Add XNNPack build on Linux ARM64 and improve Linux CPU (#22773)

### Description
1. Add XNNPack build on Linux ARM64
2. Build only one python wheel for PR request.

[AB#49763](https://aiinfra.visualstudio.com/6a833879-cd9b-44a4-a9de-adc2d818f13c/_workitems/edit/49763)



### Motivation and Context
Why I add xnnpack build on Linux ARM64  rather than Windows ARM64.
Becuase KleidiAI  doesn't support Windows

```
IF(XNNPACK_TARGET_PROCESSOR STREQUAL "arm64" AND XNNPACK_ENABLE_ARM_I8MM AND NOT CMAKE_C_COMPILER_ID STREQUAL "MSVC")
  IF (XNNPACK_ENABLE_KLEIDIAI)
    MESSAGE(STATUS "Enabling KleidiAI for Arm64")
  ENDIF()
ELSE()
  SET(XNNPACK_ENABLE_KLEIDIAI OFF)
ENDIF()
```

---------

tools/ci_build/github/azure-pipelines/linux-ci-pipeline.yml

@@ -299,6 +299,7 @@ stages:
       machine_pool: 'onnxruntime-linux-ARM64-CPU-2019'
       with_cache: true
       cmake_build_type: Release
+      python_exe_path: '/opt/python/cp310-cp310/bin/python3.10'
 
 - stage: arm64_test
   dependsOn: ['arm64_build']
@@ -307,3 +308,24 @@ stages:
     parameters:
       arch: 'aarch64'
       machine_pool: 'onnxruntime-linux-ARM64-CPU-2019'
+
+- stage: arm64_build_xnnpack
+  dependsOn: []
+  jobs:
+  - template: templates/py-linux.yml
+    parameters:
+      arch: 'aarch64'
+      machine_pool: 'onnxruntime-linux-ARM64-CPU-2019'
+      with_cache: true
+      cmake_build_type: Release
+      ep: 'XNNPack'
+      extra_build_arg: '--use_xnnpack'
+      python_exe_path: '/opt/python/cp310-cp310/bin/python3.10'
+
+- stage: arm64_test_xnnpack
+  dependsOn: ['arm64_build_xnnpack']
+  jobs:
+  - template: templates/py-packaging-linux-test-cpu.yml
+    parameters:
+      arch: 'aarch64'
+      machine_pool: 'onnxruntime-linux-ARM64-CPU-2019'

tools/ci_build/github/azure-pipelines/templates/py-linux.yml

@@ -26,8 +26,16 @@ parameters:
   type: string
   default: ''
 
+- name: ep
+  type: string
+  default: 'cpu'
+
+- name: python_exe_path
+  type: string
+  default: ''
+
 jobs:
-- job: Linux_py_Wheels_${{ parameters.arch }}
+- job: Linux_py_Wheels_${{ parameters.arch }}_${{parameters.ep}}
   timeoutInMinutes: 240
   workspace:
     clean: all
@@ -42,9 +50,15 @@ jobs:
       value: $[format('{0:dd}{0:MM}{0:yyyy}', pipeline.startTime)]
     - name: extra_build_args
       ${{ if ne(parameters.extra_build_arg, '') }}:
-        value: -x ${{ parameters.extra_build_arg }}
+        value: '-x ${{ parameters.extra_build_arg }}'
       ${{ if eq(parameters.extra_build_arg, '') }}:
         value: ''
+    - name: python_exe_path
+      ${{ if ne(parameters.python_exe_path, '') }}:
+        value: '-p ${{ parameters.python_exe_path }}'
+      ${{ if eq(parameters.python_exe_path, '') }}:
+        value: ''
+
   steps:
     - task: mspremier.PostBuildCleanup.PostBuildCleanup-task.PostBuildCleanup@3
       displayName: 'Clean Agent Directories'
@@ -78,7 +92,7 @@ jobs:
             inputs:
               targetType: filePath
               filePath: tools/ci_build/github/linux/run_python_dockerbuild.sh
-              arguments: -i onnxruntimecpubuildpython${{ parameters.arch }} -d "${{ parameters.device }}" -c ${{ parameters.cmake_build_type }} $(extra_build_args)
+              arguments: -i onnxruntimecpubuildpython${{ parameters.arch }} -d "${{ parameters.device }}" -c ${{ parameters.cmake_build_type }} $(extra_build_args) $(python_exe_path)
             ${{ if eq(parameters.with_cache, 'true') }}:
               env:
                 ADDITIONAL_DOCKER_PARAMETER: "--volume $(ORT_CACHE_DIR):/cache -e CCACHE_DIR=/cache -e ORT_BUILD_WITH_CACHE=1"
@@ -87,14 +101,14 @@ jobs:
       displayName: 'Publish Artifact: ONNXRuntime python wheel'
       inputs:
         PathtoPublish: '$(Build.BinariesDirectory)/dist'
-        ArtifactName: onnxruntime
+        ArtifactName: onnxruntime-${{ parameters.ep }}
 
