fix merged conflict

.pipelines/java_packaging.yml

@@ -160,10 +160,19 @@ stages:
         jar cmf0 META-INF/MANIFEST.MF onnxruntime-extensions-$(OrtExtVersion).jar  *
       displayName: Combine and pack JAR with Windows, Linux and MacOS Binaries
 
+    - script: |
+        cd $(Build.ArtifactStagingDirectory)
+        mkdir drop
+        cp onnxruntime-extensions-$(OrtExtVersion).jar drop/onnxruntime-extensions-$(OrtExtVersion).jar
+        cp onnxruntime-extensions-$(OrtExtVersion)-javadoc.jar drop/onnxruntime-extensions-$(OrtExtVersion)-javadoc.jar
+        cp onnxruntime-extensions-$(OrtExtVersion)-sources.jar drop/onnxruntime-extensions-$(OrtExtVersion)-sources.jar
+        cp META-INF/maven/com.microsoft.onnxruntime/onnxruntime-extensions/pom.xml drop/onnxruntime-extensions-$(OrtExtVersion).pom
+      displayName: Move files to a drop folder for publishing
+
     - task: PublishPipelineArtifact@1
       displayName: 'Publish MacOS Artifact'
       inputs:
-        targetPath: '$(Build.ArtifactStagingDirectory)'
+        targetPath: '$(Build.ArtifactStagingDirectory)/drop'
         artifact: 'drop-onnxruntime-extensions-java-cpu'
 
     - template: templates/component-governance-component-detection-steps.yml

CMakeLists.txt

@@ -285,9 +285,11 @@ function(set_msvc_c_cpp_compiler_warning_level warning_level)
   endif()
 endfunction()
 
+if (NOT ONNXRUNTIME_INCLUDE_DIR)
+  include(ext_ortlib)
+endif()
 # set default MSVC warning level to 3 for external dependencies
 set_msvc_c_cpp_compiler_warning_level(3)
-include(ext_ortlib)
 include(gsl)
 
 macro(standardize_output_folder bin_target)
@@ -681,7 +683,6 @@ endif()
 
 if(OCOS_ENABLE_GPT2_TOKENIZER OR OCOS_ENABLE_WORDPIECE_TOKENIZER)
   target_include_directories(ocos_operators PUBLIC ${nlohmann_json_SOURCE_DIR}/single_include)
-  list(APPEND ocos_libraries nlohmann_json::nlohmann_json)
 endif()
 
 # If building a shared library we can't throw an internal exception type across the library boundary as the type
@@ -695,8 +696,6 @@ if(ANDROID)
   list(APPEND ocos_libraries log)
 endif()
 
-list(APPEND ocos_libraries Microsoft.GSL::GSL)
-
 list(REMOVE_DUPLICATES OCOS_COMPILE_DEFINITIONS)
 target_compile_definitions(noexcep_operators PRIVATE ${OCOS_COMPILE_DEFINITIONS})
 if(NOT OCOS_ENABLE_CPP_EXCEPTIONS)
@@ -899,7 +898,7 @@ if (_ORTX_STANDALONE_PROJECT)
   # Run CPack to generate the NuGet package
   include(CPack)
 
-  if(OCOS_ENABLE_CTEST)
+  if(OCOS_ENABLE_CTEST AND NOT MAC_CATALYST)
     include(ext_tests)
   endif()
 endif()

base/file_sys.h

@@ -17,7 +17,20 @@ namespace ort_extensions {
 class path {
  public:
   path() = default;
-  path(const std::string& path) : path_(path){};
+  path(const std::string& path) : path_(path) {
+#ifdef _WIN32
+    w_path_ = to_wstring();
+#endif  // _WIN32
+  };
+
+#ifdef _WIN32
+  path(const std::wstring& wpath) {
+    int size_needed = WideCharToMultiByte(CP_UTF8, 0, wpath.c_str(), -1, nullptr, 0, nullptr, nullptr);
+    std::string utf8_str(size_needed, 0);
+    WideCharToMultiByte(CP_UTF8, 0, wpath.c_str(), -1, &utf8_str[0], size_needed, nullptr, nullptr);
+    path_ = utf8_str;
+  }
+#endif  // _WIN32
 
