ORT 1.19.0 Release: Cherry-Pick Round 1 (#21619)

### Description
<!-- Describe your changes. -->

PRs marked for cherry-pick.

### Motivation and Context
<!-- - Why is this change required? What problem does it solve?
- If it fixes an open issue, please link to the issue here. -->

ORT 1.19.0 Release Preparation

---------

Signed-off-by: Liqun Fu <[email protected]>
Signed-off-by: liqunfu <[email protected]>
Signed-off-by: Liqun Fu <[email protected]>
Co-authored-by: liqun Fu <[email protected]>
Co-authored-by: Jing Fang <[email protected]>
Co-authored-by: Tianlei Wu <[email protected]>
Co-authored-by: Adrian Lizarraga <[email protected]>
Co-authored-by: Changming Sun <[email protected]>
Co-authored-by: Sumit Agarwal <[email protected]>
Co-authored-by: vraspar <[email protected]>
Co-authored-by: Scott McKay <[email protected]>
Co-authored-by: Edward Chen <[email protected]>
Co-authored-by: Yi Zhang <[email protected]>
Co-authored-by: jingyanwangms <[email protected]>
Co-authored-by: Yi Zhang <[email protected]>
Co-authored-by: Chi Lo <[email protected]>
Co-authored-by: saurabh <[email protected]>
Co-authored-by: sfatimar <[email protected]>

Author: 16 people

.pipelines/nuget_config/x64/packages.config

@@ -1,6 +1,6 @@
 <?xml version="1.0" encoding="utf-8"?>
 <packages>
   <package id="python" version="3.9.7" targetFramework="native" />
-  <package id="Microsoft.AI.DirectML" version="1.15.0" targetFramework="native" />
+  <package id="Microsoft.AI.DirectML" version="1.15.1" targetFramework="native" />
   <package id="Microsoft.Windows.CppWinRT" version="2.0.201201.7" targetFramework="native" />
 </packages>

.pipelines/nuget_config/x86/packages.config

@@ -1,6 +1,6 @@
 <?xml version="1.0" encoding="utf-8"?>
 <packages>
   <package id="pythonx86" version="3.9.7" targetFramework="native" />
-  <package id="Microsoft.AI.DirectML" version="1.15.0" targetFramework="native" />
+  <package id="Microsoft.AI.DirectML" version="1.15.1" targetFramework="native" />
   <package id="Microsoft.Windows.CppWinRT" version="2.0.201201.7" targetFramework="native" />
 </packages>

cmake/external/dml.cmake

@@ -41,7 +41,7 @@ if (NOT onnxruntime_USE_CUSTOM_DIRECTML)
   set(NUGET_CONFIG ${PROJECT_SOURCE_DIR}/../NuGet.config)
   set(PACKAGES_CONFIG ${PROJECT_SOURCE_DIR}/../packages.config)
   get_filename_component(PACKAGES_DIR ${CMAKE_CURRENT_BINARY_DIR}/../packages ABSOLUTE)
-  set(DML_PACKAGE_DIR ${PACKAGES_DIR}/Microsoft.AI.DirectML.1.15.0)
+  set(DML_PACKAGE_DIR ${PACKAGES_DIR}/Microsoft.AI.DirectML.1.15.1)
 
   # Restore nuget packages, which will pull down the DirectML redist package.
   add_custom_command(

cmake/onnxruntime.cmake

@@ -38,10 +38,14 @@ function(get_c_cxx_api_headers HEADERS_VAR)
 
   # need to add header files for enabled EPs
   foreach(f ${ONNXRUNTIME_PROVIDER_NAMES})
-    file(GLOB _provider_headers CONFIGURE_DEPENDS
-      "${REPO_ROOT}/include/onnxruntime/core/providers/${f}/*.h"
-    )
-    list(APPEND _headers ${_provider_headers})
+    # The header files in include/onnxruntime/core/providers/cuda directory cannot be flattened to the same directory 
+    # with onnxruntime_c_api.h . Most other EPs probably also do not work in this way.
+    if((NOT f STREQUAL cuda) AND (NOT f STREQUAL rocm))
+      file(GLOB _provider_headers CONFIGURE_DEPENDS
+        "${REPO_ROOT}/include/onnxruntime/core/providers/${f}/*.h"
+      )
+      list(APPEND _headers ${_provider_headers})
+    endif()
   endforeach()
 
   set(${HEADERS_VAR} ${_headers} PARENT_SCOPE)

cmake/onnxruntime_mlas.cmake

@@ -555,8 +555,17 @@ else()
           ${MLAS_SRC_DIR}/intrinsics/avx2/qdwconv_avx2.cpp
           ${MLAS_SRC_DIR}/sqnbitgemm_kernel_avx2.cpp
         )
-        set_source_files_properties(${mlas_platform_srcs_avx2} PROPERTIES COMPILE_FLAGS "-mavx2 -mfma")
 
