ORT 1.23.0 cherry-pick prs 25592 - 25831

cmake/external/onnxruntime_external_deps.cmake

@@ -336,7 +336,13 @@ if (onnxruntime_ENABLE_CPUINFO)
       set(CPUINFO_SUPPORTED TRUE)
     endif()
     if (WIN32)
-      set(CPUINFO_SUPPORTED TRUE)
+      # There's an error when linking with cpuinfo on arm64ec with a vcpkg build (--use_vcpkg).
+      # TODO Fix it and then re-enable cpuinfo on arm64ec.
+      if (onnxruntime_target_platform STREQUAL "ARM64EC")
+        set(CPUINFO_SUPPORTED FALSE)
+      else()
+        set(CPUINFO_SUPPORTED TRUE)
+      endif()
     elseif (NOT ${onnxruntime_target_platform} MATCHES "^(i[3-6]86|AMD64|x86(_64)?|armv[5-8].*|aarch64|arm64)$")
       message(WARNING
         "Target processor architecture \"${onnxruntime_target_platform}\" is not supported in cpuinfo. "

cmake/onnxruntime_common.cmake

@@ -14,7 +14,7 @@ set(onnxruntime_common_src_patterns
     "${ONNXRUNTIME_ROOT}/core/platform/check_intel.h"
     "${ONNXRUNTIME_ROOT}/core/platform/check_intel.cc"
     "${ONNXRUNTIME_ROOT}/core/platform/device_discovery.h"
-    "${ONNXRUNTIME_ROOT}/core/platform/device_discovery.cc"
+    "${ONNXRUNTIME_ROOT}/core/platform/device_discovery_common.cc"
     "${ONNXRUNTIME_ROOT}/core/platform/env.h"
     "${ONNXRUNTIME_ROOT}/core/platform/env.cc"
     "${ONNXRUNTIME_ROOT}/core/platform/env_time.h"
@@ -32,26 +32,38 @@ set(onnxruntime_common_src_patterns
 
 if(WIN32)
     list(APPEND onnxruntime_common_src_patterns
-         "${ONNXRUNTIME_ROOT}/core/platform/windows/*.h"
-         "${ONNXRUNTIME_ROOT}/core/platform/windows/*.cc"
+         "${ONNXRUNTIME_ROOT}/core/platform/windows/debug_alloc.cc"
+         "${ONNXRUNTIME_ROOT}/core/platform/windows/debug_alloc.h"
+         "${ONNXRUNTIME_ROOT}/core/platform/windows/dll_load_error.cc"
+         "${ONNXRUNTIME_ROOT}/core/platform/windows/dll_load_error.h"
+         "${ONNXRUNTIME_ROOT}/core/platform/windows/env_time.cc"
+         "${ONNXRUNTIME_ROOT}/core/platform/windows/env.cc"
+         "${ONNXRUNTIME_ROOT}/core/platform/windows/env.h"
+         "${ONNXRUNTIME_ROOT}/core/platform/windows/hardware_core_enumerator.cc"
+         "${ONNXRUNTIME_ROOT}/core/platform/windows/hardware_core_enumerator.h"
+         "${ONNXRUNTIME_ROOT}/core/platform/windows/stacktrace.cc"
+         "${ONNXRUNTIME_ROOT}/core/platform/windows/telemetry.cc"
+         "${ONNXRUNTIME_ROOT}/core/platform/windows/telemetry.h"
          "${ONNXRUNTIME_ROOT}/core/platform/windows/logging/*.h"
          "${ONNXRUNTIME_ROOT}/core/platform/windows/logging/*.cc"
     )
 
 else()
     list(APPEND onnxruntime_common_src_patterns
-         "${ONNXRUNTIME_ROOT}/core/platform/posix/*.h"
-         "${ONNXRUNTIME_ROOT}/core/platform/posix/*.cc"
+         "${ONNXRUNTIME_ROOT}/core/platform/posix/env_time.cc"
+         "${ONNXRUNTIME_ROOT}/core/platform/posix/env.cc"
+         "${ONNXRUNTIME_ROOT}/core/platform/posix/stacktrace.cc"
     )
 
