Merge branch 'windows/build' into windows/online

test=develop

Author: wopeizl

.gitignore

@@ -1,3 +1,4 @@
+python/paddle/fluid/tests/unittests/reader_reset_test.recordio
 paddle/operators/check_t.save
 paddle/operators/check_tensor.ls
 paddle/operators/tensor.save

AUTHORS.md

@@ -42,6 +42,7 @@
 | QiJune | Jun Qi |
 | qingqing01 | Qing-Qing Dang |
 | reyoung | Yang Yu |
+| Sand3r- | Michal Gallus |
 | Superjom | Chun-Wei Yan |
 | tensor-tang | Jian Tang |
 | tianbingsz | Tian-Bing Xu |

cmake/inference_lib.cmake

@@ -166,8 +166,8 @@ copy(framework_lib DEPS ${framework_lib_deps}
 
 set(module "memory")
 copy(memory_lib
-        SRCS ${src_dir}/${module}/*.h ${src_dir}/${module}/detail/*.h
-        DSTS ${dst_dir}/${module} ${dst_dir}/${module}/detail
+        SRCS ${src_dir}/${module}/*.h ${src_dir}/${module}/detail/*.h ${src_dir}/${module}/allocation/*.h
+        DSTS ${dst_dir}/${module} ${dst_dir}/${module}/detail ${dst_dir}/${module}/allocation
         )
 
 set(inference_deps paddle_fluid_shared paddle_fluid)

paddle/fluid/framework/CMakeLists.txt

@@ -31,9 +31,7 @@ function(windows_symbolic TARGET)
 endfunction()
 
 add_subdirectory(ir)
-if (NOT WIN32)
 add_subdirectory(details)
-endif (NOT WIN32)
 # ddim lib
 proto_library(framework_proto SRCS framework.proto)
 
@@ -118,13 +116,8 @@ cc_test(op_proto_maker_test SRCS op_proto_maker_test.cc DEPS op_proto_maker)
 cc_library(op_info SRCS op_info.cc DEPS attribute framework_proto)
 cc_library(shape_inference SRCS shape_inference.cc DEPS ddim attribute device_context)
 
-if (NOT WIN32)
 cc_library(operator SRCS operator.cc DEPS op_info device_context tensor scope glog
     shape_inference data_transform lod_tensor profiler)
-else()
-cc_library(operator SRCS operator.cc DEPS op_info device_context tensor scope glog
-    shape_inference data_transform lod_tensor)
-endif(NOT WIN32)
 
 cc_test(operator_test SRCS operator_test.cc DEPS operator op_registry device_context)
 
@@ -179,12 +172,10 @@ else()
   cc_test(test_naive_executor SRCS naive_executor_test.cc DEPS naive_executor elementwise_add_op)
 endif()
 
-if (NOT WIN32)
 cc_library(parallel_executor SRCS parallel_executor.cc DEPS
         threaded_ssa_graph_executor scope_buffered_ssa_graph_executor
         graph build_strategy
         fast_threaded_ssa_graph_executor)
-endif() # NOT WIN32
 
 cc_library(prune SRCS prune.cc DEPS framework_proto)
 cc_test(prune_test SRCS prune_test.cc DEPS op_info prune recurrent_op device_context)

paddle/fluid/framework/details/fast_threaded_ssa_graph_executor.h

@@ -13,9 +13,9 @@
 // limitations under the License.
 
 #pragma once
+#include <ThreadPool.h>
 #include <string>
 #include <vector>
-#include "ThreadPool.h"
 #include "paddle/fluid/framework/blocking_queue.h"
 #include "paddle/fluid/framework/details/exception_holder.h"
 #include "paddle/fluid/framework/details/execution_strategy.h"

paddle/fluid/framework/operator.h

@@ -100,6 +100,7 @@ class OperatorBase {
 
   const std::string& Type() const { return type_; }
 
+  bool HasAttr(const std::string& name) const { return attrs_.count(name); }
   template <typename T>
   inline const T& Attr(const std::string& name) const {
     PADDLE_ENFORCE(attrs_.count(name) != 0, "%s should be in AttributeMap",

paddle/fluid/inference/analysis/CMakeLists.txt

@@ -7,16 +7,17 @@ set(analysis_deps # analysis_deps can be extended accross the project
 add_subdirectory(ir_passes)
 add_subdirectory(passes)
 
-cc_library(ir_pass_manager SRCS ir_pass_manager.cc DEPS graph pass ${INFER_IR_PASSES})
+cc_library(analysis_helper SRCS helper.cc DEPS framework_proto proto_desc graph paddle_fluid_api)
+
+cc_library(ir_pass_manager SRCS ir_pass_manager.cc DEPS graph pass ${INFER_IR_PASSES} analysis_helper)
 
 cc_library(argument SRCS argument.cc DEPS scope proto_desc)
 cc_library(analysis_pass SRCS analysis_pass.cc DEPS proto_desc)
 
 cc_library(analysis SRCS
   analyzer.cc
-  helper.cc
   analysis_pass
-  DEPS ${analysis_deps}
+  DEPS ${analysis_deps} analysis_helper
   )
 
 cc_test(test_dot SRCS dot_tester.cc DEPS analysis)

paddle/fluid/inference/analysis/analyzer_tester.cc

@@ -30,6 +30,7 @@ TEST(Analyzer, analysis_without_tensorrt) {
   Argument argument;
   argument.SetModelDir(FLAGS_inference_model_dir);
   argument.SetIrAnalysisPasses({"infer_clean_graph_pass"});
+  argument.SetUseGPU(false);
 
   Analyzer analyser;
   analyser.Run(&argument);
@@ -41,6 +42,7 @@ TEST(Analyzer, analysis_with_tensorrt) {
   argument.SetTensorRtWorkspaceSize(1 << 20);
   argument.SetModelDir(FLAGS_inference_model_dir);
   argument.SetIrAnalysisPasses({"infer_clean_graph_pass"});
+  argument.SetUseGPU(false);
 
   Analyzer analyser;
   analyser.Run(&argument);

paddle/fluid/inference/analysis/argument.h

@@ -116,6 +116,7 @@ struct Argument {
                       std::vector<std::string>);
 
   DECL_ARGUMENT_FIELD(use_gpu, UseGPU, bool);
+  DECL_ARGUMENT_FIELD(gpu_device_id, GPUDeviceId, int);
   DECL_ARGUMENT_FIELD(use_tensorrt, UseTensorRT, bool);
   DECL_ARGUMENT_FIELD(tensorrt_node_teller, TensorRtNodeTeller,
                       std::function<bool(const framework::ir::Node*)>);

paddle/fluid/inference/analysis/ir_passes/CMakeLists.txt

@@ -4,4 +4,6 @@ set(analysis_deps ${analysis_deps}
         subgraph_detector tensorrt_subgraph_pass
         CACHE INTERNAL "")
 
+set(pass_file ${PADDLE_BINARY_DIR}/paddle/fluid/inference/api/paddle_inference_pass.h)
+file(APPEND ${pass_file} "USE_PASS(tensorrt_subgraph_pass);\n")
 set(INFER_IR_PASSES ${INFER_IR_PASSES} tensorrt_subgraph_pass CACHE INTERNAL "")