Merge branch 'develop' into trt_auto_dynamic_shape

PaddlePaddle · Apr 14, 2023 · 6829161 · 6829161
2 parents f86d9db + ce6978c
commit 6829161
Show file tree

Hide file tree

Showing 1,107 changed files with 21,306 additions and 34,356 deletions.
diff --git a/.gitignore b/.gitignore
@@ -73,19 +73,10 @@ tools/nvcc_lazy
 
 # This file is automatically generated.
 # TODO(zhiqiang) Move this file to build directory.
-paddle/infrt/dialect/pd/ir/pd_ops.td
-paddle/infrt/dialect/phi/ir/phi_cpu_kernels.td
-paddle/infrt/dialect/phi/ir/phi_gpu_kernels.td
-tools/infrt/kernels.json
-tools/infrt/kernel_signature.json
-paddle/infrt/dialect/pd/common/pd_ops_info.h
-.lit_test_times.txt
-paddle/infrt/tests/dialect/Output
-paddle/infrt/tests/lit.cfg.py
-paddle/infrt/kernel/phi/infershaped/infershaped_kernel_launchers.cc
 paddle/fluid/pybind/eager_op_function.cc
 tools/nvcc_lazy
 
+
 # these files (directories) are generated before build system generation
 paddle/fluid/operators/generated_op*.cc
 paddle/fluid/operators/generated_sparse_op.cc

diff --git a/CMakeLists.txt b/CMakeLists.txt
@@ -257,7 +257,7 @@ option(WITH_BOX_PS "Compile with box_ps support" OFF)
 option(WITH_XBYAK "Compile with xbyak support" ON)
 option(WITH_CONTRIB "Compile the third-party contributation" OFF)
 option(WITH_PSCORE "Compile with parameter server support" ${WITH_DISTRIBUTE})
-option(WITH_HETERPS "Compile with heterps" OFF})
+option(WITH_HETERPS "Compile with heterps" OFF)
 option(WITH_INFERENCE_API_TEST
        "Test fluid inference C++ high-level api interface" OFF)
 option(WITH_INFERENCE_NVTX "Paddle inference with nvtx for profiler" OFF)
@@ -269,7 +269,6 @@ option(
   OFF)
 option(WITH_LITE "Compile Paddle Fluid with Lite Engine" OFF)
 option(WITH_CINN "Compile PaddlePaddle with CINN" OFF)
-option(WITH_INFRT "Compile PaddlePaddle with INFRT" OFF)
 option(WITH_NCCL "Compile PaddlePaddle with NCCL support" ON)
 option(WITH_RCCL "Compile PaddlePaddle with RCCL support" ON)
 option(WITH_XPU_BKCL "Compile PaddlePaddle with BAIDU KUNLUN XPU BKCL" OFF)
@@ -289,7 +288,6 @@ option(NEW_RELEASE_ALL
        OFF)
 option(NEW_RELEASE_JIT
        "PaddlePaddle next-level release strategy for backup jit package" OFF)
-option(WITH_ASCEND_INT64 "Compile with int64 kernel for ascend NPU" OFF)
 option(WITH_POCKETFFT "Compile with pocketfft support" ON)
 option(WITH_RECORD_BUILDTIME
        "Compile PaddlePaddle with record all targets build time" OFF)

diff --git a/cmake/configure.cmake b/cmake/configure.cmake
@@ -93,10 +93,6 @@ if(WITH_BOX_PS)
   add_definitions(-DPADDLE_WITH_BOX_PS)
 endif()
 
-if(WITH_ASCEND_INT64)
-  add_definitions(-DPADDLE_WITH_ASCEND_INT64)
-endif()
-
 if(WITH_XPU)
   message(STATUS "Compile with XPU!")
   add_definitions(-DPADDLE_WITH_XPU)

