Update py-onnx-runtime to 1.10; update CMS patch (#30725)

* Update py-onnx-runtime to 1.10; update CMS patch

* Update package.py

2 files changed, 146 insertions, 0 deletions

diff --git a/var/spack/repos/builtin/packages/py-onnx-runtime/cms_1_10.patch b/var/spack/repos/builtin/packages/py-onnx-runtime/cms_1_10.patch
new file mode 100644
index 0000000000..3365fb75e2
--- /dev/null
+++ b/var/spack/repos/builtin/packages/py-onnx-runtime/cms_1_10.patch
@@ -0,0 +1,142 @@
+diff --git a/cmake/CMakeLists.txt b/cmake/CMakeLists.txt
+index acbde7f56a8..eb9f7bb9fbf 100644
+--- a/cmake/CMakeLists.txt
++++ b/cmake/CMakeLists.txt
+@@ -718,7 +718,7 @@ if (onnxruntime_BUILD_BENCHMARKS)
+   endif()
+ endif()
+ 
+-if (NOT WIN32 AND NOT onnxruntime_PREFER_SYSTEM_LIB)
++if (NOT WIN32)
+   add_subdirectory(${PROJECT_SOURCE_DIR}/external/nsync EXCLUDE_FROM_ALL)
+ endif()
+ # External dependencies
+diff --git a/include/onnxruntime/core/platform/ort_mutex.h b/include/onnxruntime/core/platform/ort_mutex.h
+index e24665f5142..ddc11953fbc 100644
+--- a/include/onnxruntime/core/platform/ort_mutex.h
++++ b/include/onnxruntime/core/platform/ort_mutex.h
+@@ -101,7 +101,7 @@ std::cv_status OrtCondVar::wait_for(std::unique_lock<OrtMutex>& cond_mutex,
+   return steady_clock::now() - steady_now < rel_time ? std::cv_status::no_timeout : std::cv_status::timeout;
+ }
+ }  // namespace onnxruntime
+-#else
++#elif !defined(__aarch64__)
+ #include "nsync.h"
+ #include <mutex>               //for unique_lock
+ #include <condition_variable>  //for cv_status
+@@ -186,4 +186,11 @@ std::cv_status OrtCondVar::wait_for(std::unique_lock<OrtMutex>& cond_mutex,
+   return steady_clock::now() - steady_now < rel_time ? std::cv_status::no_timeout : std::cv_status::timeout;
+ }
+ };  // namespace onnxruntime
++#else
++#include <mutex>
++#include <condition_variable>
++namespace onnxruntime {
++using OrtMutex = std::mutex;
++using OrtCondVar = std::condition_variable;
++}  // namespace onnxruntime
+ #endif
+diff --git a/include/onnxruntime/core/session/onnxruntime_cxx_api.h b/include/onnxruntime/core/session/onnxruntime_cxx_api.h
+index 048421099bd..4430185d496 100644
+--- a/include/onnxruntime/core/session/onnxruntime_cxx_api.h
++++ b/include/onnxruntime/core/session/onnxruntime_cxx_api.h
+@@ -379,9 +379,9 @@ struct ModelMetadata : Base<OrtModelMetadata> {
+ */
+ struct Session : Base<OrtSession> {
+   explicit Session(std::nullptr_t) {}                                                                                                        ///< Create an empty Session object, must be assigned a valid one to be used
+-  Session(Env& env, const ORTCHAR_T* model_path, const SessionOptions& options);                                                             ///< Wraps OrtApi::CreateSession
+-  Session(Env& env, const ORTCHAR_T* model_path, const SessionOptions& options, OrtPrepackedWeightsContainer* prepacked_weights_container);  ///< Wraps OrtApi::CreateSessionWithPrepackedWeightsContainer
+-  Session(Env& env, const void* model_data, size_t model_data_length, const SessionOptions& options);                                        ///< Wraps OrtApi::CreateSessionFromArray
++  Session(const Env& env, const ORTCHAR_T* model_path, const SessionOptions& options);                                                             ///< Wraps OrtApi::CreateSession
++  Session(const Env& env, const ORTCHAR_T* model_path, const SessionOptions& options, OrtPrepackedWeightsContainer* prepacked_weights_container);  ///< Wraps OrtApi::CreateSessionWithPrepackedWeightsContainer
++  Session(const Env& env, const void* model_data, size_t model_data_length, const SessionOptions& options);                                        ///< Wraps OrtApi::CreateSessionFromArray
+ 
+   /** \brief Run the model returning results in an Ort allocated vector.
+   * 
+diff --git a/include/onnxruntime/core/session/onnxruntime_cxx_inline.h b/include/onnxruntime/core/session/onnxruntime_cxx_inline.h
+index 1f31dffca87..b9d2cdfc475 100644
+--- a/include/onnxruntime/core/session/onnxruntime_cxx_inline.h
++++ b/include/onnxruntime/core/session/onnxruntime_cxx_inline.h
