device_context.cc 27.8 KB
Newer Older
1
/* Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved.
Q
qijun 已提交
2 3 4 5
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
6

Q
qijun 已提交
7 8 9 10 11
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. */
Y
Yi Wang 已提交
12
#include "paddle/fluid/platform/device_context.h"
13
#include <set>
14

15
#if defined(PADDLE_WITH_CUDA) || defined(PADDLE_WITH_HIP)
16
#include "paddle/fluid/memory/allocation/cuda_device_context_allocator.h"
S
sneaxiy 已提交
17
#include "paddle/fluid/platform/cuda_device_guard.h"
18
#endif
19
#include "glog/logging.h"
20
#include "paddle/fluid/platform/profiler.h"
21

22 23 24 25 26
namespace paddle {
namespace memory {

AllocationPtr Alloc(const platform::DeviceContext& dev_ctx, size_t size) {
  auto place = dev_ctx.GetPlace();
27
  if (size == 0) {
28 29
    return Alloc(place, size);
  }
30 31

  if (platform::is_gpu_place(place)) {
32
#if defined(PADDLE_WITH_CUDA) || defined(PADDLE_WITH_HIP)
33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50
    auto* default_dev_ctx = static_cast<platform::CUDADeviceContext*>(
        platform::DeviceContextPool::Instance().Get(place));
    auto& desired_dev_ctx =
        static_cast<const platform::CUDADeviceContext&>(dev_ctx);
    if (default_dev_ctx->stream() == desired_dev_ctx.stream()) {
      return Alloc(place, size);
    } else {
      return allocation::CUDADeviceContextAllocatorPool::Instance().Alloc(
          desired_dev_ctx, size);
    }
#else
    PADDLE_THROW(platform::errors::PermissionDenied(
        "Paddle can't use CUDA device since it's not compiled with CUDA,"
        "Please recompile or reinstall Paddle with GPU support."));
#endif
  } else if (platform::is_xpu_place(place)) {
#ifdef PADDLE_WITH_XPU
    // TODO(liuyuhui): Consider xpu stream later
51 52
    return Alloc(place, size);
#else
53 54 55
    PADDLE_THROW(platform::errors::PermissionDenied(
        "Paddle can't use XPU device since it's not compiled with XPU,"
        "Please recompile or reinstall Paddle with XPU support."));
56
#endif
57 58 59
  } else {
    return Alloc(place, size);
  }
60 61 62 63 64
}

}  // namespace memory
}  // namespace paddle

Q
qijun 已提交
65 66 67
namespace paddle {
namespace platform {

68
#if defined(PADDLE_WITH_CUDA) || defined(PADDLE_WITH_HIP)
69 70 71
bool allow_tf32_cublas = true;
void SetAllowTF32Cublas(bool active) { allow_tf32_cublas = active; }
bool AllowTF32Cublas() { return allow_tf32_cublas; }
A
AshburnLee 已提交
72 73 74 75

bool allow_tf32_cudnn = true;
void SetAllowTF32Cudnn(bool active) { allow_tf32_cudnn = active; }
bool AllowTF32Cudnn() { return allow_tf32_cudnn; }
76 77
#endif  // PADDLE_WITH_CUDA

78 79 80 81 82 83 84 85 86 87 88 89 90
DeviceType Place2DeviceType(const platform::Place& place) {
  if (platform::is_cpu_place(place)) {
    return platform::DeviceType::CPU;
  } else if (platform::is_gpu_place(place)) {
    return platform::DeviceType::CUDA;
  } else if (platform::is_xpu_place(place)) {
    return platform::DeviceType::XPU;
  } else {
    PADDLE_THROW(platform::errors::Unavailable(
        "Unsupported place %s to convert into platform::DeviceType.", place));
  }
}

D
dzhwinter 已提交
91 92
DeviceContextPool* DeviceContextPool::pool = nullptr;

Y
Yu Yang 已提交
93
platform::DeviceContext* DeviceContextPool::Get(const platform::Place& place) {
94
  VLOG(6) << "DeviceContextPool Get: " << place;
D
dzhwinter 已提交
95 96
  auto it = device_contexts_.find(place);
  if (it == device_contexts_.end()) {
G
GaoWei8 已提交
97 98
    PADDLE_THROW(platform::errors::Unimplemented(
        "Place %s is not supported. Please check that your paddle compiles "
99 100
        "with WITH_GPU, WITH_XPU or WITH_ASCEND_CL option or check that "
        "your train process set the correct device id if you use Executor.",
G
GaoWei8 已提交
101
        place));
D
dzhwinter 已提交
102
  }
103
  return it->second.get().get();
D
dzhwinter 已提交
104 105
}