+    # logging files
     if (onnxruntime_USE_SYSLOG)
         list(APPEND onnxruntime_common_src_patterns
             "${ONNXRUNTIME_ROOT}/core/platform/posix/logging/*.h"
             "${ONNXRUNTIME_ROOT}/core/platform/posix/logging/*.cc"
         )
     endif()
 
-    if (CMAKE_SYSTEM_NAME STREQUAL "Android")
+    if (ANDROID)
         list(APPEND onnxruntime_common_src_patterns
             "${ONNXRUNTIME_ROOT}/core/platform/android/logging/*.h"
             "${ONNXRUNTIME_ROOT}/core/platform/android/logging/*.cc"
@@ -66,6 +78,21 @@ else()
     endif()
 endif()
 
+# platform-specific device discovery files
+if (WIN32)
+    list(APPEND onnxruntime_common_src_patterns
+         "${ONNXRUNTIME_ROOT}/core/platform/windows/device_discovery.cc")
+elseif (LINUX)
+    list(APPEND onnxruntime_common_src_patterns
+         "${ONNXRUNTIME_ROOT}/core/platform/linux/device_discovery.cc")
+elseif (APPLE)
+    list(APPEND onnxruntime_common_src_patterns
+         "${ONNXRUNTIME_ROOT}/core/platform/apple/device_discovery.cc")
+else()
+    list(APPEND onnxruntime_common_src_patterns
+         "${ONNXRUNTIME_ROOT}/core/platform/device_discovery_default.cc")
+endif()
+
 if(onnxruntime_target_platform STREQUAL "ARM64EC")
     if (MSVC)
         link_directories("$ENV{VCINSTALLDIR}/Tools/MSVC/$ENV{VCToolsVersion}/lib/ARM64EC")
@@ -216,8 +243,6 @@ endif()
 
 if (RISCV64 OR ARM64 OR ARM OR X86 OR X64 OR X86_64)
     # Link cpuinfo if supported
-    # Using it mainly in ARM with Android.
-    # Its functionality in detecting x86 cpu features are lacking, so is support for Windows.
     if (CPUINFO_SUPPORTED)
       onnxruntime_add_include_to_target(onnxruntime_common cpuinfo::cpuinfo)
       list(APPEND onnxruntime_EXTERNAL_LIBRARIES cpuinfo::cpuinfo ${ONNXRUNTIME_CLOG_TARGET_NAME})

cmake/vcpkg-ports/cpuinfo/patch_cpuinfo_h_for_arm64ec.patch

@@ -0,0 +1,22 @@
+diff --git a/include/cpuinfo.h b/include/cpuinfo.h
+index f1d35d4..9e454d2 100644
+--- a/include/cpuinfo.h
++++ b/include/cpuinfo.h
+@@ -18,7 +18,7 @@
+ #define CPUINFO_ARCH_X86 1
+ #endif
+
+-#if defined(__x86_64__) || defined(__x86_64) || defined(_M_X64) || defined(_M_AMD64)
++#if defined(__x86_64__) || defined(__x86_64) || (defined(_M_X64) && !defined(_M_ARM64EC)) || (defined(_M_AMD64) && !defined(_M_ARM64EC))
+ #define CPUINFO_ARCH_X86_64 1
+ #endif
+
+@@ -26,7 +26,7 @@
+ #define CPUINFO_ARCH_ARM 1
+ #endif
+
+-#if defined(__aarch64__) || defined(_M_ARM64)
++#if defined(__aarch64__) || defined(_M_ARM64) || defined(_M_ARM64EC)
+ #define CPUINFO_ARCH_ARM64 1
+ #endif
+

cmake/vcpkg-ports/cpuinfo/portfile.cmake

@@ -9,6 +9,8 @@ vcpkg_from_github(
     REF 8a1772a0c5c447df2d18edf33ec4603a8c9c04a6
     SHA512 b94ccbfa886221d6bb16513d074675af0a72928a9dd9485dcacdc1124a8a60aacbbe91913a1579e766dfb024f0be1d52eeead40342004ff0238a8b94a095ed08
     HEAD_REF master
+    PATCHES
+        patch_cpuinfo_h_for_arm64ec.patch
 )
 
 vcpkg_check_features(OUT_FEATURE_OPTIONS FEATURE_OPTIONS

include/onnxruntime/core/common/parse_string.h

@@ -35,13 +35,30 @@ template <typename T>
 std::enable_if_t<detail::ParseWithFromChars<T>, bool>
 TryParseStringWithClassicLocale(std::string_view str, T& value) {
   T parsed_value{};
-  const auto [ptr, ec] = std::from_chars(str.data(), str.data() + str.size(), parsed_value);
 
