7.15

.clang-format

@@ -6,11 +6,11 @@
 # The basic usage is,
 #   clang-format -i -style=file PATH/TO/SOURCE/CODE
 #
-# The -style=file implicit use ".clang-format" file located in one of 
-# parent directory. 
+# The -style=file implicit use ".clang-format" file located in one of
+# parent directory.
 # The -i means inplace change.
 #
-# The document of clang-format is 
+# The document of clang-format is
 #   http://clang.llvm.org/docs/ClangFormat.html
 #   http://clang.llvm.org/docs/ClangFormatStyleOptions.html
 ---
@@ -20,7 +20,7 @@ IndentWidth:     2
 TabWidth:        2
 ContinuationIndentWidth: 4
 AccessModifierOffset: -1  # The private/protected/public has no indent in class
-Standard:  Cpp11 
+Standard:  Cpp11
 AllowAllParametersOfDeclarationOnNextLine: true
 BinPackParameters: false
 BinPackArguments: false

.clang-tidy

@@ -9,7 +9,7 @@ bugprone-copy-constructor-init,
 -bugprone-dangling-handle,
 -bugprone-dynamic-static-initializers,
 bugprone-exception-escape,
--bugprone-fold-init-type,
+bugprone-fold-init-type,
 -bugprone-forwarding-reference-overload,
 bugprone-inaccurate-erase,
 bugprone-incorrect-roundings,
@@ -70,7 +70,7 @@ clang-analyzer-core.uninitialized.Assign,
 clang-analyzer-cplusplus.InnerPointer,
 -clang-analyzer-cplusplus.Move,
 -clang-analyzer-cplusplus.NewDelete,
--clang-analyzer-cplusplus.NewDeleteLeaks,
+clang-analyzer-cplusplus.NewDeleteLeaks,
 -clang-analyzer-cplusplus.PureVirtualCall,
 -clang-analyzer-cplusplus.SelfAssignment,
 -clang-analyzer-cplusplus.SmartPtr,