106 107 108 109 110 111 112 113 114
template <typename DevCtx, typename PlaceType>
inline void EmplaceDeviceContext(
    std::map<Place, std::shared_future<std::unique_ptr<DeviceContext>>>*
        map_ptr,
    platform::Place p) {
  using PtrType = std::unique_ptr<DeviceContext>;
  map_ptr->emplace(p, std::async(std::launch::deferred, [=] {
                     // lazy evaluation. i.e., only create device context at
                     // first `Get`
115
                     return PtrType(new DevCtx(BOOST_GET_CONST(PlaceType, p)));
116
                   }));
C
chengduozh 已提交
117 118
}

D
dzhwinter 已提交
119 120
DeviceContextPool::DeviceContextPool(
    const std::vector<platform::Place>& places) {
G
GaoWei8 已提交
121 122 123 124 125
  PADDLE_ENFORCE_GT(
      places.size(), 0,
      platform::errors::InvalidArgument("The number of platform places should "
                                        "be larger than 0. But received %d.",
                                        places.size()));
126
  std::set<Place> set;
Y
Yu Yang 已提交
127 128 129 130 131
  for (auto& p : places) {
    set.insert(p);
  }
  for (auto& p : set) {
    if (platform::is_cpu_place(p)) {
132
#ifdef PADDLE_WITH_MKLDNN
133
      EmplaceDeviceContext<MKLDNNDeviceContext, CPUPlace>(&device_contexts_, p);
134
#else
135
      EmplaceDeviceContext<CPUDeviceContext, CPUPlace>(&device_contexts_, p);
136
#endif
Y
Yu Yang 已提交
137
    } else if (platform::is_gpu_place(p)) {
138
#if defined(PADDLE_WITH_CUDA) || defined(PADDLE_WITH_HIP)
139
      EmplaceDeviceContext<CUDADeviceContext, CUDAPlace>(&device_contexts_, p);
D
dzhwinter 已提交
140
#else
G
GaoWei8 已提交
141 142 143
      PADDLE_THROW(
          platform::errors::Unimplemented("CUDAPlace is not supported. Please "
                                          "re-compile with WITH_GPU option."));
C
chengduoZH 已提交
144 145
#endif
    } else if (platform::is_cuda_pinned_place(p)) {
146
#if defined(PADDLE_WITH_CUDA) || defined(PADDLE_WITH_HIP)
147 148
      EmplaceDeviceContext<CUDAPinnedDeviceContext, CUDAPinnedPlace>(
          &device_contexts_, p);
C
chengduoZH 已提交
149
#else
G
GaoWei8 已提交
150
      PADDLE_THROW(platform::errors::Unimplemented(
G
GaoWei8 已提交
151 152
          "CUDAPlace is not supported. Please re-compile with WITH_GPU "
          "option."));
153 154 155 156 157 158 159 160
#endif
    } else if (platform::is_xpu_place(p)) {
#ifdef PADDLE_WITH_XPU
      EmplaceDeviceContext<XPUDeviceContext, XPUPlace>(&device_contexts_, p);
#else
      PADDLE_THROW(
          platform::errors::Unimplemented("XPUPlace is not supported. Please "
                                          "re-compile with WITH_XPU option."));
161 162 163 164 165 166 167 168
#endif
    } else if (platform::is_npu_place(p)) {
#ifdef PADDLE_WITH_ASCEND_CL
      EmplaceDeviceContext<NPUDeviceContext, NPUPlace>(&device_contexts_, p);
#else
      PADDLE_THROW(platform::errors::Unimplemented(
          "NPUPlace is not supported. Please "
          "re-compile with WITH_ASCEND_CL option."));
169 170 171 172 173 174 175 176 177 178
#endif
    } else if (platform::is_npu_pinned_place(p)) {
#ifdef PADDLE_WITH_ASCEND_CL
      EmplaceDeviceContext<NPUPinnedDeviceContext, NPUPinnedPlace>(
          &device_contexts_, p);
#else
      PADDLE_THROW(platform::errors::Unimplemented(
          "NPUPinnedPlace is not supported. Please re-compile with "
          "WITH_ASCEND_CL "
          "option."));
D
dzhwinter 已提交
179 180 181 182 183
#endif
    }
  }
}

184 185 186 187
CPUDeviceContext::CPUDeviceContext() {
  eigen_device_.reset(new Eigen::DefaultDevice());
}

D
dzhwinter 已提交
188
CPUDeviceContext::CPUDeviceContext(CPUPlace place) : place_(place) {
189 190 191 192 193 194 195
  eigen_device_.reset(new Eigen::DefaultDevice());
}

Eigen::DefaultDevice* CPUDeviceContext::eigen_device() const {
  return eigen_device_.get();
}

D
dzhwinter 已提交
196
Place CPUDeviceContext::GetPlace() const { return place_; }
197

198
#ifdef PADDLE_WITH_XPU
Q
QingshuChen 已提交
199 200 201 202
XPUDeviceContext::XPUDeviceContext() {
  context_ = xpu::create_context();
  xpu_version_ = get_xpu_version(place_.device);
}
203

