkspaceSize
miopenStatus_t miopenConvolutionForwardGetSolutionWorkspaceSize(miopenHandle_t handle, const miopenTensorDescriptor_t wDesc, const miopenTensorDescriptor_t xDesc, const miopenConvolutionDescriptor_t convDesc, const miopenTensorDescriptor_t yDesc, const uint64_t solution_id, size_t *workSpaceSize)
Returns the workspace size required for a particular solution id.
miopenConvolutionForward
miopenStatus_t miopenConvolutionForward(miopenHandle_t handle, const void *alpha, const miopenTensorDescriptor_t xDesc, const void *x, const miopenTensorDescriptor_t wDesc, const void *w, const miopenConvolutionDescriptor_t convDesc, miopenConvFwdAlgorithm_t algo, const void *beta, const miopenTensorDescriptor_t yDesc, void *y, void *workSpace, size_t workSpaceSize)
Execute a forward convolution layer.
miopenGetConvolutionGroupCount
miopenStatus_t miopenGetConvolutionGroupCount(miopenConvolutionDescriptor_t convDesc, int *groupCount)
Get the number of groups to be used in Group/Depthwise convolution.
miopenGetConvolutionFindMode
miopenStatus_t miopenGetConvolutionFindMode(const miopenConvolutionDescriptor_t convDesc, miopenConvolutionFindMode_t *findMode)
Reads the Find Mode attribute from the convolution descriptor.
miopenInitConvolutionDescriptor
miopenStatus_t miopenInitConvolutionDescriptor(miopenConvolutionDescriptor_t convDesc, miopenConvolutionMode_t c_mode, int pad_h, int pad_w, int stride_h, int stride_w, int dilation_h, int dilation_w)
Creates a 2-D convolution layer descriptor.
miopenConvAlgorithm_t
miopenConvAlgorithm_t
Definition miopen.h:1254
miopenConvolutionBackwardData
miopenStatus_t miopenConvolutionBackwardData(miopenHandle_t handle, const void *alpha, const miopenTensorDescriptor_t dyDesc, const void *dy, const miopenTensorDescriptor_t wDesc, const void *w, const miopenConvolutionDescriptor_t convDesc, miopenConvBwdDataAlgorithm_t algo, const void *beta, const miopenTensorDescriptor_t dxDesc, void *dx, void *workSpace, size_t workSpaceSize)
Execute a backward data convolution layer.
miopenGetConvolutionAttribute
miopenStatus_t miopenGetConvolutionAttribute(miopenConvolutionDescriptor_t convDesc, const miopenConvolutionAttrib_t attr, int *value)
Get the attribute of the convolution descriptor.
miopenGetConvolutionNdForwardOutputDim
miopenStatus_t miopenGetConvolutionNdForwardOutputDim(miopenConvolutionDescriptor_t convDesc, const miopenTensorDescriptor_t inputTensorDesc, const miopenTensorDescriptor_t filterDesc, int *nDim, int *outputTensorDimA)
Get the shape of a resulting N-dimensional tensor from a (N-2)-dimensional convolution.
miopenConvolutionBackwardWeightsCompileSolution
miopenStatus_t miopenConvolutionBackwardWeightsCompileSolution(miopenHandle_t handle, const miopenTensorDescriptor_t dyDesc, const miopenTensorDescriptor_t xDesc, const miopenConvolutionDescriptor_t convDesc, const miopenTensorDescriptor_t dwDesc, const uint64_t solution_id)
Compiles the solution provided by the user, this solution may be acquired by the miopenConvolutionBac...
miopenFindConvolutionBackwardWeightsAlgorithm
miopenStatus_t miopenFindConvolutionBackwardWeightsAlgorithm(miopenHandle_t handle, const miopenTensorDescriptor_t dyDesc, const void *dy, const miopenTensorDescriptor_t xDesc, const void *x, const miopenConvolutionDescriptor_t convDesc, const miopenTensorDescriptor_t dwDesc, void *dw, const int requestAlgoCount, int *returnedAlgoCount, miopenConvAlgoPerf_t *perfResults, void *workSpace, size_t workSpaceSize, bool exhaustiveSearch)
Search and run the backwards weights convolutional algorithms and return a list of kernel times.
miopenGetConvolutionForwardOutputDim
miopenStatus_t miopenGetConvolutionForwardOutputDim(miopenConvolutionDescriptor_t convDesc, const miopenTensorDescriptor_t inputTensorDesc, const miopenTensorDescriptor_t filterDesc, int *n, int *c, int *h, int *w)
Get the shape of a resulting 4-D tensor from a 2-D convolution.
miopenConvFwdAlgorithm_t
miopenConvFwdAlgorithm_t
Definition miopen.h:1217
miopenConvolutionForwardGetWorkSpaceSize
miopenStatus_t miopenConvolutionForwardGetWorkSpaceSize(miopenHandle_t handle, const miopenTensorDescriptor_t wDesc, const miopenTensorDescriptor_t xDesc, const miopenConvolutionDescriptor_t convDesc, const miopenTensorDescriptor_t yDesc, size_t *workSpaceSize)
Query the workspace size required for a forward convolution algorithm.
miopenCreateConvolutionDescriptor
miopenStatus_t miopenCreateConvolutionDescriptor(miopenConvolutionDescriptor_t *convDesc)
Creates a convolution layer descriptor.
miopenConvolutionBackwardWeightsGetSolution
miopenStatus_t miopenConvolutionBackwardWeightsGetSolution(miopenHandle_t handle, const miopenTensorDescriptor_t dyDesc, const miopenTensorDescriptor_t xDesc, const miopenConvolutionDescriptor_t convDesc, const miopenTensorDescriptor_t dwDesc, const size_t maxSolutionCount, size_t *solutionCount, miopenConvSolution_t *solutions)
Query the applicable solutions for a backward convolution w-r-t weights as described by input,...
miopenGetConvolutionNdDescriptor
miopenStatus_t miopenGetConvolutionNdDescriptor(miopenConvolutionDescriptor_t convDesc, int requestedSpatialDim, int *spatialDim, int *padA, int *strideA, int *dilationA, miopenConvolutionMode_t *c_mode)
Retrieves a N-dimensional convolution layer descriptor's details.
miopenConvBwdDataAlgorithm_t
miopenConvBwdDataAlgorithm_t
Definition miopen.h:1240
miopenConvolutionAttrib_t
miopenConvolutionAttrib_t
Definition miopen.h:611
miopenConvolutionBackwardWeightsGetWorkSpaceSize
miopenStatus_t miopenConvolutionBackwardWeightsGetWorkSpaceSize(miopenHandle_t handle, const miopenTensorDescriptor_t dyDesc, const miopenTensorDescriptor_t xDesc, const miopenConvolutionDescriptor_t convDesc, const miopenTensorDescriptor_t dwDesc, size_t *workSpaceSize)
Get the GPU memory required for the backward weights convolution algorithm.
miopenConvolutionBackwardWeightsImmediate
miopenStatus_t miopenConvolutionBackwardWeightsImmediate(miopenHandle_t handle, const miopenTensorDescriptor_t dyDesc, const void *dy, const miopenTensorDescriptor_t xDesc, const void *x, const miopenConvolutionDescriptor_t convDesc, const miopenTensorDescriptor_t dwDesc, void *dw, void *workSpace, size_t workSpaceSize, const uint64_t solution_id)
Executes the Backward convolution w-r-t weights operation based on the provided solution ID.
miopenFindConvolutionBackwardDataAlgorithm
miopenStatus_t miopenFindConvolutionBackwardDataAlgorithm(miopenHandle_t handle, const miopenTensorDescriptor_t dyDesc, const void *dy, const miopenTensorDescriptor_t wDesc, const void *w, const miopenConvolutionDescriptor_t convDesc, const miopenTensorDescriptor_t dxDesc, void *dx, const int requestAlgoCount, int *returnedAlgoCount, miopenConvAlgoPerf_t *perfResults, void *workSpace, size_t workSpaceSize, bool exhaustiveSearch)
Search and run the backwards data convolution algorithms and return a list of kernel times.
miopenConvolutionBackwardDataGetSolution
miopenStatus_t miopenConvolutionBackwardDataGetSolution(miopenHandle_t handle, const miopenTensorDescriptor_t dyDesc, const miopenTensorDescriptor_t wDesc, const miopenConvolutionDescriptor_t convDesc, const miopenTensorDescriptor_t dxDesc, const size_t maxSolutionCount, size_t *solutionCount, miopenConvSolution_t *solutions)
Query the applicable solutions for a backward convolution w-r-t data as described by input,...
miopenConvolutionForwardBias
miopenStatus_t miopenConvolutionForwardBias(miopenHandle_t handle, const void *alpha, const miopenTensorDescriptor_t bDesc, const void *b, const void *beta, const miopenTensorDescriptor_t yDesc, void *y)
Calculate element-wise scale and shift of a tensor via a bias tensor.
miopenConvolutionMode_t
miopenConvolutionMode_t
Definition miopen.h:415
miopenSetConvolutionAttribute
miopenStatus_t miopenSetConvolutionAttribute(miopenConvolutionDescriptor_t convDesc, const miopenConvolutionAttrib_t attr, int value)
Set the attribute of the convolution descriptor.
miopenConvolutionBackwardDataImmediate
miopenStatus_t miopenConvolutionBackwardDataImmediate(miopenHandle_t handle, const miopenTensorDescriptor_t dyDesc, const void *dy, const miopenTensorDescriptor_t wDesc, const void *w, const miopenConvolutionDescriptor_t convDesc, const miopenTensorDescriptor_t dxDesc, void *dx, void *workSpace, size_t workSpaceSize, const uint64_t solution_id)
Executes the Backward convolution w-r-t data operation based on the provided solution ID.
