update_c++17 (#53892)

* update_c++17

* update_c++17

* fix windows bug

* solve cirle depend

* solve cirle depend

* solve cirle depend

* solve cirle depend

* solve cirle depend

* fix windows bug

* fix compiler error

* fix compiler error

* update eigen3

* update eigen3

* update eigen3

* fix mac-py3 compiler error

* update C++17

* fix mac compiler error

* fix compile error

* fix coverage_compiler error

* fix coverage_ci_problem

* fix coverage_error

* fix_kunlun200 compile error

* fix kunlun200 compiler error

* fix compile error

* fix compiler error

* fix py3 failed test

* fix kunlun200 compiler error

* test

* fix test error

* fix test error

* fix test error

* test

* test

* fix mac py3 error

* fix mac py3 error

* fix mac py3 error

* fix test error

* fix test error

* fix compile error

* fix compile error

* fix compile error

* test

* test

* fix compiler error

* test

* test

* debug on ci

* fix compiler error

* fix compiler error

* test

* fix cinn compiler error

* test

* fix rocm cmpile error

* fix cinn and kunlun compile error

* update c++14

* Update flags.cmake

cmake/cuda.cmake

@@ -295,7 +295,7 @@ message(STATUS "NVCC_FLAGS_EXTRA: ${NVCC_FLAGS_EXTRA}")
 set(CUDA_PROPAGATE_HOST_FLAGS OFF)
 # Release/Debug flags set by cmake. Such as -O3 -g -DNDEBUG etc.
 # So, don't set these flags here.
-set(CMAKE_CUDA_STANDARD 14)
+set(CMAKE_CUDA_STANDARD 17)
 
 # (Note) For windows, if delete /W[1-4], /W1 will be added defaultly and conflic with -w
 # So replace /W[1-4] with /W0

cmake/external/pocketfft.cmake

@@ -26,12 +26,22 @@ set(POCKETFFT_INCLUDE_DIR ${POCKETFFT_PREFIX_DIR}/src)
 message("POCKETFFT_INCLUDE_DIR is ${POCKETFFT_INCLUDE_DIR}")
 include_directories(${POCKETFFT_INCLUDE_DIR})
 
+if(APPLE)
+  file(TO_NATIVE_PATH
+       ${PADDLE_SOURCE_DIR}/patches/pocketfft/pocketfft_hdronly.h.patch
+       native_dst)
+  set(POCKETFFT_PATCH_COMMAND
+      git checkout -- . && git checkout ${GLOO_TAG} && patch -Nd
+      ${POCKETFFT_INCLUDE_DIR}/extern_pocketfft < ${native_dst})
+endif()
+
 ExternalProject_Add(
   extern_pocketfft
   ${EXTERNAL_PROJECT_LOG_ARGS} ${SHALLOW_CLONE}
   GIT_REPOSITORY ${POCKETFFT_REPOSITORY}
   GIT_TAG ${POCKETFFT_TAG}
   PREFIX ${POCKETFFT_PREFIX_DIR}
+  PATCH_COMMAND ${POCKETFFT_PATCH_COMMAND}
   UPDATE_COMMAND ""
   CONFIGURE_COMMAND ""
   BUILD_COMMAND ""

cmake/flags.cmake

@@ -35,7 +35,11 @@ endfunction()
 
 checkcompilercxx14flag()
 if(NOT WIN32)
-  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++14")
+  if(WITH_CINN)
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++14")
+  else()
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++17")
+  endif()
 else()
   set(CMAKE_CXX_STANDARD 17)
 endif()

paddle/fluid/distributed/common/registerer.h

@@ -116,7 +116,7 @@ inline PsCoreClassMap &global_factory_map_cpp() { return global_factory_map(); }
    public:                                              \
     Any NewInstance() { return Any(new name()); }       \
   };                                                    \
-  void register_factory_##name() {                      \
+  static void register_factory_##name() {               \
     FactoryMap &map = global_factory_map_cpp()[#clazz]; \
     if (map.find(#name) == map.end()) {                 \
       map[#name] = new ObjectFactory##name();           \

