Merge remote-tracking branch 'origin/main' into fp16-conv

.vscode/settings.json

@@ -40,3 +40,4 @@
         "-build/include_subdir",
         "-runtime/references"
     ]
+}

cgmanifests/generate_cgmanifest.py

@@ -90,55 +90,6 @@ def add_github_dep(name, parsed_url):
             git_deps[dep] = name
 
 
-with open(
-    os.path.join(REPO_DIR, "tools", "ci_build", "github", "linux", "docker", "Dockerfile.manylinux2_28_cuda11"),
-) as f:
-    for line in f:
-        if not line.strip():
-            package_name = None
-            package_filename = None
-            package_url = None
-        if package_filename is None:
-            m = re.match(r"RUN\s+export\s+(.+?)_ROOT=(\S+).*", line)
-            if m is not None:
-                package_name = m.group(1)
-                package_filename = m.group(2)
-            else:
-                m = re.match(r"RUN\s+export\s+(.+?)_VERSION=(\S+).*", line)
-                if m is not None:
-                    package_name = m.group(1)
-                    package_filename = m.group(2)
-        elif package_url is None:
-            m = re.match(r"(.+?)_DOWNLOAD_URL=(\S+)", line)
-            if m is not None:
-                package_url = m.group(2)
-                if package_name == "LIBXCRYPT":
-                    package_url = m.group(2) + "/v" + package_filename + ".tar.gz"
-                elif package_name == "CMAKE":
-                    package_url = m.group(2) + "/v" + package_filename + "/cmake-" + package_filename + ".tar.gz"
-                else:
-                    package_url = m.group(2) + "/" + package_filename + ".tar.gz"
-                parsed_url = urlparse(package_url)
-                if parsed_url.hostname == "github.com":
-                    add_github_dep("manylinux dependency " + package_name, parsed_url)
-                else:
-                    registration = {
-                        "Component": {
-                            "Type": "other",
-                            "other": {
-                                "Name": package_name.lower(),
-                                "Version": package_filename.split("-")[-1],
-                                "DownloadUrl": package_url,
-                            },
-                            "comments": "manylinux dependency",
-                        }
-                    }
-                    registrations.append(registration)
-                package_name = None
-                package_filename = None
-                package_url = None
-
-
 def normalize_path_separators(path):
     return path.replace(os.path.sep, "/")
 

cgmanifests/generated/cgmanifest.json

@@ -2,112 +2,6 @@
   "$schema": "https://json.schemastore.org/component-detection-manifest.json",
   "Version": 1,
   "Registrations": [
-    {
-      "Component": {
-        "Type": "other",
-        "other": {
-          "Name": "autoconf",
-          "Version": "2.71",
-          "DownloadUrl": "http://ftp.gnu.org/gnu/autoconf/autoconf-2.71.tar.gz"
-        },
-        "comments": "manylinux dependency"
-      }
-    },
-    {
-      "Component": {
-        "Type": "other",
-        "other": {
-          "Name": "automake",
-          "Version": "1.16.5",
-          "DownloadUrl": "http://ftp.gnu.org/gnu/automake/automake-1.16.5.tar.gz"
-        },
-        "comments": "manylinux dependency"
-      }
-    },
-    {
-      "Component": {
-        "Type": "other",
-        "other": {
-          "Name": "libtool",
-          "Version": "2.4.7",
-          "DownloadUrl": "http://ftp.gnu.org/gnu/libtool/libtool-2.4.7.tar.gz"
-        },
-        "comments": "manylinux dependency"
-      }
-    },
-    {
-      "Component": {
-        "Type": "other",
-        "other": {
-          "Name": "git",
-          "Version": "2.36.2",
-          "DownloadUrl": "https://www.kernel.org/pub/software/scm/git/git-2.36.2.tar.gz"
-        },
-        "comments": "manylinux dependency"
-      }
-    },
-    {
-      "Component": {
-        "Type": "other",
-        "other": {
-          "Name": "sqlite_autoconf",
-          "Version": "3390200",
-          "DownloadUrl": "https://www.sqlite.org/2022/sqlite-autoconf-3390200.tar.gz"
-        },
-        "comments": "manylinux dependency"
-      }
-    },
-    {
-      "Component": {
-        "Type": "other",
-        "other": {
-          "Name": "openssl",
-          "Version": "1.1.1q",
-          "DownloadUrl": "https://www.openssl.org/source/openssl-1.1.1q.tar.gz"
-        },
-        "comments": "manylinux dependency"
-      }
-    },
-    {
-      "component": {
-        "type": "git",
-        "git": {
-          "commitHash": "50cf2b6dd4fdf04309445f2eec8de7051d953abf",
-          "repositoryUrl": "https://github.com/besser82/libxcrypt.git"
-        },
-        "comments": "manylinux dependency LIBXCRYPT"
-      }
-    },
-    {
-      "component": {
-        "type": "git",
-        "git": {
-          "commitHash": "a896e3d066448b3530dbcaa48869fafefd738f57",
-          "repositoryUrl": "https://github.com/emscripten-core/emsdk.git"
-        },
-        "comments": "git submodule at cmake/external/emsdk"
-      }
-    },
-    {
-      "component": {
-        "type": "git",
-        "git": {
-          "commitHash": "7a2ed51a6b682a83e345ff49fc4cfd7ca47550db",
-          "repositoryUrl": "https://github.com/google/libprotobuf-mutator.git"
-        },
-        "comments": "git submodule at cmake/external/libprotobuf-mutator"
-      }
-    },
-    {
-      "component": {
-        "type": "git",
-        "git": {
-          "commitHash": "e2525550194ce3d8a2c4a3af451c9d9b3ae6650e",
-          "repositoryUrl": "https://github.com/onnx/onnx.git"
-        },
-        "comments": "git submodule at cmake/external/onnx"
-      }
-    },
     {
       "component": {
         "type": "git",
@@ -268,6 +162,16 @@
         "comments": "mp11"
       }
     },
+    {
+      "component": {
+        "type": "git",
+        "git": {
+          "commitHash": "fdefbe85ed9c362b95b9b401cd19db068a76141f",
+          "repositoryUrl": "https://github.com/onnx/onnx.git"
+        },
+        "comments": "onnx"
+      }
+    },
     {
       "component": {
         "type": "git",