miopenSetTransposeConvNdOutputPadding
miopenStatus_t miopenSetTransposeConvNdOutputPadding(miopenConvolutionDescriptor_t convDesc, int spatialDim, const int *adjA)
Set the output padding to be used in N-dimensional Transpose convolution.
miopenSetConvolutionFindMode
miopenStatus_t miopenSetConvolutionFindMode(miopenConvolutionDescriptor_t convDesc, miopenConvolutionFindMode_t findMode)
Sets the Find Mode attribute in the convolution descriptor.
miopenSetTransposeConvOutputPadding
miopenStatus_t miopenSetTransposeConvOutputPadding(miopenConvolutionDescriptor_t convDesc, int adj_h, int adj_w)
Set the output padding to be used in 2-D Transpose convolution.
miopenConvolutionFindMode_t
miopenConvolutionFindMode_t
Definition miopen.h:640
miopenGetConvolutionSpatialDim
miopenStatus_t miopenGetConvolutionSpatialDim(miopenConvolutionDescriptor_t convDesc, int *spatialDim)
Retrieves the spatial dimension of a convolution layer descriptor.
miopenConvBwdWeightsAlgorithm_t
miopenConvBwdWeightsAlgorithm_t
Definition miopen.h:1229
miopenConvolutionForwardCompileSolution
miopenStatus_t miopenConvolutionForwardCompileSolution(miopenHandle_t handle, const miopenTensorDescriptor_t wDesc, const miopenTensorDescriptor_t xDesc, const miopenConvolutionDescriptor_t convDesc, const miopenTensorDescriptor_t yDesc, const uint64_t solution_id)
Compiles the solution provided by the user, this solution may be acquired by the miopenConvolutionFor...
miopenConvolutionBackwardDataGetSolutionCount
miopenStatus_t miopenConvolutionBackwardDataGetSolutionCount(miopenHandle_t handle, const miopenTensorDescriptor_t dyDesc, const miopenTensorDescriptor_t wDesc, const miopenConvolutionDescriptor_t convDesc, const miopenTensorDescriptor_t dxDesc, size_t *solutionCount)
Query the maximum number of solutions applicable for the given input/output and weights tensor descri...
miopenConvolutionBackwardWeights
miopenStatus_t miopenConvolutionBackwardWeights(miopenHandle_t handle, const void *alpha, const miopenTensorDescriptor_t dyDesc, const void *dy, const miopenTensorDescriptor_t xDesc, const void *x, const miopenConvolutionDescriptor_t convDesc, miopenConvBwdWeightsAlgorithm_t algo, const void *beta, const miopenTensorDescriptor_t dwDesc, void *dw, void *workSpace, size_t workSpaceSize)
Execute a backward weights convolution layer.
miopenConvolutionBackwardDataGetWorkSpaceSize
miopenStatus_t miopenConvolutionBackwardDataGetWorkSpaceSize(miopenHandle_t handle, const miopenTensorDescriptor_t dyDesc, const miopenTensorDescriptor_t wDesc, const miopenConvolutionDescriptor_t convDesc, const miopenTensorDescriptor_t dxDesc, size_t *workSpaceSize)
Query the workspace size required for a backward data convolution algorithm.
miopenFindConvolutionForwardAlgorithm
miopenStatus_t miopenFindConvolutionForwardAlgorithm(miopenHandle_t handle, const miopenTensorDescriptor_t xDesc, const void *x, const miopenTensorDescriptor_t wDesc, const void *w, const miopenConvolutionDescriptor_t convDesc, const miopenTensorDescriptor_t yDesc, void *y, const int requestAlgoCount, int *returnedAlgoCount, miopenConvAlgoPerf_t *perfResults, void *workSpace, size_t workSpaceSize, bool exhaustiveSearch)
Search and run the forward convolutional algorithms and return a list of kernel times.
miopenDestroyConvolutionDescriptor
miopenStatus_t miopenDestroyConvolutionDescriptor(miopenConvolutionDescriptor_t convDesc)
Destroys the tensor descriptor object.
miopenConvolutionBackwardWeightsGetSolutionCount
miopenStatus_t miopenConvolutionBackwardWeightsGetSolutionCount(miopenHandle_t handle, const miopenTensorDescriptor_t dyDesc, const miopenTensorDescriptor_t xDesc, const miopenConvolutionDescriptor_t convDesc, const miopenTensorDescriptor_t dwDesc, size_t *solutionCount)
Query the maximum number of solutions applicable for the given input/output and weights tensor descri...
miopenSetConvolutionGroupCount
miopenStatus_t miopenSetConvolutionGroupCount(miopenConvolutionDescriptor_t convDesc, int groupCount)
Set the number of groups to be used in Group/Depthwise convolution.
miopenConvolutionBackwardWeightsGetSolutionWorkspaceSize
miopenStatus_t miopenConvolutionBackwardWeightsGetSolutionWorkspaceSize(miopenHandle_t handle, const miopenTensorDescriptor_t dyDesc, const miopenTensorDescriptor_t xDesc, const miopenConvolutionDescriptor_t convDesc, const miopenTensorDescriptor_t dwDesc, const uint64_t solution_id, size_t *workSpaceSize)
Returns the workspace size required for a particular solution id.
miopenConvolutionForwardGetSolution
miopenStatus_t miopenConvolutionForwardGetSolution(miopenHandle_t handle, const miopenTensorDescriptor_t wDesc, const miopenTensorDescriptor_t xDesc, const miopenConvolutionDescriptor_t convDesc, const miopenTensorDescriptor_t yDesc, const size_t maxSolutionCount, size_t *solutionCount, miopenConvSolution_t *solutions)
Query the applicable solutions for a convolution configuration described by input,...
miopenGetConvolutionDescriptor
miopenStatus_t miopenGetConvolutionDescriptor(miopenConvolutionDescriptor_t convDesc, miopenConvolutionMode_t *c_mode, int *pad_h, int *pad_w, int *stride_h, int *stride_w, int *dilation_h, int *dilation_w)
Retrieves a 2-D convolution layer descriptor's details.
miopenConvolutionForwardGetSolutionCount
miopenStatus_t miopenConvolutionForwardGetSolutionCount(miopenHandle_t handle, const miopenTensorDescriptor_t wDesc, const miopenTensorDescriptor_t xDesc, const miopenConvolutionDescriptor_t convDesc, const miopenTensorDescriptor_t yDesc, size_t *solutionCount)
Query the maximum number of solutions applicable for the given input/output and weights tensor descri...
miopenConvolutionBackwardDataGetSolutionWorkspaceSize
miopenStatus_t miopenConvolutionBackwardDataGetSolutionWorkspaceSize(miopenHandle_t handle, const miopenTensorDescriptor_t dyDesc, const miopenTensorDescriptor_t wDesc, const miopenConvolutionDescriptor_t convDesc, const miopenTensorDescriptor_t dxDesc, const uint64_t solution_id, size_t *workSpaceSize)
Returns the workspace size required for a particular solution id.
miopenConvolutionBackwardBias
miopenStatus_t miopenConvolutionBackwardBias(miopenHandle_t handle, const void *alpha, const miopenTensorDescriptor_t dyDesc, const void *dy, const void *beta, const miopenTensorDescriptor_t dbDesc, void *db)
Calculates the gradient with respect to the bias.
miopenConvolutionAlgoDirect
@ miopenConvolutionAlgoDirect
Definition miopen.h:1256
miopenConvolutionAlgoWinograd
@ miopenConvolutionAlgoWinograd
Definition miopen.h:1258
miopenConvolutionAlgoFFT
@ miopenConvolutionAlgoFFT
Definition miopen.h:1257
miopenConvolutionAlgoImplicitGEMM
@ miopenConvolutionAlgoImplicitGEMM
Definition miopen.h:1259
miopenConvolutionAlgoGEMM
@ miopenConvolutionAlgoGEMM
Definition miopen.h:1255
miopenConvolutionFwdAlgoFFT
@ miopenConvolutionFwdAlgoFFT
Definition miopen.h:1220
miopenConvolutionFwdAlgoImplicitGEMM
@ miopenConvolutionFwdAlgoImplicitGEMM
Definition miopen.h:1222
miopenConvolutionFwdAlgoGEMM
@ miopenConvolutionFwdAlgoGEMM
Definition miopen.h:1218
miopenConvolutionFwdAlgoWinograd
@ miopenConvolutionFwdAlgoWinograd
Definition miopen.h:1221
miopenConvolutionFwdAlgoDirect
@ miopenConvolutionFwdAlgoDirect
Definition miopen.h:1219
miopenTransposeBwdDataAlgoGEMM
@ miopenTransposeBwdDataAlgoGEMM
Definition miopen.h:1245
miopenConvolutionBwdDataAlgoDirect
@ miopenConvolutionBwdDataAlgoDirect
Definition miopen.h:1242
miopenConvolutionBwdDataAlgoGEMM
@ miopenConvolutionBwdDataAlgoGEMM
Definition miopen.h:1241
miopenConvolutionBwdDataAlgoFFT
@ miopenConvolutionBwdDataAlgoFFT
Definition miopen.h:1243
miopenConvolutionBwdDataAlgoWinograd
@ miopenConvolutionBwdDataAlgoWinograd
Definition miopen.h:1244
miopenConvolutionBwdDataAlgoImplicitGEMM
@ miopenConvolutionBwdDataAlgoImplicitGEMM
Definition miopen.h:1247
miopenConvolutionBwdWeightsAlgoGEMM
@ miopenConvolutionBwdWeightsAlgoGEMM
Definition miopen.h:1230
miopenConvolutionBwdWeightsAlgoWinograd
@ miopenConvolutionBwdWeightsAlgoWinograd
Definition miopen.h:1232
miopenConvolutionBwdWeightsAlgoDirect
@ miopenConvolutionBwdWeightsAlgoDirect
Definition miopen.h:1231
miopenConvolutionBwdWeightsAlgoImplicitGEMM
@ miopenConvolutionBwdWeightsAlgoImplicitGEMM
Definition miopen.h:1233
miopenDestroyDropoutDescriptor
miopenStatus_t miopenDestroyDropoutDescriptor(miopenDropoutDescriptor_t dropoutDesc)
Destroys the dropout descriptor object.