   static constexpr char separator =
 #ifdef _WIN32
@@ -30,7 +43,7 @@ class path {
   std::ifstream open(ios_base::openmode mode = ios_base::in) const {
     // if Windows, need to convert the string to UTF-16
 #ifdef _WIN32
-    return std::ifstream(to_wstring(), mode);
+    return std::ifstream(w_path_, mode);
 #else
     return std::ifstream(path_, mode);
 #endif  // _WIN32
@@ -55,7 +68,7 @@ class path {
   bool is_directory() const {
 #ifdef _WIN32
     struct _stat64 info;
-    if (_wstat64(to_wstring().c_str(), &info) != 0) {
+    if (_wstat64(w_path_.c_str(), &info) != 0) {
       return false;
     }
 #else
@@ -69,8 +82,9 @@ class path {
 
  private:
   std::string path_;
-
 #ifdef _WIN32
+  std::wstring w_path_;
+
   std::wstring to_wstring() const {
     int size_needed = MultiByteToWideChar(CP_UTF8, 0, path_.c_str(), -1, nullptr, 0);
     std::wstring utf16_str(size_needed, 0);

cmake/ext_tests.cmake

@@ -1,5 +1,3 @@
-if (NOT MAC_CATALYST)
-
 if (OCOS_ENABLE_SELECTED_OPLIST)
   # currently the tests don't handle operator exclusion cleanly.
   message(FATAL_ERROR "Due to usage of OCOS_ENABLE_SELECTED_OPLIST excluding operators the tests are unable to be built and run")
@@ -55,7 +53,7 @@ function(add_test_target)
                    "${TEST_SRC_DIR}/unittest_main/test_main.cc")
     target_link_libraries(${ARG_TARGET} PRIVATE
                           ${ARG_LIBRARIES}
-                          gtest gmock)
+                          gtest)
 
     if(OCOS_USE_CUDA)
       target_link_directories(${ARG_TARGET} PRIVATE ${CUDAToolkit_LIBRARY_DIR})
@@ -93,7 +91,7 @@ function(add_test_target)
 
     target_link_libraries(${ARG_TARGET} PRIVATE
                           ${ARG_LIBRARIES}
-                          gtest gmock)
+                          gtest)
 
     set(test_data_destination_root_directory $<TARGET_FILE_DIR:${dummy_testee_target}>)
 
@@ -130,9 +128,40 @@ add_test_target(TARGET ocos_test
                 LIBRARIES ortcustomops ${ocos_libraries})
 target_compile_definitions(ocos_test PRIVATE ${OCOS_COMPILE_DEFINITIONS})
 
+if (OCOS_ENABLE_C_API)
+  file(GLOB pp_api_TEST_SRC
+    "${TEST_SRC_DIR}/pp_api_test/*.c"
+    "${TEST_SRC_DIR}/pp_api_test/*.cc"
+    "${TEST_SRC_DIR}/pp_api_test/*.h")
+
+  add_test_target(TARGET pp_api_test
+    TEST_SOURCES ${pp_api_TEST_SRC}
+    LIBRARIES onnxruntime_extensions ${ocos_libraries}
+    TEST_DATA_DIRECTORIES ${TEST_SRC_DIR}/data)
+
+  target_compile_definitions(pp_api_test PRIVATE ${OCOS_COMPILE_DEFINITIONS})
+  target_include_directories(pp_api_test PRIVATE
+    ${PROJECT_SOURCE_DIR}/
+    "$<TARGET_PROPERTY:ortcustomops,INTERFACE_INCLUDE_DIRECTORIES>"
+    "$<TARGET_PROPERTY:ocos_operators,INTERFACE_INCLUDE_DIRECTORIES>")
+
+  if (ORTX_TEST_DATA2)
+    file(TO_NATIVE_PATH "${ORTX_TEST_DATA2}/tests/data2" _TEST_DATA2)
+    add_custom_command(TARGET pp_api_test POST_BUILD
+      COMMAND ${CMAKE_COMMAND} -E create_symlink ${_TEST_DATA2} ${onnxruntime_extensions_BINARY_DIR}/data2)
+  endif()
+endif()
+
+
 # -- shared test (needs onnxruntime) --
-SET(CMAKE_FIND_ROOT_PATH_MODE_LIBRARY BOTH)
-find_library(ONNXRUNTIME onnxruntime HINTS "${ONNXRUNTIME_LIB_DIR}")
+# avoid blindling searching for onnxruntime library
+# wbhich leads to a unpredictable result
+if (NOT ONNXRUNTIME_LIB_DIR)  
+  set(ONNXRUNTIME "ONNXRUNTIME-NOTFOUND")
+else()
+  SET(CMAKE_FIND_ROOT_PATH_MODE_LIBRARY BOTH)
+  find_library(ONNXRUNTIME onnxruntime HINTS "${ONNXRUNTIME_LIB_DIR}")
+endif()
 