paddle/fluid/distributed/fleet_executor/CMakeLists.txt

@@ -25,36 +25,66 @@ cc_library(
   task_loop_thread_pool
   SRCS task_loop_thread_pool.cc task_loop_thread.cc task_loop.cc
   DEPS enforce glog)
-
-cc_library(
-  fleet_executor
-  SRCS fleet_executor.cc
-       carrier.cc
-       task_node.cc
-       runtime_graph.cc
-       dist_model.cc
-       interceptor.cc
-       compute_interceptor.cc
-       amplifier_interceptor.cc
-       cond_interceptor.cc
-       start_interceptor.cc
-       source_interceptor.cc
-       sink_interceptor.cc
-       message_service.cc
-       message_bus.cc
-       dist_model_tensor_wrapper.cc
-  DEPS proto_desc
-       standalone_executor
-       fleet_executor_desc_proto
-       interceptor_message_proto
-       task_loop_thread_pool
-       collective_helper
-       op_registry
-       executor_gc_helper
-       phi
-       glog
-       ${BRPC_DEPS})
-
+if(WITH_XPU OR WITH_ROCM)
+  cc_library(
+    fleet_executor
+    SRCS fleet_executor.cc
+         carrier.cc
+         task_node.cc
+         runtime_graph.cc
+         dist_model.cc
+         interceptor.cc
+         compute_interceptor.cc
+         amplifier_interceptor.cc
+         cond_interceptor.cc
+         start_interceptor.cc
+         source_interceptor.cc
+         sink_interceptor.cc
+         message_service.cc
+         message_bus.cc
+         dist_model_tensor_wrapper.cc
+    DEPS naive_executor
+         proto_desc
+         standalone_executor
+         fleet_executor_desc_proto
+         interceptor_message_proto
+         task_loop_thread_pool
+         collective_helper
+         executor_gc_helper
+         op_registry
+         phi
+         glog
+         ${BRPC_DEPS})
+else()
+  cc_library(
+    fleet_executor
+    SRCS fleet_executor.cc
+         carrier.cc
+         task_node.cc
+         runtime_graph.cc
+         dist_model.cc
+         interceptor.cc
+         compute_interceptor.cc
+         amplifier_interceptor.cc
+         cond_interceptor.cc
+         start_interceptor.cc
+         source_interceptor.cc
+         sink_interceptor.cc
+         message_service.cc
+         message_bus.cc
+         dist_model_tensor_wrapper.cc
+    DEPS proto_desc
+         standalone_executor
+         fleet_executor_desc_proto
+         interceptor_message_proto
+         task_loop_thread_pool
+         collective_helper
+         op_registry
+         executor_gc_helper
+         phi
+         glog
+         ${BRPC_DEPS})
+endif()
 if(WITH_DISTRIBUTE)
   set(DISTRIBUTE_COMPILE_FLAGS
       "-Wno-non-virtual-dtor -Wno-error=non-virtual-dtor -Wno-error=delete-non-virtual-dtor"

paddle/fluid/distributed/fleet_executor/test/CMakeLists.txt

@@ -44,12 +44,13 @@ cc_test_old(
   compute_interceptor_run_op_test.cc
   DEPS
   fleet_executor
-  ${BRPC_DEPS}
-  op_registry
+  naive_executor
   fill_constant_op
+  op_registry
   elementwise_add_op
   scope
-  device_context)
+  device_context
+  ${BRPC_DEPS})
 
 if(WITH_DISTRIBUTE AND NOT WITH_PSLIB)
   set_source_files_properties(

paddle/fluid/distributed/ps/service/CMakeLists.txt

@@ -41,6 +41,7 @@ brpc_library(
   PROTO
   sendrecv.proto
   DEPS
+  ps_framework_proto
   ${BRPC_DEPS})
 
 #set_property(GLOBAL PROPERTY RPC_DEPS sendrecv_rpc ${BRPC_DEPS} string_helper)
@@ -122,6 +123,7 @@ cc_library(
        scope
        phi
        ps_gpu_wrapper
+       fleet
        ${RPC_DEPS})
 
 #cc_library(

paddle/fluid/distributed/ps/table/CMakeLists.txt

@@ -36,12 +36,10 @@ set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fopenmp")
 
 set(TABLE_SRC memory_dense_table.cc barrier_table.cc common_graph_table.cc)
 #set(EXTERN_DEP rocksdb)
-
 cc_library(
   common_table
   SRCS ${TABLE_SRC}
-  DEPS ${TABLE_DEPS}
-       ${RPC_DEPS}
+  DEPS ${RPC_DEPS}
        graph_edge
        graph_node
        device_context

paddle/fluid/distributed/rpc/CMakeLists.txt

@@ -27,5 +27,5 @@ proto_library(paddle_rpc_proto SRCS rpc.proto)
 cc_library(
   paddle_rpc
   SRCS ${PADDLE_RPC_SRCS}
-  DEPS ${PADDLE_RPC_DEPS} paddle_rpc_proto)
+  DEPS ${PADDLE_RPC_DEPS} paddle_rpc_proto pybind)
 add_dependencies(paddle_rpc brpc)

paddle/fluid/distributed/test/CMakeLists.txt

@@ -74,6 +74,8 @@ cc_test_old(
   brpc_utils
   scope
   phi
+  sendrecv_rpc
+  ps_service
   ${COMMON_DEPS}
   ${RPC_DEPS})
 
@@ -117,8 +119,16 @@ cc_test_old(
 
 set_source_files_properties(
   feature_value_test.cc PROPERTIES COMPILE_FLAGS ${DISTRIBUTE_COMPILE_FLAGS})
-cc_test_old(feature_value_test SRCS feature_value_test.cc DEPS ${COMMON_DEPS}
-            table)
+
+cc_test_old(
+  feature_value_test
+  SRCS
+  feature_value_test.cc
+  DEPS
+  table
+  common_table
+  sendrecv_rpc
+  ${COMMON_DEPS})
 
 set_source_files_properties(
   sparse_sgd_rule_test.cc PROPERTIES COMPILE_FLAGS ${DISTRIBUTE_COMPILE_FLAGS})

paddle/fluid/eager/accumulation/CMakeLists.txt

@@ -2,5 +2,5 @@ if(NOT (NOT WITH_PYTHON AND ON_INFER))
   cc_library(
     accumulation_node
     SRCS accumulation_node.cc
-    DEPS gradient_accumulator phi grad_node_info)
+    DEPS gradient_accumulator phi grad_node_info final_dygraph_function)
 endif()

paddle/fluid/eager/custom_operator/CMakeLists.txt

 cc_library(
   custom_operator_node
   SRCS custom_operator_node.cc
-  DEPS phi grad_node_info custom_operator)
+  DEPS phi grad_node_info custom_operator utils)

