merge develop

paddle/fluid/operators/math/blas_impl.cu.h

@@ -62,27 +62,19 @@ struct CUBlas<float> {
                       cudaDataType_t Atype, int lda, const void *B,
                       cudaDataType_t Btype, int ldb, const float *beta, void *C,
                       cudaDataType_t Ctype, int ldc) {
-    // Because the gcc 4.8 doesn't expand template parameter pack that
-    // appears in a lambda-expression, I can not use template parameter pack
-    // here.
-    auto cublas_call = [&]() {
+// Because the gcc 4.8 doesn't expand template parameter pack that
+// appears in a lambda-expression, I can not use template parameter pack
+// here.
 #if CUDA_VERSION >= 8000
     VLOG(5) << "use_tensor_op_math: "
-              << (platform::TensorCoreAvailable() ? "True" : "False");
+            << (dev_ctx->tensor_core_available() ? "True" : "False");
+    dev_ctx->TensorCoreCublasCallIfAvailable([&](cublasHandle_t handle) {
       PADDLE_ENFORCE(platform::dynload::cublasSgemmEx(
-          dev_ctx->cublas_handle(), transa, transb, m, n, k, alpha, A, Atype,
-          lda, B, Btype, ldb, beta, C, Ctype, ldc));
+          handle, transa, transb, m, n, k, alpha, A, Atype, lda, B, Btype, ldb,
+          beta, C, Ctype, ldc));
+    });
 #else
     PADDLE_THROW("cublasSgemmEx is supported on cuda >= 8.0");
-#endif
-    };
-
-#if CUDA_VERSION >= 9000
-    // NOTES: To use Tensor Core, we should change the cublas config,
-    // but the cublas may be hold by multi-thread.
-    dev_ctx->CublasCall(cublas_call, CUBLAS_TENSOR_OP_MATH);
-#else
-    cublas_call();
 #endif
   }
 };
@@ -170,11 +162,10 @@ struct CUBlas<platform::float16> {
                       cudaDataType_t Btype, int ldb, const void *beta, void *C,
                       cudaDataType_t Ctype, int ldc,
                       cudaDataType_t computeType) {
-    auto cublas_call = [&]() {
 #if CUDA_VERSION >= 8000
     cublasGemmAlgo_t algo = CUBLAS_GEMM_DFALT;
 #if CUDA_VERSION >= 9000
-      bool use_tensor_op_math = platform::TensorCoreAvailable();
+    bool use_tensor_op_math = dev_ctx->tensor_core_available();
     if (use_tensor_op_math) {
       algo = CUBLAS_GEMM_DFALT_TENSOR_OP;
     }
@@ -182,20 +173,13 @@ struct CUBlas<platform::float16> {
             << (use_tensor_op_math ? "True" : "False");
 #endif  // CUDA_VERSION >= 9000
 
+    dev_ctx->TensorCoreCublasCallIfAvailable([&](cublasHandle_t handle) {
       PADDLE_ENFORCE(platform::dynload::cublasGemmEx(
-          dev_ctx->cublas_handle(), transa, transb, m, n, k, alpha, A, Atype,
-          lda, B, Btype, ldb, beta, C, Ctype, ldc, computeType, algo));
+          handle, transa, transb, m, n, k, alpha, A, Atype, lda, B, Btype, ldb,
+          beta, C, Ctype, ldc, computeType, algo));
+    });
 #else
     PADDLE_THROW("cublasGemmEx is supported on cuda >= 8.0");
-#endif
-    };
-
-#if CUDA_VERSION >= 9000
-    // NOTES: To use Tensor Core, we should change the cublas config,
-    // but the cublas may be hold by multi-thread.
-    dev_ctx->CublasCall(cublas_call, CUBLAS_TENSOR_OP_MATH);
-#else
-    cublas_call();
 #endif
   }
 };
@@ -223,9 +207,10 @@ void Blas<platform::CUDADeviceContext>::GEMM(CBLAS_TRANSPOSE transA,
                        CUDA_R_32F, N);
   } else {
 #endif  // CUDA_VERSION >= 8000
-
-    CUBlas<T>::GEMM(context_.cublas_handle(), cuTransB, cuTransA, N, M, K,
-                    &alpha, B, ldb, A, lda, &beta, C, N);
+    context_.CublasCall([&](cublasHandle_t handle) {
+      CUBlas<T>::GEMM(handle, cuTransB, cuTransA, N, M, K, &alpha, B, ldb, A,
+                      lda, &beta, C, N);
+    });
 
