Merge branch 'develop' of https://github.com/PaddlePaddle/Paddle into refine_double_buffer_fix

doc/CMakeLists.txt

@@ -1,2 +1,9 @@
+add_custom_target(paddle_apis ALL
+                  DEPENDS paddle_v2_apis paddle_fluid_apis)
+
+add_custom_target(paddle_docs ALL
+                  DEPENDS paddle_v2_docs paddle_v2_docs_cn
+                  paddle_fluid_docs paddle_fluid_docs_cn)
+
 add_subdirectory(v2)
 add_subdirectory(fluid)

doc/fluid/CMakeLists.txt

@@ -27,6 +27,8 @@ sphinx_add_target(paddle_fluid_docs
                   ${CMAKE_CURRENT_SOURCE_DIR}
                   ${SPHINX_HTML_DIR_EN})
 
+add_dependencies(paddle_fluid_docs gen_proto_py)
+
 # configured documentation tools and intermediate build results
 set(BINARY_BUILD_DIR_CN "${CMAKE_CURRENT_BINARY_DIR}/cn/_build")
 
@@ -47,3 +49,7 @@ sphinx_add_target(paddle_fluid_docs_cn
                   ${SPHINX_CACHE_DIR_CN}
                   ${CMAKE_CURRENT_SOURCE_DIR}
                   ${SPHINX_HTML_DIR_CN})
+
+add_dependencies(paddle_fluid_docs_cn gen_proto_py)
+
+add_subdirectory(api)

doc/fluid/api/CMakeLists.txt

@@ -0,0 +1,22 @@
+# configured documentation tools and intermediate build results
+set(BINARY_BUILD_DIR_EN "${CMAKE_CURRENT_BINARY_DIR}/en/_build")
+
+# Sphinx cache with pickled ReST documents
+set(SPHINX_CACHE_DIR_EN "${CMAKE_CURRENT_BINARY_DIR}/en/_doctrees")
+
+# HTML output director
+set(SPHINX_HTML_DIR_EN "${CMAKE_CURRENT_BINARY_DIR}/en/html")
+
+configure_file(
+    "${CMAKE_CURRENT_SOURCE_DIR}/../../templates/conf.py.en.in"
+    "${BINARY_BUILD_DIR_EN}/conf.py"
+    @ONLY)
+
+sphinx_add_target(paddle_fluid_apis
+                  html
+                  ${BINARY_BUILD_DIR_EN}
+                  ${SPHINX_CACHE_DIR_EN}
+                  ${CMAKE_CURRENT_SOURCE_DIR}
+                  ${SPHINX_HTML_DIR_EN})
+
+add_dependencies(paddle_fluid_apis  gen_proto_py framework_py_proto copy_paddle_pybind)

doc/v2/CMakeLists.txt

@@ -20,13 +20,15 @@ configure_file(
     "${BINARY_BUILD_DIR_EN}/conf.py"
     @ONLY)
 
-sphinx_add_target(paddle_docs
+sphinx_add_target(paddle_v2_docs
                   html
                   ${BINARY_BUILD_DIR_EN}
                   ${SPHINX_CACHE_DIR_EN}
                   ${CMAKE_CURRENT_SOURCE_DIR}
                   ${SPHINX_HTML_DIR_EN})
 
+add_dependencies(paddle_v2_docs gen_proto_py)
+
 # configured documentation tools and intermediate build results
 set(BINARY_BUILD_DIR_CN "${CMAKE_CURRENT_BINARY_DIR}/cn/_build")
 
@@ -41,11 +43,13 @@ configure_file(
     "${BINARY_BUILD_DIR_CN}/conf.py"
     @ONLY)
 
-sphinx_add_target(paddle_docs_cn
+sphinx_add_target(paddle_v2_docs_cn
                   html
                   ${BINARY_BUILD_DIR_CN}
                   ${SPHINX_CACHE_DIR_CN}
                   ${CMAKE_CURRENT_SOURCE_DIR}
                   ${SPHINX_HTML_DIR_CN})
 
+add_dependencies(paddle_v2_docs_cn gen_proto_py)
+
 add_subdirectory(api)

doc/v2/api/CMakeLists.txt

@@ -12,9 +12,11 @@ configure_file(
     "${BINARY_BUILD_DIR_EN}/conf.py"
     @ONLY)
 
-sphinx_add_target(paddle_api_docs
+sphinx_add_target(paddle_v2_apis
                   html
                   ${BINARY_BUILD_DIR_EN}
                   ${SPHINX_CACHE_DIR_EN}
                   ${CMAKE_CURRENT_SOURCE_DIR}
                   ${SPHINX_HTML_DIR_EN})
+
+add_dependencies(paddle_v2_apis  gen_proto_py framework_py_proto copy_paddle_pybind)

paddle/fluid/operators/detail/grpc_server.cc

@@ -174,13 +174,13 @@ void AsyncGRPCServer::ShutdownQueue() {
   std::unique_lock<std::mutex> lock(cq_mutex_);
   cq_send_->Shutdown();
   cq_get_->Shutdown();
-  is_shut_down_ = true;
 }
 