paddle/fluid/framework/CMakeLists.txt

@@ -695,14 +695,14 @@ endif()
 cc_library(
   executor_gc_helper
   SRCS executor_gc_helper.cc
-  DEPS scope
+  DEPS while_op_helper
+       recurrent_op_helper
+       conditional_block_op_helper
+       scope
        proto_desc
        operator
        garbage_collector
-       op_registry
-       while_op_helper
-       recurrent_op_helper
-       conditional_block_op_helper)
+       op_registry)
 if(WITH_DISTRIBUTE)
   if(WITH_PSLIB)
     cc_library(
@@ -729,7 +729,10 @@ if(WITH_DISTRIBUTE)
            section_worker.cc
            device_worker_factory.cc
            data_set.cc
-      DEPS op_registry
+      DEPS fleet_executor
+           fleet_wrapper
+           recurrent_op_helper
+           op_registry
            device_context
            scope
            framework_proto
@@ -737,7 +740,6 @@ if(WITH_DISTRIBUTE)
            glog
            fs
            shell
-           fleet_wrapper
            heter_wrapper
            ps_gpu_wrapper
            box_wrapper
@@ -753,7 +755,6 @@ if(WITH_DISTRIBUTE)
            timer
            monitor
            heter_service_proto
-           fleet_executor
            ${BRPC_DEP})
     set(DISTRIBUTE_COMPILE_FLAGS "")
     if(CMAKE_CXX_COMPILER_VERSION VERSION_GREATER 7.0)
@@ -806,7 +807,8 @@ if(WITH_DISTRIBUTE)
            heter_section_worker.cc
            device_worker_factory.cc
            data_set.cc
-      DEPS op_registry
+      DEPS recurrent_op_helper
+           op_registry
            device_context
            scope
            framework_proto
@@ -883,7 +885,8 @@ if(WITH_DISTRIBUTE)
            section_worker.cc
            device_worker_factory.cc
            data_set.cc
-      DEPS op_registry
+      DEPS recurrent_op_helper
+           op_registry
            device_context
            scope
            framework_proto
@@ -944,7 +947,8 @@ elseif(WITH_PSLIB)
          section_worker.cc
          device_worker_factory.cc
          data_set.cc
-    DEPS op_registry
+    DEPS recurrent_op_helper
+         op_registry
          device_context
          scope
          framework_proto
@@ -992,7 +996,8 @@ else()
          section_worker.cc
          device_worker_factory.cc
          data_set.cc
-    DEPS op_registry
+    DEPS recurrent_op_helper
+         op_registry
          device_context
          scope
          framework_proto

paddle/fluid/framework/details/CMakeLists.txt

@@ -75,28 +75,30 @@ if(WITH_GPU)
   nv_library(
     all_reduce_op_handle
     SRCS all_reduce_op_handle.cc
-    DEPS op_handle_base
+    DEPS variable_visitor
+         op_handle_base
          scope
          lod_tensor
          phi
          memory
-         dynload_cuda
-         variable_visitor)
+         dynload_cuda)
   nv_library(
     fused_all_reduce_op_handle
     SRCS fused_all_reduce_op_handle.cc
-    DEPS op_handle_base
+    DEPS all_reduce_op_handle
+         variable_visitor
+         op_handle_base
          scope
          lod_tensor
          phi
          memory
          dynload_cuda
-         variable_visitor
          place)
   nv_library(
     grad_merge_all_reduce_op_handle
     SRCS grad_merge_all_reduce_op_handle.cc
-    DEPS op_handle_base
+    DEPS fused_all_reduce_op_handle
+         op_handle_base
          scope
          lod_tensor
          phi
@@ -104,8 +106,7 @@ if(WITH_GPU)
          dynload_cuda
          variable_visitor
          place
-         all_reduce_op_handle
-         fused_all_reduce_op_handle)
+         all_reduce_op_handle)
 
   if(WITH_DGC)
     nv_library(
@@ -159,18 +160,20 @@ elseif(WITH_ROCM)
   hip_library(
     fused_all_reduce_op_handle
     SRCS fused_all_reduce_op_handle.cc
-    DEPS op_handle_base
+    DEPS all_reduce_op_handle
+         op_handle_base
+         variable_visitor
          scope
          lod_tensor
          phi
          memory
          dynload_cuda
-         variable_visitor
          place)
   hip_library(
     grad_merge_all_reduce_op_handle
     SRCS grad_merge_all_reduce_op_handle.cc
-    DEPS op_handle_base
+    DEPS fused_all_reduce_op_handle
+         op_handle_base
          scope
          lod_tensor
          phi
@@ -178,8 +181,7 @@ elseif(WITH_ROCM)
          dynload_cuda
          variable_visitor
          place
-         all_reduce_op_handle
-         fused_all_reduce_op_handle)
+         all_reduce_op_handle)
 
   if(WITH_DISTRIBUTE)
     hip_library(
@@ -212,7 +214,8 @@ else()
   cc_library(
     fused_all_reduce_op_handle
     SRCS fused_all_reduce_op_handle.cc
-    DEPS op_handle_base
+    DEPS all_reduce_op_handle
+         op_handle_base
          scope
          lod_tensor
          phi
@@ -222,15 +225,15 @@ else()
   cc_library(
     grad_merge_all_reduce_op_handle
     SRCS grad_merge_all_reduce_op_handle.cc
-    DEPS op_handle_base
+    DEPS fused_all_reduce_op_handle
+         op_handle_base
          scope
          lod_tensor
          phi
          memory
          variable_visitor
          place
-         all_reduce_op_handle
-         fused_all_reduce_op_handle)
+         all_reduce_op_handle)
   if(WITH_DISTRIBUTE)
     cc_library(
       reduce_op_handle