204
XPUDeviceContext::~XPUDeviceContext() {}
205 206 207 208 209 210 211 212 213 214 215 216 217 218 219

XPUDeviceContext::XPUDeviceContext(XPUPlace place) : place_(place) {
  int dev_id = -1;
  int ret = xpu_current_device(&dev_id);
  PADDLE_ENFORCE_EQ(ret, XPU_SUCCESS,
                    platform::errors::External(
                        "XPU API return wrong value[%d], please check whether "
                        "Baidu Kunlun Card is properly installed.",
                        ret));
  ret = xpu_set_device(place.device);
  PADDLE_ENFORCE_EQ(ret, XPU_SUCCESS,
                    platform::errors::External(
                        "XPU API return wrong value[%d], please check whether "
                        "Baidu Kunlun Card is properly installed.",
                        ret));
220 221 222

  LOG_FIRST_N(WARNING, 1) << "Please NOTE: xpu device: " << place_.device;

223
  context_ = xpu::create_context();
224 225 226
  const int MAX_XPU_NUM = 16;
  static void* l3ptrs[MAX_XPU_NUM] = {nullptr};

227 228 229 230 231
  int l3_size = 13.5 * 1024 * 1024;
  if (std::getenv("XPU_PADDLE_L3_SIZE") != nullptr) {
    l3_size = atoi(std::getenv("XPU_PADDLE_L3_SIZE"));
  }

232 233 234 235 236 237 238 239 240 241 242 243 244
  auto selected_xpus = GetXPUSelectedDevices();
  for (unsigned int i = 0; i < selected_xpus.size(); i++) {
    if (place.device == selected_xpus[i]) {
      if (l3ptrs[place.device] == nullptr) {
        xpu_malloc(static_cast<void**>(&l3ptrs[place.device]), l3_size,
                   XPU_MEM_L3);
      }
      if (l3ptrs[place.device] != nullptr) {
        context_->_l3_mgr.set(l3ptrs[place.device], l3_size);
        VLOG(3) << "xpu place " << place.device << " set l3 size " << l3_size;
      }
      break;
    }
245
  }
246

247 248 249 250 251 252 253 254 255 256 257 258 259 260 261
  ret = xpu_set_device(dev_id);
  PADDLE_ENFORCE_EQ(ret, XPU_SUCCESS,
                    platform::errors::External(
                        "XPU API return wrong value[%d], please check whether "
                        "Baidu Kunlun Card is properly installed.",
                        ret));
}

void XPUDeviceContext::Wait() const {
  int ret = xpu_set_device(place_.device);
  PADDLE_ENFORCE_EQ(ret, XPU_SUCCESS,
                    platform::errors::External(
                        "XPU API return wrong value[%d], please check whether "
                        "Baidu Kunlun Card is properly installed.",
                        ret));
262
  xpu_wait(context_->xpu_stream);
263 264 265 266 267 268 269
}

Place XPUDeviceContext::GetPlace() const { return place_; }

xpu::Context* XPUDeviceContext::x_context() const { return context_; }
#endif

270 271 272 273 274 275 276 277 278 279 280 281 282 283 284
#ifdef PADDLE_WITH_ASCEND_CL
NPUDeviceContext::NPUDeviceContext(NPUPlace place) : place_(place) {
  NPUDeviceGuard guard(place_.device);
  // PADDLE_ENFORCE_NPU_SUCCESS(aclrtCreateContext(&context_, place_.device));
  // NOTE(zhiqiu): Usually, no need to create context explicitly,
  // ACL creates a default context which contains 1 default stream
  // and 1 sync strean after aclrtSetDevice.
  PADDLE_ENFORCE_NPU_SUCCESS(aclrtGetCurrentContext(&context_));
  stream_.reset(new stream::NPUStream(place));
}

NPUDeviceContext::~NPUDeviceContext() {
  // NPUDeviceGuard guard(place_.device);
  // PADDLE_ENFORCE_NPU_SUCCESS(aclrtDestroyContext(context_));
}
285

286
void NPUDeviceContext::Wait() const {
287 288 289
  platform::RecordEvent record_event("NPUDeviceContext/wait");
  VLOG(4) << "NPU context(" << this << ")  Wait";
  stream_->Wait();
290 291 292 293 294 295 296
}

aclrtStream NPUDeviceContext::stream() const { return stream_->raw_stream(); }

Place NPUDeviceContext::GetPlace() const { return place_; }

aclrtContext NPUDeviceContext::context() const { return context_; }
297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312

NPUPinnedDeviceContext::NPUPinnedDeviceContext() {
  eigen_device_.reset(new Eigen::DefaultDevice());
}

NPUPinnedDeviceContext::NPUPinnedDeviceContext(NPUPinnedPlace place)
    : place_(place) {
  eigen_device_.reset(new Eigen::DefaultDevice());
}

Eigen::DefaultDevice* NPUPinnedDeviceContext::eigen_device() const {
  return eigen_device_.get();
}