.github/CODEOWNERS

@@ -1 +1,76 @@
+# This file is migrated from CI script, it's an effort of modenizing our dev infra.
+# Code owners are expected to take responsibility for review patches to respective file.
+
+/CMakeLists.txt @wanghuancoder @Aurelius84 @XiaoguangHu01 @qili93
+paddle/fluid/distributed/collective @sneaxiy @ForFishes
+paddle/fluid/eager/autograd_meta.cc @JiabinYang @phlrain
+paddle/fluid/eager/autograd_meta.h @JiabinYang @phlrain
+paddle/fluid/eager/backward.cc @JiabinYang @phlrain
+paddle/fluid/eager/backward.h @JiabinYang @phlrain
+paddle/fluid/eager/grad_node_info.cc @JiabinYang @phlrain
+paddle/fluid/eager/grad_node_info.h @JiabinYang @phlrain
+paddle/fluid/eager/grad_tensor_holder.cc @JiabinYang @phlrain
+paddle/fluid/eager/grad_tensor_holder.h @JiabinYang @phlrain
+paddle/fluid/eager/tensor_wrapper.h @JiabinYang @phlrain
+paddle/fluid/framework/block_desc.h @XiaoguangHu01 @zhiqiu @Xreki @qili93 @Aurelius84
+paddle/fluid/framework/details/op_registry.h @XiaoguangHu01 @zhiqiu @Xreki @qili93 @Aurelius84
+paddle/fluid/framework/framework.proto @XiaoguangHu01 @zhiqiu @Xreki @qili93 @Aurelius84
+paddle/fluid/framework/grad_op_desc_maker.h @XiaoguangHu01 @zhiqiu @Xreki @qili93 @Aurelius84
+paddle/fluid/framework/ir/graph.h @XiaoguangHu01 @zhiqiu @Xreki @qili93 @Aurelius84
+paddle/fluid/framework/ir/node.h @XiaoguangHu01 @zhiqiu @Xreki @qili93 @Aurelius84
+paddle/fluid/framework/lod_tensor.h @XiaoguangHu01 @zhiqiu @Xreki @qili93 @Aurelius84
+paddle/fluid/framework/op_desc.h @XiaoguangHu01 @zhiqiu @Xreki @qili93 @Aurelius84
+paddle/fluid/framework/operator.h @XiaoguangHu01 @zhiqiu @Xreki @qili93 @Aurelius84
+paddle/fluid/framework/scope.h @XiaoguangHu01 @zhiqiu @Xreki @qili93 @Aurelius84
+paddle/fluid/framework/selected_rows.h @XiaoguangHu01 @zhiqiu @Xreki @qili93 @Aurelius84
+paddle/fluid/framework/tensor.h @XiaoguangHu01 @zhiqiu @Xreki @qili93 @Aurelius84
+paddle/fluid/framework/unused_var_check.cc @zhiqiu @phlrain
+paddle/fluid/framework/var_desc.h @XiaoguangHu01 @zhiqiu @Xreki @qili93 @Aurelius84
+paddle/fluid/operators/distributed/send_recv.proto.in @gongweibao @seiriosPlus
+paddle/fluid/prim/api/api.yaml @cxxly @xiaoguoguo626807 @Charles-hit @cyber-pioneer @JiabinYang
+paddle/fluid/prim/api/composite_backward/composite_backward_api.h @cxxly @xiaoguoguo626807 @Charles-hit @cyber-pioneer @JiabinYang
+paddle/fluid/prim/api/composite_backward/composite_double_backward_api.h @cxxly @xiaoguoguo626807 @Charles-hit @cyber-pioneer @JiabinYang
+paddle/fluid/prim/api/manual_prim/prim_manual_api.h @cxxly @xiaoguoguo626807 @Charles-hit @cyber-pioneer @JiabinYang
+paddle/phi/api/include/tensor.h @phlrain @zyfncg @YuanRisheng
+paddle/phi/core/attribute.h @phlrain @zyfncg @YuanRisheng
+paddle/phi/core/dense_tensor.h @phlrain @zyfncg @YuanRisheng
+paddle/phi/core/device_context.h @phlrain @zyfncg @YuanRisheng
+paddle/phi/core/infermeta_utils.h @phlrain @zyfncg @YuanRisheng
+paddle/phi/core/kernel_context.h @phlrain @zyfncg @YuanRisheng
+paddle/phi/core/kernel_factory.h @phlrain @zyfncg @YuanRisheng
+paddle/phi/core/kernel_registry.h @phlrain @zyfncg @YuanRisheng
+paddle/phi/core/kernel_utils.h @phlrain @zyfncg @YuanRisheng
+paddle/phi/core/meta_tensor.h @phlrain @zyfncg @YuanRisheng
+paddle/phi/core/tensor_base.h @phlrain @zyfncg @YuanRisheng
+paddle/phi/core/tensor_meta.h @phlrain @zyfncg @YuanRisheng
+paddle/phi/infermeta/spmd_rules @LiYuRio @ForFishes @zhiqiu
+paddle/scripts/paddle_build.bat @zhwesky2010 @wanghuancoder @Aurelius84
+paddle/scripts/paddle_build.sh @risemeup1 @zhangbo9674 @XieYunshen
+pyproject.toml @SigureMo @gouzil
+python/paddle/autograd/backward_utils.py @Aurelius84 @cxxly @xiaoguoguo626807 @changeyoung98 @cyber-pioneer
+python/paddle/autograd/ir_backward.py @Aurelius84 @cxxly @xiaoguoguo626807 @changeyoung98
+python/paddle/base/backward.py @XiaoguangHu01 @zhiqiu @Xreki @qili93 @Aurelius84
+python/paddle/base/compiler.py @XiaoguangHu01 @zhiqiu @Xreki @qili93 @Aurelius84
+python/paddle/base/dygraph/layers.py @JiabinYang @phlrain
+python/paddle/base/framework.py @XiaoguangHu01 @zhiqiu @Xreki @qili93 @Aurelius84
+python/paddle/base/__init__.py @phlrain @Aurelius84 @qili93
+python/paddle/base/tests/unittests/white_list/check_op_sequence_batch_1_input_white_list.py @Aurelius84 @phlrain
+python/paddle/base/tests/unittests/white_list/check_op_sequence_instance_0_input_white_list.py @Aurelius84 @phlrain
+python/paddle/base/tests/unittests/white_list/check_shape_white_list.py @hong19860320 @Aurelius84 @phlrain
+python/paddle/base/tests/unittests/white_list/compile_vs_runtime_white_list.py @Aurelius84 @phlrain
+python/paddle/base/tests/unittests/white_list/no_check_set_white_list.py @Aurelius84 @phlrain
+python/paddle/base/tests/unittests/white_list/no_grad_set_white_list.py @Aurelius84 @phlrain
+python/paddle/base/tests/unittests/white_list/op_accuracy_white_list.py @juncaipeng @zhangting2020 @Aurelius84
+python/paddle/base/tests/unittests/white_list/op_threshold_white_list.py @juncaipeng @zhangting2020 @Aurelius84
+python/paddle/distributed/fleet/__init__.py @sneaxiy @raindrops2sea
+python/paddle/distributed/fleet/launch.py @sneaxiy @raindrops2sea
+python/paddle/distributed/__init__.py @sneaxiy @raindrops2sea
+python/paddle/incubate/autograd/composite_rules.py @cyber-pioneer @xiaoguoguo626807 @Charles-hit @JiabinYang
+python/paddle/incubate/autograd/primitives.py @cyber-pioneer @xiaoguoguo626807 @Charles-hit @JiabinYang
+python/paddle/_typing @SigureMo @zrr1999 @gouzil
 python/requirements.txt @phlrain @jzhang533 @kolinwei
+test/dygraph_to_static @SigureMo @Aurelius84 @gouzil
+test/sot @SigureMo @Aurelius84 @gouzil
+tools/parallel_UT_rule.py @zhwesky2010 @wanghuancoder @Aurelius84
+tools/windows/run_unittests.sh @zhwesky2010 @wanghuancoder @Aurelius84
+.pre-commit-config.yaml @SigureMo @gouzil