miopenRNGType_t
miopenRNGType_t
Definition miopen.h:5803
miopenDropoutGetStatesSize
miopenStatus_t miopenDropoutGetStatesSize(miopenHandle_t handle, size_t *stateSizeInBytes)
Query the amount of memory required to store the states of the random number generators.
miopenDropoutGetReserveSpaceSize
miopenStatus_t miopenDropoutGetReserveSpaceSize(const miopenTensorDescriptor_t xDesc, size_t *reserveSpaceSizeInBytes)
Query the amount of memory required to run dropout.
miopenDropoutForward
miopenStatus_t miopenDropoutForward(miopenHandle_t handle, const miopenDropoutDescriptor_t dropoutDesc, const miopenTensorDescriptor_t noise_shape, const miopenTensorDescriptor_t xDesc, const void *x, const miopenTensorDescriptor_t yDesc, void *y, void *reserveSpace, size_t reserveSpaceSizeInBytes)
Execute forward dropout operation.
miopenGetDropoutDescriptor
miopenStatus_t miopenGetDropoutDescriptor(miopenDropoutDescriptor_t dropoutDesc, miopenHandle_t handle, float *dropout, void **states, unsigned long long *seed, bool *use_mask, bool *state_evo, miopenRNGType_t *rng_mode)
Get the details of the dropout descriptor.
miopenCreateDropoutDescriptor
miopenStatus_t miopenCreateDropoutDescriptor(miopenDropoutDescriptor_t *dropoutDesc)
Creates the dropout descriptor object.
miopenDropoutBackward
miopenStatus_t miopenDropoutBackward(miopenHandle_t handle, const miopenDropoutDescriptor_t dropoutDesc, const miopenTensorDescriptor_t noise_shape, const miopenTensorDescriptor_t dyDesc, const void *dy, const miopenTensorDescriptor_t dxDesc, void *dx, void *reserveSpace, size_t reserveSpaceSizeInBytes)
Execute backward dropout operation.
miopenRestoreDropoutDescriptor
miopenStatus_t miopenRestoreDropoutDescriptor(miopenDropoutDescriptor_t dropoutDesc, miopenHandle_t handle, float dropout, void *states, size_t stateSizeInBytes, unsigned long long seed, bool use_mask, bool state_evo, miopenRNGType_t rng_mode)
Restore the dropout descriptor to a saved state.
miopenSetDropoutDescriptor
miopenStatus_t miopenSetDropoutDescriptor(miopenDropoutDescriptor_t dropoutDesc, miopenHandle_t handle, float dropout, void *states, size_t stateSizeInBytes, unsigned long long seed, bool use_mask, bool state_evo, miopenRNGType_t rng_mode)
Initialize the dropout descriptor.
MIOPEN_RNG_PSEUDO_XORWOW
@ MIOPEN_RNG_PSEUDO_XORWOW
Definition miopen.h:5804
miopenFindResultsOrder_t
miopenFindResultsOrder_t
Definition miopen.h:6234
miopenGetSolutionSolverId
miopenStatus_t miopenGetSolutionSolverId(miopenSolution_t solution, uint64_t *solverId)
Reads id of the solver referred by the solution.
miopenSetSoftmaxDescriptor
miopenStatus_t miopenSetSoftmaxDescriptor(miopenSoftmaxDescriptor_t softmaxDesc, float alpha, float beta, miopenSoftmaxAlgorithm_t algorithm, miopenSoftmaxMode_t mode)
Sets the softmax descriptor details.
miopenGetSolutionTime
miopenStatus_t miopenGetSolutionTime(miopenSolution_t solution, float *time)
Reads the time spent to execute the solution the last it was run.
miopenGetMhaDescriptor
miopenStatus_t miopenGetMhaDescriptor(miopenMhaDescriptor_t mhaDesc, float *scale)
Gets the Mha descriptor details.
miopenMhaMask_t
miopenMhaMask_t
Initializes a problem object describing a Mha operation.
Definition miopen.h:6262
miopenSetMhaDescriptor
miopenStatus_t miopenSetMhaDescriptor(miopenMhaDescriptor_t mhaDesc, float scale)
Sets the Mha descriptor details.
miopenSetFindOptionWorkspaceLimit
miopenStatus_t miopenSetFindOptionWorkspaceLimit(miopenFindOptions_t options, size_t value)
Sets the workspace limit find option. Default value is maximum of size_t.
miopenSetFindOptionPreallocatedTensor
miopenStatus_t miopenSetFindOptionPreallocatedTensor(miopenFindOptions_t options, miopenTensorArgumentId_t id, void *buffer)
Attaches a preallocated tensor to find options. If not used, buffers are allocated by MIOpen internal...
miopenCreateSoftmaxProblem
miopenStatus_t miopenCreateSoftmaxProblem(miopenProblem_t *problem, miopenSoftmaxDescriptor_t operatorDesc, miopenProblemDirection_t direction)
Initializes a problem object describing a softmax operation.
miopenDestroyFindOptions
miopenStatus_t miopenDestroyFindOptions(miopenFindOptions_t options)
Destroys miopenFindOptions object.
miopenSetFindOptionAttachBinaries
miopenStatus_t miopenSetFindOptionAttachBinaries(miopenFindOptions_t options, unsigned attach)
Forces library to attach kernel binaries to solutions for later saving. This allows zero lookup miope...
miopenFuseProblems
miopenStatus_t miopenFuseProblems(miopenProblem_t problem1, miopenProblem_t problem2)
Fuse two problems into a single one. Problems can be either regular, or fused. No problems are dispos...
miopenGetSolutionSize
miopenStatus_t miopenGetSolutionSize(miopenSolution_t solution, size_t *size)
Reads the expected size of a solution.
miopenTensorArgumentId_t
miopenTensorArgumentId_t
Definition miopen.h:6153
miopenFindSolutions
miopenStatus_t miopenFindSolutions(miopenHandle_t handle, miopenProblem_t problem, miopenFindOptions_t options, miopenSolution_t *solutions, size_t *numSolutions, size_t maxSolutions)
Finds solutions to a problem by running different applicable solutions. Memory is automatically alloc...
miopenSetFindOptionTuning
miopenStatus_t miopenSetFindOptionTuning(miopenFindOptions_t options, int value)
Sets the tuning find option. Default value is zero.
miopenGetSolutionWorkspaceSize
miopenStatus_t miopenGetSolutionWorkspaceSize(miopenSolution_t solution, size_t *workspaceSize)
Reads the amount of workspace required to execute the solution.
miopenSetFindOptionResultsOrder
miopenStatus_t miopenSetFindOptionResultsOrder(miopenFindOptions_t options, miopenFindResultsOrder_t value)
Sets the results order find option. Default value is miopenFindResultsOrderByTime.
miopenRunSolution
miopenStatus_t miopenRunSolution(miopenHandle_t handle, miopenSolution_t solution, size_t nInputs, const miopenTensorArgument_t *tensors, void *workspace, size_t workspaceSize)
Runs the solution using the passed in buffers.
miopenCreateMhaProblem
miopenStatus_t miopenCreateMhaProblem(miopenProblem_t *problem, miopenMhaDescriptor_t operatorDesc, miopenProblemDirection_t direction)
miopenProblemDirection_t
miopenProblemDirection_t
Definition miopen.h:6140
miopenCreateBatchnormProblem
miopenStatus_t miopenCreateBatchnormProblem(miopenProblem_t *problem, miopenBatchNormMode_t mode, bool runningMeanVariance, miopenProblemDirection_t direction)
Initializes a problem object describing an activation operation.
miopenCreateMhaDescriptor
miopenStatus_t miopenCreateMhaDescriptor(miopenMhaDescriptor_t *mhaDesc)
Creates the mha descriptor object.
miopenDestroySolution
miopenStatus_t miopenDestroySolution(miopenSolution_t solution)
Destroys solution object.
miopenLoadSolution
miopenStatus_t miopenLoadSolution(miopenSolution_t *solution, const char *data, size_t size)
Loads solution object from binary data.
miopenSaveSolution
miopenStatus_t miopenSaveSolution(miopenSolution_t solution, char *data)
Saves a solution object as binary data.
miopenGetSolverIdConvAlgorithm
miopenStatus_t miopenGetSolverIdConvAlgorithm(uint64_t solverId, miopenConvAlgorithm_t *result)
Gets the convolution algorithm implemented by a solver.
miopenSetFindOptionPreallocatedWorkspace
miopenStatus_t miopenSetFindOptionPreallocatedWorkspace(miopenFindOptions_t options, void *buffer, size_t size)
Attaches the preallocated workspace to find options. Allocated by the library by default.
miopenCreateActivationProblem
miopenStatus_t miopenCreateActivationProblem(miopenProblem_t *problem, miopenActivationDescriptor_t operatorDesc, miopenProblemDirection_t direction)
Initializes a problem object describing an activation operation.