 if("${ONNXRUNTIME}" STREQUAL "ONNXRUNTIME-NOTFOUND")
   message(WARNING "The prebuilt onnxruntime library was not found, extensions_test will be skipped.")
@@ -197,25 +226,10 @@ else()
         COMMAND ${CMAKE_COMMAND} -E copy_if_different ${ONNXRUNTIME} ${CMAKE_BINARY_DIR}/lib
       )
     endif()
-  endblock()
-
-  if (OCOS_ENABLE_C_API)
-    file(GLOB pp_api_TEST_SRC
-      "${TEST_SRC_DIR}/pp_api_test/*.c"
-      "${TEST_SRC_DIR}/pp_api_test/*.cc"
-      "${TEST_SRC_DIR}/pp_api_test/*.h")
-
-    add_test_target(TARGET pp_api_test
-      TEST_SOURCES ${pp_api_TEST_SRC}
-      LIBRARIES onnxruntime_extensions ${ocos_libraries}
-      TEST_DATA_DIRECTORIES ${TEST_SRC_DIR}/data)
-
-    target_compile_definitions(pp_api_test PRIVATE ${OCOS_COMPILE_DEFINITIONS})
-    target_include_directories(pp_api_test PRIVATE
-      ${PROJECT_SOURCE_DIR}/
-      "$<TARGET_PROPERTY:ortcustomops,INTERFACE_INCLUDE_DIRECTORIES>"
-      "$<TARGET_PROPERTY:ocos_operators,INTERFACE_INCLUDE_DIRECTORIES>")
-  endif()
-endif()
 
+    if (OCOS_ENABLE_C_API)
+      # avoid copying the same data directory at the same time.
+      add_dependencies(extensions_test pp_api_test)
+    endif()
+  endblock()
 endif()

cmake/externals/googletest.cmake

@@ -4,8 +4,8 @@ FetchContent_Declare(
   URL_HASH    SHA1=06096d3900c356e468ba060a609642c635131106
 )
 
+set(BUILD_GMOCK OFF CACHE BOOL "" FORCE)
+set(INSTALL_GTEST OFF CACHE BOOL "" FORCE)
 FetchContent_MakeAvailable(googletest)
-set_target_properties(gmock PROPERTIES FOLDER "externals/gtest")
-set_target_properties(gmock_main PROPERTIES FOLDER "externals/gtest")
 set_target_properties(gtest PROPERTIES FOLDER "externals/gtest")
 set_target_properties(gtest_main PROPERTIES FOLDER "externals/gtest")

cmake/externals/gsl.cmake

@@ -14,5 +14,13 @@ else()
     )
 endif()
 
-FetchContent_MakeAvailable(GSL)
-get_target_property(GSL_INCLUDE_DIR Microsoft.GSL::GSL INTERFACE_INCLUDE_DIRECTORIES)
+FetchContent_GetProperties(GSL)
+string(TOLOWER "GSL" lcName)
+if(NOT ${lcName}_POPULATED)
+  FetchContent_Populate(GSL)
+#  add_subdirectory(${GSL_SOURCE_DIR} ${GSL_BINARY_DIR} EXCLUDE_FROM_ALL)
+endif()
+
+set(GSL_INCLUDE_DIR ${gsl_SOURCE_DIR}/include)
+
+#get_target_property(GSL_INCLUDE_DIR Microsoft.GSL::GSL INTERFACE_INCLUDE_DIRECTORIES)

cmake/externals/json.cmake

@@ -7,5 +7,4 @@ set(JSON_BuildTests OFF CACHE INTERNAL "")
 FetchContent_GetProperties(nlohmann_json)
 if(NOT nlohmann_json_POPULATED)
   FetchContent_Populate(nlohmann_json)
-  add_subdirectory(${nlohmann_json_SOURCE_DIR} ${nlohmann_json_BINARY_DIR} EXCLUDE_FROM_ALL)
 endif()