.gitignore

@@ -87,8 +87,6 @@ paddle/fluid/operators/generated_sparse_op.cc
 paddle/fluid/operators/generated_static_op.cc
 paddle/fluid/operators/generated_fused_op.cc
 paddle/fluid/operators/ops_signature/generated_*.cc
-paddle/phi/api/yaml/parsed_apis/
-paddle/fluid/operators/generator/parsed_ops/
 paddle/fluid/pybind/tmp_eager_op_function_impl.h
 paddle/fluid/pybind/eager_op_function_impl.h
 paddle/fluid/pybind/eager_op_function_impl.h
@@ -107,3 +105,8 @@ paddle/phi/kernels/fusion/cutlass/cutlass_kernels/fpA_intB_gemm/autogen/*
 paddle/phi/kernels/fusion/cutlass/cutlass_kernels/fpA_intB_gemm/autogen_tmp/*
 paddle/fluid/pybind/static_op_function.*
 paddle/fluid/pybind/ops_api.cc
+python/paddle/tensor/tensor.pyi
+paddle/phi/kernels/fusion/cutlass/conv2d/build
+paddle/phi/kernels/fusion/cutlass/conv2d/cutlass
+paddle/phi/kernels/fusion/cutlass/gemm_epilogue/build
+paddle/phi/kernels/fusion/cutlass/gemm_epilogue/cutlass

.gitmodules

@@ -122,3 +122,6 @@
 	path = third_party/nlohmann_json
 	url = https://github.com/nlohmann/json.git
 	ignore = dirty
+[submodule "third_party/yaml-cpp"]
+	path = third_party/yaml-cpp
+	url = https://github.com/jbeder/yaml-cpp

.pre-commit-config.yaml

@@ -20,7 +20,6 @@ repos:
     -   id: sort-simple-yaml
         files: (ops|backward|op_[a-z_]+)\.yaml$
     -   id: trailing-whitespace
-        files: (.*\.(py|bzl|md|rst|c|cc|cxx|cpp|cu|h|hpp|hxx|xpu|kps|cmake|yaml|yml|hook)|BUILD|.*\.BUILD|WORKSPACE|CMakeLists\.txt)$
 -   repo: https://github.com/Lucas-C/pre-commit-hooks.git
     rev: v1.5.1
     hooks:
@@ -44,7 +43,7 @@ repos:
         name: copyright_checker
         entry: python ./tools/codestyle/copyright.py
         language: system
-        files: \.(c|cc|cxx|cpp|cu|h|hpp|hxx|proto|xpu|kps|py|sh)$
+        files: \.(c|cc|cxx|cpp|cu|h|hpp|hxx|proto|xpu|kps|py|pyi|sh)$
         exclude: |
             (?x)^(
                 paddle/utils/.*|
@@ -55,9 +54,8 @@ repos:
     rev: 23.3.0
     hooks:
     -   id: black
-        files: (.*\.(py|pyi|bzl)|BUILD|.*\.BUILD|WORKSPACE)$
 -   repo: https://github.com/astral-sh/ruff-pre-commit
-    rev: v0.3.5
+    rev: v0.5.0
     hooks:
     -   id: ruff
         args: [--fix, --exit-non-zero-on-fix, --no-cache]