paddle/fluid/framework/ir/fused_multi_transformer_encoder_pass_tester.cc

@@ -14,7 +14,9 @@ limitations under the License. */
 #include "paddle/fluid/framework/ir/fused_multi_transformer_encoder_pass.h"  // NOLINT
 #include "paddle/fluid/framework/ir/pass_tester_helper.h"
 #include "paddle/fluid/framework/op_version_registry.h"
-
+#ifndef UNUSED
+#define UNUSED __attribute__((unused))
+#endif
 namespace paddle {
 namespace framework {
 namespace ir {
@@ -180,7 +182,7 @@ TEST(FusedMultiTransformerEncoderPass, basic) {
   // FFN: post LayerNorm
   auto* ffn_ln_scale = layers.data("ffn_ln_scale", {1024}, true);
   auto* ffn_ln_bias = layers.data("ffn_ln_bias", {1024}, true);
-  layers.layer_norm(ffn_out, ffn_ln_scale, ffn_ln_bias)[0];
+  UNUSED auto res = layers.layer_norm(ffn_out, ffn_ln_scale, ffn_ln_bias)[0];
 
   std::unique_ptr<ir::Graph> graph(new ir::Graph(layers.main_program()));
   graph->Set("__param_scope__", CreateParamScope());
@@ -355,7 +357,7 @@ TEST(MultiDevicesFusedMultiTransformerEncoderPass, basic) {
   // FFN: post LayerNorm
   auto* ffn_ln_scale = layers.data("ffn_ln_scale", {1024}, true);
   auto* ffn_ln_bias = layers.data("ffn_ln_bias", {1024}, true);
-  layers.layer_norm(ffn_out, ffn_ln_scale, ffn_ln_bias)[0];
+  UNUSED auto res = layers.layer_norm(ffn_out, ffn_ln_scale, ffn_ln_bias)[0];
 
   std::unique_ptr<ir::Graph> graph(new ir::Graph(layers.main_program()));
   graph->Set("__param_scope__", CreateParamScope());

paddle/fluid/framework/new_executor/interpreter/CMakeLists.txt

@@ -37,4 +37,4 @@ set(INTERPRETER_DEPS
 cc_library(
   interpreter
   SRCS ${INTERPRETER_SRCS}
-  DEPS ${INTERPRETER_DEPS})
+  DEPS standalone_executor ${INTERPRETER_DEPS})

paddle/fluid/framework/paddle2cinn/CMakeLists.txt

@@ -39,7 +39,9 @@ cc_library(
        cinn_cache_key
        cinn_graph_symbolization
        cinn
-       cinn_launch_context)
+       cinn_launch_context
+       python
+       pybind)
 
 if(WITH_TESTING)
   # cc_test_old(cinn_lib_test SRCS cinn_lib_test.cc DEPS cinn)

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

@@ -7,7 +7,8 @@ if(WITH_GPU AND TENSORRT_FOUND)
   cc_library(
     tensorrt_subgraph_pass
     SRCS tensorrt_subgraph_pass.cc
-    DEPS subgraph_util tensorrt_op_teller infer_io_utils)
+    DEPS convert_to_mixed_precision subgraph_util tensorrt_op_teller
+         infer_io_utils)
 
   set(analysis_deps
       ${analysis_deps} subgraph_util tensorrt_subgraph_pass

paddle/fluid/inference/api/CMakeLists.txt

@@ -15,12 +15,25 @@ if(WITH_MKLDNN)
       ${mkldnn_quantizer_cfg}
       PARENT_SCOPE)
 endif()
+if(WIN32)
+  cc_library(
+    analysis_config
+    SRCS analysis_config.cc
+    DEPS ${mkldnn_quantizer_cfg} lod_tensor paddle_pass_builder table_printer
+         utf8proc)
+else()
+  cc_library(
+    analysis_config
+    SRCS analysis_config.cc
+    DEPS analysis_helper
+         processgroup_comm_utils
+         ${mkldnn_quantizer_cfg}
+         lod_tensor
+         paddle_pass_builder
+         table_printer
+         utf8proc)
+endif()
 
-cc_library(
-  analysis_config
-  SRCS analysis_config.cc
-  DEPS ${mkldnn_quantizer_cfg} lod_tensor paddle_pass_builder table_printer
-       utf8proc)
 cc_library(
   paddle_infer_contrib
   SRCS paddle_infer_contrib.cc
@@ -28,14 +41,14 @@ cc_library(
 cc_library(paddle_pass_builder SRCS paddle_pass_builder.cc)
 
 set(paddle_inference_api_deps
-    lod_tensor
-    scope
     reset_tensor_array
     analysis_config
     paddle_infer_contrib
     zero_copy_tensor
     trainer_desc_proto
-    custom_operator)
+    custom_operator
+    lod_tensor
+    scope)
 
 if(WITH_CRYPTO)
   list(APPEND paddle_inference_api_deps paddle_crypto)
@@ -43,11 +56,17 @@ endif()
 if(WITH_CUSTOM_DEVICE)
   set(paddle_inference_api_deps ${paddle_inference_api_deps} phi)
 endif()
-
-cc_library(
-  paddle_inference_api
-  SRCS api.cc api_impl.cc helper.cc
-  DEPS ${paddle_inference_api_deps})
+if(WIN32)
+  cc_library(
+    paddle_inference_api
+    SRCS api.cc api_impl.cc helper.cc
+    DEPS executor ${paddle_inference_api_deps})
+else()
+  cc_library(
+    paddle_inference_api
+    SRCS api.cc api_impl.cc helper.cc
+    DEPS executor paddle_inference_io ${paddle_inference_api_deps})
+endif()
 