cmake/externals/opencv.cmake

@@ -107,6 +107,16 @@ set(BUILD_TESTS       OFF CACHE INTERNAL "")
 set(CV_TRACE          OFF CACHE INTERNAL "")
 
 set(CV_DISABLE_OPTIMIZATION     ON CACHE INTERNAL "")
+set(BUILD_PERF_TESTS            OFF CACHE INTERNAL "")
+set(BUILD_opencv_java_bindings_generator	  OFF CACHE INTERNAL "")
+set(BUILD_opencv_js_bindings_generator	  OFF CACHE INTERNAL "")
+set(BUILD_opencv_objc_bindings_generator	  OFF CACHE INTERNAL "")
+set(BUILD_opencv_python_bindings_generator  OFF CACHE INTERNAL "")
+set(BUILD_opencv_python_tests  OFF CACHE INTERNAL "")
+
+set(WITH_ADE                   OFF CACHE INTERNAL "")
+set(VIDEOIO_ENABLE_PLUGINS	OFF CACHE INTERNAL "")
+set(HIGHGUI_ENABLE_PLUGINS  OFF CACHE INTERNAL "")
 
 if(IOS)
   # copy what OpenCV's platforms/ios/build_framework.py does and set CPU_BASELINE=DETECT
@@ -157,13 +167,3 @@ endif()
 
 # unset it to avoid affecting other projects.
 unset(EXECUTABLE_OUTPUT_PATH CACHE)
-
-if (CMAKE_SYSTEM_NAME MATCHES "Windows")
-    set(opencv_projs gen_opencv_java_source gen_opencv_js_source gen_opencv_python_source)
-    list(APPEND opencv_projs gen_opencv_objc_source gen_opencv_objc_source_ios gen_opencv_objc_source_osx)
-    list(APPEND opencv_projs opencv_highgui_plugins opencv_videoio_plugins)
-    foreach(p ${opencv_projs})
-        set_target_properties(${p} PROPERTIES FOLDER "externals/opencv")
-        set_target_properties(${p} PROPERTIES EXCLUDE_FROM_ALL TRUE EXCLUDE_FROM_DEFAULT_BUILD TRUE)
-    endforeach()
-endif()

operators/cuda/add_mul.h

@@ -0,0 +1,46 @@
+// Copyright (c) Microsoft Corporation. All rights reserved.
+// Licensed under the MIT License.
+
+#pragma once
+#include "ocos.h"
+#include "add_mul_impl.cuh"
+#include "ortx_common.h"
+
+namespace contrib {
+
+template <typename T, bool addition>
+struct AddOrMulSharedInput {
+  template <typename TDict>
+  OrtxStatus OnModelAttach(const TDict& /*dict*/) {
+    return {};
+  }
+  OrtxStatus Compute(Ort::Custom::CUDAKernelContext* ctx,
+                     const ortc::Tensor<T>& tensor_a,
+                     const ortc::Tensor<T>& tensor_b,
+                     const ortc::Tensor<T>& tensor_c,
+                     ortc::Tensor<T>& output_ab,
+                     ortc::Tensor<T>& output_ac) const {
+    const T* input_data_a = tensor_a.Data();
+    const T* input_data_b = tensor_b.Data();
+    const T* input_data_c = tensor_c.Data();
+
+    auto length_a = tensor_a.NumberOfElement();
+    auto length_b = tensor_b.NumberOfElement();
+    auto length_c = tensor_c.NumberOfElement();
+
+    T* output_data_ab = output_ab.Allocate(length_a <= length_b ? tensor_b.Shape() : tensor_a.Shape());
+    T* output_data_ac = output_ab.Allocate(length_a <= length_c ? tensor_c.Shape() : tensor_a.Shape());
+
+    if (0 == input_data_a || 0 == input_data_b || 0 == input_data_c) {
+      return {};
+    }
+    LaunchAddOrMulSharedInputKernel<T>(reinterpret_cast<cudaStream_t>(ctx->GetCudaStream()),
+                                       input_data_a, input_data_b, input_data_c,
+                                       output_data_ab, output_data_ac,
+                                       length_a, length_b, length_c,
+                                       addition);
+    return {};
+  }
+};
+
+}  // namespace contrib