CMakeLists.txt

@@ -54,8 +54,7 @@ option(WITH_XPU "Compile PaddlePaddle with BAIDU KUNLUN XPU" OFF)
 option(WITH_XPU_KP "Compile PaddlePaddle with BAIDU XPU compiler " OFF)
 option(WITH_XPU_XFT "Compile PaddlePaddle with BAIDU XPU-XFT" OFF)
 option(WITH_XPU_PLUGIN "Compile PaddlePaddle with BAIDU XPU plugin" OFF)
-option(WITH_XPU_XHPC "Compile PaddlePaddle with BAIDU XPU-HPC library"
-       ${WITH_XPU})
+option(WITH_XPU_XRE5 "Compile PaddlePaddle with BAIDU XPU XRE 5" OFF)
 option(WITH_WIN_DUMP_DBG "Compile with windows core dump debug mode" OFF)
 option(WITH_ROCM "Compile PaddlePaddle with ROCM platform" OFF)
 option(WITH_IPU "Compile PaddlePaddle with Graphcore IPU" OFF)
@@ -91,14 +90,17 @@ endif()
 if(WITH_GPU AND WITH_XPU_XFT)
   message(FATAL_ERROR "Error when compile GPU and XPU-XFT at the same time")
 endif()
-if(WITH_GPU AND WITH_XPU_XHPC)
-  message(FATAL_ERROR "Error when compile GPU and XPU-HPC at the same time")
+if(WITH_GPU AND WITH_XPU_XRE5)
+  message(FATAL_ERROR "Error when compile GPU and XPU-XRE5 at the same time")
 endif()
 if(WITH_GPU AND WITH_ROCM)
   message(FATAL_ERROR "Error when compile CUDA and ROCM at the same time")
 endif()
 
 if(WITH_GPU AND NOT APPLE)
+  if(WITH_PIP_CUDA_LIBRARIES AND CMAKE_SYSTEM_NAME STREQUAL "Windows")
+    add_definitions(-DPADDLE_WITH_PIP_CUDA_LIBRARIES)
+  endif()
   #(Note risemeup1): The cudart dynamic library libcudart.so is used by set CUDA_USE_STATIC_CUDA_RUNTIME and CMAKE_CUDA_FLAGS
   if(CMAKE_SYSTEM_NAME STREQUAL "Linux" AND CMAKE_SYSTEM_PROCESSOR STREQUAL
                                             "x86_64")
@@ -107,8 +109,8 @@ if(WITH_GPU AND NOT APPLE)
         CACHE BOOL "" FORCE)
     set(CMAKE_CUDA_FLAGS "--cudart shared")
     if(WITH_PIP_CUDA_LIBRARIES)
-      #(Note risemeup1): Flag 'WITH_PIP_CUDA_LIBRARIES' will be used in dynamic_loader.cc to search for CUDA-related .so files through the Python libraries provided by NVIDIA.
-      add_definitions(-DWITH_PIP_CUDA_LIBRARIES)
+      #(Note risemeup1): Flag 'PADDLE_WITH_PIP_CUDA_LIBRARIES' will be used in dynamic_loader.cc to search for CUDA-related .so files through the Python libraries provided by NVIDIA.
+      add_definitions(-DPADDLE_WITH_PIP_CUDA_LIBRARIES)
     endif()
   endif()
   enable_language(CUDA)
@@ -465,12 +467,12 @@ if(NOT WITH_XPU AND WITH_XPU_BKCL)
       CACHE STRING "Disable BKCL when compiling without XPU" FORCE)
 endif()
 