+@@ -538,16 +538,16 @@ inline SessionOptions& SessionOptions::AppendExecutionProvider_OpenVINO(const Or
+   return *this;
+ }
+ 
+-inline Session::Session(Env& env, const ORTCHAR_T* model_path, const SessionOptions& options) {
++inline Session::Session(const Env& env, const ORTCHAR_T* model_path, const SessionOptions& options) {
+   ThrowOnError(GetApi().CreateSession(env, model_path, options, &p_));
+ }
+ 
+-inline Session::Session(Env& env, const ORTCHAR_T* model_path, const SessionOptions& options,
++inline Session::Session(const Env& env, const ORTCHAR_T* model_path, const SessionOptions& options,
+                         OrtPrepackedWeightsContainer* prepacked_weights_container) {
+   ThrowOnError(GetApi().CreateSessionWithPrepackedWeightsContainer(env, model_path, options, prepacked_weights_container, &p_));
+ }
+ 
+-inline Session::Session(Env& env, const void* model_data, size_t model_data_length, const SessionOptions& options) {
++inline Session::Session(const Env& env, const void* model_data, size_t model_data_length, const SessionOptions& options) {
+   ThrowOnError(GetApi().CreateSessionFromArray(env, model_data, model_data_length, options, &p_));
+ }
+ 
+diff --git a/onnxruntime/core/mlas/lib/platform.cpp b/onnxruntime/core/mlas/lib/platform.cpp
+index de7fee8c07a..6d97cf07a05 100644
+--- a/onnxruntime/core/mlas/lib/platform.cpp
++++ b/onnxruntime/core/mlas/lib/platform.cpp
+@@ -16,6 +16,7 @@ Module Name:
+ --*/
+ 
+ #include "mlasi.h"
++#include <string>
+ 
+ #if defined(MLAS_TARGET_POWER) && defined(__linux__)
+ #include <sys/auxv.h>
+@@ -197,8 +198,11 @@ Return Value:
+         //
+ 
+         uint64_t xcr0 = MlasReadExtendedControlRegister(_XCR_XFEATURE_ENABLED_MASK);
++        const char *cpu_opt = std::getenv("MLAS_DYNAMIC_CPU_ARCH");
++        if (cpu_opt == nullptr) cpu_opt = "99";
++        auto opt = std::stoi(cpu_opt);
+ 
+-        if ((xcr0 & 0x6) == 0x6) {
++        if (opt > 0 && (xcr0 & 0x6) == 0x6) {
+ 
+             this->GemmFloatKernel = MlasGemmFloatKernelAvx;
+ 
+@@ -231,7 +235,7 @@ Return Value:
+             __cpuid_count(7, 0, Cpuid7[0], Cpuid7[1], Cpuid7[2], Cpuid7[3]);
+ #endif
+ 
+-            if (((Cpuid1[2] & 0x1000) != 0) && ((Cpuid7[1] & 0x20) != 0)) {
++            if (opt > 1 && ((Cpuid1[2] & 0x1000) != 0) && ((Cpuid7[1] & 0x20) != 0)) {
+ 
+                 this->GemmU8S8Dispatch = &MlasGemmU8S8DispatchAvx2;
+                 this->GemmU8S8Kernel = MlasGemmU8S8KernelAvx2;
+@@ -290,7 +294,7 @@ Return Value:
+                 // operating system supports saving AVX512F state.
+                 //
+ 
+-                if (((Cpuid7[1] & 0x10000) != 0) && ((xcr0 & 0xE0) == 0xE0)) {
++                if (opt > 2 && ((Cpuid7[1] & 0x10000) != 0) && ((xcr0 & 0xE0) == 0xE0)) {
+ 
+                     this->GemmFloatKernel = MlasGemmFloatKernelAvx512F;
+                     this->GemmDoubleKernel = MlasGemmDoubleKernelAvx512F;
+diff --git a/onnxruntime/core/platform/posix/ort_mutex.cc b/onnxruntime/core/platform/posix/ort_mutex.cc
+index 8a5d41eb360..89111c9daa5 100644
+--- a/onnxruntime/core/platform/posix/ort_mutex.cc
++++ b/onnxruntime/core/platform/posix/ort_mutex.cc
+@@ -1,6 +1,7 @@
+ // Copyright (c) Microsoft Corporation. All rights reserved.
+ // Licensed under the MIT License.
+ 
++#if !defined(__aarch64__)
+ #include "core/common/common.h"
+ #include "core/platform/ort_mutex.h"
+ #include <assert.h>
+@@ -40,4 +41,5 @@ void OrtCondVar::wait(std::unique_lock<OrtMutex>& lk) {
+   nsync::nsync_cv_wait(&native_cv_object, lk.mutex()->native_handle());
+ }
+ 
+-}  // namespace onnxruntime
+\ No newline at end of file
++}  // namespace onnxruntime
++#endif
diff --git a/var/spack/repos/builtin/packages/py-onnx-runtime/package.py b/var/spack/repos/builtin/packages/py-onnx-runtime/package.py
index d2eef4d69c..e1475eb5cf 100644
--- a/var/spack/repos/builtin/packages/py-onnx-runtime/package.py
+++ b/var/spack/repos/builtin/packages/py-onnx-runtime/package.py
@@ -19,6 +19,7 @@ class PyOnnxRuntime(CMakePackage, PythonPackage):
     homepage = "https://github.com/microsoft/onnxruntime"
     git      = "https://github.com/microsoft/onnxruntime.git"
 
