Fix c++17 bug (#54228)

* “update”

* update

* update

* update

* update

* test

* update

* test

* fix_c++17_bug

* fix coverage compile error

* test

* test

* test

* fix C++17 error

* fix c++17 error

* fix c++17 error

* test

* test

* test

* test

* fix cinn compile error

* compile to compiler

* set cinn c++14

---------
Co-authored-by: huangjiyi <947613776@qq.com>
Co-authored-by: Nhuangjiyi <43315610+huangjiyi@users.noreply.github.com>

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/flags.cmake

@@ -38,7 +38,7 @@ if(NOT WIN32)
   if(WITH_CINN)
     set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++14")
   else()
-    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++14")
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++17")
   endif()
 else()
   set(CMAKE_CXX_STANDARD 17)

paddle/fluid/distributed/fleet_executor/CMakeLists.txt

@@ -25,66 +25,35 @@ cc_library(
   task_loop_thread_pool
   SRCS task_loop_thread_pool.cc task_loop_thread.cc task_loop.cc
   DEPS enforce glog)
-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()
+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})
 if(WITH_DISTRIBUTE)
   set(DISTRIBUTE_COMPILE_FLAGS
       "-Wno-non-virtual-dtor -Wno-error=non-virtual-dtor -Wno-error=delete-non-virtual-dtor"

paddle/fluid/framework/CMakeLists.txt

@@ -729,8 +729,7 @@ if(WITH_DISTRIBUTE)
            section_worker.cc
            device_worker_factory.cc
            data_set.cc
-      DEPS fleet_executor
-           fleet_wrapper
+      DEPS fleet_wrapper
            recurrent_op_helper
            op_registry
            device_context
@@ -837,7 +836,6 @@ if(WITH_DISTRIBUTE)
            fleet
            heter_server
            brpc
-           fleet_executor
            phi)
     set(DISTRIBUTE_COMPILE_FLAGS "")
     if(CMAKE_CXX_COMPILER_VERSION VERSION_GREATER 7.0)
@@ -907,8 +905,7 @@ if(WITH_DISTRIBUTE)
            graph_to_program_pass
            variable_helper
            timer
-           monitor
-           fleet_executor)
+           monitor)
   endif()
 elseif(WITH_PSLIB)
   set(DISTRIBUTE_COMPILE_FLAGS "")
@@ -969,7 +966,6 @@ elseif(WITH_PSLIB)
          variable_helper
          timer
          monitor
-         fleet_executor
          ${BRPC_DEP})
 else()
   cc_library(
@@ -1017,8 +1013,7 @@ else()
          graph_to_program_pass
          variable_helper
          timer
-         monitor
-         fleet_executor)
+         monitor)
 endif()
 
 target_link_libraries(executor while_op_helper executor_gc_helper

paddle/fluid/framework/paddle2cinn/CMakeLists.txt

@@ -40,6 +40,7 @@ cc_library(
        cinn_graph_symbolization
        cinn
        cinn_launch_context
+       parallel_executor
        python
        pybind)
 
@@ -56,6 +57,7 @@ if(WITH_TESTING)
     SRCS
     build_cinn_pass_test.cc
     DEPS
+    fleet_executor
     build_cinn_pass
     cinn_compiler
     op_registry
@@ -72,6 +74,7 @@ if(WITH_TESTING)
     SRCS
     cinn_zero_tensor_trick_pass_test.cc
     DEPS
+    fleet_executor
     build_cinn_pass
     cinn_compiler
     op_registry

paddle/fluid/inference/api/CMakeLists.txt

@@ -80,14 +80,20 @@ if(WITH_ONNXRUNTIME)
          infer_io_utils
          model_utils
          onnxruntime
-         paddle2onnx)
+         paddle2onnx
+         fleet_executor)
 else()
   cc_library(
     analysis_predictor
     SRCS analysis_predictor.cc resource_manager.cc infer_context.cc
          ${mkldnn_quantizer_src}
-    DEPS ${inference_deps} zero_copy_tensor ir_pass_manager op_compatible_info
-         infer_io_utils model_utils)
+    DEPS ${inference_deps}
+         zero_copy_tensor
+         ir_pass_manager
+         op_compatible_info
+         infer_io_utils
+         model_utils
+         fleet_executor)
 endif()
 
 if(WITH_ONNXRUNTIME AND WIN32)