 if(WIN32)
   target_link_libraries(paddle_inference_api phi)

paddle/fluid/jit/CMakeLists.txt

@@ -46,4 +46,7 @@ cc_library(
        jit_function_schema
        jit_interpreter_engine
        jit_predictor_engine
-       jit_function)
+       jit_function
+       executor
+       recurrent_op_helper
+       recurrent_op)

paddle/fluid/operators/CMakeLists.txt

@@ -126,7 +126,12 @@ if (WITH_GPU OR WITH_ROCM)
 endif()
 
 op_library(lstm_op DEPS ${OP_HEADER_DEPS})
-op_library(recurrent_op DEPS ${OP_HEADER_DEPS})
+if (WITH_ROCM)
+    op_library(recurrent_op DEPS executor ${OP_HEADER_DEPS})
+else()
+    op_library(recurrent_op DEPS ${OP_HEADER_DEPS})
+endif()
+
 
 set(COMMON_OP_DEPS ${OP_HEADER_DEPS})
 

paddle/fluid/operators/controlflow/CMakeLists.txt

@@ -9,7 +9,7 @@ register_operators(EXCLUDES conditional_block_op DEPS naive_executor
 cc_library(
   conditional_block_op
   SRCS conditional_block_op.cc
-  DEPS standalone_executor executor)
+  DEPS executor standalone_executor)
 cc_library(
   op_variant
   SRCS op_variant.cc
@@ -17,15 +17,15 @@ cc_library(
 cc_library(
   conditional_block_op_helper
   SRCS conditional_block_op_helper.cc
-  DEPS operator op_variant conditional_block_op)
+  DEPS op_variant operator conditional_block_op)
 cc_library(
   recurrent_op_helper
   SRCS recurrent_op_helper.cc
-  DEPS operator op_variant recurrent_op)
+  DEPS recurrent_op op_variant operator)
 cc_library(
   while_op_helper
   SRCS while_op_helper.cc
-  DEPS operator op_variant)
+  DEPS op_variant operator)
 
 if(WITH_UNITY_BUILD)
   target_link_libraries(paddle_operators_controlflow_unity conditional_block_op)

paddle/phi/core/dense_tensor.h

@@ -268,8 +268,8 @@ class DenseTensor : public TensorBase,
   };
 
  protected:
-  std::shared_ptr<InplaceVersion> inplace_version_counter_{
-      std::make_shared<InplaceVersion>()};
+  std::shared_ptr<InplaceVersion> inplace_version_counter_ =
+      std::make_shared<InplaceVersion>();
 
 /* @jim19930609: This is a hack
 In general, it is badly designed to fuse MKLDNN-specific objects into a

paddle/phi/kernels/funcs/reduce_functor.h

@@ -258,7 +258,7 @@ struct AMaxOrAMinGradFunctor {
     // reduce_all
     if (size == static_cast<int>(x_numel)) {
       auto equal_number = mask.sum()
-                              .reshape(Eigen::array<int, 1>({1}))
+                              .reshape(Eigen::array<int, 1>({{1}}))
                               .broadcast(Eigen::array<int, 1>({size}));
       dx->device(place) =
           dy->broadcast(dim).reshape(x->dimensions()) * mask / equal_number;

patches/pocketfft/pocketfft_hdronly.h.patch

+diff --git a/pocketfft_hdronly.h b/pocketfft_hdronly.h
+index d75ada6..d3c352d 100644
+--- a/pocketfft_hdronly.h
++++ b/pocketfft_hdronly.h
+@@ -149,17 +149,17 @@ template<> struct VLEN<double> { static constexpr size_t val=2; };
+ #endif
+ #endif
+ 
+-#if __cplusplus >= 201703L
+-inline void *aligned_alloc(size_t align, size_t size)
+-  {
+-  // aligned_alloc() requires that the requested size is a multiple of "align"
+-  void *ptr = ::aligned_alloc(align,(size+align-1)&(~(align-1)));
+-  if (!ptr) throw std::bad_alloc();
+-  return ptr;
+-  }
+-inline void aligned_dealloc(void *ptr)
+-    { free(ptr); }
+-#else // portable emulation
++// #if __cplusplus >= 201703L
++// inline void *aligned_alloc(size_t align, size_t size)
++//   {
++//   // aligned_alloc() requires that the requested size is a multiple of "align"
++//   void *ptr = ::aligned_alloc(align,(size+align-1)&(~(align-1)));
++//   if (!ptr) throw std::bad_alloc();
++//   return ptr;
++//   }
++// inline void aligned_dealloc(void *ptr)
++//     { free(ptr); }
++// #else // portable emulation
+ inline void *aligned_alloc(size_t align, size_t size)
+   {
+   align = std::max(align, alignof(max_align_t));
+@@ -172,7 +172,7 @@ inline void *aligned_alloc(size_t align, size_t size)
+   }
+ inline void aligned_dealloc(void *ptr)
+   { if (ptr) free((reinterpret_cast<void**>(ptr))[-1]); }
+-#endif
++// #endif
+ 
+ template<typename T> class arr
+   {

test/cpp/eager/data_structure_tests/CMakeLists.txt

@@ -4,12 +4,36 @@ cc_test_old(test_egr_ds_auotgrad_meta SRCS autograd_meta_test.cc DEPS
             ${eager_deps})
 