miopenDestroyProblem
miopenStatus_t miopenDestroyProblem(miopenProblem_t problem)
Destroys a problem object.
miopenCreateFindOptions
miopenStatus_t miopenCreateFindOptions(miopenFindOptions_t *options)
Initializes miopenFindOptions object.
miopenGetSoftmaxDescriptor
miopenStatus_t miopenGetSoftmaxDescriptor(const miopenSoftmaxDescriptor_t softmaxDesc, float *alpha, float *beta, miopenSoftmaxAlgorithm_t *algorithm, miopenSoftmaxMode_t *mode)
Gets the softmax layer descriptor details.
miopenCreateConvProblem
miopenStatus_t miopenCreateConvProblem(miopenProblem_t *problem, miopenConvolutionDescriptor_t operatorDesc, miopenProblemDirection_t direction)
Initializes a problem object describing a convolution operation.
miopenCreateBiasProblem
miopenStatus_t miopenCreateBiasProblem(miopenProblem_t *problem, miopenProblemDirection_t direction)
Initializes a problem object describing an bias operation.
miopenSetProblemTensorDescriptor
miopenStatus_t miopenSetProblemTensorDescriptor(miopenProblem_t problem, miopenTensorArgumentId_t id, const miopenTensorDescriptor_t descriptor)
Sets a tensor descriptor for the specified argument.
miopenCreateSoftmaxDescriptor
miopenStatus_t miopenCreateSoftmaxDescriptor(miopenSoftmaxDescriptor_t *softmaxDesc)
Creates the Softmax descriptor object.
miopenFindResultsOrderByWorkspaceSize
@ miopenFindResultsOrderByWorkspaceSize
Definition miopen.h:6236
miopenFindResultsOrderByTime
@ miopenFindResultsOrderByTime
Definition miopen.h:6235
miopenMhaMaskCausal
@ miopenMhaMaskCausal
Definition miopen.h:6264
miopenMhaMaskNone
@ miopenMhaMaskNone
Definition miopen.h:6263
miopenTensorMhaDescaleS
@ miopenTensorMhaDescaleS
Definition miopen.h:6165
miopenTensorMhaO
@ miopenTensorMhaO
Definition miopen.h:6171
miopenTensorMhaAmaxDV
@ miopenTensorMhaAmaxDV
Definition miopen.h:6189
miopenTensorBatchnormScaleDiff
@ miopenTensorBatchnormScaleDiff
Definition miopen.h:6212
miopenTensorMhaMask
@ miopenTensorMhaMask
Definition miopen.h:6223
miopenTensorMhaDescaleDS
@ miopenTensorMhaDescaleDS
Definition miopen.h:6179
miopenTensorMhaDropoutSeed
@ miopenTensorMhaDropoutSeed
Definition miopen.h:6169
miopenTensorBatchnormSavedMean
@ miopenTensorBatchnormSavedMean
Definition miopen.h:6209
miopenTensorActivationDY
@ miopenTensorActivationDY
Definition miopen.h:6197
miopenTensorBatchnormDX
@ miopenTensorBatchnormDX
Definition miopen.h:6217
miopenTensorMhaDescaleV
@ miopenTensorMhaDescaleV
Definition miopen.h:6164
miopenTensorMhaK
@ miopenTensorMhaK
Definition miopen.h:6159
miopenTensorConvolutionX
@ miopenTensorConvolutionX
Definition miopen.h:6155
miopenTensorMhaAmaxDK
@ miopenTensorMhaAmaxDK
Definition miopen.h:6188
miopenTensorMhaScaleDS
@ miopenTensorMhaScaleDS
Definition miopen.h:6180
miopenTensorBatchnormBias
@ miopenTensorBatchnormBias
Definition miopen.h:6215
miopenTensorMhaM
@ miopenTensorMhaM
Definition miopen.h:6174
miopenTensorBatchnormRunningMean
@ miopenTensorBatchnormRunningMean
Definition miopen.h:6207
miopenTensorMhaDropoutProbability
@ miopenTensorMhaDropoutProbability
Definition miopen.h:6168
miopenTensorSoftmaxY
@ miopenTensorSoftmaxY
Definition miopen.h:6202
miopenTensorSoftmaxDY
@ miopenTensorSoftmaxDY
Definition miopen.h:6204
miopenTensorMhaDescaleO
@ miopenTensorMhaDescaleO
Definition miopen.h:6177
miopenTensorMhaScaleO
@ miopenTensorMhaScaleO
Definition miopen.h:6167
miopenScalarBatchnormExpAvgFactor
@ miopenScalarBatchnormExpAvgFactor
Definition miopen.h:6225
miopenTensorBatchnormScale
@ miopenTensorBatchnormScale
Definition miopen.h:6211
miopenTensorBatchnormRunningVariance
@ miopenTensorBatchnormRunningVariance
Definition miopen.h:6208
miopenTensorActivationDX
@ miopenTensorActivationDX
Definition miopen.h:6196
miopenTensorMhaScaleDK
@ miopenTensorMhaScaleDK
Definition miopen.h:6182
miopenTensorMhaV
@ miopenTensorMhaV
Definition miopen.h:6161
miopenTensorMhaAmaxS
@ miopenTensorMhaAmaxS
Definition miopen.h:6173
miopenTensorBatchnormBiasDiff
@ miopenTensorBatchnormBiasDiff
Definition miopen.h:6216
miopenTensorBatchnormSavedVariance
@ miopenTensorBatchnormSavedVariance
Definition miopen.h:6210
miopenTensorActivationX
@ miopenTensorActivationX
Definition miopen.h:6194
miopenTensorMhaScaleDV
@ miopenTensorMhaScaleDV
Definition miopen.h:6183
miopenTensorMhaAmaxO
@ miopenTensorMhaAmaxO
Definition miopen.h:6172
miopenScalarBatchnormEpsilon
@ miopenScalarBatchnormEpsilon
Definition miopen.h:6226
miopenTensorMhaScaleDQ
@ miopenTensorMhaScaleDQ
Definition miopen.h:6181
miopenTensorSoftmaxDX
@ miopenTensorSoftmaxDX
Definition miopen.h:6203
miopenTensorMhaAmaxDS
@ miopenTensorMhaAmaxDS
Definition miopen.h:6190
miopenTensorMhaDV
@ miopenTensorMhaDV
Definition miopen.h:6186
miopenTensorMhaQ
@ miopenTensorMhaQ
Definition miopen.h:6160
miopenTensorMhaAmaxDQ
@ miopenTensorMhaAmaxDQ
Definition miopen.h:6187
miopenTensorConvolutionY
@ miopenTensorConvolutionY
Definition miopen.h:6157
miopenTensorBatchnormEstimatedMean
@ miopenTensorBatchnormEstimatedMean
Definition miopen.h:6213
miopenTensorBatchnormDY
@ miopenTensorBatchnormDY
Definition miopen.h:6218
miopenTensorMhaZInv
@ miopenTensorMhaZInv
Definition miopen.h:6175
miopenTensorMhaDescaleQ
@ miopenTensorMhaDescaleQ
Definition miopen.h:6163
miopenTensorMhaBias
@ miopenTensorMhaBias
Definition miopen.h:6191
miopenTensorBatchnormEstimatedVariance
@ miopenTensorBatchnormEstimatedVariance
Definition miopen.h:6214
miopenTensorArgumentIsScalar
@ miopenTensorArgumentIsScalar
Definition miopen.h:6221
miopenTensorArgumentIdInvalid
@ miopenTensorArgumentIdInvalid
Definition miopen.h:6154
miopenTensorMhaDO
@ miopenTensorMhaDO
Definition miopen.h:6176
miopenTensorMhaDescaleDO
@ miopenTensorMhaDescaleDO
Definition miopen.h:6178
miopenTensorConvolutionW
@ miopenTensorConvolutionW
Definition miopen.h:6156
miopenTensorMhaDescaleK
@ miopenTensorMhaDescaleK
Definition miopen.h:6162
miopenTensorMhaDropoutOffset
@ miopenTensorMhaDropoutOffset
Definition miopen.h:6170
miopenTensorBiasY
@ miopenTensorBiasY
Definition miopen.h:6199
miopenTensorMhaDQ
@ miopenTensorMhaDQ
Definition miopen.h:6184
miopenTensorSoftmaxX
@ miopenTensorSoftmaxX
Definition miopen.h:6201
miopenTensorBatchnormY
@ miopenTensorBatchnormY
Definition miopen.h:6206
miopenTensorMhaScaleS
@ miopenTensorMhaScaleS
Definition miopen.h:6166
miopenTensorBias
@ miopenTensorBias
Definition miopen.h:6200
miopenTensorActivationY
@ miopenTensorActivationY
Definition miopen.h:6195
miopenTensorBatchnormX
@ miopenTensorBatchnormX
Definition miopen.h:6205
miopenTensorBiasX
@ miopenTensorBiasX
Definition miopen.h:6198
miopenTensorMhaDK
@ miopenTensorMhaDK
Definition miopen.h:6185
miopenProblemDirectionBackwardWeights
@ miopenProblemDirectionBackwardWeights
Definition miopen.h:6143
miopenProblemDirectionInference
@ miopenProblemDirectionInference
Definition miopen.h:6145
miopenProblemDirectionForward
@ miopenProblemDirectionForward
Definition miopen.h:6141
miopenProblemDirectionBackward
@ miopenProblemDirectionBackward
Definition miopen.h:6142
miopenGetGetitemWorkspaceSize
miopenStatus_t miopenGetGetitemWorkspaceSize(miopenHandle_t handle, uint32_t indexCount, const miopenTensorDescriptor_t *indexDescs, size_t *sizeInBytes)
Helper function to query the minimum workspace size required by the getitem call.