cmake/CMakeLists.txt

@@ -39,7 +39,12 @@ include(CMakeDependentOption)
 include(FetchContent)
 include(CheckFunctionExists)
 
+# TODO: update this once all system adapt c++20
+if(CMAKE_SYSTEM_NAME STREQUAL "Darwin")
+set(CMAKE_CXX_STANDARD 20)
+else()
 set(CMAKE_CXX_STANDARD 17)
+endif()
 
 set_property(GLOBAL PROPERTY USE_FOLDERS ON)
 # NOTE: POSITION INDEPENDENT CODE hurts performance, and it only make sense on POSIX systems
@@ -68,6 +73,11 @@ option(onnxruntime_ENABLE_PYTHON "Enable python buildings" OFF)
 # Enable it may cause LNK1169 error
 option(onnxruntime_ENABLE_MEMLEAK_CHECKER "Experimental: Enable memory leak checker in Windows debug build" OFF)
 option(onnxruntime_USE_CUDA "Build with CUDA support" OFF)
+# Enable ONNX Runtime CUDA EP's internal unit tests that directly access the EP's internal functions instead of through
+# OpKernels. When the option is ON, we will have two copies of GTest library in the same process. It is not a typical
+# use. If you hit any problem with that, please do not report it to GTest. Turn OFF the following build option instead.
+cmake_dependent_option(onnxruntime_ENABLE_CUDA_EP_INTERNAL_TESTS "Build with CUDA unit tests" OFF "onnxruntime_USE_CUDA;onnxruntime_BUILD_UNIT_TESTS;LINUX" OFF)
+
 option(onnxruntime_ENABLE_CUDA_LINE_NUMBER_INFO "When building with CUDA support, generate device code line number information." OFF)
 option(onnxruntime_USE_OPENVINO "Build with OpenVINO support" OFF)
 option(onnxruntime_USE_COREML "Build with CoreML support" OFF)
@@ -141,10 +151,11 @@ option(onnxruntime_DISABLE_SPARSE_TENSORS "Disable sparse tensors data types" OF
 option(onnxruntime_DISABLE_OPTIONAL_TYPE "Disable optional type" OFF)
 option(onnxruntime_DISABLE_FLOAT8_TYPES "Disable float 8 types" OFF)
 option(onnxruntime_MINIMAL_BUILD "Exclude as much as possible from the build. Support ORT format models. No support for ONNX format models." OFF)
-cmake_dependent_option(onnxruntime_DISABLE_RTTI "Disable RTTI" ON "NOT onnxruntime_ENABLE_PYTHON" OFF)
+cmake_dependent_option(onnxruntime_DISABLE_RTTI "Disable RTTI" ON "NOT onnxruntime_ENABLE_PYTHON;NOT onnxruntime_USE_CUDA" OFF)
 # For now onnxruntime_DISABLE_EXCEPTIONS will only work with onnxruntime_MINIMAL_BUILD, more changes (ONNX, non-CPU EP, ...) are required to run this standalone
 cmake_dependent_option(onnxruntime_DISABLE_EXCEPTIONS "Disable exception handling. Requires onnxruntime_MINIMAL_BUILD currently." ON "onnxruntime_MINIMAL_BUILD;NOT onnxruntime_ENABLE_PYTHON" OFF)
-option(onnxruntime_DISABLE_ABSEIL "Do not link to Abseil. Redefine Inlined containers to STD containers." OFF)
+# Even when onnxruntime_DISABLE_ABSEIL is ON, ONNX Runtime still needs to link to abseil.
+option(onnxruntime_DISABLE_ABSEIL "Do not use Abseil data structures in ONNX Runtime source code. Redefine Inlined containers to STD containers." OFF)
 