 // This URL explains why shutdown is complicate:
 void AsyncGRPCServer::ShutDown() {
-  server_->Shutdown();
+  is_shut_down_ = true;
   ShutdownQueue();
+  server_->Shutdown();
 }
 
 void AsyncGRPCServer::TryToRegisterNewSendOne() {
@@ -213,14 +213,14 @@ void AsyncGRPCServer::HandleRequest(::grpc::ServerCompletionQueue* cq,
   bool ok = false;
   while (true) {
     if (!cq->Next(&tag, &ok)) {
-      LOG(INFO) << cq_name << " get CompletionQueue shutdown!";
+      LOG(INFO) << cq_name << " CompletionQueue shutdown!";
       break;
     }
 
     PADDLE_ENFORCE(tag);
     // FIXME(typhoonzero): de-couple the barriers with recv_op
-    if (cq_name == "cq_get") WaitCond(1);
-    if (cq_name == "cq_send") WaitCond(0);
+    if (!is_shut_down_ && cq_name == "cq_get") WaitCond(1);
+    if (!is_shut_down_ && cq_name == "cq_send") WaitCond(0);
 
     RequestBase* base = (RequestBase*)tag;
     // reference:

paddle/fluid/operators/listen_and_serv_op.cc

@@ -88,7 +88,6 @@ class ListenAndServOp : public framework::OperatorBase {
 
   void Stop() override {
     rpc_service_->Push(LISTEN_TERMINATE_MESSAGE);
-    rpc_service_->ShutDown();
     server_thread_->join();
   }
 

paddle/fluid/operators/lrn_op.cc

@@ -214,7 +214,10 @@ class LRNOpMaker : public framework::OpProtoAndCheckerMaker {
         "Defaults to \"NHWC\". Specify the data format of the output data, "
         "the input will be transformed automatically. ")
         .SetDefault("AnyLayout");
-    AddAttr<bool>("is_test", "").SetDefault(false);
+    AddAttr<bool>("is_test",
+                  "Turns on memory optimization that optimizes away "
+                  "unnecessary memory allocations. Used by MKLDNN.")
+        .SetDefault(false);
 
     AddComment(R"DOC(
 Local Response Normalization Operator.

paddle/fluid/operators/lrn_op.h

@@ -121,6 +121,10 @@ class LRNGradKernel : public framework::OpKernel<T> {
     T alpha = ctx.Attr<T>("alpha");
     T beta = ctx.Attr<T>("beta");
 
+    PADDLE_ENFORCE(
+        !ctx.Attr<bool>("is_test"),
+        "is_test attribute should be set to False in training phase.");
+
     LRNGradFunctor<DeviceContext, T> f;
     f(ctx, x, out, mid, x_g, out_g, N, C, H, W, n, alpha, beta);
   }

paddle/fluid/operators/send_recv_op_test.cc

@@ -122,7 +122,8 @@ void StartServerNet(bool is_sparse) {
 
   // sub program run in listen_and_serv_op, for simple test we use sum
   f::ProgramDesc program;
-  f::BlockDesc *optimize_block = program.MutableBlock(0);
+  const auto &root_block = program.Block(0);
+  auto *optimize_block = program.AppendBlock(root_block);
   // X for server side tensors, RX for received tensers, must be of same shape.
   AddOp("sum", {{"X", {"x0", "x1"}}}, {{"Out", {"Out"}}}, {}, optimize_block);
 

paddle/fluid/platform/profiler_test.cc

@@ -13,7 +13,9 @@ See the License for the specific language governing permissions and
 limitations under the License. */
 
 #include "paddle/fluid/platform/profiler.h"
+#ifdef PADDLE_WITH_CUDA
 #include "cuda_runtime.h"
+#endif
 #include "gtest/gtest.h"
 
 TEST(Event, CpuElapsedTime) {
@@ -159,6 +161,7 @@ TEST(RecordEvent, RecordEvent) {
   DisableProfiler(EventSortingKey::kTotal, "/tmp/profiler");
 }
 
+#ifdef PADDLE_WITH_CUDA
 TEST(TMP, stream_wait) {
   cudaStream_t stream;
   cudaStreamCreate(&stream);
@@ -166,3 +169,4 @@ TEST(TMP, stream_wait) {
   cudaStreamSynchronize(stream);
   cudaStreamSynchronize(stream);
 }
+#endif