miopenGetitemBackward
miopenStatus_t miopenGetitemBackward(miopenHandle_t handle, void *workspace, size_t workspaceSizeInBytes, const miopenTensorDescriptor_t dyDesc, const void *dy, uint32_t indexCount, const miopenTensorDescriptor_t *indexDescs, const void *const *indexs, const miopenTensorDescriptor_t dxDesc, void *dx, const miopenTensorDescriptor_t errorDesc, void *error, uint32_t dimCount, const int32_t *dims, uint32_t sliceCount, const int32_t *slices, uint32_t offset)
Execute a getitem backward layer.
miopenGroupNormForward
miopenStatus_t miopenGroupNormForward(miopenHandle_t handle, miopenNormMode_t mode, const miopenTensorDescriptor_t xDesc, const void *x, const miopenTensorDescriptor_t weightDesc, const void *weight, const miopenTensorDescriptor_t biasDesc, const void *bias, const uint64_t num_groups, const float epsilon, const miopenTensorDescriptor_t yDesc, void *y, const miopenTensorDescriptor_t meanDesc, void *mean, const miopenTensorDescriptor_t rstdDesc, void *rstd)
Execute a groupnorm forward layer.
miopenGetErrorString
const char * miopenGetErrorString(miopenStatus_t error)
Get character string for an error code.
miopenCreateWithStream
miopenStatus_t miopenCreateWithStream(miopenHandle_t *handle, miopenAcceleratorQueue_t stream)
Create a MIOpen handle with an accelerator stream.
miopenDestroy
miopenStatus_t miopenDestroy(miopenHandle_t handle)
Destroys the MIOpen handle.
miopenTuningPolicy_t
miopenTuningPolicy_t
Definition miopen.h:8021
miopenGetTuningPolicy
miopenStatus_t miopenGetTuningPolicy(miopenHandle_t handle, miopenTuningPolicy_t *value)
Get tuning policy from a handle.
miopenDeallocatorFunction
void(* miopenDeallocatorFunction)(void *context, void *memory)
Custom deallocator function.
Definition miopen.h:144
miopenMathType_t
miopenMathType_t
Definition miopen.h:102
miopenAllocatorFunction
void *(* miopenAllocatorFunction)(void *context, size_t sizeBytes)
Custom allocator function.
Definition miopen.h:134
miopenGetStream
miopenStatus_t miopenGetStream(miopenHandle_t handle, miopenAcceleratorQueue_t *streamID)
Get the previously created accelerator command queue.
miopenEnableProfiling
miopenStatus_t miopenEnableProfiling(miopenHandle_t handle, bool enable)
Enable profiling to retrieve kernel time.
miopenGetVersion
miopenStatus_t miopenGetVersion(size_t *major, size_t *minor, size_t *patch)
Method to return version of MIOpen.
miopenSetTuningPolicy
miopenStatus_t miopenSetTuningPolicy(miopenHandle_t handle, miopenTuningPolicy_t newValue)
Update tuning policy for a specific handle. API alternative for MIOPEN_FIND_ENFORCE environment varia...
miopenStatus_t
miopenStatus_t
Definition miopen.h:87
miopenSetAllocator
miopenStatus_t miopenSetAllocator(miopenHandle_t handle, miopenAllocatorFunction allocator, miopenDeallocatorFunction deallocator, void *allocatorContext)
Set allocator for previously created miopenHandle.
miopenF8RoundingMode_t
miopenF8RoundingMode_t
Definition miopen.h:111
miopenCreate
miopenStatus_t miopenCreate(miopenHandle_t *handle)
Method to create the MIOpen handle object.
miopenGetKernelTime
miopenStatus_t miopenGetKernelTime(miopenHandle_t handle, float *time)
Get time for last kernel launched.
miopenSetStream
miopenStatus_t miopenSetStream(miopenHandle_t handle, miopenAcceleratorQueue_t streamID)
Set accelerator command queue previously created.
miopenMathDefault
@ miopenMathDefault
Definition miopen.h:104
miopenMathPedantic
@ miopenMathPedantic
Definition miopen.h:105
miopenStatusUnsupportedOp
@ miopenStatusUnsupportedOp
Definition miopen.h:96
miopenStatusGpuOperationsSkipped
@ miopenStatusGpuOperationsSkipped
Definition miopen.h:97
miopenStatusUnknownError
@ miopenStatusUnknownError
Definition miopen.h:95
miopenStatusSuccess
@ miopenStatusSuccess
Definition miopen.h:88
miopenStatusVersionMismatch
@ miopenStatusVersionMismatch
Definition miopen.h:98
miopenStatusAllocFailed
@ miopenStatusAllocFailed
Definition miopen.h:92
miopenStatusNotImplemented
@ miopenStatusNotImplemented
Definition miopen.h:94
miopenStatusBadParm
@ miopenStatusBadParm
Definition miopen.h:91
miopenStatusNotInitialized
@ miopenStatusNotInitialized
Definition miopen.h:89
miopenStatusInternalError
@ miopenStatusInternalError
Definition miopen.h:93
miopenStatusInvalidValue
@ miopenStatusInvalidValue
Definition miopen.h:90
miopenF8RoundingModeStandard
@ miopenF8RoundingModeStandard
Definition miopen.h:112
miopenF8RoundingModeStochastic
@ miopenF8RoundingModeStochastic
Definition miopen.h:113
miopenKthvalueForward
miopenStatus_t miopenKthvalueForward(miopenHandle_t handle, miopenTensorDescriptor_t inputDesc, const void *input, miopenTensorDescriptor_t outputDesc, void *output, miopenTensorDescriptor_t indicesDesc, size_t *indices, size_t k, int32_t dim=-1, bool keepDim=false)
Execute a Kthvalue forward layer.
miopenLayerNormForward
miopenStatus_t miopenLayerNormForward(miopenHandle_t handle, miopenNormMode_t mode, const miopenTensorDescriptor_t xDesc, const void *x, const miopenTensorDescriptor_t weightDesc, const void *weight, const miopenTensorDescriptor_t biasDesc, const void *bias, const float epsilon, const int32_t normalized_dim, const miopenTensorDescriptor_t yDesc, void *y, const miopenTensorDescriptor_t meanDesc, void *mean, const miopenTensorDescriptor_t rstdDesc, void *rstd)
Execute a layernorm forward layer.
miopenT5LayerNormBackward
miopenStatus_t miopenT5LayerNormBackward(miopenHandle_t handle, miopenNormMode_t mode, void *workspace, size_t workspaceSizeInBytes, const miopenTensorDescriptor_t dyDesc, const void *dy, const miopenTensorDescriptor_t xDesc, const void *x, const miopenTensorDescriptor_t weightDesc, const void *weight, const miopenTensorDescriptor_t rstdDesc, const void *rstd, const miopenTensorDescriptor_t dxDesc, void *dx, const miopenTensorDescriptor_t dwDesc, void *dw)
Execute a T5layernorm backward layer.
miopenLayerNormBackward
miopenStatus_t miopenLayerNormBackward(miopenHandle_t handle, miopenNormMode_t mode, void *workspace, size_t workspaceSizeInBytes, const miopenTensorDescriptor_t dyDesc, const void *dy, const miopenTensorDescriptor_t xDesc, const void *x, const miopenTensorDescriptor_t weightDesc, const void *weight, const miopenTensorDescriptor_t meanDesc, const void *mean, const miopenTensorDescriptor_t rstdDesc, const void *rstd, const int32_t normalized_dim, const miopenTensorDescriptor_t dxDesc, void *dx, const miopenTensorDescriptor_t dwDesc, void *dw, const miopenTensorDescriptor_t dbDesc, void *db)
Execute a layernorm backward layer.
miopenT5LayerNormForward
miopenStatus_t miopenT5LayerNormForward(miopenHandle_t handle, miopenNormMode_t mode, const miopenTensorDescriptor_t xDesc, const void *x, const miopenTensorDescriptor_t weightDesc, const void *weight, const float epsilon, const miopenTensorDescriptor_t yDesc, void *y, const miopenTensorDescriptor_t rstdDesc, void *rstd)
Execute a T5layernorm forward layer.
miopenNormMode_t
miopenNormMode_t
Definition miopen.h:471
miopenGetLayerNormBackwardWorkspaceSize
miopenStatus_t miopenGetLayerNormBackwardWorkspaceSize(miopenHandle_t handle, miopenNormMode_t mode, const miopenTensorDescriptor_t dyDesc, const miopenTensorDescriptor_t xDesc, const miopenTensorDescriptor_t weightDesc, const miopenTensorDescriptor_t meanDesc, const miopenTensorDescriptor_t rstdDesc, const int32_t normalized_dim, const miopenTensorDescriptor_t dxDesc, const miopenTensorDescriptor_t dwDesc, const miopenTensorDescriptor_t dbDesc, size_t *sizeInBytes)
Helper function to query the minimum workspace size required by the layernorm backward call.
miopenGetT5LayerNormBackwardWorkspaceSize
miopenStatus_t miopenGetT5LayerNormBackwardWorkspaceSize(miopenHandle_t handle, miopenNormMode_t mode, const miopenTensorDescriptor_t dyDesc, const miopenTensorDescriptor_t xDesc, const miopenTensorDescriptor_t weightDesc, const miopenTensorDescriptor_t rstdDesc, const miopenTensorDescriptor_t dxDesc, const miopenTensorDescriptor_t dwDesc, size_t *sizeInBytes)
Helper function to query the minimum workspace size required by the T5layernorm backward call.