     - task: PublishPipelineArtifact@0
       displayName: 'Publish Test Binaries'
       inputs:
-        artifactName: 'drop-linux-cpu-${{ parameters.arch }}'
+        artifactName: 'drop-linux-cpu-${{ parameters.arch }}-${{ parameters.ep }}'
         targetPath: '$(Build.BinariesDirectory)/${{ parameters.cmake_build_type }}'
 
     - template: component-governance-component-detection-steps.yml
       parameters :
-        condition : 'succeeded'
+        condition : 'succeeded'

tools/ci_build/github/azure-pipelines/templates/py-packaging-linux-test-cpu.yml

@@ -28,6 +28,10 @@ parameters:
   type: number
   default: 120
 
+- name: ep
+  type: string
+  default: 'cpu'
+
 jobs:
 - job: Linux_Test_CPU${{ parameters.extra_job_id }}_${{ parameters.arch }}
   timeoutInMinutes: ${{ parameters.timeout }}
@@ -43,30 +47,30 @@ jobs:
   # The public ADO project
   - ${{ if eq(variables['System.CollectionId'], 'f3ad12f2-e480-4533-baf2-635c95467d29') }}:
     - download: current   # pipeline resource identifier.
-      artifact: 'drop-linux-cpu-${{ parameters.arch }}'
+      artifact: 'drop-linux-cpu-${{ parameters.arch }}-${{parameters.ep}}'
 
     - download: current   # pipeline resource identifier.
-      artifact: 'onnxruntime${{ parameters.python_wheel_suffix }}'
+      artifact: 'onnxruntime${{ parameters.python_wheel_suffix }}-${{ parameters.ep }}'
 
     - bash: |
         set -e -x
-        mv "$(Pipeline.Workspace)/drop-linux-cpu-${{ parameters.arch }}" $(Build.BinariesDirectory)/${{parameters.cmake_build_type}}
-        mv "$(Pipeline.Workspace)/onnxruntime${{ parameters.python_wheel_suffix }}" "$(Build.BinariesDirectory)/whl"
+        mv "$(Pipeline.Workspace)/drop-linux-cpu-${{ parameters.arch }}-${{parameters.ep}}" $(Build.BinariesDirectory)/${{parameters.cmake_build_type}}
+        mv "$(Pipeline.Workspace)/onnxruntime${{ parameters.python_wheel_suffix }}-${{parameters.ep}}" "$(Build.BinariesDirectory)/whl"
         cp -r "$(Build.BinariesDirectory)/whl" $(Build.BinariesDirectory)/tmp
         find "$(Build.BinariesDirectory)/tmp" -name '*.whl' -exec bash -c 'unzip -d "${1%.*}" "$1"' _ {} \;
   # The private ADO project
   - ${{ if eq(variables['System.CollectionId'], 'bc038106-a83b-4dab-9dd3-5a41bc58f34c') }}:
     - download: build   # pipeline resource identifier.
-      artifact: 'drop-linux-cpu-${{ parameters.arch }}'
+      artifact: 'drop-linux-cpu-${{ parameters.arch }}-${{parameters.ep}}'
 
     - download: build   # pipeline resource identifier.
-      artifact: 'onnxruntime${{ parameters.python_wheel_suffix }}'
+      artifact: 'onnxruntime${{ parameters.python_wheel_suffix }}-${{ parameters.ep }}'
 
     - bash: |
         set -e -x
         ls $(Pipeline.Workspace)/build
-        mv "$(Pipeline.Workspace)/build/drop-linux-cpu-${{ parameters.arch }}" $(Build.BinariesDirectory)/${{parameters.cmake_build_type}}
-        mv "$(Pipeline.Workspace)/build/onnxruntime${{ parameters.python_wheel_suffix }}" "$(Build.BinariesDirectory)/whl"
+        mv "$(Pipeline.Workspace)/build/drop-linux-cpu-${{ parameters.arch }}-${{parameters.ep}}" $(Build.BinariesDirectory)/${{parameters.cmake_build_type}}
+        mv "$(Pipeline.Workspace)/build/onnxruntime${{ parameters.python_wheel_suffix }}-${{parameters.ep}}" "$(Build.BinariesDirectory)/whl"
         cp -r "$(Build.BinariesDirectory)/whl" $(Build.BinariesDirectory)/tmp
         find "$(Build.BinariesDirectory)/tmp" -name '*.whl' -exec bash -c 'unzip -d "${1%.*}" "$1"' _ {} \;
 

