refine cmake (#4212)

.gitignore

@@ -47,7 +47,6 @@ Makefile
 
 cmake_install.cmake
 
-
 *.cbp
 
 paddle-mobile.cbp

CMakeLists.txt

@@ -70,46 +70,46 @@ find_package(Threads REQUIRED)
 include(simd)
 
 ################################ Exposed Configurations #######################################
-lite_option(WITH_DSO         "Compile PaddlePaddle with dynamic linked CUDA" ON)
-lite_option(WITH_AVX         "Compile PaddlePaddle with AVX intrinsics"      ON IF ${AVX_FOUND})
-lite_option(WITH_TESTING     "Compile PaddlePaddle with unit testing"        OFF)
-lite_option(WITH_MKL         "Compile PaddlePaddle with MKL support."        ON IF ${AVX_FOUND})
-lite_option(WITH_ARM_DOTPROD "Compile PaddlePaddle with ARM dot production"  ON)
-lite_option(WITH_SYSTEM_BLAS   "Use system blas library"           OFF)
+lite_option(WITH_DSO            "Compile PaddlePaddle with dynamic linked CUDA" ON)
+lite_option(WITH_AVX            "Compile PaddlePaddle with AVX intrinsics"      ON IF ${AVX_FOUND})
+lite_option(WITH_TESTING        "Compile PaddlePaddle with unit testing"        OFF)
+lite_option(WITH_MKL            "Compile PaddlePaddle with MKL support."        ON IF ${AVX_FOUND})
+lite_option(WITH_ARM_DOTPROD    "Compile PaddlePaddle with ARM dot production"  ON)
+lite_option(WITH_SYSTEM_BLAS    "Use system blas library"           OFF)
 
 # for lite, both server and mobile framework.
-lite_option(LITE_WITH_JAVA "Enable Java JNI lib in lite mode" OFF)
-lite_option(LITE_WITH_PYTHON "Enable Python api lib in lite mode" OFF)
-lite_option(LITE_WITH_CUDA "Enable CUDA in lite mode" OFF)
-lite_option(LITE_WITH_X86  "Enable X86 in lite mode"  ON)
-lite_option(LITE_WITH_ARM  "Enable ARM in lite mode"  OFF)
-lite_option(LITE_WITH_NPU  "Enable NPU in lite mode"  OFF)
-lite_option(LITE_WITH_RKNPU  "Enable RKNPU in lite mode"  OFF)
-lite_option(LITE_WITH_MLU  "Enable MLU in lite mode"  OFF)
+lite_option(LITE_WITH_JAVA      "Enable Java JNI lib in lite mode" OFF)
+lite_option(LITE_WITH_PYTHON    "Enable Python api lib in lite mode" OFF)
+lite_option(LITE_WITH_CUDA      "Enable CUDA in lite mode" OFF)
+lite_option(LITE_WITH_X86       "Enable X86 in lite mode"  ON)
+lite_option(LITE_WITH_ARM       "Enable ARM in lite mode"  OFF)
+lite_option(LITE_WITH_NPU       "Enable NPU in lite mode"  OFF)
+lite_option(LITE_WITH_RKNPU     "Enable RKNPU in lite mode"  OFF)
+lite_option(LITE_WITH_MLU       "Enable MLU in lite mode"  OFF)
 lite_option(LITE_WITH_HUAWEI_ASCEND_NPU  "Enable HUAWEI_ASCEND_NPU in lite mode"  OFF)
-lite_option(LITE_WITH_XPU  "Enable XPU in lite mode"  OFF)
-lite_option(LITE_WITH_XTCL  "Enable XPU via XTCL"  OFF IF LITE_WITH_XPU)
-lite_option(LITE_WITH_BM   "Enable BM in lite mode"   OFF)
-lite_option(LITE_WITH_APU  "Enable APU in lite mode"  OFF)
-lite_option(LITE_WITH_TRAIN "Enable training operators and kernels in lite" OFF)
-lite_option(LITE_WITH_OPENMP "Enable OpenMP in lite framework" ON)
-lite_option(LITE_WITH_OPENCL   "Enable OpenCL support in lite" OFF)
-lite_option(LITE_WITH_FPGA   "Enable FPGA support in lite" OFF)
+lite_option(LITE_WITH_XPU       "Enable XPU in lite mode"  OFF)
+lite_option(LITE_WITH_XTCL      "Enable XPU via XTCL"  OFF IF LITE_WITH_XPU)
+lite_option(LITE_WITH_BM        "Enable BM in lite mode"   OFF)
+lite_option(LITE_WITH_APU       "Enable APU in lite mode"  OFF)
+lite_option(LITE_WITH_TRAIN     "Enable training operators and kernels in lite" OFF)
+lite_option(LITE_WITH_OPENMP    "Enable OpenMP in lite framework" ON)
+lite_option(LITE_WITH_OPENCL    "Enable OpenCL support in lite" OFF)
+lite_option(LITE_WITH_FPGA      "Enable FPGA support in lite" OFF)
 lite_option(LITE_WITH_LIGHT_WEIGHT_FRAMEWORK  "Enable light-weight framework" OFF)
-lite_option(LITE_WITH_PROFILE  "Enable profile mode in lite framework"  OFF)
+lite_option(LITE_WITH_PROFILE   "Enable profile mode in lite framework"  OFF)
 lite_option(LITE_WITH_PRECISION_PROFILE "Enable precision profile in profile mode ON in lite" OFF)
-lite_option(LITE_WITH_LOG "Enable log printing or not." ON)
+lite_option(LITE_WITH_LOG       "Enable log printing or not." ON)
 lite_option(LITE_WITH_EXCEPTION "Enable throwing the exception when error occurs in lite" OFF)
-lite_option(LITE_WITH_NVTX "Enable nvtx or not, please enable LITE_WITH_CUDA first." OFF)
+lite_option(LITE_WITH_NVTX      "Enable nvtx or not, please enable LITE_WITH_CUDA first." OFF)
 lite_option(LITE_ON_TINY_PUBLISH "Publish tiny predictor lib." OFF)
 lite_option(LITE_ON_MODEL_OPTIMIZE_TOOL "Build the model optimize tool" OFF)
 # publish options
-lite_option(LITE_BUILD_EXTRA "Enable extra algorithm support in Lite, both kernels and operators" OFF)
-lite_option(LITE_BUILD_TAILOR "Enable tailoring library according to model" OFF)
+lite_option(LITE_BUILD_EXTRA    "Enable extra algorithm support in Lite, both kernels and operators" OFF)
+lite_option(LITE_BUILD_TAILOR   "Enable tailoring library according to model" OFF)
 # cv build options
-lite_option(LITE_WITH_CV  "Enable build cv image in lite" OFF)
-lite_option(LITE_WITH_STATIC_CUDA  "Statically link cuda libraries." OFF)
-lite_option(CUDA_WITH_FP16 "Compile with cuda half support" OFF)
+lite_option(LITE_WITH_CV        "Enable build cv image in lite" OFF)
+lite_option(LITE_WITH_STATIC_CUDA "Statically link cuda libraries." OFF)
+lite_option(CUDA_WITH_FP16      "Compile with cuda half support" OFF)
 lite_option(LITE_WITH_ARM_CLANG "when arm lang is clang, its ON." OFF)
 