diff --git a/cmake/cuda.cmake b/cmake/cuda.cmake
@@ -7,28 +7,33 @@ if(WITH_NV_JETSON)
   set(paddle_known_gpu_archs "53 62 72")
   set(paddle_known_gpu_archs10 "53 62 72")
   set(paddle_known_gpu_archs11 "53 62 72 87")
+  set(paddle_known_gpu_archs12 "53 62 72 87 90")
 elseif(NEW_RELEASE_ALL)
   message("Using New Release Strategy - All Arches Packge")
   add_definitions(-DNEW_RELEASE_ALL)
-  set(paddle_known_gpu_archs "35 50 52 60 61 70 75 80 86")
-  set(paddle_known_gpu_archs10 "35 50 52 60 61 70 75")
+  set(paddle_known_gpu_archs "50 52 60 61 70 75 80 86 90")
+  set(paddle_known_gpu_archs10 "50 52 60 61 70 75")
   set(paddle_known_gpu_archs11 "50 60 61 70 75 80")
+  set(paddle_known_gpu_archs12 "50 60 61 70 75 80 90")
 elseif(NEW_RELEASE_PYPI)
   message("Using New Release Strategy - Cubin Packge")
   add_definitions(-DNEW_RELEASE_PYPI)
-  set(paddle_known_gpu_archs "35 50 52 60 61 70 75 80 86")
+  set(paddle_known_gpu_archs "50 52 60 61 70 75 80 86 90")
   set(paddle_known_gpu_archs10 "")
   set(paddle_known_gpu_archs11 "61 70 75 80")
+  set(paddle_known_gpu_archs12 "61 70 75 80 90")
 elseif(NEW_RELEASE_JIT)
   message("Using New Release Strategy - JIT Packge")
   add_definitions(-DNEW_RELEASE_JIT)
-  set(paddle_known_gpu_archs "35 50 52 60 61 70 75 80 86")
-  set(paddle_known_gpu_archs10 "35 50 60 70 75")
-  set(paddle_known_gpu_archs11 "35 50 60 70 75 80")
+  set(paddle_known_gpu_archs "50 52 60 61 70 75 80 86 90")
+  set(paddle_known_gpu_archs10 "50 60 70 75")
+  set(paddle_known_gpu_archs11 "50 60 70 75 80")
+  set(paddle_known_gpu_archs12 "50 60 70 75 80 90")
 else()
-  set(paddle_known_gpu_archs "35 50 52 60 61 70 75 80")
+  set(paddle_known_gpu_archs "50 52 60 61 70 75 80 90")
   set(paddle_known_gpu_archs10 "50 52 60 61 70 75")
   set(paddle_known_gpu_archs11 "52 60 61 70 75 80")
+  set(paddle_known_gpu_archs12 "52 60 61 70 75 80 90")
 endif()
 
 ######################################################################################
@@ -58,7 +63,8 @@ function(detect_installed_gpus out_variable)
       "}\n")
 
     execute_process(
-      COMMAND "${CUDA_NVCC_EXECUTABLE}" "--run" "${cufile}"
+      COMMAND "${CUDA_NVCC_EXECUTABLE}" "-ccbin=${CMAKE_C_COMPILER}" "--run"
+              "${cufile}"
       WORKING_DIRECTORY "${PROJECT_BINARY_DIR}/CMakeFiles/"
       RESULT_VARIABLE nvcc_res
       OUTPUT_VARIABLE nvcc_out
@@ -99,12 +105,12 @@ endfunction()
 function(select_nvcc_arch_flags out_variable out_arch_bin)
   # List of arch names
   set(archs_names
-      "Kepler"
       "Maxwell"
       "Pascal"
       "Volta"
       "Turing"
       "Ampere"
+      "Hopper"
       "All"
       "Manual")
   set(archs_name_default "Auto")
@@ -143,9 +149,7 @@ function(select_nvcc_arch_flags out_variable out_arch_bin)
     unset(CUDA_ARCH_PTX CACHE)
   endif()
 
-  if(${CUDA_ARCH_NAME} STREQUAL "Kepler")
-    set(cuda_arch_bin "30 35")
-  elseif(${CUDA_ARCH_NAME} STREQUAL "Maxwell")
+  if(${CUDA_ARCH_NAME} STREQUAL "Maxwell")
     if(WITH_NV_JETSON)
       set(cuda_arch_bin "53")
     else()
@@ -175,6 +179,8 @@ function(select_nvcc_arch_flags out_variable out_arch_bin)
         set(cuda_arch_bin "80 86")
       endif()
     endif()
+  elseif(${CUDA_ARCH_NAME} STREQUAL "Hopper")
+    set(cuda_arch_bin "90")
   elseif(${CUDA_ARCH_NAME} STREQUAL "All")
     set(cuda_arch_bin ${paddle_known_gpu_archs})
   elseif(${CUDA_ARCH_NAME} STREQUAL "Auto")
@@ -265,6 +271,11 @@ elseif(${CMAKE_CUDA_COMPILER_VERSION} LESS 12.0) # CUDA 11.2+
   set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -D_MWAITXINTRIN_H_INCLUDED")
   set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -D__STRICT_ANSI__")
   set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -Wno-deprecated-gpu-targets")
+elseif(${CMAKE_CUDA_COMPILER_VERSION} LESS 13.0) # CUDA 12.0+
+  set(paddle_known_gpu_archs "${paddle_known_gpu_archs12} 86")
+  set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -D_MWAITXINTRIN_H_INCLUDED")
+  set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -D__STRICT_ANSI__")
+  set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -Wno-deprecated-gpu-targets")
 endif()
 