tools/ci_build/github/linux/build_linux_python_package.sh

@@ -7,7 +7,7 @@ mkdir -p /build/dist
 
 EXTRA_ARG=""
 ENABLE_CACHE=false
-# Put 3.10 at the last because Ubuntu 22.04 use python 3.10 and we will upload the intermediate build files of this 
+# Put 3.10 at the last because Ubuntu 22.04 use python 3.10 and we will upload the intermediate build files of this
 # config to Azure DevOps Artifacts and download them to a Ubuntu 22.04 machine to run the tests.
 PYTHON_EXES=("/opt/python/cp311-cp311/bin/python3.11" "/opt/python/cp312-cp312/bin/python3.12" "/opt/python/cp313-cp313/bin/python3.13" "/opt/python/cp313-cp313t/bin/python3.13t" "/opt/python/cp310-cp310/bin/python3.10")
 while getopts "d:p:x:c:e" parameter_Option
@@ -15,7 +15,7 @@ do case "${parameter_Option}"
 in
 #GPU|CPU|NPU.
 d) BUILD_DEVICE=${OPTARG};;
-p) PYTHON_EXES=${OPTARG};;
+p) PYTHON_EXES=("${OPTARG}");;
 x) EXTRA_ARG=${OPTARG};;
 c) BUILD_CONFIG=${OPTARG};;
 e) ENABLE_CACHE=true;;
@@ -89,9 +89,11 @@ export CMAKE_ARGS="-DONNX_GEN_PB_TYPE_STUBS=ON -DONNX_WERROR=OFF"
 for PYTHON_EXE in "${PYTHON_EXES[@]}"
 do
   rm -rf /build/"$BUILD_CONFIG"
-  ${PYTHON_EXE} -m pip install -r /onnxruntime_src/tools/ci_build/github/linux/python/requirements.txt  
-  ${PYTHON_EXE} /onnxruntime_src/tools/ci_build/build.py "${BUILD_ARGS[@]}"
-
+  # that's a workaround for the issue that there's no python3 in the docker image
+  # like xnnpack's cmakefile, it uses pythone3 to run a external command
+  python3_dir=$(dirname "$PYTHON_EXE")
+  ${PYTHON_EXE} -m pip install -r /onnxruntime_src/tools/ci_build/github/linux/python/requirements.txt
+  PATH=$python3_dir:$PATH ${PYTHON_EXE} /onnxruntime_src/tools/ci_build/build.py "${BUILD_ARGS[@]}"
   cp /build/"$BUILD_CONFIG"/dist/*.whl /build/dist
 done
 

tools/ci_build/github/linux/run_python_dockerbuild.sh

@@ -2,14 +2,15 @@
 set -e -x
 BUILD_CONFIG="Release"
 
-while getopts "i:d:x:c:" parameter_Option
+while getopts "i:d:x:c:p:" parameter_Option
 do case "${parameter_Option}"
 in
 i) DOCKER_IMAGE=${OPTARG};;
 d) DEVICE=${OPTARG};;
 x) BUILD_EXTR_PAR=${OPTARG};;
 c) BUILD_CONFIG=${OPTARG};;
-*) echo "Usage: $0 -i <docker_image> -d <GPU|CPU> [-x <extra_build_arg>] [-c <build_config>]"
+p) PYTHON_EXES=${OPTARG};;
+*) echo "Usage: $0 -i <docker_image> -d <GPU|CPU> [-x <extra_build_arg>] [-c <build_config>] [-p <python_exe_path>]"
    exit 1;;
 esac
 done
@@ -33,7 +34,7 @@ docker run --rm \
     -e ORT_DISABLE_PYTHON_PACKAGE_LOCAL_VERSION \
     -e DEFAULT_TRAINING_PACKAGE_DEVICE \
     $ADDITIONAL_DOCKER_PARAMETER \
-    $DOCKER_IMAGE tools/ci_build/github/linux/build_linux_python_package.sh $DOCKER_SCRIPT_OPTIONS
+    $DOCKER_IMAGE tools/ci_build/github/linux/build_linux_python_package.sh -c $BUILD_CONFIG -p $PYTHON_EXES $DOCKER_SCRIPT_OPTIONS
 
 sudo rm -rf "${BUILD_BINARIESDIRECTORY}/${BUILD_CONFIG}/onnxruntime" "${BUILD_BINARIESDIRECTORY}/${BUILD_CONFIG}/pybind11" \
     "${BUILD_BINARIESDIRECTORY}/${BUILD_CONFIG}/models" "${BUILD_BINARIESDIRECTORY}/${BUILD_CONFIG}/_deps" \