-  if (ec != std::errc{}) {
+  std::from_chars_result conversion_result{};
+  if constexpr (std::is_integral_v<T> && std::is_unsigned_v<T>) {
+    // For unsigned integral types, also handle hex values, i.e., those beginning with "0x".
+    // std::from_chars() does not accept the "0x" prefix.
+    const bool has_hex_prefix = str.size() >= 2 &&
+                                str[0] == '0' &&
+                                (str[1] == 'x' || str[1] == 'X');
+
+    if (has_hex_prefix) {
+      str = str.substr(2);
+    }
+
+    const int base = has_hex_prefix ? 16 : 10;
+    conversion_result = std::from_chars(str.data(), str.data() + str.size(), parsed_value, base);
+  } else {
+    conversion_result = std::from_chars(str.data(), str.data() + str.size(), parsed_value);
+  }
+
+  if (conversion_result.ec != std::errc{}) {
     return false;
   }
 
-  if (ptr != str.data() + str.size()) {
+  if (conversion_result.ptr != str.data() + str.size()) {
     return false;
   }
 

include/onnxruntime/core/framework/execution_provider.h

@@ -36,6 +36,7 @@ class GraphOptimizerRegistry;
 #include "core/framework/framework_provider_common.h"
 #include "core/framework/stream_handles.h"
 #include "core/framework/tuning_context.h"
+#include "core/session/onnxruntime_c_api.h"
 
 struct OrtEpDevice;
 struct OrtRunOptions;
@@ -322,6 +323,29 @@ class IExecutionProvider {
   virtual common::Status Compile(const std::vector<FusedNodeAndGraph>& fused_nodes_and_graphs,
                                  std::vector<NodeComputeInfo>& node_compute_funcs);
 
+  /**
+   * Get the compatibility info for a compiled model.
+   *
+   * The execution provider determines this value, which denotes the compatibility of the compiled model with the EP.
+   * This is stored in the model metadata under a key associated with the EP type.
+   */
+  virtual std::string GetCompiledModelCompatibilityInfo(const onnxruntime::GraphViewer& graph_viewer) const {
+    // graph_viewer and model_metadata are not used in the default implementation.
+    ORT_UNUSED_PARAMETER(graph_viewer);
+    // Default implementation returns empty string
+    return std::string();
+  }
+
+  /**
+   * Validate the compatibility of a compiled model with this execution provider.
+   */
+  virtual common::Status ValidateCompiledModelCompatibilityInfo(const std::string& /*compatibility_info*/,
+                                                                OrtCompiledModelCompatibility& model_compatibility) const {
+    // Default implementation indicates this EP does not support model compatibility validation
+    model_compatibility = OrtCompiledModelCompatibility_EP_NOT_APPLICABLE;
+    return Status::OK();
+  }
+
 #endif
 
   void SetLogger(const logging::Logger* logger) {

include/onnxruntime/core/graph/indexed_sub_graph.h

@@ -31,7 +31,7 @@ struct IndexedSubGraph {
     std::string domain;  ///< Domain of customized SubGraph/FunctionProto
     int since_version;   ///< Since version of customized SubGraph/FunctionProto.
 