-if(NOT WITH_XPU AND WITH_XPU_XHPC)
-  message(
-    WARNING "Disable XHPC when compiling without XPU. Force WITH_XPU_XHPC=OFF.")
-  set(WITH_XPU_XHPC
+if(NOT WITH_XPU AND WITH_XPU_XRE5)
+  message((WARNING
+           "Disable XRE5 when compiling without XPU. Force WITH_XPU_XRE5=OFF"))
+  set(WITH_XPU_XRE5
       OFF
-      CACHE STRING "Disable XHPC when compiling without XPU" FORCE)
+      CACHE STRING "Disable XRE5 when compiling without XPU" FORCE)
 endif()
 
 if(WITH_NCCL)

cmake/PaddleConfig.cmake.in

@@ -12,7 +12,7 @@
 get_filename_component(PADDLE_INSTALL_PREFIX "${CMAKE_CURRENT_LIST_FILE}/../.." ABSOLUTE)
 
 # include directories
-set(PADDLE_INCLUDE_DIRS 
+set(PADDLE_INCLUDE_DIRS
     ${PADDLE_INSTALL_PREFIX}/include
     ${PADDLE_INSTALL_PREFIX}/include/third_party
 )

cmake/cblas.cmake

@@ -15,9 +15,11 @@
 # Find the CBlas and lapack libraries
 #
 # It will search MKLML, OpenBlas, reference-cblas, extern-openblas in order.
+# On APPLE, accelerate framework (apple's blas implementation) will be
+# used, if applicable.
 #
 # If any cblas implementation found, the following variable will be set.
-#    CBLAS_PROVIDER  # one of MKLML, OPENBLAS, REFERENCE
+#    CBLAS_PROVIDER  # one of MKLML, ACCELERATE, OPENBLAS, REFERENCE
 #    CBLAS_INC_DIR   # the include directory for cblas.
 #    CBLAS_LIBS      # a list of libraries should be linked by paddle.
 #                    # Each library should be full path to object file.
@@ -45,6 +47,24 @@ if(WITH_MKLML)
                  "(include: ${CBLAS_INC_DIR}, library: ${CBLAS_LIBRARIES})")
 endif()
 
+## find accelerate on apple
+if(APPLE AND NOT DEFINED CBLAS_PROVIDER)
+  find_library(ACCELERATE_FRAMEWORK Accelerate)
+  if(ACCELERATE_FRAMEWORK)
+    message(STATUS "Accelerate framework found " "${ACCELERATE_FRAMEWORK}")
+
+    set(CBLAS_PROVIDER ACCELERATE)
+    # no need to setup include dir if it's accelerate
+    # set(CBLAS_INC_DIR "")
+    set(CBLAS_LIBRARIES ${ACCELERATE_FRAMEWORK})
+
+    add_definitions(-DPADDLE_USE_ACCELERATE)
+    add_definitions(-DLAPACK_FOUND)
+  else()
+    message(WARNING "Accelerate framework not found")
+  endif()
+endif()
+
 ## Then find openblas.
 if(NOT DEFINED CBLAS_PROVIDER)
   set(OPENBLAS_ROOT

cmake/ccache.cmake

@@ -17,19 +17,21 @@ if(NOT WIN32)
   endif()
 elseif("${CMAKE_GENERATOR}" STREQUAL "Ninja")
   # (Note:zhouwei25) Only Ninja Generator can support sccache now
-  find_program(SCCACHE_PATH sccache)
+  if(WITH_SCCACHE)
+    find_program(SCCACHE_PATH sccache)
 
-  if(SCCACHE_PATH)
-    execute_process(COMMAND sccache -V OUTPUT_VARIABLE sccache_version)
-    message(
-      STATUS
-        "sccache is founded, use [${SCCACHE_PATH}] to speed up compile on Windows."
-    )
+    if(SCCACHE_PATH)
+      execute_process(COMMAND sccache -V OUTPUT_VARIABLE sccache_version)
+      message(
+        STATUS
+          "sccache is founded, use [${SCCACHE_PATH}] to speed up compile on Windows."
+      )
 