 #if CUDA_VERSION >= 8000
   }
@@ -266,9 +251,12 @@ inline void Blas<platform::CUDADeviceContext>::GEMM(
       CUDA_R_16F, lda, &h_beta, C, CUDA_R_16F, N, CUDA_R_32F);
 #else
   // CUDA 7.5 does not support cublasGemmEx, hence we fall back to use hgemm
-  CUBlas<platform::float16>::GEMM(context_.cublas_handle(), cuTransB, cuTransA,
-                                  N, M, K, &h_alpha, h_B, ldb, h_A, lda,
-                                  &h_beta, h_C, N);
+
+  context_.CublasCall([&](cublasHandle_t handle) {
+    CUBlas<platform::float16>::GEMM(handle, cuTransB, cuTransA, N, M, K,
+                                    &h_alpha, h_B, ldb, h_A, lda, &h_beta, h_C,
+                                    N);
+  });
 #endif  // CUDA_VERSION >= 8000
 }
 
@@ -292,8 +280,10 @@ void Blas<platform::CUDADeviceContext>::GEMM(bool transA, bool transB, int M,
   } else {
 #endif  // CUDA_VERSION >= 8000
 
-    CUBlas<T>::GEMM(context_.cublas_handle(), cuTransB, cuTransA, N, M, K,
-                    &alpha, B, ldb, A, lda, &beta, C, ldc);
+    context_.CublasCall([&](cublasHandle_t handle) {
+      CUBlas<T>::GEMM(handle, cuTransB, cuTransA, N, M, K, &alpha, B, ldb, A,
+                      lda, &beta, C, ldc);
+    });
 
 #if CUDA_VERSION >= 8000
   }
@@ -311,16 +301,19 @@ inline void Blas<platform::CUDADeviceContext>::GEMM(
   cublasOperation_t cuTransA = transA ? CUBLAS_OP_T : CUBLAS_OP_N;
   cublasOperation_t cuTransB = transB ? CUBLAS_OP_T : CUBLAS_OP_N;
 
-  CUBlas<platform::float16>::GEMM(context_.cublas_handle(), cuTransB, cuTransA,
-                                  N, M, K, &alpha, B, ldb, A, lda, &beta, C,
-                                  ldc);
+  context_.CublasCall([&](cublasHandle_t handle) {
+    CUBlas<platform::float16>::GEMM(handle, cuTransB, cuTransA, N, M, K, &alpha,
+                                    B, ldb, A, lda, &beta, C, ldc);
+  });
 }
 
 template <>
 template <typename T>
 void Blas<platform::CUDADeviceContext>::AXPY(int n, T alpha, const T *x,
                                              T *y) const {
-  CUBlas<T>::AXPY(context_.cublas_handle(), n, &alpha, x, 1, y, 1);
+  context_.CublasCall([&](cublasHandle_t handle) {
+    CUBlas<T>::AXPY(handle, n, &alpha, x, 1, y, 1);
+  });
 }
 
 template <>
@@ -330,8 +323,9 @@ void Blas<platform::CUDADeviceContext>::GEMV(bool trans_a, int M, int N,
                                              T beta, T *C) const {
   cublasOperation_t cuTransA = !trans_a ? CUBLAS_OP_T : CUBLAS_OP_N;
 
-  CUBlas<T>::GEMV(context_.cublas_handle(), cuTransA, N, M, &alpha, A, N, B, 1,
-                  &beta, C, 1);
+  context_.CublasCall([&](cublasHandle_t handle) {
+    CUBlas<T>::GEMV(handle, cuTransA, N, M, &alpha, A, N, B, 1, &beta, C, 1);
+  });
 }
 
 template <>
@@ -353,28 +347,28 @@ void Blas<platform::CUDADeviceContext>::BatchedGEMM(
 