Place NPUPinnedDeviceContext::GetPlace() const { return place_; }

313 314 315
#endif

#if defined(PADDLE_WITH_CUDA) || defined(PADDLE_WITH_HIP)
Q
init  
qijun 已提交
316 317 318 319 320 321 322
class EigenCudaStreamDevice : public Eigen::StreamInterface {
 public:
  EigenCudaStreamDevice() : scratch_(nullptr), semaphore_(nullptr) {
    Eigen::initializeDeviceProp();
  }
  ~EigenCudaStreamDevice() override {}

323
  void Reinitialize(const gpuStream_t* cuda_stream, CUDAPlace place) {
Q
init  
qijun 已提交
324 325 326 327 328
    stream_ = cuda_stream;
    place_ = place;
    device_prop_ = &Eigen::m_deviceProperties[place.device];
  }

329
  const gpuStream_t& stream() const override { return *stream_; }
Q
init  
qijun 已提交
330

331 332 333
#ifdef PADDLE_WITH_HIP
  const hipDeviceProp_t& deviceProperties() const override {
#else
Q
init  
qijun 已提交
334
  const cudaDeviceProp& deviceProperties() const override {
335
#endif
Q
init  
qijun 已提交
336 337 338 339
    return *device_prop_;
  }

  void* allocate(size_t num_bytes) const override {
S
sneaxiy 已提交
340 341 342
    if (UNLIKELY(num_bytes == 0)) {
      return nullptr;
    }
343 344 345
    auto buf = memory::Alloc(place_, num_bytes);
    VLOG(4) << "Eigen allocated at " << buf->ptr() << ", size" << buf->size()
            << " requested " << num_bytes;
346
    void* retv = buf->ptr();
S
sneaxiy 已提交
347 348 349 350
    {
      std::lock_guard<std::mutex> lock(mtx_);
      allocations_.emplace(retv, std::move(buf));
    }
351
    return retv;
Q
init  
qijun 已提交
352 353
  }

S
sneaxiy 已提交
354 355 356 357 358 359
  void deallocate(void* buffer) const override {
    if (LIKELY(buffer)) {
      std::lock_guard<std::mutex> lock(mtx_);
      allocations_.erase(buffer);
    }
  }
Q
init  
qijun 已提交
360 361 362

  void* scratchpad() const override {
    if (scratch_ == NULL) {
Z
Zhang Ting 已提交
363
      scratch_ = allocate(Eigen::kGpuScratchSize + sizeof(unsigned int));
Q
init  
qijun 已提交
364 365 366 367 368 369
    }
    return scratch_;
  }

  unsigned int* semaphore() const override {
    if (semaphore_ == NULL) {
Z
Zhang Ting 已提交
370
      char* scratch = static_cast<char*>(scratchpad()) + Eigen::kGpuScratchSize;
Q
init  
qijun 已提交
371
      semaphore_ = reinterpret_cast<unsigned int*>(scratch);
372 373 374 375
#ifdef PADDLE_WITH_HIP
      PADDLE_ENFORCE_CUDA_SUCCESS(
          hipMemsetAsync(semaphore_, 0, sizeof(unsigned int), *stream_));
#else
376
      PADDLE_ENFORCE_CUDA_SUCCESS(
Q
init  
qijun 已提交
377
          cudaMemsetAsync(semaphore_, 0, sizeof(unsigned int), *stream_));
378
#endif
Q
init  
qijun 已提交
379 380 381 382 383
    }
    return semaphore_;
  }

 private:
D
dzhwinter 已提交
384
  CUDAPlace place_;
385 386 387 388
  const gpuStream_t* stream_;  // not owned;
#ifdef PADDLE_WITH_HIP
  const hipDeviceProp_t* device_prop_;
#else
Q
init  
qijun 已提交
389
  const cudaDeviceProp* device_prop_;  // not owned;
390
#endif
Q
qijun 已提交
391
  mutable void* scratch_;
Q
init  
qijun 已提交
392
  mutable unsigned int* semaphore_;
S
sneaxiy 已提交
393
  mutable std::mutex mtx_;  // to protect allocations_
Y
Yu Yang 已提交
394
  mutable std::unordered_map<void*, memory::AllocationPtr> allocations_;
Q
init  
qijun 已提交
395 396
};

397 398 399 400 401 402 403 404 405
void CudnnWorkspaceHandle::ReallocWorkspace(size_t required_workspace_bytes) {
  if (required_workspace_bytes <= WorkspaceSize()) {
    return;
  }
  // reset allocation first before re-allocate to save memory
  allocation_.reset();
  allocation_ = memory::Alloc(device_context_, required_workspace_bytes);
}

406 407 408 409 410 411 412 413 414 415 416 417
thread_local std::unordered_map<const CUDADeviceContext*,
                                std::shared_ptr<CUDAContext>>
    CUDADeviceContext::thread_ctx_;
thread_local std::mutex CUDADeviceContext::ctx_mtx_;

void CUDAContext::InitEigenContext() {
  eigen_stream_.reset(new EigenCudaStreamDevice());
  eigen_stream_->Reinitialize(&RawStream(), place_);
  eigen_device_.reset(new Eigen::GpuDevice(eigen_stream_.get()));
}