 option(onnxruntime_EXTENDED_MINIMAL_BUILD "onnxruntime_MINIMAL_BUILD with support for execution providers that compile kernels." OFF)
 option(onnxruntime_MINIMAL_BUILD_CUSTOM_OPS "Add custom operator kernels support to a minimal build." OFF)
@@ -264,10 +275,6 @@ if (onnxruntime_ENABLE_TRAINING_APIS)
   endif()
 endif()
 
-if (onnxruntime_USE_CUDA)
-  set(onnxruntime_DISABLE_RTTI OFF)
-endif()
-
 if (onnxruntime_USE_ROCM)
   if (WIN32)
     message(FATAL_ERROR "ROCM does not support build in Windows!")

cmake/deps.txt

@@ -24,7 +24,7 @@ microsoft_gsl;https://github.com/microsoft/GSL/archive/refs/tags/v4.0.0.zip;cf36
 microsoft_wil;https://github.com/microsoft/wil/archive/refs/tags/v1.0.230629.1.zip;e4a542a323c070376f7c2d1973d0f7ddbc1d2fa5
 mimalloc;https://github.com/microsoft/mimalloc/archive/refs/tags/v2.1.1.zip;d5ee7d34223d0567892db5179849939c8769dc41
 mp11;https://github.com/boostorg/mp11/archive/refs/tags/boost-1.82.0.zip;9bc9e01dffb64d9e0773b2e44d2f22c51aace063
-onnx;https://github.com/onnx/onnx/archive/e2525550194ce3d8a2c4a3af451c9d9b3ae6650e.zip;782f23d788185887f520a90535513e244218e928
+onnx;https://github.com/onnx/onnx/archive/14303de049144035dfd94ace5f7a3b44773b1aad.zip;250eab9690392b248d75b56e605fb49eca373442
 #use the commit of supporting all the plugins and TRT 8.6-GA (https://github.com/onnx/onnx-tensorrt/commit/0462dc31ae78f48744b6141ae376df1f96d3f459)
 onnx_tensorrt;https://github.com/onnx/onnx-tensorrt/archive/0462dc31ae78f48744b6141ae376df1f96d3f459.zip;5ff086361956cceb81ed17453a1fd8db2aa4328d
 protobuf;https://github.com/protocolbuffers/protobuf/archive/refs/tags/v21.12.zip;7cf2733949036c7d52fda017badcab093fe73bfa
@@ -44,4 +44,4 @@ tensorboard;https://github.com/tensorflow/tensorboard/archive/373eb09e4c5d2b3cc2
 cutlass;https://github.com/NVIDIA/cutlass/archive/refs/tags/v3.0.0.zip;0f95b3c1fc1bd1175c4a90b2c9e39074d1bccefd
 utf8_range;https://github.com/protocolbuffers/utf8_range/archive/72c943dea2b9240cd09efde15191e144bc7c7d38.zip;9925739c9debc0efa2adcb194d371a35b6a03156
 extensions;https://github.com/microsoft/onnxruntime-extensions/archive/94142d8391c9791ec71c38336436319a2d4ac7a0.zip;4365ac5140338b4cb75a39944a4be276e3829b3c
-composable_kernel;https://github.com/ROCmSoftwarePlatform/composable_kernel/archive/d52ec01652b7d620386251db92455968d8d90bdc.zip;6b5ce8edf3625f8817086c194fbf94b664e1b0e0
+composable_kernel;https://github.com/ROCmSoftwarePlatform/composable_kernel/archive/d52ec01652b7d620386251db92455968d8d90bdc.zip;6b5ce8edf3625f8817086c194fbf94b664e1b0e0