paddle/fluid/inference/tensorrt/CMakeLists.txt

@@ -28,10 +28,23 @@ nv_test(
   test_tensorrt
   SRCS test_tensorrt.cc
   DEPS dynload_cuda device_context dynamic_loader)
-nv_test(
-  test_tensorrt_engine
-  SRCS test_engine.cc test_dynamic_engine.cc
-  DEPS dynload_cuda tensorrt_engine tensorrt_plugin)
+if(WIN32)
+  nv_test(
+    test_tensorrt_engine
+    SRCS test_engine.cc test_dynamic_engine.cc
+    DEPS dynload_cuda tensorrt_engine tensorrt_plugin)
+elseif(WITH_CINN)
+  nv_test(
+    test_tensorrt_engine
+    SRCS test_engine.cc test_dynamic_engine.cc
+    DEPS fleet_executor cinn_compiler dynload_cuda tensorrt_engine
+         tensorrt_plugin python)
+else()
+  nv_test(
+    test_tensorrt_engine
+    SRCS test_engine.cc test_dynamic_engine.cc
+    DEPS fleet_executor dynload_cuda tensorrt_engine tensorrt_plugin python)
+endif()
 nv_test(
   test_arg_mapping_context
   SRCS test_arg_mapping_context.cc

paddle/fluid/inference/tensorrt/convert/CMakeLists.txt

@@ -137,11 +137,35 @@ nv_test(
   SRCS test_op_converter.cc
   DEPS paddle_framework ${GLOB_OPERATOR_DEPS} tensorrt_engine
        tensorrt_converter)
-
-nv_test(
-  test_custom_plugin_creater
-  SRCS test_custom_plugin_creater.cc
-  DEPS paddle_framework tensorrt_converter phi custom_operator init_phi)
+if(WIN32)
+  nv_test(
+    test_custom_plugin_creater
+    SRCS test_custom_plugin_creater.cc
+    DEPS paddle_framework tensorrt_converter phi custom_operator init_phi)
+elseif(WITH_CINN)
+  nv_test(
+    test_custom_plugin_creater
+    SRCS test_custom_plugin_creater.cc
+    DEPS paddle_framework
+         tensorrt_converter
+         phi
+         custom_operator
+         init_phi
+         fleet_executor
+         cinn_compiler
+         python)
+else()
+  nv_test(
+    test_custom_plugin_creater
+    SRCS test_custom_plugin_creater.cc
+    DEPS paddle_framework
+         tensorrt_converter
+         phi
+         custom_operator
+         init_phi
+         fleet_executor
+         python)
+endif()
 
 if(WITH_ONNXRUNTIME AND WIN32)
   # Copy onnxruntime for some c++ test in Windows, since the test will

paddle/fluid/inference/utils/CMakeLists.txt

@@ -11,7 +11,15 @@ cc_library(
   model_utils
   SRCS model_utils.cc
   DEPS proto_desc enforce)
-cc_test_old(infer_io_utils_tester SRCS io_utils_tester.cc DEPS infer_io_utils)
+
+cc_test_old(
+  infer_io_utils_tester
+  SRCS
+  io_utils_tester.cc
+  DEPS
+  infer_io_utils
+  fleet_executor
+  python)
 
 if(WITH_ONNXRUNTIME AND WIN32)
   # Copy onnxruntime for some c++ test in Windows, since the test will

paddle/fluid/operators/CMakeLists.txt

@@ -93,7 +93,7 @@ endif()
 set(OP_HEADER_DEPS ${OP_HEADER_DEPS} phi phi_utils static_prim_api get_expected_kernel_func)
 
 register_operators(EXCLUDES py_func_op dgc_op generated_op1 generated_op2 generated_op3 generated_op4 load_combine_op lstm_op run_program_op quantize_linear_op
-        recurrent_op save_combine_op sparse_attention_op sync_batch_norm_op activation_op ${OP_MKL_DEPS} DEPS ${OP_HEADER_DEPS})
+        recurrent_op save_combine_op sparse_attention_op sync_batch_norm_op activation_op ${OP_MKL_DEPS} DEPS ${OP_HEADER_DEPS} processgroup_comm_utils)
 
 op_library(generated_op UNITY SRCS generated_op1.cc generated_op2.cc generated_op3.cc generated_op4.cc DEPS ${OP_HEADER_DEPS})
 op_library(run_program_op DEPS executor_cache ${OP_HEADER_DEPS})
@@ -119,9 +119,9 @@ else()
 endif()
 