CUDAContext::CUDAContext(const CUDAPlace& place,
418 419
                         const stream::Priority& priority,
                         const stream::StreamFlag& flag) {
420 421
  place_ = place;
  CUDADeviceGuard guard(place_.device);
422
  stream_.reset(new stream::CUDAStream(place, priority, flag));
423 424 425
  InitEigenContext();
  InitCuBlasContext();
  InitCuDNNContext();
426
#ifndef PADDLE_WITH_HIP
G
Guo Sheng 已提交
427
  InitCuSolverContext();
428
#endif
429 430 431 432 433 434
}

CUDAContext::~CUDAContext() {
  CUDADeviceGuard guard(place_.device);
  DestoryCuDNNContext();
  DestoryCuBlasContext();
435
#ifndef PADDLE_WITH_HIP
G
Guo Sheng 已提交
436
  DestoryCuSolverContext();
437
#endif
438 439
}

440
CUDADeviceContext::CUDADeviceContext(CUDAPlace place) : place_(place) {
Y
Yu Yang 已提交
441
  CUDADeviceGuard guard(place_.device);
C
chengduo 已提交
442 443 444
  compute_capability_ = GetCUDAComputeCapability(place_.device);
  multi_process_ = GetCUDAMultiProcessors(place_.device);
  max_threads_per_mp_ = GetCUDAMaxThreadsPerMultiProcessor(place_.device);
445
  max_grid_dim_size_ = GetGpuMaxGridDimSize(place_.device);
446
  max_threads_per_block_ = GetCUDAMaxThreadsPerBlock(place_.device);
447

C
chengduo 已提交
448 449 450
  driver_version_ = GetCUDADriverVersion(place_.device);
  runtime_version_ = GetCUDARuntimeVersion(place_.device);

451
  LOG_FIRST_N(WARNING, 1) << "Please NOTE: device: " << place_.device
452 453 454
                          << ", GPU Compute Capability: "
                          << compute_capability_ / 10 << "."
                          << compute_capability_ % 10
C
chengduo 已提交
455
                          << ", Driver API Version: " << driver_version_ / 1000
456
                          << "." << (driver_version_ % 100) / 10
C
chengduo 已提交
457 458 459
                          << ", Runtime API Version: "
                          << runtime_version_ / 1000 << "."
                          << (runtime_version_ % 100) / 10;
460 461 462 463 464 465 466 467
#ifdef PADDLE_WITH_HIP
  size_t version_major, version_minor, version_patch;
  PADDLE_ENFORCE_CUDA_SUCCESS(dynload::miopenGetVersion(
      &version_major, &version_minor, &version_patch));
  LOG_FIRST_N(WARNING, 1) << "device: " << place_.device
                          << ", MIOpen Version: " << version_major << "."
                          << version_minor << "." << version_patch;
#else
468 469 470
  size_t cudnn_dso_ver = dynload::cudnnGetVersion();
  LOG_FIRST_N(WARNING, 1) << "device: " << place_.device
                          << ", cuDNN Version: " << cudnn_dso_ver / 1000 << "."
471
                          << (cudnn_dso_ver % 1000) / 100 << ".";
472
#endif
S
sneaxiy 已提交
473 474
  {
    // Check CUDA/CUDNN version compatiblity
475 476
    auto local_cuda_version =
        (driver_version_ / 1000) * 10 + (driver_version_ % 100) / 10;
477 478 479
#ifdef PADDLE_WITH_HIP
    auto compile_cuda_version = (HIP_VERSION / 100) * 10 + (HIP_VERSION % 10);
#else
480 481
    auto compile_cuda_version =
        (CUDA_VERSION / 1000) * 10 + (CUDA_VERSION % 100) / 10;
482
#endif
S
sneaxiy 已提交
483 484 485 486 487 488 489 490 491 492 493 494
    if (local_cuda_version < compile_cuda_version) {
      LOG_FIRST_N(WARNING, 1)
          << "WARNING: device: " << place_.device
          << ". The installed Paddle is compiled with CUDA "
          << compile_cuda_version / 10 << "." << compile_cuda_version % 10
          << ", but CUDA runtime version in your machine is "
          << local_cuda_version / 10 << "." << local_cuda_version % 10
          << ", which may cause serious incompatible bug. "
          << "Please recompile or reinstall Paddle with compatible CUDA "
             "version.";
    }
  }
495
  default_ctx_.reset(new CUDAContext(place_));
496 497 498 499
}

CUDADeviceContext::~CUDADeviceContext() {
  SetDeviceId(place_.device);
500
#if defined(PADDLE_WITH_NCCL) || defined(PADDLE_WITH_RCCL)
501 502 503 504
  if (nccl_comm_) {
    PADDLE_ENFORCE_CUDA_SUCCESS(dynload::ncclCommDestroy(nccl_comm_));
  }
#endif
505 506
}

L
liaogang 已提交
507
Place CUDADeviceContext::GetPlace() const { return place_; }
508