 #if CUDA_VERSION >= 9010
   if (FLAGS_enable_cublas_tensor_op_math && std::is_same<T, float>::value) {
-    auto cublas_call = [&]() {
     cublasGemmAlgo_t algo = CUBLAS_GEMM_DFALT;
-      bool use_tensor_op_math = platform::TensorCoreAvailable();
+    bool use_tensor_op_math = context_.tensor_core_available();
     if (use_tensor_op_math) {
       algo = CUBLAS_GEMM_DFALT_TENSOR_OP;
     }
     VLOG(5) << "use_tensor_op_math: "
             << (use_tensor_op_math ? "True" : "False");
 
+    context_.TensorCoreCublasCallIfAvailable([&](cublasHandle_t handle) {
       PADDLE_ENFORCE(platform::dynload::cublasGemmStridedBatchedEx(
-          context_.cublas_handle(), cuTransB, cuTransA, N, M, K, &alpha, B,
-          CUDA_R_32F, ldb, strideB, A, CUDA_R_32F, lda, strideA, &beta, C,
-          CUDA_R_32F, ldc, strideC, batchCount, CUDA_R_32F, algo));
-    };
-    auto &dev_ctx = const_cast<platform::CUDADeviceContext &>(context_);
-    dev_ctx.CublasCall(cublas_call, CUBLAS_TENSOR_OP_MATH);
+          handle, cuTransB, cuTransA, N, M, K, &alpha, B, CUDA_R_32F, ldb,
+          strideB, A, CUDA_R_32F, lda, strideA, &beta, C, CUDA_R_32F, ldc,
+          strideC, batchCount, CUDA_R_32F, algo));
+    });
   } else {
 #endif  // CUDA_VERSION >= 9010
 
-    CUBlas<T>::GEMM_STRIDED_BATCH(context_.cublas_handle(), cuTransB, cuTransA,
-                                  N, M, K, &alpha, B, ldb, strideB, A, lda,
-                                  strideA, &beta, C, ldc, strideC, batchCount);
+    context_.CublasCall([&](cublasHandle_t handle) {
+      CUBlas<T>::GEMM_STRIDED_BATCH(handle, cuTransB, cuTransA, N, M, K, &alpha,
+                                    B, ldb, strideB, A, lda, strideA, &beta, C,
+                                    ldc, strideC, batchCount);
+    });
 
 #if CUDA_VERSION >= 9010
   }

paddle/fluid/platform/cuda_helper.h

+// Copyright (c) 2019 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
+
+#include <mutex>  // NOLINT
+
+#include "paddle/fluid/platform/dynload/cublas.h"
+#include "paddle/fluid/platform/macros.h"
+
+#if CUDA_VERSION < 9000
+enum cublasMath_t { CUBLAS_DEFAULT_MATH = 0 };
+#endif
+
+namespace paddle {
+namespace platform {
+
+class CublasHandleHolder {
+ public:
+  CublasHandleHolder(cudaStream_t stream, cublasMath_t math_type) {
+    PADDLE_ENFORCE(dynload::cublasCreate(&handle_));
+    PADDLE_ENFORCE(dynload::cublasSetStream(handle_, stream));
+#if CUDA_VERSION >= 9000
+    if (math_type == CUBLAS_TENSOR_OP_MATH) {
+      PADDLE_ENFORCE(
+          dynload::cublasSetMathMode(handle_, CUBLAS_TENSOR_OP_MATH));
+    }
+#endif
+  }
+
+  ~CublasHandleHolder() { PADDLE_ENFORCE(dynload::cublasDestroy(handle_)); }
+
+  template <typename Callback>
+  inline void Call(Callback &&callback) const {
+    std::lock_guard<std::mutex> guard(mtx_);
+    callback(handle_);
+  }
+
+ private:
+  DISABLE_COPY_AND_ASSIGN(CublasHandleHolder);
+
+  cublasHandle_t handle_;
+  mutable std::mutex mtx_;
+};
+
+}  // namespace platform
+}  // namespace paddle

