cusolver.h 4.4 KB
Newer Older
G
Guo Sheng 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
/* Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved.

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

    http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License. */
#pragma once

16
#include <cuda.h>
G
Guo Sheng 已提交
17 18 19 20
#include <cusolverDn.h>
#include <mutex>  // NOLINT

#include "paddle/fluid/platform/dynload/dynamic_loader.h"
W
wanghuancoder 已提交
21
#include "paddle/fluid/platform/port.h"
G
Guo Sheng 已提交
22 23 24 25 26 27

namespace paddle {
namespace platform {
namespace dynload {
extern std::once_flag cusolver_dso_flag;
extern void *cusolver_dso_handle;
28

G
Guo Sheng 已提交
29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50
#define DECLARE_DYNAMIC_LOAD_CUSOLVER_WRAP(__name)                   \
  struct DynLoad__##__name {                                         \
    template <typename... Args>                                      \
    cusolverStatus_t operator()(Args... args) {                      \
      using cusolverFunc = decltype(&::__name);                      \
      std::call_once(cusolver_dso_flag, []() {                       \
        cusolver_dso_handle =                                        \
            paddle::platform::dynload::GetCusolverDsoHandle();       \
      });                                                            \
      static void *p_##__name = dlsym(cusolver_dso_handle, #__name); \
      return reinterpret_cast<cusolverFunc>(p_##__name)(args...);    \
    }                                                                \
  };                                                                 \
  extern DynLoad__##__name __name

#define CUSOLVER_ROUTINE_EACH(__macro)  \
  __macro(cusolverDnCreate);            \
  __macro(cusolverDnDestroy);           \
  __macro(cusolverDnSetStream);         \
  __macro(cusolverDnSpotrf_bufferSize); \
  __macro(cusolverDnDpotrf_bufferSize); \
  __macro(cusolverDnSpotrf);            \
51
  __macro(cusolverDnDpotrf);            \
Z
zhiboniu 已提交
52 53 54 55
  __macro(cusolverDnSpotrs);            \
  __macro(cusolverDnDpotrs);            \
  __macro(cusolverDnCpotrs);            \
  __macro(cusolverDnZpotrs);            \
56 57 58 59 60 61 62 63
  __macro(cusolverDnSsyevd_bufferSize); \
  __macro(cusolverDnDsyevd_bufferSize); \
  __macro(cusolverDnCheevd_bufferSize); \
  __macro(cusolverDnZheevd_bufferSize); \
  __macro(cusolverDnSsyevd);            \
  __macro(cusolverDnDsyevd);            \
  __macro(cusolverDnCheevd);            \
  __macro(cusolverDnZheevd);
G
Guo Sheng 已提交
64 65 66 67 68 69

CUSOLVER_ROUTINE_EACH(DECLARE_DYNAMIC_LOAD_CUSOLVER_WRAP);

#if CUDA_VERSION >= 9020
#define CUSOLVER_ROUTINE_EACH_R1(__macro) \
  __macro(cusolverDnSpotrfBatched);       \
70
  __macro(cusolverDnDpotrfBatched);       \
Z
zhiboniu 已提交
71 72
  __macro(cusolverDnSpotrsBatched);       \
  __macro(cusolverDnDpotrsBatched);       \
73
  __macro(cusolverDnSgesvdj_bufferSize);  \
74 75 76 77 78 79 80 81
  __macro(cusolverDnSgeqrf_bufferSize);   \
  __macro(cusolverDnDgeqrf_bufferSize);   \
  __macro(cusolverDnCgeqrf_bufferSize);   \
  __macro(cusolverDnZgeqrf_bufferSize);   \
  __macro(cusolverDnSorgqr_bufferSize);   \
  __macro(cusolverDnDorgqr_bufferSize);   \
  __macro(cusolverDnCungqr_bufferSize);   \
  __macro(cusolverDnZungqr_bufferSize);   \
82 83 84 85
  __macro(cusolverDnDestroyGesvdjInfo);   \
  __macro(cusolverDnCreateGesvdjInfo);    \
  __macro(cusolverDnDgesvdj_bufferSize);  \
  __macro(cusolverDnSgesvdj);             \
86 87 88 89 90 91 92 93 94
  __macro(cusolverDnDgesvdj);             \
  __macro(cusolverDnSgeqrf);              \
  __macro(cusolverDnDgeqrf);              \
  __macro(cusolverDnCgeqrf);              \
  __macro(cusolverDnZgeqrf);              \
  __macro(cusolverDnSorgqr);              \
  __macro(cusolverDnDorgqr);              \
  __macro(cusolverDnCungqr);              \
  __macro(cusolverDnZungqr);
G
Guo Sheng 已提交
95 96 97 98

CUSOLVER_ROUTINE_EACH_R1(DECLARE_DYNAMIC_LOAD_CUSOLVER_WRAP)
#endif

99 100 101 102 103 104 105 106 107 108 109 110
#if CUDA_VERSION >= 9020
#define CUSOLVER_ROUTINE_EACH_R2(__macro) \
  __macro(cusolverDnCreateSyevjInfo);     \
  __macro(cusolverDnSsyevj_bufferSize);   \
  __macro(cusolverDnDsyevj_bufferSize);   \
  __macro(cusolverDnSsyevj);              \
  __macro(cusolverDnDsyevj);              \
  __macro(cusolverDnDestroySyevjInfo);

CUSOLVER_ROUTINE_EACH_R2(DECLARE_DYNAMIC_LOAD_CUSOLVER_WRAP)
#endif

111
#undef DECLARE_DYNAMIC_LOAD_CUSOLVER_WRAP
G
Guo Sheng 已提交
112 113 114
}  // namespace dynload
}  // namespace platform
}  // namespace paddle