 if (WITH_GPU OR WITH_ROCM)
-    op_library(sync_batch_norm_op)
+    op_library(sync_batch_norm_op DEPS processgroup_comm_utils)
     if ((NOT WIN32) AND (NOT WITH_ROCM) AND (NOT PADDLE_WITH_ARM) AND (NOT ${CMAKE_CUDA_COMPILER_VERSION} VERSION_LESS 11.3) )
-        op_library(sparse_attention_op)
+        op_library(sparse_attention_op DEPS processgroup_comm_utils)
     endif()
 endif()
 

paddle/fluid/operators/tensorrt/CMakeLists.txt

-op_library(tensorrt_engine_op DEPS tensorrt_engine tensorrt_converter
-           infer_io_utils analysis_helper)
-nv_test(
-  test_tensorrt_engine_op
-  SRCS tensorrt_engine_op_test.cc
-  DEPS tensorrt_engine_op analysis)
+set(tensorrt_engine_op_deps tensorrt_engine tensorrt_converter infer_io_utils
+                            analysis_helper)
+
+op_library(tensorrt_engine_op DEPS ${tensorrt_engine_op_deps})
+if(NOT WIN32)
+  nv_test(
+    test_tensorrt_engine_op
+    SRCS tensorrt_engine_op_test.cc
+    DEPS tensorrt_engine_op analysis fleet_executor python)
+else()
+  nv_test(
+    test_tensorrt_engine_op
+    SRCS tensorrt_engine_op_test.cc
+    DEPS tensorrt_engine_op analysis fleet_executor)
+endif()
 
 if(WITH_ONNXRUNTIME AND WIN32)
   # Copy onnxruntime for some c++ test in Windows, since the test will

paddle/scripts/paddle_build.sh

@@ -3751,7 +3751,6 @@ EOF
 
     ccache -z
     cd ..
-
     if [ "${PYTHON_EXECUTABLE}" != "" ];then
         if [ "$SYSTEM" == "Darwin" ]; then
             ${PYTHON_EXECUTABLE} setup.py $2 --plat-name=macosx_10_9_x86_64;build_error=$?
@@ -3767,7 +3766,7 @@ EOF
     fi
     # ci will collect ccache hit rate
     collect_ccache_hits
-
+    
     if [ "$build_error" != 0 ];then
         exit 7;
     fi

test/cpp/eager/data_structure_tests/CMakeLists.txt

-cc_test_old(test_egr_ds_eager_tensor SRCS eager_tensor_test.cc DEPS
-            ${eager_deps})
-cc_test_old(test_egr_ds_auotgrad_meta SRCS autograd_meta_test.cc DEPS
-            ${eager_deps})
+cc_test_old(
+  test_egr_ds_eager_tensor
+  SRCS
+  eager_tensor_test.cc
+  DEPS
+  fleet_executor
+  final_dygraph_function
+  ${eager_deps}
+  python)
+cc_test_old(
+  test_egr_ds_auotgrad_meta
+  SRCS
+  autograd_meta_test.cc
+  DEPS
+  fleet_executor
+  final_dygraph_function
+  ${eager_deps}
+  python)
 
 if(NOT ((NOT WITH_PYTHON) AND ON_INFER))
+  if(WITH_CINN)
+    set(eager_deps ${eager_deps} cinn_compiler python)
+  endif()
   cc_test_old(
     test_egr_ds_grad_tensor_holder
     SRCS
     grad_tensor_holder_test.cc
     DEPS
+    fleet_executor
     conditional_block_op
     ${eager_deps}
     ${generated_deps})
@@ -17,6 +35,7 @@ if(NOT ((NOT WITH_PYTHON) AND ON_INFER))
     SRCS
     grad_node_info_test.cc
     DEPS
+    fleet_executor
     conditional_block_op
     ${eager_deps}
     ${generated_deps})
@@ -25,6 +44,7 @@ if(NOT ((NOT WITH_PYTHON) AND ON_INFER))
     SRCS
     accumulation_node_test.cc
     DEPS
+    fleet_executor
     conditional_block_op
     ${eager_deps}
     ${generated_deps})
@@ -33,6 +53,7 @@ if(NOT ((NOT WITH_PYTHON) AND ON_INFER))
     SRCS
     tensor_wrapper_test.cc
     DEPS
+    fleet_executor
     conditional_block_op
     ${eager_deps}
     ${generated_deps})