-    ONNX_NAMESPACE::OperatorStatus status;  ///< Status of customized SubGraph/FunctionProto.
+    ONNX_NAMESPACE::OperatorStatus status{ONNX_NAMESPACE::OperatorStatus::STABLE};  ///< Status of customized SubGraph/FunctionProto.
 
     std::vector<std::string> inputs;                 ///< Inputs of customized SubGraph/FunctionProto.
     std::vector<std::string> outputs;                ///< Outputs of customized SubGraph/FunctionProto.

include/onnxruntime/core/providers/nv_tensorrt_rtx/nv_provider_options.h

@@ -32,9 +32,8 @@ constexpr const char* kProfilesMinShapes = "nv_profile_min_shapes";
 constexpr const char* kProfilesMaxShapes = "nv_profile_max_shapes";
 constexpr const char* kProfilesOptShapes = "nv_profile_opt_shapes";
 constexpr const char* kCudaGraphEnable = "nv_cuda_graph_enable";
-constexpr const char* kONNXBytestream = "nv_onnx_bytestream";
-constexpr const char* kONNXBytestreamSize = "nv_onnx_bytestream_size";
 constexpr const char* kMultiProfileEnable = "nv_multi_profile_enable";
+constexpr const char* kUseExternalDataInitializer = "nv_use_external_data_initializer";
 
 }  // namespace provider_option_names
 namespace run_option_names {

include/onnxruntime/core/session/environment.h

@@ -199,23 +199,6 @@ class Environment {
 
   using OrtAllocatorUniquePtr = std::unique_ptr<OrtAllocator, std::function<void(OrtAllocator*)>>;
 
-  // if the user calls CreateSharedAllocator and wraps the plugin EP's allocator with an arena we end up with
-  // OrtAllocator from EP -> wrapped in IAllocatorImplWrappingOrtAllocator -> inside a BFCArena IAllocator.
-  // we can put that in shared_allocators_ for sessions to use, but to have an OrtAllocator available in
-  // shared_ort_allocators_ that can be used outside of a session we need to additionally wrap that in an
-  // OrtAllocatorImplWrappingIAllocator. way too many levels of indirection but that is what it is currently.
-  // we need something to own that final OrtAllocator, so we add it to arena_ort_allocators_.
-  //
-  // TODO: we could split out the BFCArena implementation so it can be plugged into either an IAllocator
-  // or an OrtAllocator instance to reduce the indirection a little.
-  // with that we get an OrtAllocator from the EP, wrap it with an OrtAllocator based BFCArena, and wrap that with the
-  // IAllocatorImplWrappingOrtAllocator which takes ownership of the OrtAllocator and is in shared_allocators_.
-  //
-  // Alternatively we can disable wrapping an EP's allocator with a BFCArena and say the EP should provide the arena
-  // implementation directly. They're free to copy BFCArena as it came from TF originally. Or we could provide a
-  // cut-and-paste BFCArena implementation that works using the EP API that can be included in the EP source.
-  std::unordered_map<const OrtMemoryInfo*, std::unique_ptr<OrtAllocatorImplWrappingIAllocator>> arena_ort_allocators_;
-
 #if !defined(ORT_MINIMAL_BUILD)
   // register EPs that are built into the ORT binary so they can take part in AutoEP selection
   // added to ep_libraries

include/onnxruntime/core/session/onnxruntime_c_api.h

@@ -5829,7 +5829,7 @@ struct OrtApi {
    *
    * \since Version 1.23.
    */
-  ORT_API2_STATUS(Graph_GetNodes, const OrtGraph* graph,
+  ORT_API2_STATUS(Graph_GetNodes, _In_ const OrtGraph* graph,
                   _Out_writes_(num_nodes) const OrtNode** nodes, _In_ size_t num_nodes);
 
   /** \brief Get the parent node for the given graph, if any exists.
@@ -6469,6 +6469,17 @@ struct OrtApi {
                   _In_reads_(num_tensors) OrtValue* const* dst_tensors,
                   _In_opt_ OrtSyncStream* stream,
                   _In_ size_t num_tensors);
+
+  /** \brief Get ::OrtModelMetadata from an ::OrtGraph
+   *
+   * \param[in] graph The OrtGraph instance.
+   * \param[out] out Newly created ::OrtModelMetadata. Must be freed using OrtApi::ReleaseModelMetadata.
+   *
+   * \snippet{doc} snippets.dox OrtStatus Return Value
+   *
+   * \since Version 1.23.
+   */
+  ORT_API2_STATUS(Graph_GetModelMetadata, _In_ const OrtGraph* graph, _Outptr_ OrtModelMetadata** out);
 };
 
 /*