cmake/deps_update_and_upload.py

@@ -0,0 +1,56 @@
+# in case deps.txt is updated, run this file to update and upload the dependencies so that CI can use them.
+# Before running the script, increase the version number found at:
+# https://aiinfra.visualstudio.com/Lotus/_artifacts/feed/Lotus/UPack/onnxruntime_build_dependencies/versions
+# Run without --do-upload once to verify downloading. Use --do-upload when you are ready to publish.
+# python cmake/deps_update_and_upload.py --root-path C:/temp/onnxruntime_deps --version 1.0.82 --do-upload
+# update version number in tools\ci_build\github\azure-pipelines\templates\download-deps.yml
+import re
+import subprocess
+import os
+import argparse
+import tempfile
+
+parser = argparse.ArgumentParser(description="Update dependencies and publish to Azure Artifacts")
+parser.add_argument(
+    "--root-path", type=str, default=tempfile.gettempdir(), help="Target root path for downloaded files"
+)
+parser.add_argument("--version", type=str, default="1.0.82", help="Package version to publish")
+parser.add_argument("--do-upload", action="store_true", help="Upload the package to Azure Artifacts")
+args = parser.parse_args()
+
+with open("cmake/deps.txt") as file:
+    text = file.read()
+
+lines = [line for line in text.split("\n") if not line.startswith("#") and ";" in line]
+
+root_path = args.root_path
+
+for line in lines:
+    url = re.sub("^[^;]+?;https://([^;]+?);.*", r"https://\1", line)
+    filename = re.sub("^[^;]+?;https://([^;]+?);.*", r"\1", line)
+    full_path = os.path.join(root_path, filename)
+    subprocess.run(["curl", "-sSL", "--create-dirs", "-o", full_path, url])
+
+package_name = "onnxruntime_build_dependencies"
+version = args.version
+
+# Check if the user is logged in to Azure
+result = subprocess.run("az account show", shell=True, capture_output=True, text=True)
+if "No subscriptions found" in result.stderr:
+    # Prompt the user to log in to Azure
+    print("You are not logged in to Azure. Please log in to continue.")
+    subprocess.run("az login", shell=True)
+
+# Publish the package to Azure Artifacts if --no-upload is not specified
+
+cmd = f'az artifacts universal publish --organization https://dev.azure.com/onnxruntime --feed onnxruntime --name {package_name} --version {version} --description "onnxruntime build time dependencies" --path {root_path}'
+if args.do_upload:
+    subprocess.run(cmd, shell=True)
+else:
+    print("would have run: " + cmd)
+
+cmd = f'az artifacts universal publish --organization https://dev.azure.com/aiinfra --feed Lotus --name {package_name} --version {version} --description "onnxruntime build time dependencies" --path {root_path}'
+if args.do_upload:
+    subprocess.run(cmd, shell=True)
+else:
+    print("would have run: " + cmd)

cmake/external/onnxruntime_external_deps.cmake

@@ -37,8 +37,12 @@ if (onnxruntime_BUILD_UNIT_TESTS)
     set(gtest_disable_pthreads ON)
   endif()
   set(INSTALL_GTEST OFF CACHE BOOL "" FORCE)
-  # Set it to ON will cause crashes in onnxruntime_test_all when onnxruntime_USE_CUDA is ON
-  set(GTEST_HAS_ABSL OFF CACHE BOOL "" FORCE)
+  if (CMAKE_SYSTEM_NAME STREQUAL "iOS")
+    # Needs to update onnxruntime/test/xctest/xcgtest.mm
+    set(GTEST_HAS_ABSL OFF CACHE BOOL "" FORCE)
+  else()
+    set(GTEST_HAS_ABSL ON CACHE BOOL "" FORCE)
+  endif()
   # gtest and gmock
   FetchContent_Declare(
     googletest

cmake/onnxruntime_config.h.in

@@ -22,5 +22,5 @@
 #cmakedefine HAS_UNUSED_BUT_SET_VARIABLE
 #cmakedefine HAS_UNUSED_VARIABLE
 #cmakedefine HAS_USELESS_CAST
-#cmakedefine ORT_BUILD_INFO u8"@ORT_BUILD_INFO@"
-#cmakedefine ORT_VERSION u8"@ORT_VERSION@"
+#cmakedefine ORT_BUILD_INFO "@ORT_BUILD_INFO@"
+#cmakedefine ORT_VERSION "@ORT_VERSION@"