test/cpp/eager/performance_tests/CMakeLists.txt

 if(NOT (NOT WITH_PYTHON AND ON_INFER))
+  if(WITH_CINN)
+    set(eager_deps ${eager_deps} cinn_compiler python)
+  endif()
   cc_library(
     performance_benchmark_utils
     SRCS benchmark_utils.cc
@@ -17,6 +20,7 @@ if(NOT (NOT WITH_PYTHON AND ON_INFER))
     SRCS
     benchmark_eager_cpu.cc
     DEPS
+    fleet_executor
     conditional_block_op
     performance_benchmark_utils
     ${eager_deps}
@@ -26,6 +30,7 @@ if(NOT (NOT WITH_PYTHON AND ON_INFER))
     SRCS
     benchmark_fluid_cpu.cc
     DEPS
+    fleet_executor
     dygraph_function
     performance_benchmark_utils
     ${eager_deps}
@@ -36,17 +41,19 @@ if(NOT (NOT WITH_PYTHON AND ON_INFER))
     SRCS
     benchmark_eager_cuda.cc
     DEPS
+    fleet_executor
     conditional_block_op
     performance_benchmark_utils
     ${eager_deps}
-    ${fluid_deps})
+    ${fluid_deps}
+    fleet_executor)
   cc_test_old(
     test_egr_performance_benchmark_fluid_cuda
     SRCS
     benchmark_fluid_cuda.cc
     DEPS
+    fleet_executor
     conditional_block_op
-    standalone_executor
     performance_benchmark_utils
     dygraph_function
     ${eager_deps}

test/cpp/fluid/CMakeLists.txt

@@ -67,7 +67,9 @@ cc_test(
   save_load_combine_op_test
   SRCS save_load_combine_op_test.cc
   DEPS save_combine_op load_combine_op)
-
+if(WITH_CINN)
+  set(CINN_DEPS cinn_compiler python)
+endif()
 if(WITH_GPU)
   nv_test(
     dropout_op_test
@@ -81,7 +83,7 @@ if(WITH_GPU)
   nv_test(
     feed_forward_test
     SRCS feed_forward_test.cu
-    DEPS elementwise_add_op matmul_op tensor phi)
+    DEPS fleet_executor elementwise_add_op matmul_op tensor phi ${CINN_DEPS})
 elseif(WITH_ROCM)
   hip_test(
     dropout_op_test
@@ -103,16 +105,33 @@ 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
-  executor
-  recurrent_op_helper
-  recurrent_op
-  elementwise_add_op
-  ${COMMON_OP_DEPS})
+if(WITH_CINN)
+  cc_test_old(
+    op_debug_string_test
+    SRCS
+    op_debug_string_test.cc
+    DEPS
+    executor
+    fleet_executor
+    cinn_compiler
+    recurrent_op_helper
+    recurrent_op
+    elementwise_add_op
+    ${COMMON_OP_DEPS}
+    python)
+else()
+  cc_test_old(
+    op_debug_string_test
+    SRCS
+    op_debug_string_test.cc
+    DEPS
+    executor
+    fleet_executor
+    recurrent_op_helper
+    recurrent_op
+    elementwise_add_op
+    ${COMMON_OP_DEPS})
+endif()
 
 if(WITH_GPU)
   cc_test(

test/cpp/fluid/cinn/CMakeLists.txt

@@ -4,6 +4,7 @@ if(WITH_TESTING)
     SRCS
     cinn_launch_context_test.cc
     DEPS
+    fleet_executor
     phi
     lod_tensor
     scope

test/cpp/fluid/fused/CMakeLists.txt

+if(WITH_CINN)
+  set(CINN_DEPS cinn_compiler python)
+endif()
 if(WITH_GPU OR WITH_ROCM)
   # fusion_group
   if(NOT APPLE AND NOT WIN32)
@@ -30,13 +33,15 @@ if(WITH_GPU OR WITH_ROCM)
     nv_test(
       test_fused_layernorm_residual_dropout_bias
       SRCS fused_layernorm_residual_dropout_bias_test.cu
-      DEPS tensor
+      DEPS fleet_executor
+           tensor
            op_registry
            dropout_op
            generated_op
            device_context
            phi
-           memory)
+           memory
+           ${CINN_DEPS})
   endif()
   # resnet_unit needs cudnn 8.0 above
   if((NOT WITH_ROCM) AND (NOT ${CUDNN_VERSION} VERSION_LESS 8000))