 if(NOT ((NOT WITH_PYTHON) AND ON_INFER))
-  cc_test_old(test_egr_ds_grad_tensor_holder SRCS grad_tensor_holder_test.cc
-              DEPS ${eager_deps} ${generated_deps})
-  cc_test_old(test_egr_ds_grad_node_info SRCS grad_node_info_test.cc DEPS
-              ${eager_deps} ${generated_deps})
-  cc_test_old(test_egr_ds_accumulation_node SRCS accumulation_node_test.cc DEPS
-              ${eager_deps} ${generated_deps})
-  cc_test_old(test_egr_ds_tensor_wrapper SRCS tensor_wrapper_test.cc DEPS
-              ${eager_deps} ${generated_deps})
+  cc_test_old(
+    test_egr_ds_grad_tensor_holder
+    SRCS
+    grad_tensor_holder_test.cc
+    DEPS
+    conditional_block_op
+    ${eager_deps}
+    ${generated_deps})
+  cc_test_old(
+    test_egr_ds_grad_node_info
+    SRCS
+    grad_node_info_test.cc
+    DEPS
+    conditional_block_op
+    ${eager_deps}
+    ${generated_deps})
+  cc_test_old(
+    test_egr_ds_accumulation_node
+    SRCS
+    accumulation_node_test.cc
+    DEPS
+    conditional_block_op
+    ${eager_deps}
+    ${generated_deps})
+  cc_test_old(
+    test_egr_ds_tensor_wrapper
+    SRCS
+    tensor_wrapper_test.cc
+    DEPS
+    conditional_block_op
+    ${eager_deps}
+    ${generated_deps})
 endif()

test/cpp/eager/performance_tests/CMakeLists.txt

@@ -17,6 +17,7 @@ if(NOT (NOT WITH_PYTHON AND ON_INFER))
     SRCS
     benchmark_eager_cpu.cc
     DEPS
+    conditional_block_op
     performance_benchmark_utils
     ${eager_deps}
     ${fluid_deps})
@@ -25,6 +26,7 @@ if(NOT (NOT WITH_PYTHON AND ON_INFER))
     SRCS
     benchmark_fluid_cpu.cc
     DEPS
+    dygraph_function
     performance_benchmark_utils
     ${eager_deps}
     ${fluid_deps})
@@ -34,6 +36,7 @@ if(NOT (NOT WITH_PYTHON AND ON_INFER))
     SRCS
     benchmark_eager_cuda.cc
     DEPS
+    conditional_block_op
     performance_benchmark_utils
     ${eager_deps}
     ${fluid_deps})
@@ -42,7 +45,10 @@ if(NOT (NOT WITH_PYTHON AND ON_INFER))
     SRCS
     benchmark_fluid_cuda.cc
     DEPS
+    conditional_block_op
+    standalone_executor
     performance_benchmark_utils
+    dygraph_function
     ${eager_deps}
     ${fluid_deps})
 endif()

test/cpp/fluid/CMakeLists.txt

@@ -103,8 +103,16 @@ cc_test(
   SRCS share_buffer_op_test.cc
   DEPS lod_tensor device_context generated_static_op)
 
-cc_test_old(op_debug_string_test SRCS op_debug_string_test.cc DEPS
-            elementwise_add_op ${COMMON_OP_DEPS})
+cc_test_old(
+  op_debug_string_test
+  SRCS
+  op_debug_string_test.cc
+  DEPS
+  executor
+  recurrent_op_helper
+  recurrent_op
+  elementwise_add_op
+  ${COMMON_OP_DEPS})
 
 if(WITH_GPU)
   cc_test(

test/cpp/fluid/elementwise/CMakeLists.txt

 cc_test(
   test_elementwise_add_op_inplace
   SRCS test_elementwise_add_op_inplace.cc
-  DEPS op_registry elementwise_add_op scope device_context enforce executor)
+  DEPS executor op_registry elementwise_add_op scope device_context enforce)
 cc_test(
   test_elementwise_div_grad_grad
   SRCS test_elementwise_div_grad_grad.cc
-  DEPS op_registry elementwise_div_op scope device_context enforce executor)
+  DEPS executor op_registry elementwise_div_op scope device_context enforce)
 cc_test(
   test_elementwise_add_grad_grad
   SRCS test_elementwise_add_grad_grad.cc
-  DEPS op_registry elementwise_add_op scope device_context enforce executor)
+  DEPS executor op_registry elementwise_add_op scope device_context enforce)

test/cpp/fluid/mkldnn/CMakeLists.txt

 cc_test(
   test_mkldnn_op_inplace
   SRCS test_mkldnn_op_inplace.cc
-  DEPS op_registry
+  DEPS executor
+       op_registry
        elementwise_add_op
        activation_op
        phi
        scope
        device_context
        enforce
-       executor
        generated_static_op)
 
 set(TEST_MKLDNN_CACHING_DEPS
@@ -36,6 +36,11 @@ cc_test_old(
   SRCS
   test_mkldnn_op_nhwc.cc
   DEPS
+  conditional_block_op
+  standalone_executor
+  executor
+  recurrent_op_helper
+  recurrent_op
   op_registry
   pool_op
   shape_op
@@ -48,5 +53,4 @@ cc_test_old(
   fused_transpose_op
   scope
   device_context
-  enforce
-  executor)
+  enforce)