 if(NOT ${CMAKE_CUDA_COMPILER_VERSION} LESS 10.0)

diff --git a/cmake/experiments/cuda_module_loading_lazy.cmake b/cmake/experiments/cuda_module_loading_lazy.cmake
@@ -31,9 +31,9 @@ if(LINUX)
     message("cuda 11.7+ already support lazy module loading")
     return()
   endif()
-  if(${CUDA_VERSION} VERSION_LESS "11.2" AND ${CMAKE_CXX_COMPILER_VERSION}
+  if(${CUDA_VERSION} VERSION_LESS "12.0" AND ${CMAKE_CXX_COMPILER_VERSION}
                                              VERSION_GREATER_EQUAL 12.0)
-    message("cuda less than 11.2 doesn't support gcc12")
+    message("cuda less than 12.0 doesn't support gcc12")
     return()
   endif()
 

diff --git a/cmake/external/ascend.cmake b/cmake/external/ascend.cmake
diff --git a/cmake/external/gtest.cmake b/cmake/external/gtest.cmake
@@ -69,37 +69,72 @@ if(NOT WIN32 AND ${CMAKE_CXX_COMPILER_VERSION} VERSION_GREATER 12.0)
   set(GTEST_PATCH_COMMAND patch -d ${GTEST_SOURCE_DIR}/googletest/src <
                           ${native_src})
 endif()
-ExternalProject_Add(
-  extern_gtest
-  ${EXTERNAL_PROJECT_LOG_ARGS} ${SHALLOW_CLONE}
-  GIT_REPOSITORY ${GTEST_REPOSITORY}
-  GIT_TAG ${GTEST_TAG}
-  DEPENDS ${GTEST_DEPENDS}
-  PREFIX ${GTEST_PREFIX_DIR}
-  UPDATE_COMMAND ""
-  PATCH_COMMAND ${GTEST_PATCH_COMMAND}
-  CMAKE_ARGS -DCMAKE_CXX_COMPILER=${CMAKE_CXX_COMPILER}
-             -DCMAKE_C_COMPILER=${CMAKE_C_COMPILER}
-             -DCMAKE_CXX_FLAGS=${GTEST_CMAKE_CXX_FLAGS}
-             -DCMAKE_CXX_FLAGS_RELEASE=${CMAKE_CXX_FLAGS_RELEASE}
-             -DCMAKE_CXX_FLAGS_DEBUG=${CMAKE_CXX_FLAGS_DEBUG}
-             -DCMAKE_C_FLAGS=${GTEST_CMAKE_C_FLAGS}
-             -DCMAKE_C_FLAGS_DEBUG=${CMAKE_C_FLAGS_DEBUG}
-             -DCMAKE_C_FLAGS_RELEASE=${CMAKE_C_FLAGS_RELEASE}
-             -DCMAKE_INSTALL_PREFIX=${GTEST_INSTALL_DIR}
-             -DCMAKE_POSITION_INDEPENDENT_CODE=ON
-             -DBUILD_GMOCK=ON
-             -Dgtest_disable_pthreads=ON
-             -Dgtest_force_shared_crt=ON
-             -DCMAKE_BUILD_TYPE=${THIRD_PARTY_BUILD_TYPE}
-             ${EXTERNAL_OPTIONAL_ARGS}
-  CMAKE_CACHE_ARGS
-    -DCMAKE_INSTALL_PREFIX:PATH=${GTEST_INSTALL_DIR}
-    -DCMAKE_POSITION_INDEPENDENT_CODE:BOOL=ON
-    -DCMAKE_BUILD_TYPE:STRING=${THIRD_PARTY_BUILD_TYPE}
-  BUILD_BYPRODUCTS ${GTEST_LIBRARIES}
-  BUILD_BYPRODUCTS ${GTEST_MAIN_LIBRARIES}
-  BUILD_BYPRODUCTS ${GMOCK_LIBRARIES})
+if(WIN32)
+  ExternalProject_Add(
+    extern_gtest
+    ${EXTERNAL_PROJECT_LOG_ARGS} ${SHALLOW_CLONE}
+    GIT_REPOSITORY ${GTEST_REPOSITORY}
+    GIT_TAG ${GTEST_TAG}
+    DEPENDS ${GTEST_DEPENDS}
+    PREFIX ${GTEST_PREFIX_DIR}
+    UPDATE_COMMAND ""
+    PATCH_COMMAND ${GTEST_PATCH_COMMAND}
+    CMAKE_ARGS -DCMAKE_CXX_COMPILER=${CMAKE_CXX_COMPILER}