509
void CUDADeviceContext::Wait() const { context()->Stream()->Wait(); }
510

K
Kexin Zhao 已提交
511
int CUDADeviceContext::GetComputeCapability() const {
C
chengduo 已提交
512
  return compute_capability_;
K
Kexin Zhao 已提交
513 514
}

515
int CUDADeviceContext::GetMaxPhysicalThreadCount() const {
C
chengduo 已提交
516
  return multi_process_ * max_threads_per_mp_;
517 518
}

519 520 521 522 523 524
int CUDADeviceContext::GetSMCount() const { return multi_process_; }

int CUDADeviceContext::GetMaxThreadsPerBlock() const {
  return max_threads_per_block_;
}

525
Eigen::GpuDevice* CUDADeviceContext::eigen_device() const {
526
  return context()->EigenDevice().get();
527 528
}

529
bool CUDADeviceContext::tensor_core_available() const {
530
  return context()->CublasTensorCoreHandle() != nullptr;
S
sneaxiy 已提交
531 532
}

533 534 535 536
dim3 CUDADeviceContext::GetCUDAMaxGridDimSize() const {
  return max_grid_dim_size_;
}

537 538 539
#ifdef PADDLE_WITH_HIP
miopenHandle_t CUDADeviceContext::cudnn_handle() const {
#else
540
cudnnHandle_t CUDADeviceContext::cudnn_handle() const {
541
#endif
542 543
  return context()->CudnnHandle();
}
544

545 546 547 548 549
#ifdef PADDLE_WITH_HIP
rocblas_handle CUDADeviceContext::cublas_handle() const {
  return context()->CublasHandle()->GetCublasHandle();
}
#else
550 551 552
cublasHandle_t CUDADeviceContext::cublas_handle() const {
  return context()->CublasHandle()->GetCublasHandle();
}
553
#endif
554

S
sneaxiy 已提交
555
CudnnWorkspaceHandle CUDADeviceContext::cudnn_workspace_handle() const {
556
  return CudnnWorkspaceHandle(*this, &cudnn_handle_mtx_);
557
}
558

559
#ifndef PADDLE_WITH_HIP
G
Guo Sheng 已提交
560 561 562
cusolverDnHandle_t CUDADeviceContext::cusolver_dn_handle() const {
  return context()->CusolverDnHandle();
}
563
#endif
G
Guo Sheng 已提交
564

565
gpuStream_t CUDADeviceContext::stream() const { return context()->RawStream(); }
Q
qijun 已提交
566

C
chengduoZH 已提交
567 568 569 570 571 572 573 574 575 576 577 578 579 580
CUDAPinnedDeviceContext::CUDAPinnedDeviceContext() {
  eigen_device_.reset(new Eigen::DefaultDevice());
}

CUDAPinnedDeviceContext::CUDAPinnedDeviceContext(CUDAPinnedPlace place)
    : place_(place) {
  eigen_device_.reset(new Eigen::DefaultDevice());
}

Eigen::DefaultDevice* CUDAPinnedDeviceContext::eigen_device() const {
  return eigen_device_.get();
}

Place CUDAPinnedDeviceContext::GetPlace() const { return place_; }
L
Luo Tao 已提交
581
#endif
Q
qijun 已提交
582

T
tensor-tang 已提交
583 584
#ifdef PADDLE_WITH_MKLDNN
MKLDNNDeviceContext::MKLDNNDeviceContext(CPUPlace place)
585
    : CPUDeviceContext(place), p_blobmap_() {
586
  p_blobmap_.reset(new BlobMap());
587
  p_exec_items_.reset(new ExecShape());
588
  p_mutex_.reset(new std::mutex());
T
tensor-tang 已提交
589 590
}

591 592
MKLDNNDeviceContextThreadLocals::Body::Body()
    : cur_engine(mkldnn::engine::kind::cpu, 0), cur_stream(cur_engine) {
593 594 595 596 597 598
  cur_mkldnn_session_id = kMKLDNNSessionID_Default;
  cur_input_shape_str = "";
  cur_input_shape_cache_capacity = 1;
  cur_paddle_data_layout = paddle::framework::DataLayout::kNCHW;
}

599 600 601 602 603 604 605 606 607 608 609 610
// When Thread finish we clear oneDNN cache
// This is needed when we have one executor used by many threads
// e.g. test_analyzer_detect. Thread ID is not part of caching key
// (for naive executor) so we need to clear cache when one thread finish
// and other is to start inference
// TODO(jczaja): Ideally it would be good to clear only part of cache
// related to thread that is to be terminated
MKLDNNDeviceContextThreadLocals::Body::~Body() {
  auto cpu_place = paddle::platform::CPUPlace();
  platform::DeviceContextPool& pool = platform::DeviceContextPool::Instance();
  platform::MKLDNNDeviceContext* dev_ctx =
      (platform::MKLDNNDeviceContext*)pool.Get(cpu_place);
611
  dev_ctx->ResetBlobMap(exec_ptr_);
612 613
}