test/cpp/fluid/lite/CMakeLists.txt

-cc_test_old(test_lite_engine_op SRCS lite_engine_op_test.cc DEPS lite_engine_op
-            analysis)
+cc_test_old(
+  test_lite_engine_op
+  SRCS
+  lite_engine_op_test.cc
+  DEPS
+  fleet_executor
+  lite_engine_op
+  analysis)

test/cpp/fluid/mkldnn/CMakeLists.txt

@@ -29,26 +29,54 @@ cc_test(
   test_mkldnn_caching
   SRCS test_mkldnn_caching.cc
   DEPS ${TEST_MKLDNN_CACHING_DEPS})
-
-cc_test_old(
-  test_mkldnn_op_nhwc
-  SRCS
-  test_mkldnn_op_nhwc.cc
-  DEPS
-  conditional_block_op
-  standalone_executor
-  executor
-  recurrent_op_helper
-  recurrent_op
-  op_registry
-  generated_static_op
-  crop_op
-  activation_op
-  generated_op
-  generated_static_op
-  phi
-  transpose_op
-  fused_transpose_op
-  scope
-  device_context
-  enforce)
+if(WITH_CINN)
+  cc_test_old(
+    test_mkldnn_op_nhwc
+    SRCS
+    test_mkldnn_op_nhwc.cc
+    DEPS
+    fleet_executor
+    conditional_block_op
+    standalone_executor
+    executor
+    recurrent_op_helper
+    cinn_compiler
+    recurrent_op
+    op_registry
+    generated_static_op
+    crop_op
+    activation_op
+    generated_op
+    generated_static_op
+    phi
+    transpose_op
+    fused_transpose_op
+    scope
+    device_context
+    enforce
+    python)
+else()
+  cc_test_old(
+    test_mkldnn_op_nhwc
+    SRCS
+    test_mkldnn_op_nhwc.cc
+    DEPS
+    fleet_executor
+    conditional_block_op
+    standalone_executor
+    executor
+    recurrent_op_helper
+    recurrent_op
+    op_registry
+    generated_static_op
+    crop_op
+    activation_op
+    generated_op
+    generated_static_op
+    phi
+    transpose_op
+    fused_transpose_op
+    scope
+    device_context
+    enforce)
+endif()

test/cpp/fluid/prim_ops/CMakeLists.txt

@@ -36,4 +36,16 @@ set(PRIM_OP_SRCS
     ${PRIM_OP_PATH}cast_p_op.cc
     ${PRIM_OP_PATH}rsqrt_p_op.cc
     ${PRIM_OP_PATH}uniform_random_p_op.cc)
-cc_test_old(prim_op_test SRCS prim_op_test.cc ${PRIM_OP_SRCS} DEPS op_registry)
+if(WITH_CINN)
+  set(CINN_DEPS cinn_compiler)
+endif()
+cc_test_old(
+  prim_op_test
+  SRCS
+  prim_op_test.cc
+  ${PRIM_OP_SRCS}
+  DEPS
+  fleet_executor
+  ${CINN_DEPS}
+  op_registry
+  python)

test/cpp/prim/CMakeLists.txt

@@ -14,12 +14,15 @@ set(prim_eager_deps
 
 set(prim_generated_deps final_dygraph_function final_dygraph_node
                         dygraph_function dygraph_node)
-
+if(WITH_CINN)
+  set(CINN_DEPS cinn_compiler)
+endif()
 cc_test_old(
   test_comp_static
   SRCS
   test_static_prim.cc
   DEPS
+  fleet_executor
   static_utils
   static_prim_api
   generated_op
@@ -33,9 +36,14 @@ cc_test_old(
   phi
   static_global_utils
   static_tensor_operants
-  generated_static_op)
+  generated_static_op
+  ${CINN_DEPS}
+  python)
 
 if(NOT (NOT WITH_PYTHON AND ON_INFER))
+  if(WITH_CINN)
+    set(prim_eager_deps ${prim_eager_deps} cinn_compiler)
+  endif()
   cc_library(
     init_env_utils
     SRCS init_env_utils.cc
@@ -46,9 +54,11 @@ if(NOT (NOT WITH_PYTHON AND ON_INFER))
     SRCS
     test_eager_prim.cc
     DEPS
+    fleet_executor
     ${prim_eager_deps}
     ${prim_generated_deps}
     prim_utils
     static_global_utils
-    init_env_utils)
+    init_env_utils
+    python)
 endif()