 # TODO(Superjomn) Remove WITH_ANAKIN option if not needed latter.

cmake/configure.cmake

@@ -84,7 +84,8 @@ if (WITH_MKLML AND MKLML_IOMP_LIB)
         set(OPENMP_FLAGS "")
     else(WIN32)
         set(OPENMP_FLAGS "-fopenmp")
-    endif(WIN32)
+    endif(WIN32 OR APPLE)
+
     set(CMAKE_C_CREATE_SHARED_LIBRARY_FORBIDDEN_FLAGS ${OPENMP_FLAGS})
     set(CMAKE_CXX_CREATE_SHARED_LIBRARY_FORBIDDEN_FLAGS ${OPENMP_FLAGS})
     set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${OPENMP_FLAGS}")

lite/CMakeLists.txt

 include(lite)
 
-message(WARNING "Lite enabled!")
 message(STATUS "LIGHT_FRAMEWORK:\t${LITE_WITH_LIGHT_WEIGHT_FRAMEWORK}")
 message(STATUS "LITE_WITH_CUDA:\t${LITE_WITH_CUDA}")
 message(STATUS "LITE_WITH_X86:\t${LITE_WITH_X86}")

lite/api/CMakeLists.txt

@@ -119,20 +119,35 @@ if(LITE_WITH_HUAWEI_ASCEND_NPU)
     set(cxx_api_deps ${cxx_api_deps} ${huawei_ascend_npu_deps})
 endif()
 