miopenAddLayerNormForward
miopenStatus_t miopenAddLayerNormForward(miopenHandle_t handle, miopenNormMode_t mode, const miopenTensorDescriptor_t xDesc, const void *x, const miopenTensorDescriptor_t x2Desc, const void *x2, const miopenTensorDescriptor_t weightDesc, const void *weight, const miopenTensorDescriptor_t biasDesc, const void *bias, const float epsilon, const int32_t normalized_dim, const miopenTensorDescriptor_t yDesc, void *y, const miopenTensorDescriptor_t meanDesc, void *mean, const miopenTensorDescriptor_t rstdDesc, void *rstd)
Execute a add and layernorm forward layer.
miopenSet2dPoolingDescriptor
miopenStatus_t miopenSet2dPoolingDescriptor(miopenPoolingDescriptor_t poolDesc, miopenPoolingMode_t mode, int windowHeight, int windowWidth, int pad_h, int pad_w, int stride_h, int stride_w)
Sets a 2-D pooling layer descriptor details.
miopenSetPoolingWorkSpaceIndexMode
miopenStatus_t miopenSetPoolingWorkSpaceIndexMode(miopenPoolingDescriptor_t poolDesc, miopenPoolingWorkspaceIndexMode_t workspace_index)
Set workspace index mode for pooling layer. The default mode is miopenPoolingWorkSpaceIndexMask.
miopenGetPoolingForwardOutputDim
miopenStatus_t miopenGetPoolingForwardOutputDim(const miopenPoolingDescriptor_t poolDesc, const miopenTensorDescriptor_t tensorDesc, int *n, int *c, int *h, int *w)
Gets the shape of the output tensor for 2-D pooling.
miopenPoolingGetWorkSpaceSize
miopenStatus_t miopenPoolingGetWorkSpaceSize(const miopenTensorDescriptor_t yDesc, size_t *workSpaceSize)
Get the amount of GPU memory required for pooling.
miopenSetNdPoolingDescriptor
miopenStatus_t miopenSetNdPoolingDescriptor(miopenPoolingDescriptor_t poolDesc, const miopenPoolingMode_t mode, int nbDims, const int *windowDimA, const int *padA, const int *stridesA)
Set details of a N-D pooling layer descriptor.
miopenPoolingForward
miopenStatus_t miopenPoolingForward(miopenHandle_t handle, const miopenPoolingDescriptor_t poolDesc, const void *alpha, const miopenTensorDescriptor_t xDesc, const void *x, const void *beta, const miopenTensorDescriptor_t yDesc, void *y, bool do_backward, void *workSpace, size_t workSpaceSize)
Execute a forward pooling layer.
miopenPoolingMode_t
miopenPoolingMode_t
Definition miopen.h:438
miopenPoolingGetWorkSpaceSizeV2
miopenStatus_t miopenPoolingGetWorkSpaceSizeV2(const miopenPoolingDescriptor_t poolDesc, const miopenTensorDescriptor_t yDesc, size_t *workSpaceSize)
Get the amount of GPU memory required for pooling.
miopenGetPoolingWorkSpaceIndexMode
miopenStatus_t miopenGetPoolingWorkSpaceIndexMode(miopenPoolingDescriptor_t poolDesc, miopenPoolingWorkspaceIndexMode_t *workspace_index)
Get workspace index mode for pooling layer.
miopenIndexType_t
miopenIndexType_t
Definition miopen.h:391
miopenGetPoolingIndexType
miopenStatus_t miopenGetPoolingIndexType(miopenPoolingDescriptor_t poolDesc, miopenIndexType_t *index_type)
Get the index data type for pooling layer. The index type to any of the miopenIndexType_t sizes; 8,...
miopenPoolingWorkspaceIndexMode_t
miopenPoolingWorkspaceIndexMode_t
Definition miopen.h:451
miopenGetPoolingNdForwardOutputDim
miopenStatus_t miopenGetPoolingNdForwardOutputDim(const miopenPoolingDescriptor_t poolDesc, const miopenTensorDescriptor_t tensorDesc, int dims, int *tensorDimArr)
Gets the shape of the output tensor for N-D pooling.
miopenGetNdPoolingDescriptor
miopenStatus_t miopenGetNdPoolingDescriptor(const miopenPoolingDescriptor_t poolDesc, int nbDimsRequested, miopenPoolingMode_t *mode, int *nbDims, int *windowDimA, int *padA, int *stridesA)
Get details of a N-D pooling layer descriptor.
miopenCreatePoolingDescriptor
miopenStatus_t miopenCreatePoolingDescriptor(miopenPoolingDescriptor_t *poolDesc)
Creates a pooling layer descriptor.
miopenSetPoolingIndexType
miopenStatus_t miopenSetPoolingIndexType(miopenPoolingDescriptor_t poolDesc, miopenIndexType_t index_type)
Set index data type for pooling layer. The default indexing type is uint8_t. Users can set the index ...
miopenGet2dPoolingDescriptor
miopenStatus_t miopenGet2dPoolingDescriptor(const miopenPoolingDescriptor_t poolDesc, miopenPoolingMode_t *mode, int *windowHeight, int *windowWidth, int *pad_h, int *pad_w, int *stride_h, int *stride_w)
Gets a 2-D pooling layer descriptor details.
miopenDestroyPoolingDescriptor
miopenStatus_t miopenDestroyPoolingDescriptor(miopenPoolingDescriptor_t poolDesc)
Destroys the pooling descriptor object.
miopenPoolingBackward
miopenStatus_t miopenPoolingBackward(miopenHandle_t handle, const miopenPoolingDescriptor_t poolDesc, const void *alpha, const miopenTensorDescriptor_t yDesc, const void *y, const miopenTensorDescriptor_t dyDesc, const void *dy, const miopenTensorDescriptor_t xDesc, const void *x, const void *beta, const miopenTensorDescriptor_t dxDesc, void *dx, void *workSpace)
Execute a backward pooling layer.
miopenReduceCalculationForward
miopenStatus_t miopenReduceCalculationForward(miopenHandle_t handle, miopenReduceCalculationNanPropagation_t nanPropagation, void *workspace, size_t workspaceSizeInBytes, const miopenTensorDescriptor_t xDesc, const void *x, const int32_t dim, const miopenReduceCalculationOp_t reduceCalculationOp, const miopenTensorDescriptor_t reduceDesc, void *y)
Execute a reducecalculation forward layer.
miopenReduceCalculationOp_t
miopenReduceCalculationOp_t
Definition miopen.h:6667
miopenGetReduceCalculationWorkspaceSize
miopenStatus_t miopenGetReduceCalculationWorkspaceSize(miopenHandle_t handle, const miopenTensorDescriptor_t xDesc, const int32_t dim, const miopenReduceCalculationOp_t reduceCalculationOp, const miopenTensorDescriptor_t reduceDesc, size_t *sizeInBytes)
Helper function to query the minimum workspace size required by the ReduceTensor call.
MIOPEN_REDUCE_CALCULATION_PROD
@ MIOPEN_REDUCE_CALCULATION_PROD
Definition miopen.h:6668
MIOPEN_REDUCE_CALCULATION_SUM
@ MIOPEN_REDUCE_CALCULATION_SUM
Definition miopen.h:6670
miopenSoftmaxBackward_V2
miopenStatus_t miopenSoftmaxBackward_V2(miopenHandle_t handle, const void *alpha, const miopenTensorDescriptor_t yDesc, const void *y, const miopenTensorDescriptor_t dyDesc, const void *dy, const void *beta, const miopenTensorDescriptor_t dxDesc, void *dx, miopenSoftmaxAlgorithm_t algorithm, miopenSoftmaxMode_t mode)
Execute a softmax backwards layer with expanded modes and algorithms.
miopenSoftmaxForward
miopenStatus_t miopenSoftmaxForward(miopenHandle_t handle, const void *alpha, const miopenTensorDescriptor_t xDesc, const void *x, const void *beta, const miopenTensorDescriptor_t yDesc, void *y)
Execute a softmax forward layer.
miopenSoftmaxForward_V2
miopenStatus_t miopenSoftmaxForward_V2(miopenHandle_t handle, const void *alpha, const miopenTensorDescriptor_t xDesc, const void *x, const void *beta, const miopenTensorDescriptor_t yDesc, void *y, miopenSoftmaxAlgorithm_t algorithm, miopenSoftmaxMode_t mode)
Execute a softmax forward layer with expanded modes and algorithms.
miopenSoftmaxBackward
miopenStatus_t miopenSoftmaxBackward(miopenHandle_t handle, const void *alpha, const miopenTensorDescriptor_t yDesc, const void *y, const miopenTensorDescriptor_t dyDesc, const void *dy, const void *beta, const miopenTensorDescriptor_t dxDesc, void *dx)
Execute a softmax backwards layer.
miopenSoftmaxMode_t
miopenSoftmaxMode_t
Definition miopen.h:534
miopenSoftmaxAlgorithm_t
miopenSoftmaxAlgorithm_t
Definition miopen.h:523
miopenDestroySeqTensorDescriptor
miopenStatus_t miopenDestroySeqTensorDescriptor(miopenSeqTensorDescriptor_t tensorDesc)
Destroys the sequence data tensor descriptor.
miopenGetTensorDescriptor
miopenStatus_t miopenGetTensorDescriptor(miopenTensorDescriptor_t tensorDesc, miopenDataType_t *dataType, int *dimsA, int *stridesA)
Get the details of the N-dimensional tensor descriptor.