614 615 616 617 618 619 620 621 622 623
void MKLDNNDeviceContextThreadLocals::Body::set_cur_mkldnn_session_id(
    size_t sid) {
  cur_mkldnn_session_id = sid;
}
size_t MKLDNNDeviceContextThreadLocals::Body::get_cur_mkldnn_session_id(void) {
  return cur_mkldnn_session_id;
}

void MKLDNNDeviceContextThreadLocals::Body::set_cur_input_shape_str(
    std::string input_shape_str) {
624 625
  cur_input_shape_str = input_shape_str;
}
626 627
void MKLDNNDeviceContextThreadLocals::Body::set_cur_input_shape_cache_capacity(
    int input_shape_cache_capacity) {
628 629
  cur_input_shape_cache_capacity = input_shape_cache_capacity;
}
S
Sylwester Fraczek 已提交
630

631 632
void MKLDNNDeviceContextThreadLocals::Body::set_cur_paddle_data_layout(
    framework::DataLayout dl) {
633 634 635
  cur_paddle_data_layout = dl;
}

636 637
framework::DataLayout
MKLDNNDeviceContextThreadLocals::Body::get_cur_paddle_data_layout(void) {
638 639 640
  return cur_paddle_data_layout;
}

641 642 643 644 645 646 647 648 649
void MKLDNNDeviceContextThreadLocals::Body::log_lib_version(void) {
  if (!said_once) {
    said_once = true;
    auto dv = dnnl::version();
    LOG(INFO) << "oneDNN v" << dv->major << "." << dv->minor << "."
              << dv->patch;
  }
}

650 651 652 653 654 655 656 657
const mkldnn::engine& MKLDNNDeviceContextThreadLocals::Body::get_engine(void) {
  return cur_engine;
}

mkldnn::stream& MKLDNNDeviceContextThreadLocals::Body::get_stream(void) {
  return cur_stream;
}

658
void MKLDNNDeviceContext::ResetBlobMap(void* ptr) {
659 660 661
  std::lock_guard<decltype(*p_mutex_)> lock(*p_mutex_);
  if (!block_next_cache_clearing_) {
    VLOG(3) << "Clearing DNNL cache.";
662 663 664 665 666 667
    // If no specific executor pointer then clear
    // everything. For executor pointer then clear only
    // objects allocated when using given executor
    if (ptr == nullptr) {
      p_blobmap_->clear();
    } else {
668 669 670 671 672
      // Iterate through all shapes and release
      // for each shape and active executor all entries
      // of this executor
      for (auto& s : *p_exec_items_) {
        for (auto& v : (*s.second)[ptr]) {
673
          (v.first)->erase(v.second);
674 675
        }
        s.second->erase(ptr);
676 677
      }
    }
678 679 680 681 682 683
  } else {
    VLOG(3) << "Prevented Clearing DNNL cache.";
    block_next_cache_clearing_ = false;
  }
}

684 685
void MKLDNNDeviceContext::RemoveShapeEntriesWithExecutor(void) const {
  p_exec_items_->erase(p_exec_items_->begin());
686 687
}

688 689
void MKLDNNDeviceContext::LinkEntryWithExecutor(BlobPtr_t<KeyBlob> pblob,
                                                KeyBlob::iterator it) const {
690
  // Take current input shape from TLS
691 692
  // Take current executor addess from TLS
  // and for this executor's items add the one defined with arguments
693 694 695 696 697 698 699 700 701
  auto key_it = p_exec_items_
                    ->insert(std::make_pair(tls().cur_input_shape_str,
                                            std::make_shared<ExecMap>()))
                    .first;
  (*key_it->second)[tls().get_curr_exec()].push_back(std::make_pair(pblob, it));

  VLOG(3) << "LinkEntryWithExecutor, shapes: " << p_exec_items_->size()
          << " curr exec size: "
          << (*key_it->second)[tls().get_curr_exec()].size() << "\n";
702 703
}

704 705 706 707
void MKLDNNDeviceContext::BlockNextCacheClearing() {
  std::lock_guard<decltype(*p_mutex_)> lock(*p_mutex_);
  VLOG(3) << "Next DNNL cache clearing has been blocked.";
  block_next_cache_clearing_ = true;
708
}
709

710
size_t MKLDNNDeviceContext::GetShapeBlobSize() const {
711
  std::lock_guard<decltype(*p_mutex_)> lock(*p_mutex_);
712
  BlobMap* pMap = p_blobmap_.get();
713
  auto map_it = pMap->find(tls().cur_mkldnn_session_id);
714
  if (map_it == pMap->end()) {
715 716 717
    PADDLE_THROW(platform::errors::NotFound(
        "MKLDNNDeviceContext don't find cur_mkldnn_session_id: %d.",
        tls().cur_mkldnn_session_id));
718 719 720 721
  }
  return map_it->second->size();
}

722
void MKLDNNDeviceContext::SetBlob(const std::string& name,
723
                                  BlobPtr_t<void> data) const {
724
  BlobMap* pMap = p_blobmap_.get();
725
  BlobPtr_t<ShapeBlob> sBlob = nullptr;
726
  BlobPtr_t<KeyBlob> pBlob = nullptr;
727