test/cpp/inference/analysis/CMakeLists.txt

@@ -60,6 +60,7 @@ if(NOT APPLE AND NOT WIN32)
     EXTRA_DEPS
     reset_tensor_array
     paddle_inference_shared
+    python
     ARGS
     --inference_model_dir=${WORD2VEC_MODEL_DIR})
 elseif(WIN32)

test/cpp/inference/api/CMakeLists.txt

@@ -126,6 +126,8 @@ if(WITH_TESTING AND WITH_INFERENCE_API_TEST)
       ${filename}
       EXTRA_DEPS
       paddle_inference_shared
+      python
+      pybind
       ARGS
       --infer_model=${install_dir}/model
       --infer_data=${install_dir}/data.txt
@@ -139,6 +141,8 @@ if(WITH_TESTING AND WITH_INFERENCE_API_TEST)
       ${filename}
       EXTRA_DEPS
       paddle_inference_shared
+      python
+      pybind
       ARGS
       --infer_model=${install_dir}/model
       --infer_data=${install_dir}/data.txt
@@ -156,6 +160,7 @@ if(WITH_TESTING AND WITH_INFERENCE_API_TEST)
       ${filename}
       EXTRA_DEPS
       paddle_inference_shared
+      python
       ARGS
       --infer_model=${install_dir}/mobilenet_v2_models/1
       --infer_model2=${install_dir}/mobilenet_v2_models/xx
@@ -164,7 +169,7 @@ if(WITH_TESTING AND WITH_INFERENCE_API_TEST)
 
   function(inference_analysis_api_test_build TARGET_NAME filename)
     inference_analysis_test_build(${TARGET_NAME} SRCS ${filename} EXTRA_DEPS
-                                  paddle_inference_shared)
+                                  paddle_inference_shared python)
   endfunction()
 
   function(inference_analysis_api_int8_test_run TARGET_NAME test_binary
@@ -224,7 +229,7 @@ if(WITH_TESTING AND WITH_INFERENCE_API_TEST)
   function(inference_analysis_api_test_with_fake_data_build TARGET_NAME
            filename)
     inference_analysis_test_build(${TARGET_NAME} SRCS ${filename} EXTRA_DEPS
-                                  paddle_inference_shared)
+                                  paddle_inference_shared python)
   endfunction()
 
   function(inference_analysis_api_test_with_fake_data_run TARGET_NAME
@@ -403,6 +408,7 @@ if(WITH_TESTING AND WITH_INFERENCE_API_TEST)
     analyzer_dam_tester.cc
     EXTRA_DEPS
     paddle_inference_shared
+    python
     ARGS
     --infer_model=${DAM_SMALL_INSTALL_DIR}/model
     --infer_data=${DAM_SMALL_INSTALL_DIR}/data.txt)
@@ -458,6 +464,8 @@ if(WITH_TESTING AND WITH_INFERENCE_API_TEST)
     analyzer_ernie_tester.cc
     EXTRA_DEPS
     paddle_inference_shared
+    python
+    pybind
     ARGS
     --infer_model=${ERNIE_INSTALL_DIR}/model
     --infer_data=${ERNIE_INSTALL_DIR}/data.txt
@@ -503,6 +511,8 @@ if(WITH_TESTING AND WITH_INFERENCE_API_TEST)
     analyzer_transformer_compare_tester.cc
     EXTRA_DEPS
     paddle_inference_shared
+    python
+    pybind
     ARGS
     --infer_model=${TRANSFORMER_INSTALL_DIR}/model
     --infer_data=${TRANSFORMER_INSTALL_DIR}/data.txt
@@ -514,6 +524,8 @@ if(WITH_TESTING AND WITH_INFERENCE_API_TEST)
     analyzer_transformer_fuse_tester.cc
     EXTRA_DEPS
     paddle_inference_shared
+    python
+    pybind
     ARGS
     --infer_model=${TRANSFORMER_INSTALL_DIR}/model
     --infer_data=${TRANSFORMER_INSTALL_DIR}/data.txt
@@ -525,6 +537,7 @@ if(WITH_TESTING AND WITH_INFERENCE_API_TEST)
     analyzer_transformer_profile_tester.cc
     EXTRA_DEPS
     paddle_inference_shared
+    python
     ARGS
     --infer_model=${TRANSFORMER_INSTALL_DIR}/model
     --infer_data=${TRANSFORMER_INSTALL_DIR}/data.txt
@@ -543,6 +556,7 @@ if(WITH_TESTING AND WITH_INFERENCE_API_TEST)
     analyzer_vit_ocr_tester.cc
     EXTRA_DEPS
     paddle_inference_shared
+    python
     ARGS
     --infer_model=${VIT_OCR_INSTALL_DIR}/vit_ocr/model
     --infer_data=${VIT_OCR_INSTALL_DIR}/vit_ocr/datavit.txt)
@@ -566,6 +580,8 @@ if(WITH_TESTING AND WITH_INFERENCE_API_TEST)
     analyzer_detect_functional_mkldnn_tester.cc
     EXTRA_DEPS
     paddle_inference_shared
+    python
+    pybind
     ARGS
     --infer_model=${DENSEBOX_INSTALL_DIR}/model
     --infer_data=${DENSEBOX_INSTALL_DIR}/detect_input_50.txt
@@ -1161,6 +1177,8 @@ if(WITH_TESTING AND WITH_INFERENCE_API_TEST)
     lite_mul_model_test.cc
     EXTRA_DEPS
     paddle_inference_shared
+    python
+    pybind
     ARGS
     --infer_model=${LITE_MODEL_INSTALL_DIR})
   inference_analysis_test(
@@ -1169,6 +1187,8 @@ if(WITH_TESTING AND WITH_INFERENCE_API_TEST)
     lite_resnet50_test.cc
     EXTRA_DEPS
     paddle_inference_shared
+    python
+    pybind
     ARGS
     --infer_model=${RESNET50_MODEL_DIR})
 