cmake/onnxruntime_mlas.cmake

@@ -581,7 +581,7 @@ set_target_properties(onnxruntime_mlas PROPERTIES FOLDER "ONNXRuntime")
 if (WIN32)
   target_compile_options(onnxruntime_mlas PRIVATE "$<$<COMPILE_LANGUAGE:CXX>:/wd6385>" "$<$<COMPILE_LANGUAGE:CXX>:/wd4127>")
   if (onnxruntime_ENABLE_STATIC_ANALYSIS)
-    target_compile_options(onnxruntime_mlas PRIVATE  "$<$<COMPILE_LANGUAGE:CXX>:/analyze:stacksize" 131072>)
+    target_compile_options(onnxruntime_mlas PRIVATE  "$<$<COMPILE_LANGUAGE:CXX>:/analyze:stacksize 131072>")
   endif()
 endif()
 

cmake/onnxruntime_providers.cmake

@@ -460,13 +460,17 @@ if (onnxruntime_USE_CUDA)
   if (onnxruntime_REDUCED_OPS_BUILD)
     substitute_op_reduction_srcs(onnxruntime_providers_cuda_src)
   endif()
-  # cuda_provider_interface.cc is removed from the object target: onnxruntime_providers_cuda_obj and
-  # add to the lib onnxruntime_providers_cuda separatedly.
-  # onnxruntime_providers_cuda_ut can share all the object files with onnxruntime_providers_cuda except cuda_provider_interface.cc.
-  set(cuda_provider_interface_src ${ONNXRUNTIME_ROOT}/core/providers/cuda/cuda_provider_interface.cc)
-  list(REMOVE_ITEM onnxruntime_providers_cuda_src ${cuda_provider_interface_src})
-  onnxruntime_add_object_library(onnxruntime_providers_cuda_obj ${onnxruntime_providers_cuda_src})
-  onnxruntime_add_shared_library_module(onnxruntime_providers_cuda ${cuda_provider_interface_src} $<TARGET_OBJECTS:onnxruntime_providers_cuda_obj>)
+  if(onnxruntime_ENABLE_CUDA_EP_INTERNAL_TESTS)
+    # cuda_provider_interface.cc is removed from the object target: onnxruntime_providers_cuda_obj and
+    # add to the lib onnxruntime_providers_cuda separatedly.
+    # onnxruntime_providers_cuda_ut can share all the object files with onnxruntime_providers_cuda except cuda_provider_interface.cc.
+    set(cuda_provider_interface_src ${ONNXRUNTIME_ROOT}/core/providers/cuda/cuda_provider_interface.cc)
+    list(REMOVE_ITEM onnxruntime_providers_cuda_src ${cuda_provider_interface_src})
+    onnxruntime_add_object_library(onnxruntime_providers_cuda_obj ${onnxruntime_providers_cuda_src})
+    onnxruntime_add_shared_library_module(onnxruntime_providers_cuda ${cuda_provider_interface_src} $<TARGET_OBJECTS:onnxruntime_providers_cuda_obj>)
+  else()
+    onnxruntime_add_shared_library_module(onnxruntime_providers_cuda ${onnxruntime_providers_cuda_src})
+  endif()
   # config_cuda_provider_shared_module can be used to config onnxruntime_providers_cuda_obj, onnxruntime_providers_cuda & onnxruntime_providers_cuda_ut.
   # This function guarantees that all 3 targets have the same configurations.
   function(config_cuda_provider_shared_module target)
@@ -600,7 +604,9 @@ if (onnxruntime_USE_CUDA)
       target_compile_definitions(${target} PRIVATE ENABLE_ATEN)
     endif()
   endfunction()
-  config_cuda_provider_shared_module(onnxruntime_providers_cuda_obj)
+  if(onnxruntime_ENABLE_CUDA_EP_INTERNAL_TESTS)
+    config_cuda_provider_shared_module(onnxruntime_providers_cuda_obj)
+  endif()
   config_cuda_provider_shared_module(onnxruntime_providers_cuda)
 
   install(TARGETS onnxruntime_providers_cuda
@@ -1750,7 +1756,7 @@ if (onnxruntime_USE_TVM)
 
   target_include_directories(onnxruntime_providers_tvm PRIVATE
           ${TVM_INCLUDES}
-          ${PYTHON_INLCUDE_DIRS})
+          ${PYTHON_INCLUDE_DIRS})
   onnxruntime_add_include_to_target(onnxruntime_providers_tvm onnxruntime_common onnxruntime_framework onnx onnx_proto ${PROTOBUF_LIB} flatbuffers::flatbuffers Boost::mp11 safeint_interface)
 
   add_dependencies(onnxruntime_providers_tvm ${onnxruntime_EXTERNAL_DEPENDENCIES})