paddle/fluid/platform/device_context.cc

@@ -245,8 +245,15 @@ CUDADeviceContext::CUDADeviceContext(CUDAPlace place)
   eigen_stream_.reset(new EigenCudaStreamDevice());
   eigen_stream_->Reinitialize(&stream_, place);
   eigen_device_.reset(new Eigen::GpuDevice(eigen_stream_.get()));
-  PADDLE_ENFORCE(dynload::cublasCreate(&cublas_handle_));
-  PADDLE_ENFORCE(dynload::cublasSetStream(cublas_handle_, stream_));
+  cublas_handle_.reset(new CublasHandleHolder(stream_, CUBLAS_DEFAULT_MATH));
+
+  if (TensorCoreAvailable()) {
+#if CUDA_VERSION >= 9000
+    cublas_tensor_core_handle_.reset(
+        new CublasHandleHolder(stream_, CUBLAS_TENSOR_OP_MATH));
+#endif
+  }
+
   if (dynload::HasCUDNN()) {
     cudnn_holder_.reset(new CudnnHolder(&stream_, place));
   }
@@ -306,7 +313,8 @@ CUDADeviceContext::~CUDADeviceContext() {
   SetDeviceId(place_.device);
   Wait();
   WaitStreamCallback();
-  PADDLE_ENFORCE(dynload::cublasDestroy(cublas_handle_));
+  cublas_handle_.reset();
+  cublas_tensor_core_handle_.reset();
   eigen_stream_.reset();
   eigen_device_.reset();
   PADDLE_ENFORCE(cudaStreamDestroy(stream_));
@@ -335,8 +343,8 @@ Eigen::GpuDevice* CUDADeviceContext::eigen_device() const {
   return eigen_device_.get();
 }
 
-cublasHandle_t CUDADeviceContext::cublas_handle() const {
-  return cublas_handle_;
+bool CUDADeviceContext::tensor_core_available() const {
+  return cublas_tensor_core_handle_ != nullptr;
 }
 
 cudnnHandle_t CUDADeviceContext::cudnn_handle() const {

paddle/fluid/platform/device_context.h

@@ -20,6 +20,7 @@ limitations under the License. */
 #include "paddle/fluid/memory/malloc.h"
 #include "paddle/fluid/platform/temporary_allocator.h"
 #ifdef PADDLE_WITH_CUDA
+#include "paddle/fluid/platform/cuda_helper.h"
 #include "paddle/fluid/platform/dynload/cublas.h"
 #include "paddle/fluid/platform/dynload/cudnn.h"
 #include "paddle/fluid/platform/gpu_info.h"
@@ -209,39 +210,6 @@ class CudnnWorkspaceHandle {
   std::unique_ptr<std::lock_guard<std::mutex>> guard_;
 };
 
-#if CUDA_VERSION >= 9000
-class ScopedCublasMathMode {
- public:
-  ScopedCublasMathMode(cublasHandle_t handle, cublasMath_t new_math_mode)
-      : handle_(handle) {
-    need_reset = false;
-    PADDLE_ENFORCE(
-        platform::dynload::cublasGetMathMode(handle_, &old_math_mode_),
-        "Failed to get old cublas math mode");
-    if (old_math_mode_ != new_math_mode) {
-      PADDLE_ENFORCE(
-          platform::dynload::cublasSetMathMode(handle_, new_math_mode),
-          "Failed to set old cublas math mode");
-      need_reset = true;
-    }
-  }
-
-  ~ScopedCublasMathMode() {
-    if (need_reset) {
-      PADDLE_ENFORCE(
-          platform::dynload::cublasSetMathMode(handle_, old_math_mode_),
-          "Failed to set old cublas math mode");
-    }
-  }
-
- private:
-  cublasHandle_t handle_;
-  cublasMath_t old_math_mode_;
-  bool need_reset;
-};
-
-#endif
-
 class CUDADeviceContext : public DeviceContext {
  public:
   explicit CUDADeviceContext(CUDAPlace place);
@@ -262,8 +230,25 @@ class CUDADeviceContext : public DeviceContext {
   /*! \brief  Return eigen device in the device context. */
   Eigen::GpuDevice* eigen_device() const;
 