miopenSetNdTensorDescriptorWithLayout
miopenStatus_t miopenSetNdTensorDescriptorWithLayout(miopenTensorDescriptor_t tensorDesc, miopenDataType_t dataType, miopenTensorLayout_t tensorLayout, const int *lens, int num_lens)
Set shape of ND tensor with specific layout.
miopenSetTensor
miopenStatus_t miopenSetTensor(miopenHandle_t handle, const miopenTensorDescriptor_t yDesc, void *y, const void *alpha)
Fills a tensor with a single value.
miopenOpTensor
miopenStatus_t miopenOpTensor(miopenHandle_t handle, miopenTensorOp_t tensorOp, const void *alpha1, const miopenTensorDescriptor_t aDesc, const void *A, const void *alpha2, const miopenTensorDescriptor_t bDesc, const void *B, const void *beta, const miopenTensorDescriptor_t cDesc, void *C)
Execute element-wise tensor operations.
miopenGetTensorNumBytes
miopenStatus_t miopenGetTensorNumBytes(miopenTensorDescriptor_t tensorDesc, size_t *numBytes)
Returns number of bytes associated with tensor descriptor.
miopenGet4dTensorDescriptor
miopenStatus_t miopenGet4dTensorDescriptor(miopenTensorDescriptor_t tensorDesc, miopenDataType_t *dataType, int *n, int *c, int *h, int *w, int *nStride, int *cStride, int *hStride, int *wStride)
Get the details of the tensor descriptor.
miopenTensorLayout_t
miopenTensorLayout_t
Definition miopen.h:374
miopenTransformTensor
miopenStatus_t miopenTransformTensor(miopenHandle_t handle, const void *alpha, const miopenTensorDescriptor_t xDesc, const void *x, const void *beta, const miopenTensorDescriptor_t yDesc, void *y)
Copies one tensor to another tensor with a different layout/scale.
miopenCreateTensorDescriptor
miopenStatus_t miopenCreateTensorDescriptor(miopenTensorDescriptor_t *tensorDesc)
Create a Tensor Descriptor.
miopenSetTensorDescriptorV2
miopenStatus_t miopenSetTensorDescriptorV2(miopenTensorDescriptor_t tensorDesc, miopenDataType_t dataType, int nbDims, const size_t *dimsA, const size_t *stridesA)
Set shape of N-dimensional tensor.
miopenScaleTensor
miopenStatus_t miopenScaleTensor(miopenHandle_t handle, const miopenTensorDescriptor_t yDesc, void *y, const void *alpha)
Scales all elements in a tensor by a single value.
miopenSetTensorCastType
miopenStatus_t miopenSetTensorCastType(miopenTensorDescriptor_t tensorDesc, miopenDataType_t cast_type)
Set the tensor cast type.
miopenSet4dTensorDescriptor
miopenStatus_t miopenSet4dTensorDescriptor(miopenTensorDescriptor_t tensorDesc, miopenDataType_t dataType, int n, int c, int h, int w)
Set shape of 4D tensor.
miopenSet4dTensorDescriptorEx
miopenStatus_t miopenSet4dTensorDescriptorEx(miopenTensorDescriptor_t tensorDesc, miopenDataType_t dataType, int n, int c, int h, int w, int nStride, int cStride, int hStride, int wStride)
Set shape and stride of 4D tensor.
miopenDataType_t
miopenDataType_t
Definition miopen.h:354
miopenCreateSeqTensorDescriptor
miopenStatus_t miopenCreateSeqTensorDescriptor(miopenSeqTensorDescriptor_t *tensorDesc)
Create a Tensor Descriptor for sequence data.
miopenTensorOp_t
miopenTensorOp_t
Definition miopen.h:403
miopenGetTensorDescriptorSize
miopenStatus_t miopenGetTensorDescriptorSize(miopenTensorDescriptor_t tensorDesc, int *size)
Set shape of N-dimensional tensor.
miopenDestroyTensorDescriptor
miopenStatus_t miopenDestroyTensorDescriptor(miopenTensorDescriptor_t tensorDesc)
Destroys the tensor descriptor.
miopenSetTensorDescriptor
miopenStatus_t miopenSetTensorDescriptor(miopenTensorDescriptor_t tensorDesc, miopenDataType_t dataType, int nbDims, const int *dimsA, const int *stridesA)
Set shape of N-dimensional tensor.
miopenPaddingMode_t
miopenPaddingMode_t
Definition miopen.h:427
miopenPaddingDefault
@ miopenPaddingDefault
Definition miopen.h:428
miopenPaddingSame
@ miopenPaddingSame
Definition miopen.h:429
miopenPaddingValid
@ miopenPaddingValid
Definition miopen.h:430
miopenReduceCalculationNanPropagation_t
miopenReduceCalculationNanPropagation_t
Definition miopen.h:6651
MIOPEN_REDUCE_CALCULATION_PROPAGATE_NAN
@ MIOPEN_REDUCE_CALCULATION_PROPAGATE_NAN
Definition miopen.h:6653
MIOPEN_REDUCE_CALCULATION_NOT_PROPAGATE_NAN
@ MIOPEN_REDUCE_CALCULATION_NOT_PROPAGATE_NAN
Definition miopen.h:6652
miopenAlphaBetaCase_t
miopenAlphaBetaCase_t
Enum for specifying the alpha-beta case for convolution operations.
Definition miopen.h:6974
SCALE
@ SCALE
Definition miopen.h:6981
BILINEAR
@ BILINEAR
Definition miopen.h:6982
DEFAULT
@ DEFAULT
Definition miopen.h:6980
ERROR_STATE
@ ERROR_STATE
Definition miopen.h:6983
MIOPEN_DECLARE_OBJECT
#define MIOPEN_DECLARE_OBJECT(name)
Definition miopen.h:57
miopenNanPropagation_t
miopenNanPropagation_t
Definition miopen.h:579
MIOPEN_PROPAGATE_NAN
@ MIOPEN_PROPAGATE_NAN
Definition miopen.h:581
MIOPEN_NOT_PROPAGATE_NAN
@ MIOPEN_NOT_PROPAGATE_NAN
Definition miopen.h:580
MIOPEN_LOSS_REDUCTION_MEAN
@ MIOPEN_LOSS_REDUCTION_MEAN
Definition miopen.h:7841
MIOPEN_LOSS_REDUCTION_SUM
@ MIOPEN_LOSS_REDUCTION_SUM
Definition miopen.h:7840
MIOPEN_LOSS_REDUCTION_NONE
@ MIOPEN_LOSS_REDUCTION_NONE
Definition miopen.h:7839
miopenTuningPolicyDbClean
@ miopenTuningPolicyDbClean
Definition miopen.h:8027
miopenTuningPolicyDbUpdate
@ miopenTuningPolicyDbUpdate
Definition miopen.h:8023
miopenTuningPolicyNone
@ miopenTuningPolicyNone
Definition miopen.h:8022
miopenTuningPolicySearch
@ miopenTuningPolicySearch
Definition miopen.h:8024
miopenTuningPolicySearchDbUpdate
@ miopenTuningPolicySearchDbUpdate
Definition miopen.h:8026
miopenPoolingAverage
@ miopenPoolingAverage
Definition miopen.h:440
miopenPoolingAverageInclusive
@ miopenPoolingAverageInclusive
Definition miopen.h:441
miopenPoolingMax
@ miopenPoolingMax
Definition miopen.h:439
miopenTensorCHWN
@ miopenTensorCHWN
Definition miopen.h:377
miopenTensorNHWC
@ miopenTensorNHWC
Definition miopen.h:376
miopenTensorNDHWC
@ miopenTensorNDHWC
Definition miopen.h:383
miopenTensorNCHW
@ miopenTensorNCHW
Definition miopen.h:375
miopenTensorCHWNc4
@ miopenTensorCHWNc4
Definition miopen.h:380
miopenTensorNCHWc8
@ miopenTensorNCHWc8
Definition miopen.h:379
miopenTensorNCDHW
@ miopenTensorNCDHW
Definition miopen.h:382
miopenTensorCHWNc8
@ miopenTensorCHWNc8
Definition miopen.h:381
miopenTensorNCHWc4
@ miopenTensorNCHWc4
Definition miopen.h:378
MIOPEN_REDUCE_TENSOR_FLATTENED_INDICES
@ MIOPEN_REDUCE_TENSOR_FLATTENED_INDICES
Definition miopen.h:591
MIOPEN_REDUCE_TENSOR_NO_INDICES
@ MIOPEN_REDUCE_TENSOR_NO_INDICES
Definition miopen.h:590
miopenActivationLEAKYRELU
@ miopenActivationLEAKYRELU
Definition miopen.h:510
miopenActivationPASTHRU
@ miopenActivationPASTHRU
Definition miopen.h:501
miopenActivationABS
@ miopenActivationABS
Definition miopen.h:506
miopenActivationLOGISTIC
@ miopenActivationLOGISTIC
Definition miopen.h:502
miopenActivationCLIPPEDRELU