-    set(CMAKE_C_COMPILER_LAUNCHER ${SCCACHE_PATH})
-    set(CMAKE_CXX_COMPILER_LAUNCHER ${SCCACHE_PATH})
-    # (Note:zhouwei25) sccache for cuda compiler has bug so that it can't be hit
-    # refer to https://github.com/mozilla/sccache/issues/1017, so we fix it
-    set(CMAKE_CUDA_COMPILER_LAUNCHER ${SCCACHE_PATH})
+      set(CMAKE_C_COMPILER_LAUNCHER ${SCCACHE_PATH})
+      set(CMAKE_CXX_COMPILER_LAUNCHER ${SCCACHE_PATH})
+      # (Note:zhouwei25) sccache for cuda compiler has bug so that it can't be hit
+      # refer to https://github.com/mozilla/sccache/issues/1017, so we fix it
+      set(CMAKE_CUDA_COMPILER_LAUNCHER ${SCCACHE_PATH})
+    endif()
   endif()
 endif()

cmake/cinn.cmake

@@ -104,6 +104,11 @@ if(WITH_GPU)
                                              /usr/lib /usr/lib64 REQUIRED)
 endif()
 
+if(WITH_ROCM)
+  message(STATUS "CINN Compile with ROCM support")
+  add_definitions(-DCINN_WITH_HIP)
+endif()
+
 set(cinnapi_src CACHE INTERNAL "" FORCE)
 set(core_src CACHE INTERNAL "" FORCE)
 set(core_includes CACHE INTERNAL "" FORCE)
@@ -156,14 +161,13 @@ cinn_cc_library(
   DEPS
   glog
   ${llvm_libs}
-  cinn_framework_proto
   param_proto
   auto_schedule_proto
   schedule_desc_proto
+  tile_config_proto
   absl
   isl
   ginac
-  pybind
   op_fusion
   cinn_op_dialect
   ${jitify_deps})
@@ -172,8 +176,7 @@ add_dependencies(cinnapi GEN_LLVM_RUNTIME_IR_HEADER ${core_deps})
 target_link_libraries(cinnapi op_dialect pir phi)
 add_dependencies(cinnapi op_dialect pir phi)
 
-target_link_libraries(cinnapi ${PYTHON_LIBRARIES})
-
+add_dependencies(cinnapi python)
 if(WITH_MKL)
   target_link_libraries(cinnapi cinn_mklml)
   add_dependencies(cinnapi cinn_mklml)
@@ -216,10 +219,10 @@ function(gen_cinncore LINKTYPE)
     DEPS
     glog
     ${llvm_libs}
-    cinn_framework_proto
     param_proto
     auto_schedule_proto
     schedule_desc_proto
+    tile_config_proto
     absl
     isl
     ginac
@@ -303,19 +306,8 @@ if(PUBLISH_LIBS)
   add_custom_command(
     TARGET cinncore_static
     POST_BUILD
-    COMMAND
-      cmake -E copy ${PROJECT_SOURCE_DIR}/tools/cinn/tutorials_demo/demo.cc
-      ${CMAKE_BINARY_DIR}/dist/demo.cc
-    COMMAND
-      cmake -E copy
-      ${PROJECT_SOURCE_DIR}/tools/cinn/tutorials_demo/build_demo.sh
-      ${CMAKE_BINARY_DIR}/dist/build_demo.sh
     COMMAND cmake -E copy ${CMAKE_BINARY_DIR}/libcinncore_static.a
             ${CMAKE_BINARY_DIR}/dist/cinn/lib/libcinncore_static.a
-    COMMAND
-      cmake -E copy
-      ${CMAKE_BINARY_DIR}/paddle/cinn/frontend/paddle/libcinn_framework_proto.a
-      ${CMAKE_BINARY_DIR}/dist/cinn/lib/libcinn_framework_proto.a
     COMMAND
       cmake -E copy ${CMAKE_BINARY_DIR}/paddle/cinn/hlir/pe/libparam_proto.a
       ${CMAKE_BINARY_DIR}/dist/cinn/lib/libparam_proto.a

cmake/cinn/core.cmake

@@ -19,7 +19,10 @@ function(cinn_cc_library TARGET_NAME)
     endif()
 
     if(cinn_cc_library_DEPS)
-      # Don't need link libwarpctc.so
+      if("${cinn_cc_library_DEPS};" MATCHES "python;")
+        list(REMOVE_ITEM cinn_cc_library_DEPS python)
+        add_dependencies(${TARGET_NAME} python)
+      endif()
       target_link_libraries(${TARGET_NAME} ${cinn_cc_library_DEPS})
       add_dependencies(${TARGET_NAME} ${cinn_cc_library_DEPS})
     endif()