728
  int sid = tls().get_cur_mkldnn_session_id();
T
tensor-tang 已提交
729

730
  std::lock_guard<decltype(*p_mutex_)> lock(*p_mutex_);
T
tensor-tang 已提交
731

732 733
  // Find ShapeBlob for current mkldnn session id.
  auto map_it = pMap->find(sid);
734 735 736

  if (map_it == pMap->end()) {
    // 1st time to set blob in current thread
737
    sBlob = std::make_shared<ShapeBlob>();
738 739
    (*pMap)[sid] = sBlob;
    VLOG(2) << "SetBlob: sid=" << sid << ", add new sid\n";
740
  } else {
741
    sBlob = map_it->second;
742
  }
T
tensor-tang 已提交
743

744
  // Find KeyBlob for current input shape
745
  auto key_it = sBlob->find(tls().cur_input_shape_str);
746

747
  if (key_it == sBlob->end()) {
748 749
    // In cache clearing mode, cur_input_shape_cache_capacity defines
    // max pblob capacity
750 751
    if ((static_cast<size_t>(sid) ==
         MKLDNNDeviceContextThreadLocals::kMKLDNNSessionID_CacheClearing) &&
752
        sBlob->size() &&
753
        (sBlob->size() >=
754
         static_cast<size_t>(tls().cur_input_shape_cache_capacity))) {
755 756 757 758
      VLOG(2) << "sid=" << sid
              << ", remove all blobs of shape: " << sBlob->begin()->first;
      sBlob->erase(sBlob->begin()->first);
      RemoveShapeEntriesWithExecutor();
759
    }
760
    pBlob = std::make_shared<KeyBlob>();
761
    (*sBlob)[tls().cur_input_shape_str] = pBlob;
762
  } else {
763
    pBlob = key_it->second;
764 765
  }

766
  // Find Blob via name
767 768 769 770
  auto blob_it = pBlob->find(name);
  if (blob_it == pBlob->end()) {
    auto el =
        pBlob->insert(std::make_pair(name, data));  //  (*pBlob)[name] = data;
771 772 773
    // Register new element in per executor map
    // to have easily erased when executor terminated
    LinkEntryWithExecutor(pBlob, el.first);
774 775 776
  } else {
    blob_it->second = data;  // set data to existing blob
  }
777
  VLOG(2) << "SetBlob: sid=" << sid << ", add blob=" << name << "\n";
778
  // lock will be automatically released when out of scope
779
  return;
T
tensor-tang 已提交
780 781
}

782
unsigned int MKLDNNDeviceContext::GetCachedObjectsNumber(void) const {
783 784 785
  unsigned int num_entries = 0;
  for (auto const& l3 : *p_blobmap_) {
    for (auto const& l2 : *(l3.second)) {
786
      num_entries += (l2.second)->size();
787 788 789 790 791
    }
  }
  return num_entries;
}

792
MKLDNNDeviceContext::BlobPtr_t<void> MKLDNNDeviceContext::GetBlob(
793
    const std::string& name) const {
794
  BlobMap* pMap = p_blobmap_.get();
795
  BlobPtr_t<ShapeBlob> sBlob = nullptr;
796
  BlobPtr_t<KeyBlob> pBlob = nullptr;
T
tensor-tang 已提交
797

798
  int sid = tls().get_cur_mkldnn_session_id();
T
tensor-tang 已提交
799

800
  std::lock_guard<decltype(*p_mutex_)> lock(*p_mutex_);
801

802 803
  // Find ShapeBlob for current mkldnn session id firstly
  auto map_it = pMap->find(sid);
804
  if (map_it == pMap->end()) {
805
    VLOG(2) << "GetBlob: sid=" << sid << ", miss sid\n";
806 807 808 809 810
    return nullptr;
  }
  sBlob = map_it->second;

  // Find KeyBlob for current input shape secondly
811
  auto sBlob_it = sBlob->find(tls().cur_input_shape_str);
812
  if (sBlob_it == sBlob->end()) {
813
    VLOG(2) << "GetBlob: sid=" << tls().cur_input_shape_str
814 815 816 817
            << ", miss input_shape_str\n";
    return nullptr;
  }
  pBlob = sBlob_it->second;
818 819

  // Find Blob via name
820
  auto key_it = pBlob->find(name);
821

822
  if (key_it == pBlob->end()) {
823
    VLOG(2) << "GetBlob sid=" << sid << ", miss blob=" << name << "\n";
824 825
    return nullptr;
  }
826

827
  VLOG(2) << "GetBlob sid=" << sid << ", get blob=" << name << "\n";
828 829
  // lock will be automatically released when out of scope
  return key_it->second;
T
tensor-tang 已提交
830 831 832
}

#endif
Q
qijun 已提交
833
}  // namespace platform
Q
qijun 已提交
834
}  // namespace paddle