+message(STATUS "CMAKE_CXX_COMPILER_ID: ${CMAKE_CXX_COMPILER_ID}")
+message(STATUS "CMAKE_CXX_COMPILER_VERSION: ${CMAKE_CXX_COMPILER_VERSION}")
+
+if(NOT "${CMAKE_CXX_COMPILER_ID}" STREQUAL "GNU" OR CMAKE_CXX_COMPILER_VERSION VERSION_GREATER "10")
+          message(STATUS "Using -mavx2 -mfma -mavxvnni flags")
+          set_source_files_properties(${mlas_platform_srcs_avx2} PROPERTIES COMPILE_FLAGS "-mavx2 -mfma -mavxvnni")
+else()
+          message(STATUS "Using -mavx2 -mfma flags")
+          set_source_files_properties(${mlas_platform_srcs_avx2} PROPERTIES COMPILE_FLAGS "-mavx2 -mfma")
+endif()
         set(mlas_platform_srcs_avx512f
           ${MLAS_SRC_DIR}/x86_64/DgemmKernelAvx512F.S
           ${MLAS_SRC_DIR}/x86_64/SgemmKernelAvx512F.S
@@ -575,7 +584,7 @@ else()
           ${MLAS_SRC_DIR}/x86_64/ConvSymKernelAvx512Core.S
           ${MLAS_SRC_DIR}/sqnbitgemm_kernel_avx512.cpp
         )
-        set_source_files_properties(${mlas_platform_srcs_avx512core} PROPERTIES COMPILE_FLAGS "-mavx512bw -mavx512dq -mavx512vl")
+        set_source_files_properties(${mlas_platform_srcs_avx512core} PROPERTIES COMPILE_FLAGS "-mfma -mavx512vnni -mavx512bw -mavx512dq -mavx512vl")
 
         set(mlas_platform_srcs_avx512vnni
           ${MLAS_SRC_DIR}/sqnbitgemm_kernel_avx512vnni.cpp

cmake/onnxruntime_providers_cpu.cmake

@@ -219,6 +219,7 @@ if (onnxruntime_ENABLE_TRAINING)
 endif()
 
 install(FILES ${PROJECT_SOURCE_DIR}/../include/onnxruntime/core/providers/cpu/cpu_provider_factory.h  DESTINATION ${CMAKE_INSTALL_INCLUDEDIR}/onnxruntime/)
+install(FILES ${PROJECT_SOURCE_DIR}/../include/onnxruntime/core/providers/resource.h ${PROJECT_SOURCE_DIR}/../include/onnxruntime/core/providers/custom_op_context.h DESTINATION ${CMAKE_INSTALL_INCLUDEDIR}/onnxruntime/core/providers)
 set_target_properties(onnxruntime_providers PROPERTIES LINKER_LANGUAGE CXX)
 set_target_properties(onnxruntime_providers PROPERTIES FOLDER "ONNXRuntime")
 

cmake/onnxruntime_providers_cuda.cmake

@@ -289,8 +289,15 @@
     config_cuda_provider_shared_module(onnxruntime_providers_cuda_obj)
   endif()
   config_cuda_provider_shared_module(onnxruntime_providers_cuda)
-
+  # Cannot use glob because the file cuda_provider_options.h should not be exposed out.
+  set(ONNXRUNTIME_CUDA_PROVIDER_PUBLIC_HEADERS
+        "${REPO_ROOT}/include/onnxruntime/core/providers/cuda/cuda_context.h"
+        "${REPO_ROOT}/include/onnxruntime/core/providers/cuda/cuda_resource.h"
+      )
+  set_target_properties(onnxruntime_providers_cuda PROPERTIES
+    PUBLIC_HEADER "${ONNXRUNTIME_CUDA_PROVIDER_PUBLIC_HEADERS}")
   install(TARGETS onnxruntime_providers_cuda
+          PUBLIC_HEADER DESTINATION ${CMAKE_INSTALL_INCLUDEDIR}/onnxruntime/core/providers/cuda
           ARCHIVE  DESTINATION ${CMAKE_INSTALL_LIBDIR}
           LIBRARY  DESTINATION ${CMAKE_INSTALL_LIBDIR}
           RUNTIME  DESTINATION ${CMAKE_INSTALL_BINDIR})