@@ -1178,6 +1198,7 @@ if(WITH_TESTING AND WITH_INFERENCE_API_TEST)
     analyzer_capi_exp_tester.cc
     EXTRA_DEPS
     ${INFERENCE_C_EXTRA_DEPS}
+    python
     ARGS
     --infer_model=${RESNET50_MODEL_DIR}/model)
   if(WIN32)
@@ -1192,6 +1213,8 @@ if(WITH_TESTING AND WITH_INFERENCE_API_TEST)
     analyzer_capi_exp_pd_config_tester.cc
     EXTRA_DEPS
     ${INFERENCE_C_EXTRA_DEPS}
+    python
+    pybind
     ARGS
     --infer_model=${MOBILENET_INSTALL_DIR}/model)
   if(WIN32)
@@ -1207,6 +1230,7 @@ if(WITH_TESTING AND WITH_INFERENCE_API_TEST)
     analyzer_capi_exp_pd_tensor_tester.cc
     EXTRA_DEPS
     ${INFERENCE_C_EXTRA_DEPS}
+    python
     ARGS
     --infer_model=${MOBILENET_INSTALL_DIR}/model)
   if(WIN32)
@@ -1223,6 +1247,8 @@ if(WITH_TESTING AND WITH_INFERENCE_API_TEST)
       analyzer_capi_exp_pd_threads_tester.cc
       EXTRA_DEPS
       ${INFERENCE_C_EXTRA_DEPS}
+      python
+      pybind
       ARGS
       --infer_model=${MOBILENET_INSTALL_DIR}/model)
     if(WIN32)
@@ -1240,6 +1266,7 @@ if(WITH_TESTING AND WITH_INFERENCE_API_TEST)
     analyzer_zerocopy_tensor_tester.cc
     EXTRA_DEPS
     paddle_inference_shared
+    python
     ARGS
     --infer_model=${OCR_INSTALL_DIR}/model)
 
@@ -1250,6 +1277,7 @@ if(WITH_TESTING AND WITH_INFERENCE_API_TEST)
       analyzer_dist_model_tester.cc
       EXTRA_DEPS
       paddle_inference_shared
+      python
       ARGS
       --infer_model=${OCR_INSTALL_DIR}/model)
   endif()
@@ -1264,6 +1292,7 @@ if(WITH_TESTING AND WITH_INFERENCE_API_TEST)
       analyzer_dist_model_xpu_tester.cc
       EXTRA_DEPS
       paddle_inference_shared
+      python
       ARGS
       --infer_model=${OCR_INSTALL_DIR}/model)
   endif()
@@ -1274,6 +1303,7 @@ if(WITH_TESTING AND WITH_INFERENCE_API_TEST)
     analyzer_paddle_tensor_tester.cc
     EXTRA_DEPS
     paddle_inference_shared
+    python
     ARGS
     --infer_model=${OCR_INSTALL_DIR}/model
     --infer_data=${OCR_INSTALL_DIR}/data.txt
@@ -1302,6 +1332,8 @@ if(WITH_TESTING AND WITH_INFERENCE_API_TEST)
     analyzer_capi_exp_ner_tester.cc
     EXTRA_DEPS
     ${INFERENCE_C_EXTRA_DEPS}
+    python
+    pybind
     ARGS
     --infer_model=${CHINESE_NER_INSTALL_DIR}/model)
   if(WIN32)
@@ -1459,6 +1491,7 @@ if(WITH_TESTING AND WITH_INFERENCE_API_TEST)
       xpu_runtime_config_resnet50_test.cc
       EXTRA_DEPS
       paddle_inference_shared
+      python
       ARGS
       --infer_model=${RESNET50_MODEL_DIR})
   endif()
@@ -1474,6 +1507,7 @@ if(WITH_TESTING AND WITH_INFERENCE_API_TEST)
         api_impl_tester.cc
         DEPS
         paddle_inference_shared
+        python
         ARGS
         --word2vec_dirname=${WORD2VEC_MODEL_DIR}
         --book_dirname=${IMG_CLS_RESNET_INSTALL_DIR})
@@ -1497,6 +1531,7 @@ if(WITH_TESTING AND WITH_INFERENCE_API_TEST)
       analysis_predictor_tester.cc
       DEPS
       paddle_inference_shared
+      python
       ARGS
       --dirname=${WORD2VEC_MODEL_DIR})
   elseif(WIN32)
@@ -1508,6 +1543,7 @@ if(WITH_TESTING AND WITH_INFERENCE_API_TEST)
       analysis_predictor
       benchmark
       ${inference_deps}
+      pybind
       ARGS
       --dirname=${WORD2VEC_MODEL_DIR})
   endif()
@@ -1517,7 +1553,8 @@ if(WITH_TESTING AND WITH_INFERENCE_API_TEST)
       cc_test(
         test_mkldnn_quantizer
         SRCS mkldnn_quantizer_tester.cc
-        DEPS paddle_inference_shared ARGS --dirname=${WORD2VEC_MODEL_DIR})
+        DEPS paddle_inference_shared python ARGS
+             --dirname=${WORD2VEC_MODEL_DIR})
     elseif(WIN32)
       cc_test(
         test_mkldnn_quantizer