Merge pull request #863 from intel/sync_msft_25112025

Sync with Microsoft ONNX Runtime - 25/11/2025

Author: Jaswanth51

cmake/deps.txt

@@ -31,7 +31,7 @@ googletest;https://github.com/google/googletest/archive/refs/tags/v1.17.0.zip;f6
 googlexnnpack;https://github.com/google/XNNPACK/archive/3cf85e705098622d59056dcb8f5f963ea7bb0a00.zip;6f6bbba627241f89463ca845febaf063982b34fe
 json;https://github.com/nlohmann/json/archive/refs/tags/v3.11.3.zip;5e88795165cc8590138d1f47ce94ee567b85b4d6
 microsoft_gsl;https://github.com/microsoft/GSL/archive/refs/tags/v4.0.0.zip;cf368104cd22a87b4dd0c80228919bb2df3e2a14
-microsoft_wil;https://github.com/microsoft/wil/archive/refs/tags/v1.0.230629.1.zip;e4a542a323c070376f7c2d1973d0f7ddbc1d2fa5
+microsoft_wil;https://github.com/microsoft/wil/archive/refs/tags/v1.0.250325.1.zip;826c8bd47c2258ec61b8b218e031e5b33d27f761
 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/refs/tags/v1.19.1.zip;c5215b5697dcdfd71799f001b8c4054a6bba6b09

cmake/onnxruntime_unittests.cmake

@@ -1528,8 +1528,14 @@ endif()
       onnxruntime_add_shared_library(onnxruntime_runtime_path_test_shared_library
                                      ${onnxruntime_runtime_path_test_shared_library_src})
 
-      target_link_libraries(onnxruntime_runtime_path_test_shared_library PRIVATE
-                            onnxruntime_common cpuinfo ${CMAKE_DL_LIBS})
+      if (CMAKE_SYSTEM_NAME MATCHES "AIX")
+        target_link_libraries(onnxruntime_runtime_path_test_shared_library PRIVATE
+                              onnxruntime_common ${CMAKE_DL_LIBS})
+        set_target_properties(onnxruntime_runtime_path_test_shared_library PROPERTIES AIX_SHARED_LIBRARY_ARCHIVE OFF)
+      else()
+        target_link_libraries(onnxruntime_runtime_path_test_shared_library PRIVATE
+                              onnxruntime_common cpuinfo ${CMAKE_DL_LIBS})
+      endif()
       target_include_directories(onnxruntime_runtime_path_test_shared_library PRIVATE ${ONNXRUNTIME_ROOT})
 
       if(UNIX)

docs/OperatorKernels.md

@@ -88,7 +88,7 @@ Do not modify directly.*
 |Conv|*in* X:**T**<br> *in* W:**T**<br> *in* B:**T**<br> *out* Y:**T**|22+|**T** = tensor(float)|
 |||[11, 21]|**T** = tensor(float)|
 |||[1, 10]|**T** = tensor(float)|
-|ConvInteger|*in* x:**T1**<br> *in* w:**T2**<br> *in* x_zero_point:**T1**<br> *in* w_zero_point:**T2**<br> *out* y:**T3**|10+|**T1** = tensor(uint8)<br/> **T2** = tensor(uint8)<br/> **T3** = tensor(int32)|
+|ConvInteger|*in* x:**T1**<br> *in* w:**T2**<br> *in* x_zero_point:**T1**<br> *in* w_zero_point:**T2**<br> *out* y:**T3**|10+|**T1** = tensor(int8), tensor(uint8)<br/> **T2** = tensor(int8), tensor(uint8)<br/> **T3** = tensor(int32)|
 |ConvTranspose|*in* X:**T**<br> *in* W:**T**<br> *in* B:**T**<br> *out* Y:**T**|22+|**T** = tensor(float)|
 |||[11, 21]|**T** = tensor(float)|
 |||[1, 10]|**T** = tensor(float)|

include/onnxruntime/core/framework/allocator.h

@@ -38,6 +38,13 @@ struct OrtArenaCfg {
   int max_dead_bytes_per_chunk;           // use -1 to allow ORT to choose the default
   int initial_growth_chunk_size_bytes;    // use -1 to allow ORT to choose the default
   int64_t max_power_of_two_extend_bytes;  // use -1 to allow ORT to choose the default
+  // Use CudaMemPool based arena if available (starting with cuda 11.2)
+  int use_cuda_mempool = -1;
+  // Amount of reserved memory in bytes to hold onto before trying
+  // to release memory back to the OS.
+  uint64_t cuda_mempool_release_threshold = 0;
+  // Bytes to keep on shrink for CudaMemPool, 0 is to attempt to release all, allocated space not affected.
+  size_t cuda_mempool_bytes_to_keep_on_shrink = 0;
 