@ miopenActivationCLIPPEDRELU
Definition miopen.h:508
miopenActivationRELU
@ miopenActivationRELU
Definition miopen.h:504
miopenActivationPOWER
@ miopenActivationPOWER
Definition miopen.h:507
miopenActivationELU
@ miopenActivationELU
Definition miopen.h:512
miopenActivationCLAMP
@ miopenActivationCLAMP
Definition miopen.h:515
miopenActivationSOFTRELU
@ miopenActivationSOFTRELU
Definition miopen.h:505
miopenActivationTANH
@ miopenActivationTANH
Definition miopen.h:503
MIOPEN_REDUCE_TENSOR_MUL
@ MIOPEN_REDUCE_TENSOR_MUL
Definition miopen.h:556
MIOPEN_REDUCE_TENSOR_MAX
@ MIOPEN_REDUCE_TENSOR_MAX
Definition miopen.h:560
MIOPEN_REDUCE_TENSOR_AVG
@ MIOPEN_REDUCE_TENSOR_AVG
Definition miopen.h:564
MIOPEN_REDUCE_TENSOR_NORM1
@ MIOPEN_REDUCE_TENSOR_NORM1
Definition miopen.h:566
MIOPEN_REDUCE_TENSOR_AMAX
@ MIOPEN_REDUCE_TENSOR_AMAX
Definition miopen.h:562
MIOPEN_REDUCE_TENSOR_MIN
@ MIOPEN_REDUCE_TENSOR_MIN
Definition miopen.h:558
MIOPEN_REDUCE_TENSOR_ADD
@ MIOPEN_REDUCE_TENSOR_ADD
Definition miopen.h:555
MIOPEN_REDUCE_TENSOR_NORM2
@ MIOPEN_REDUCE_TENSOR_NORM2
Definition miopen.h:568
MIOPEN_CONVOLUTION_ATTRIB_DETERMINISTIC
@ MIOPEN_CONVOLUTION_ATTRIB_DETERMINISTIC
Definition miopen.h:616
MIOPEN_CONVOLUTION_ATTRIB_FP8_ROUNDING_MODE
@ MIOPEN_CONVOLUTION_ATTRIB_FP8_ROUNDING_MODE
Definition miopen.h:620
MIOPEN_CONVOLUTION_ATTRIB_MATH_TYPE
@ MIOPEN_CONVOLUTION_ATTRIB_MATH_TYPE
Definition miopen.h:629
MIOPEN_CONVOLUTION_ATTRIB_FP16_ALT_IMPL
@ MIOPEN_CONVOLUTION_ATTRIB_FP16_ALT_IMPL
Definition miopen.h:612
miopenIndexUint16
@ miopenIndexUint16
Definition miopen.h:393
miopenIndexUint64
@ miopenIndexUint64
Definition miopen.h:395
miopenIndexUint32
@ miopenIndexUint32
Definition miopen.h:394
miopenIndexUint8
@ miopenIndexUint8
Definition miopen.h:392
miopenFloat
@ miopenFloat
Definition miopen.h:356
miopenBFloat8_fnuz
@ miopenBFloat8_fnuz
Definition miopen.h:364
miopenBFloat16
@ miopenBFloat16
Definition miopen.h:360
miopenInt8
@ miopenInt8
Definition miopen.h:358
miopenInt32
@ miopenInt32
Definition miopen.h:357
miopenHalf
@ miopenHalf
Definition miopen.h:355
miopenInt64
@ miopenInt64
Definition miopen.h:365
miopenDouble
@ miopenDouble
Definition miopen.h:362
miopenFloat8_fnuz
@ miopenFloat8_fnuz
Definition miopen.h:363
MIOPEN_8BIT_INDICES
@ MIOPEN_8BIT_INDICES
Definition miopen.h:603
MIOPEN_32BIT_INDICES
@ MIOPEN_32BIT_INDICES
Definition miopen.h:600
MIOPEN_16BIT_INDICES
@ MIOPEN_16BIT_INDICES
Definition miopen.h:602
MIOPEN_64BIT_INDICES
@ MIOPEN_64BIT_INDICES
Definition miopen.h:601
miopenPoolingWorkspaceIndexImage
@ miopenPoolingWorkspaceIndexImage
Definition miopen.h:453
miopenPoolingWorkspaceIndexMask
@ miopenPoolingWorkspaceIndexMask
Definition miopen.h:452
miopenDepthwise
@ miopenDepthwise
Definition miopen.h:419
miopenGroupConv
@ miopenGroupConv
Definition miopen.h:418
miopenTranspose
@ miopenTranspose
Definition miopen.h:417
miopenConvolution
@ miopenConvolution
Definition miopen.h:416
MIOPEN_ELEMENTWISE_AFFINE_FUSED_ADD
@ MIOPEN_ELEMENTWISE_AFFINE_FUSED_ADD
Definition miopen.h:475
MIOPEN_WEIGHT_BIAS
@ MIOPEN_WEIGHT_BIAS
Definition miopen.h:473
MIOPEN_ELEMENTWISE_AFFINE
@ MIOPEN_ELEMENTWISE_AFFINE
Definition miopen.h:472
MIOPEN_WEIGHT_BIAS_T5
@ MIOPEN_WEIGHT_BIAS_T5
Definition miopen.h:481
MIOPEN_ELEMENTWISE_AFFINE_T5
@ MIOPEN_ELEMENTWISE_AFFINE_T5
Definition miopen.h:479
MIOPEN_WEIGHT_BIAS_FUSED_ADD
@ MIOPEN_WEIGHT_BIAS_FUSED_ADD
Definition miopen.h:477
MIOPEN_REDUCE_EXTREME_ARGMAX
@ MIOPEN_REDUCE_EXTREME_ARGMAX
Definition miopen.h:6730
MIOPEN_REDUCE_EXTREME_MIN
@ MIOPEN_REDUCE_EXTREME_MIN
Definition miopen.h:6732
MIOPEN_REDUCE_EXTREME_MAX
@ MIOPEN_REDUCE_EXTREME_MAX
Definition miopen.h:6734
MIOPEN_REDUCE_EXTREME_ARGMIN
@ MIOPEN_REDUCE_EXTREME_ARGMIN
Definition miopen.h:6728
miopenTensorOpMin
@ miopenTensorOpMin
Definition miopen.h:406
miopenTensorOpAdd
@ miopenTensorOpAdd
Definition miopen.h:404
miopenTensorOpMul
@ miopenTensorOpMul
Definition miopen.h:405
miopenTensorOpMax
@ miopenTensorOpMax
Definition miopen.h:407
miopenBNSpatial
@ miopenBNSpatial
Definition miopen.h:492
miopenBNPerActivation
@ miopenBNPerActivation
Definition miopen.h:491
miopenConvolutionFindModeDynamicHybrid
@ miopenConvolutionFindModeDynamicHybrid
Definition miopen.h:652
miopenConvolutionFindModeDefault
@ miopenConvolutionFindModeDefault
Definition miopen.h:659
miopenConvolutionFindModeTrustVerifyFull
@ miopenConvolutionFindModeTrustVerifyFull
Definition miopen.h:658
miopenConvolutionFindModeTrustVerify
@ miopenConvolutionFindModeTrustVerify
Definition miopen.h:657
miopenConvolutionFindModeHybrid
@ miopenConvolutionFindModeHybrid
Definition miopen.h:647
miopenConvolutionFindModeFast
@ miopenConvolutionFindModeFast
Definition miopen.h:643
miopenConvolutionFindModeNormal
@ miopenConvolutionFindModeNormal
Definition miopen.h:641
miopenLRNWithinChannel
@ miopenLRNWithinChannel
Definition miopen.h:462
miopenLRNCrossChannel
@ miopenLRNCrossChannel
Definition miopen.h:463
MIOPEN_SOFTMAX_MODE_INSTANCE
@ MIOPEN_SOFTMAX_MODE_INSTANCE
Definition miopen.h:535
MIOPEN_SOFTMAX_MODE_CHANNEL
@ MIOPEN_SOFTMAX_MODE_CHANNEL
Definition miopen.h:536
MIOPEN_SOFTMAX_FAST
@ MIOPEN_SOFTMAX_FAST
Definition miopen.h:524
MIOPEN_SOFTMAX_ACCURATE
@ MIOPEN_SOFTMAX_ACCURATE
Definition miopen.h:525
MIOPEN_SOFTMAX_LOG
@ MIOPEN_SOFTMAX_LOG
Definition miopen.h:526
miopenConvAlgoPerf_t
Perf struct for forward, backward filter, or backward data algorithms.
Definition miopen.h:1269
miopenConvAlgoPerf_t::fwd_algo
miopenConvFwdAlgorithm_t fwd_algo
Definition miopen.h:1272
miopenConvAlgoPerf_t::bwd_data_algo
miopenConvBwdDataAlgorithm_t bwd_data_algo
Definition miopen.h:1276
miopenConvAlgoPerf_t::memory
size_t memory
Definition miopen.h:1280
miopenConvAlgoPerf_t::bwd_weights_algo
miopenConvBwdWeightsAlgorithm_t bwd_weights_algo
Definition miopen.h:1273
miopenConvAlgoPerf_t::time
float time
Definition miopen.h:1279
miopenConvSolution_t
Performance struct for forward, backward filter, or backward data algorithms in immediate mode.
Definition miopen.h:1293
miopenConvSolution_t::algorithm
miopenConvAlgorithm_t algorithm
Definition miopen.h:1300
miopenConvSolution_t::solution_id
uint64_t solution_id
Definition miopen.h:1299
miopenConvSolution_t::workspace_size
size_t workspace_size
Definition miopen.h:1297
miopenConvSolution_t::time
float time
Definition miopen.h:1294
miopenTensorArgument_t
Values of a tensor or scalar argument for the miopenRunSolution function.
Definition miopen.h:6465
miopenTensorArgument_t::id
miopenTensorArgumentId_t id
Definition miopen.h:6468
miopenTensorArgument_t::buffer
void * buffer
Definition miopen.h:6478
miopenTensorArgument_t::descriptor
miopenTensorDescriptor_t * descriptor
Definition miopen.h:6474