-  /*! \brief  Return cublas handle in the device context. */
-  cublasHandle_t cublas_handle() const;
+  /*! \brief  Call cublas function safely. */
+  template <typename Callback>
+  inline void CublasCall(Callback&& callback) const {
+    cublas_handle_->Call(std::forward<Callback>(callback));
+  }
+
+  /*! \brief  Check whether tensor core is supported */
+  bool tensor_core_available() const;
+
+  /*! \brief  Call cublas function with Tensor Core safely. If
+      Tensor Core is not available, use DEFAULT_MATH instead. */
+  template <typename Callback>
+  inline void TensorCoreCublasCallIfAvailable(Callback&& callback) const {
+    if (cublas_tensor_core_handle_) {
+      cublas_tensor_core_handle_->Call(std::forward<Callback>(callback));
+    } else {
+      cublas_handle_->Call(std::forward<Callback>(callback));
+    }
+  }
 
   /*! \brief  Return cudnn  handle in the device context. */
   cudnnHandle_t cudnn_handle() const;
@@ -282,7 +267,6 @@ class CUDADeviceContext : public DeviceContext {
 
   template <typename Callback>
   void RecordEvent(cudaEvent_t ev, Callback callback) {
-    std::lock_guard<std::mutex> guard(mtx_);
     callback();
     PADDLE_ENFORCE(cudaEventRecord(ev, stream_));
   }
@@ -294,18 +278,6 @@ class CUDADeviceContext : public DeviceContext {
 
   void WaitStreamCallback() const { callback_manager_->Wait(); }
 
-#if CUDA_VERSION >= 9000
-  /*! \brief CublasCall may need to change cublas's config,
-   *  but the cublas may be hold by multi-thread, so we should
-   *  add lock here. */
-  template <typename Callback>
-  void CublasCall(Callback callback, cublasMath_t new_math) {
-    std::lock_guard<std::mutex> guard(cublas_mtx_);
-    ScopedCublasMathMode scoped_cublas_math(cublas_handle_, new_math);
-    callback();
-  }
-#endif
-
  private:
   CUDAPlace place_;
 
@@ -313,7 +285,9 @@ class CUDADeviceContext : public DeviceContext {
   std::unique_ptr<EigenCudaStreamDevice> eigen_stream_;
   std::unique_ptr<CudnnHolder> cudnn_holder_;
   cudaStream_t stream_;
-  cublasHandle_t cublas_handle_;
+
+  std::unique_ptr<CublasHandleHolder> cublas_handle_;
+  std::unique_ptr<CublasHandleHolder> cublas_tensor_core_handle_;
 
   int compute_capability_;
   int runtime_version_;
@@ -321,12 +295,10 @@ class CUDADeviceContext : public DeviceContext {
   int multi_process_;
   int max_threads_per_mp_;
 
-  mutable std::mutex mtx_;
-
   // StreamCallbackManager is thread-safe
   std::unique_ptr<StreamCallbackManager> callback_manager_;
 
-  mutable std::mutex cublas_mtx_;
+  DISABLE_COPY_AND_ASSIGN(CUDADeviceContext);
 };
 
 template <>

paddle/fluid/platform/device_context_test.cu

@@ -43,9 +43,6 @@ TEST(Device, CUDADeviceContext) {
     ASSERT_NE(nullptr, gpu_device);
     cudnnHandle_t cudnn_handle = device_context->cudnn_handle();
     ASSERT_NE(nullptr, cudnn_handle);
-    cublasHandle_t cublas_handle = device_context->cublas_handle();
-    ASSERT_NE(nullptr, cublas_handle);
-    ASSERT_NE(nullptr, device_context->stream());
     delete device_context;
   }
 }