Merge pull request #11762 from tensor-tang/refine/set_num_threads

move SetNumThreads to platform

paddle/fluid/inference/io.cc

@@ -20,7 +20,7 @@ limitations under the License. */
 #include "paddle/fluid/framework/block_desc.h"
 #include "paddle/fluid/framework/feed_fetch_type.h"
 #include "paddle/fluid/framework/op_registry.h"
-#include "paddle/fluid/operators/math/blas.h"
+#include "paddle/fluid/platform/cpu_helper.h"
 #include "paddle/fluid/pybind/pybind.h"
 
 DEFINE_string(devices, "", "The devices to be used which is joined by comma.");
@@ -33,7 +33,7 @@ namespace inference {
 
 void Init(const std::vector<std::string> argv) {
   framework::InitGflags(argv);
-  operators::math::SetNumThreads(FLAGS_math_num_threads);
+  platform::SetNumThreads(FLAGS_math_num_threads);
   // init devices
   std::vector<int> devices;
   std::string token;

paddle/fluid/inference/tests/book/test_inference_nlp.cc

@@ -19,7 +19,7 @@ limitations under the License. */
 #include "gflags/gflags.h"
 #include "gtest/gtest.h"
 #include "paddle/fluid/inference/tests/test_helper.h"
-#include "paddle/fluid/operators/math/blas.h"
+#include "paddle/fluid/platform/cpu_helper.h"
 #ifdef PADDLE_WITH_MKLML
 #include <omp.h>
 #endif
@@ -164,7 +164,7 @@ TEST(inference, nlp) {
   // only use 1 thread number per std::thread
   omp_set_dynamic(0);
   omp_set_num_threads(1);
-  paddle::operators::math::SetNumThreads(1);
+  paddle::platform::SetNumThreads(1);
 #endif
 
   double start_ms = 0, stop_ms = 0;

paddle/fluid/operators/math/blas.h

@@ -29,18 +29,6 @@ namespace paddle {
 namespace operators {
 namespace math {
 
-static void SetNumThreads(int num_threads) {
-#ifdef PADDLE_USE_OPENBLAS
-  int real_num_threads = num_threads > 1 ? num_threads : 1;
-  openblas_set_num_threads(real_num_threads);
-#elif defined(PADDLE_WITH_MKLML)
-  int real_num_threads = num_threads > 1 ? num_threads : 1;
-  platform::dynload::MKL_Set_Num_Threads(real_num_threads);
-#else
-  PADDLE_ENFORCE(false, "To be implemented.");
-#endif
-}
-
 /**
  * Matrix Descriptor of a memory buffer.
  *

paddle/fluid/platform/CMakeLists.txt

@@ -28,6 +28,9 @@ cc_test(place_test SRCS place_test.cc DEPS place glog gflags)
 
 add_subdirectory(dynload)
 
+cc_library(cpu_helper SRCS cpu_helper.cc DEPS cblas enforce)
+cc_test(cpu_helper_test SRCS cpu_helper_test.cc DEPS cpu_helper)
+
 IF(WITH_GPU)
     set(GPU_CTX_DEPS dynload_cuda dynamic_loader)
 ELSE()
@@ -43,7 +46,7 @@ ENDIF()
 # memcpy depends on device_context, here add deps individually for
 # avoiding cycle dependencies
 cc_library(device_context SRCS device_context.cc init.cc DEPS malloc
-    place eigen3 stringpiece ${GPU_CTX_DEPS} ${MKLDNN_CTX_DEPS})
+    place eigen3 stringpiece cpu_helper ${GPU_CTX_DEPS} ${MKLDNN_CTX_DEPS})
 nv_test(device_context_test SRCS device_context_test.cu DEPS device_context gpu_info)
 
 cc_test(init_test SRCS init_test.cc DEPS device_context)

paddle/fluid/platform/cpu_helper.cc

+/* 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. */
+
+#include "paddle/fluid/platform/cpu_helper.h"
+#include "paddle/fluid/platform/enforce.h"
+
+#ifdef PADDLE_WITH_MKLML
+#include "paddle/fluid/platform/dynload/mklml.h"
+#endif
+
+#ifdef PADDLE_USE_OPENBLAS
+#include <cblas.h>
+#endif
+
+namespace paddle {
+namespace platform {
+
+void SetNumThreads(int num_threads) {
+#ifdef PADDLE_USE_OPENBLAS
+  int real_num_threads = num_threads > 1 ? num_threads : 1;
+  openblas_set_num_threads(real_num_threads);
+#elif defined(PADDLE_WITH_MKLML)
+  int real_num_threads = num_threads > 1 ? num_threads : 1;
+  platform::dynload::MKL_Set_Num_Threads(real_num_threads);
+#else
+  PADDLE_ENFORCE(false, "To be implemented.");
+#endif
+}
+
+}  // namespace platform
+}  // namespace paddle

paddle/fluid/platform/cpu_helper.h

+/* 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
+
+#include <stddef.h>
+
+namespace paddle {
+namespace platform {
+
+//! Set the number of threads in use.
+void SetNumThreads(int num_threads);
+
+}  // namespace platform
+}  // namespace paddle

paddle/fluid/platform/cpu_helper_test.cc

+/* Copyright (c) 2018 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. */
+
+#include "paddle/fluid/platform/cpu_helper.h"
+
+#include "gtest/gtest.h"
+
+TEST(CpuHelper, SetNumThread) {
+  paddle::platform::SetNumThreads(1);
+  paddle::platform::SetNumThreads(4);
+}

paddle/fluid/platform/init.cc

@@ -17,7 +17,7 @@ limitations under the License. */
 #include <string>
 
 #include "paddle/fluid/framework/operator.h"
-#include "paddle/fluid/operators/math/blas.h"
+#include "paddle/fluid/platform/cpu_helper.h"
 #include "paddle/fluid/platform/device_context.h"
 #include "paddle/fluid/platform/init.h"
 #include "paddle/fluid/platform/place.h"
@@ -115,7 +115,7 @@ void InitDevices(bool init_p2p, const std::vector<int> devices) {
   places.emplace_back(platform::CPUPlace());
   platform::DeviceContextPool::Init(places);
 #ifndef PADDLE_WITH_MKLDNN
-  operators::math::SetNumThreads(1);
+  platform::SetNumThreads(1);
 #endif
 }