   bool IsValid() {
     return arena_extend_strategy >= -1 && arena_extend_strategy <= 1 &&
@@ -55,6 +62,9 @@ struct OrtArenaCfg {
     static constexpr const char* InitialGrowthChunkSizeBytes = "arena.initial_growth_chunk_size_bytes";
     static constexpr const char* MaxPowerOfTwoExtendBytes = "arena.max_power_of_two_extend_bytes";
     static constexpr const char* MaxMem = "arena.max_mem";
+    static constexpr const char* UseCudaMemPool = "arena.use_cuda_mempool";
+    static constexpr const char* CudaMempoolReleaseThreshold = "arena.cuda_mempool_release_threshold";
+    static constexpr const char* CudaMempoolBytesToKeepOnShrink = "arena.cuda_mempool_bytes_to_keep_on_shrink";
   };
 
   static onnxruntime::common::Status FromKeyValuePairs(const OrtKeyValuePairs& kvps, OrtArenaCfg& cfg);
@@ -348,4 +358,13 @@ void AllocatorDefaultFree(void* p);
 void* AllocatorDefaultAllocAligned(size_t size, size_t alignment);
 void AllocatorDefaultFreeAligned(void* p, size_t alignment);
 
+class IArena : public IAllocator {
+ public:
+  using IAllocator::IAllocator;
+  virtual Status Shrink() = 0;
+  // Only implemented when IsStreamAware() returns true
+  virtual void ReleaseStreamBuffers(Stream* /*stream*/) {}
+  static IArena* SafeArenaCast(IAllocator* allocator);
+};
+
 }  // namespace onnxruntime

include/onnxruntime/core/session/onnxruntime_c_api.h

@@ -6591,6 +6591,23 @@ struct OrtApi {
    * \since Version 1.24
    */
   ORT_API_T(bool, TensorTypeAndShape_HasShape, _In_ const OrtTensorTypeAndShapeInfo* info);
+
+  /** \brief Get all config entries from ::OrtKernelInfo.
+   *
+   * Gets all configuration entries from the ::OrtKernelInfo object as key-value pairs.
+   * Config entries are set on the ::OrtSessionOptions and are accessible in custom operator kernels.
+   *
+   * Used in the CreateKernel callback of an OrtCustomOp to access all session configuration entries
+   * during kernel construction.
+   *
+   * \param[in] info An instance of ::OrtKernelInfo.
+   * \param[out] out A pointer to a newly created OrtKeyValuePairs instance containing all config entries.
+   *                 Note: the user should call OrtApi::ReleaseKeyValuePairs.
+   *
+   * \snippet{doc} snippets.dox OrtStatus Return Value
+   * \since Version 1.24
+   */
+  ORT_API2_STATUS(KernelInfo_GetConfigEntries, _In_ const OrtKernelInfo* info, _Outptr_ OrtKeyValuePairs** out);
 };
 
 /*

include/onnxruntime/core/session/onnxruntime_cxx_api.h

@@ -2768,6 +2768,8 @@ struct KernelInfoImpl : Base<T> {
 
   std::string GetNodeName() const;
   Logger GetLogger() const;
+
+  KeyValuePairs GetConfigEntries() const;
 };
 
 }  // namespace detail

include/onnxruntime/core/session/onnxruntime_cxx_inline.h

@@ -2822,6 +2822,13 @@ inline Logger KernelInfoImpl<T>::GetLogger() const {
   return Logger{out};
 }
 
+template <typename T>
+inline KeyValuePairs KernelInfoImpl<T>::GetConfigEntries() const {
+  OrtKeyValuePairs* out = nullptr;
+  Ort::ThrowOnError(GetApi().KernelInfo_GetConfigEntries(this->p_, &out));
+  return KeyValuePairs{out};
+}
+
 inline void attr_utils::GetAttr(const OrtKernelInfo* p, const char* name, float& out) {
   Ort::ThrowOnError(GetApi().KernelInfoGetAttribute_float(p, name, &out));
 }

include/onnxruntime/core/session/onnxruntime_ep_device_ep_metadata_keys.h

@@ -9,6 +9,9 @@
 // Key for the execution provider version string. This should be available for all plugin EPs.
 static const char* const kOrtEpDevice_EpMetadataKey_Version = "version";
 
+// Key for the execution provider OS driver version.
+static const char* const kOrtEpDevice_EpMetadataKey_OSDriverVersion = "os_driver_version";
+
 // Prefix for execution provider compatibility information stored in model metadata.
 // Used when generating EP context models to store compatibility strings for each EP.
 // Full key format: "ep_compatibility_info.<EP_TYPE>"