+               -DCMAKE_C_COMPILER=${CMAKE_C_COMPILER}
+               -DCMAKE_CXX_FLAGS=${GTEST_CMAKE_CXX_FLAGS}
+               -DCMAKE_CXX_FLAGS_RELEASE=${CMAKE_CXX_FLAGS_RELEASE}
+               -DCMAKE_CXX_FLAGS_DEBUG=${CMAKE_CXX_FLAGS_DEBUG}
+               -DCMAKE_C_FLAGS=${GTEST_CMAKE_C_FLAGS}
+               -DCMAKE_C_FLAGS_DEBUG=${CMAKE_C_FLAGS_DEBUG}
+               -DCMAKE_C_FLAGS_RELEASE=${CMAKE_C_FLAGS_RELEASE}
+               -DCMAKE_INSTALL_PREFIX=${GTEST_INSTALL_DIR}
+               -DCMAKE_POSITION_INDEPENDENT_CODE=ON
+               -DBUILD_GMOCK=ON
+               -Dgtest_disable_pthreads=ON
+               -Dgtest_force_shared_crt=ON
+               -DCMAKE_BUILD_TYPE=${THIRD_PARTY_BUILD_TYPE}
+               ${EXTERNAL_OPTIONAL_ARGS}
+    CMAKE_CACHE_ARGS
+      -DCMAKE_INSTALL_PREFIX:PATH=${GTEST_INSTALL_DIR}
+      -DCMAKE_POSITION_INDEPENDENT_CODE:BOOL=ON
+      -DCMAKE_BUILD_TYPE:STRING=${THIRD_PARTY_BUILD_TYPE}
+    BUILD_BYPRODUCTS ${GTEST_LIBRARIES}
+    BUILD_BYPRODUCTS ${GTEST_MAIN_LIBRARIES}
+    BUILD_BYPRODUCTS ${GMOCK_LIBRARIES})
+else()
+  ExternalProject_Add(
+    extern_gtest
+    ${EXTERNAL_PROJECT_LOG_ARGS} ${SHALLOW_CLONE}
+    GIT_REPOSITORY ${GTEST_REPOSITORY}
+    GIT_TAG ${GTEST_TAG}
+    DEPENDS ${GTEST_DEPENDS}
+    PREFIX ${GTEST_PREFIX_DIR}
+    UPDATE_COMMAND ""
+    PATCH_COMMAND ${GTEST_PATCH_COMMAND}
+    CMAKE_ARGS -DCMAKE_CXX_COMPILER=${CMAKE_CXX_COMPILER}
+               -DCMAKE_C_COMPILER=${CMAKE_C_COMPILER}
+               -DCMAKE_CXX_FLAGS=${GTEST_CMAKE_CXX_FLAGS}
+               -DCMAKE_CXX_FLAGS_RELEASE=${CMAKE_CXX_FLAGS_RELEASE}
+               -DCMAKE_CXX_FLAGS_DEBUG=${CMAKE_CXX_FLAGS_DEBUG}
+               -DCMAKE_C_FLAGS=${GTEST_CMAKE_C_FLAGS}
+               -DCMAKE_C_FLAGS_DEBUG=${CMAKE_C_FLAGS_DEBUG}
+               -DCMAKE_C_FLAGS_RELEASE=${CMAKE_C_FLAGS_RELEASE}
+               -DCMAKE_INSTALL_PREFIX=${GTEST_INSTALL_DIR}
+               -DCMAKE_POSITION_INDEPENDENT_CODE=ON
+               -DBUILD_GMOCK=ON
+               -Dgtest_disable_pthreads=ON
+               -Dgtest_force_shared_crt=ON
+               -DCMAKE_BUILD_TYPE=${THIRD_PARTY_BUILD_TYPE}
+               ${EXTERNAL_OPTIONAL_ARGS}
+    CMAKE_CACHE_ARGS
+      -DCMAKE_INSTALL_PREFIX:PATH=${GTEST_INSTALL_DIR}
+      -DCMAKE_POSITION_INDEPENDENT_CODE:BOOL=ON
+      -DCMAKE_BUILD_TYPE:STRING=${THIRD_PARTY_BUILD_TYPE}
+    CMAKE_GENERATOR "Unix Makefiles"
+    BUILD_BYPRODUCTS ${GTEST_LIBRARIES}
+    BUILD_BYPRODUCTS ${GTEST_MAIN_LIBRARIES}
+    BUILD_BYPRODUCTS ${GMOCK_LIBRARIES})
+endif()
 
 add_library(gtest STATIC IMPORTED GLOBAL)
 set_property(TARGET gtest PROPERTY IMPORTED_LOCATION ${GTEST_LIBRARIES})