cmake/onnxruntime_providers_rocm.cmake

@@ -223,8 +223,13 @@
   if (onnxruntime_ENABLE_ATEN)
     target_compile_definitions(onnxruntime_providers_rocm PRIVATE ENABLE_ATEN)
   endif()
-
+  file(GLOB ONNXRUNTIME_ROCM_PROVIDER_PUBLIC_HEADERS CONFIGURE_DEPENDS
+        "${REPO_ROOT}/include/onnxruntime/core/providers/rocm/*.h"
+      )
+  set_target_properties(onnxruntime_providers_rocm PROPERTIES
+    PUBLIC_HEADER "${ONNXRUNTIME_ROCM_PROVIDER_PUBLIC_HEADERS}")
   install(TARGETS onnxruntime_providers_rocm
+          PUBLIC_HEADER DESTINATION ${CMAKE_INSTALL_INCLUDEDIR}/onnxruntime/core/providers/rocm
           ARCHIVE  DESTINATION ${CMAKE_INSTALL_LIBDIR}
           LIBRARY  DESTINATION ${CMAKE_INSTALL_LIBDIR}
           RUNTIME  DESTINATION ${CMAKE_INSTALL_BINDIR})

csharp/src/Microsoft.ML.OnnxRuntime/targets/net8.0-ios/targets.xml

@@ -1,7 +1,7 @@
 <?xml version="1.0" encoding="utf-8"?>
 <Project ToolsVersion="4.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
   <ItemGroup Condition="('$(OutputType)'!='Library' OR '$(IsAppExtension)'=='True')">
-    <NativeReference Include="$(MSBuildThisFileDirectory)..\..\runtimes\ios\native\onnxruntime.xcframework">
+    <NativeReference Include="$(MSBuildThisFileDirectory)..\..\runtimes\ios\native\onnxruntime.xcframework.zip">
       <Kind>Static</Kind>
       <IsCxx>True</IsCxx>
       <SmartLink>True</SmartLink>
@@ -10,4 +10,4 @@
       <WeakFrameworks>CoreML</WeakFrameworks>
     </NativeReference>
   </ItemGroup>
-</Project>
+</Project>

include/onnxruntime/core/optimizer/graph_transformer_utils.h

@@ -3,12 +3,15 @@
 
 #pragma once
 
+#include <string>
 #include <memory>
+#include <unordered_map>
 #include <unordered_set>
 #include <vector>
 
 #include "core/common/inlined_containers.h"
 #include "core/framework/session_options.h"
+#include "core/framework/tensor.h"
 #include "core/optimizer/graph_transformer.h"
 #include "core/platform/threadpool.h"
 
@@ -51,7 +54,8 @@ InlinedVector<std::unique_ptr<GraphTransformer>> GenerateTransformers(
     const SessionOptions& session_options,
     const IExecutionProvider& execution_provider /*required by constant folding*/,
     const InlinedHashSet<std::string>& rules_and_transformers_to_disable = {},
-    concurrency::ThreadPool* intra_op_thread_pool = nullptr);
+    concurrency::ThreadPool* intra_op_thread_pool = nullptr,
+    std::unordered_map<std::string, std::unique_ptr<Tensor>>* p_buffered_tensors = nullptr);
 
 #endif  // !defined(ORT_MINIMAL_BUILD)
 
@@ -81,7 +85,8 @@ InlinedVector<std::unique_ptr<GraphTransformer>> GenerateTransformersForMinimalB
     const SatApplyContextVariant& apply_context,
     const IExecutionProvider& cpu_execution_provider,
     const InlinedHashSet<std::string>& rules_and_transformers_to_disable = {},
-    concurrency::ThreadPool* intra_op_thread_pool = nullptr);
+    concurrency::ThreadPool* intra_op_thread_pool = nullptr,
+    std::unordered_map<std::string, std::unique_ptr<Tensor>>* p_buffered_tensors = nullptr);
 
 #endif  // !defined(ORT_MINIMAL_BUILD) || defined(ORT_EXTENDED_MINIMAL_BUILD)