-message(STATUS "get ops ${ops}")
-message(STATUS "get X86 kernels ${x86_kernels}")
-message(STATUS "get CUDA kernels ${cuda_kernels}")
-message(STATUS "get Host kernels ${host_kernels}")
-message(STATUS "get ARM kernels ${arm_kernels}")
-message(STATUS "get OpenCL kernels ${opencl_kernels}")
-message(STATUS "get NPU kernels ${npu_kernels}")
-message(STATUS "get APU kernels ${apu_kernels}")
-message(STATUS "get XPU kernels ${xpu_kernels}")
-message(STATUS "get RKNPU kernels ${rknpu_kernels}")
-message(STATUS "get FPGA kernels ${fpga_kernels}")
-message(STATUS "get BM kernels ${bm_kernels}")
-message(STATUS "get MLU kernels ${mlu_kernels}")
-message(STATUS "get HUAWEI_ASCEND_NPU kernels ${huawei_ascend_npu_kernels}")
+list(LENGTH ops num_ops)
+list(LENGTH x86_kernels num_x86_kernels)
+list(LENGTH cuda_kernels num_cuda_kernels)
+list(LENGTH host_kernels num_host_kernels)
+list(LENGTH arm_kernels num_arm_kernels)
+list(LENGTH opencl_kernels num_opencl_kernels)
+list(LENGTH npu_kernels num_npu_kernels)
+list(LENGTH apu_kernels num_apu_kernels)
+list(LENGTH xpu_kernels num_xpu_kernels)
+list(LENGTH rknpu_kernels num_rknpu_kernels)
+list(LENGTH fpga_kernels num_fpga_kernels)
+list(LENGTH bm_kernels num_bm_kernels)
+list(LENGTH mlu_kernels num_mlu_kernels)
+list(LENGTH huawei_ascend_npu_kernels num_huawei_ascend_npu_kernels)
+
+message(STATUS "Collected ${num_ops} ops")
+message(STATUS "Collected ${num_x86_kernels} X86 kernels")
+message(STATUS "Collected ${num_cuda_kernels} CUDA kernels")
+message(STATUS "Collected ${num_host_kernels} Host kernels")
+message(STATUS "Collected ${num_arm_kernels} ARM kernels")
+message(STATUS "Collected ${num_opencl_kernels} OpenCL kernels")
+message(STATUS "Collected ${num_npu_kernels} NPU kernels")
+message(STATUS "Collected ${num_apu_kernels} APU kernels")
+message(STATUS "Collected ${num_xpu_kernels} XPU kernels")
+message(STATUS "Collected ${num_rknpu_kernels} RKNPU kernels")
+message(STATUS "Collected ${num_fpga_kernels} FPGA kernels")
+message(STATUS "Collected ${num_bm_kernels} BM kernels")
+message(STATUS "Collected ${num_mlu_kernels} MLU kernels")
+message(STATUS "Collected ${num_huawei_ascend_npu_kernels} HUAWEI_ASCEND_NPU kernels")
 
 # for full api
 if (NOT LITE_ON_TINY_PUBLISH)

lite/core/mir/CMakeLists.txt

@@ -107,26 +107,3 @@ if (NOT LITE_WITH_LIGHT_WEIGHT_FRAMEWORK)
     #     pattern_matcher_high_api proto_desc mir_pass_manager fc_op mul_op elementwise_ops
     #     mir_passes compatible_pb program ${ops})
 endif()
-
-message(STATUS "----> Ops lite: ${ops}")
-message(STATUS "----> Host kernels: ${host_kernels}")
-message(STATUS "----> X86 kernels: ${x86_kernels}")
-
-# lite_cc_test(test_lite_fc_fuse SRCS fusion/fc_fuse_pass_test.cc
-#    DEPS cxx_api mir_passes
-#    ${ops} ${host_kernels} ${x86_kernels} ${arm_kernels}
-#    ARGS --model_dir=${LITE_MODEL_DIR}/lite_fc_model
-#         --optimized_model=${LITE_MODEL_DIR}/lite_fc_model_opt SERIAL)
-
-# lite_download_and_uncompress(${LITE_MODEL_DIR} ${LITE_URL} "lite_fc_model.tar.gz")
-# add_dependencies(test_lite_fc_fuse extern_lite_download_lite_fc_model_tar_gz)
-
-
-# lite_cc_test(test_lite_conv_elementwise_add_activation_fuse
-#              SRCS fusion/conv_elementwise_add_activation_fuse_pass_test.cc
-#              DEPS cxx_api mir_passes
-#              ${ops} ${host_kernels} ${x86_kernels})
-# lite_cc_test(test_lite_elementwise_add_activation_fuse
-#              SRCS fusion/elementwise_add_activation_fuse_pass_test.cc
-#              DEPS cxx_api mir_passes
-#              ${ops} ${host_kernels} ${x86_kernels})