+    version('1.10.0', tag='v1.10.0', submodules=True)
     version('1.7.2',  tag='v1.7.2',  submodules=True)
 
     variant('cuda', default=False, description='Build with CUDA support')
@@ -35,6 +36,7 @@ class PyOnnxRuntime(CMakePackage, PythonPackage):
     depends_on('py-cerberus', type=('build', 'run'))
     depends_on('py-wheel', type='build')
     depends_on('py-onnx', type=('build', 'run'))
+    depends_on('py-flatbuffers', type=('build', 'run'))
     depends_on('zlib')
     depends_on('libpng')
     depends_on('py-pybind11', type='build')
@@ -47,6 +49,8 @@ class PyOnnxRuntime(CMakePackage, PythonPackage):
     # Adopted from CMS experiment's fork of onnxruntime
     # https://github.com/cms-externals/onnxruntime/compare/5bc92df...d594f80
     patch('cms.patch', level=1, when='@1.7.2')
+    # https://github.com/cms-externals/onnxruntime/compare/0d9030e...7a6355a
+    patch('cms_1_10.patch', whe='@1.10')
     # https://github.com/microsoft/onnxruntime/issues/4234#issuecomment-698077636
     patch('libiconv.patch', level=0, when='@1.7.2')
     # https://github.com/microsoft/onnxruntime/commit/de4089f8cbe0baffe56a363cc3a41595cc8f0809.patch