Backmerging with Msft commits

cmake/CMakeLists.txt

@@ -180,7 +180,6 @@ option(onnxruntime_ENABLE_TRAINING_OPS "Include training operators but no traini
 option(onnxruntime_ENABLE_TRAINING_E2E_TESTS "Enable training end-to-end tests." OFF)
 option(onnxruntime_ENABLE_CPU_FP16_OPS "Build with advanced instruction sets" ON)
 option(onnxruntime_USE_NCCL "Build with NCCL support" OFF)
-option(onnxruntime_USE_MPI "Build with MPI support" OFF)
 
 # WebAssembly options
 option(onnxruntime_BUILD_WEBASSEMBLY_STATIC_LIB "Enable this option to create WebAssembly static library" OFF)
@@ -1267,6 +1266,7 @@ function(onnxruntime_set_compile_flags target_name)
       # Unsupported by Clang 18 yet.
       list(REMOVE_ITEM ORT_HIP_WARNING_FLAGS -Wno-dangling-reference)
 
+      list(REMOVE_ITEM ORT_HIP_WARNING_FLAGS -Wno-interference-size)
       # float16.h:90:12: error: ‘tmp’ is used uninitialized
       list(APPEND ORT_HIP_WARNING_FLAGS -Wno-uninitialized)
       list(APPEND ORT_HIP_WARNING_FLAGS -Wno-deprecated-copy)
@@ -1650,29 +1650,6 @@ if (onnxruntime_ENABLE_DLPACK)
 endif()
 
 if (UNIX OR onnxruntime_USE_NCCL)
-  # MPI is INDEPENDENT of NCCL for now. You can build NCLL without MPI and launch multi-GPU with your own launcher.
-  if (onnxruntime_USE_MPI)
-    if (EXISTS "${onnxruntime_MPI_HOME}")
-      set(MPI_HOME "${onnxruntime_MPI_HOME}")
-    elseif (EXISTS "/bert_ort/openmpi")
-      set(MPI_HOME "/bert_ort/openmpi")
-    endif()
-    find_package(MPI)
-
-    if (MPI_CXX_FOUND)
-      message( STATUS "MPI Version: ${MPI_CXX_VERSION}")
-      message( STATUS "MPI (include: ${MPI_CXX_INCLUDE_DIRS}, library: ${MPI_CXX_LIBRARIES})" )
-      mark_as_advanced(MPI_CXX_INCLUDE_DIRS MPI_CXX_LIBRARIES)
-      list(APPEND onnxruntime_EXTERNAL_LIBRARIES ${MPI_CXX_LIBRARIES} ${MPI_CXX_LINK_FLAGS})
-    else ()
-      message(
-            FATAL_ERROR
-            "MPI is not found. Please define onnxruntime_MPI_HOME to specify the path of MPI. Otherwise, NCCL will be disabled."
-            "or you can remove --use_mpi from build args to disable MPI."
-          )
-    endif()
-  endif()
-
   # Find NCCL
   if (onnxruntime_USE_NCCL)
     if (onnxruntime_USE_CUDA)
@@ -1742,13 +1719,9 @@ if (UNIX OR onnxruntime_USE_NCCL)
   endif()
 else()
   set(onnxruntime_USE_NCCL OFF)
-  set(onnxruntime_USE_MPI OFF)
-  message( WARNING "MPI and NCCL are disabled because build is on Windows or USE_NCCL is set to OFF." )
+  message( WARNING "NCCL is disabled because build is on Windows or USE_NCCL is set to OFF." )
 endif()
 
-if (onnxruntime_USE_MPI)
-  add_definitions(-DUSE_MPI=1)
-endif()
 
 # Default version parts for Microsoft.AI.MachineLearning.dll, onnxruntime.dll, onnxruntime_providers_openvino.dll and onnxruntime_providers_shared.dll in non-ADO pipeline local builds
 set(VERSION_MAJOR_PART   0 CACHE STRING "First part of numeric file/product version.")

cmake/external/onnxruntime_external_deps.cmake

@@ -719,6 +719,11 @@ if (onnxruntime_USE_WEBGPU)
       # - (private) Remove hard-coded CMAKE_OSX_DEPLOYMENT_TARGET in Dawn's CMake files
       #   https://github.com/microsoft/onnxruntime/pull/23729
       #
+      # - (private) Reduce unsafe buffer usage warning in aligned_storage.h
+      #   https://github.com/microsoft/onnxruntime/pull/24308
+      #   The patch disables the UNSAFE_BUFFER_USAGE warning around the AlignedStorage struct in aligned_storage.h. This is done
+      #   by using TINT_BEGIN_DISABLE_WARNING and TINT_END_DISABLE_WARNING macros, which helps in warnings related to unsafe buffer usage
+      #   usage when compiling the code, making the build process cleaner and faster.
       #
       PATCH_COMMAND ${Patch_EXECUTABLE} --binary --ignore-whitespace -p1 < ${PROJECT_SOURCE_DIR}/patches/dawn/dawn.patch
       EXCLUDE_FROM_ALL

cmake/onnxruntime_framework.cmake

@@ -77,9 +77,6 @@ if (onnxruntime_ENABLE_TRAINING_OPS)
     onnxruntime_add_include_to_target(onnxruntime_framework Python::Module dlpack::dlpack)
   endif()
 endif()
-if (onnxruntime_USE_MPI)
-  target_include_directories(onnxruntime_framework PUBLIC ${MPI_CXX_INCLUDE_DIRS})
-endif()
 
 if (onnxruntime_ENABLE_ATEN)
   onnxruntime_add_include_to_target(onnxruntime_framework dlpack::dlpack)

cmake/onnxruntime_providers_cpu.cmake

@@ -206,7 +206,7 @@ if (onnxruntime_ENABLE_TRAINING)
     onnxruntime_add_include_to_target(onnxruntime_providers Python::Module)
   endif()
 
-  if (onnxruntime_USE_NCCL OR onnxruntime_USE_MPI)
+  if (onnxruntime_USE_NCCL)
     target_include_directories(onnxruntime_providers PUBLIC ${MPI_CXX_INCLUDE_DIRS})
   endif()
 endif()

cmake/onnxruntime_providers_cuda.cmake

@@ -252,10 +252,6 @@
       target_include_directories(${target} PRIVATE ${ORTTRAINING_ROOT} ${MPI_CXX_INCLUDE_DIRS})
     endif()
 
-    if(onnxruntime_USE_MPI)
-      target_link_libraries(${target} PRIVATE ${MPI_LIBRARIES} ${MPI_CXX_LINK_FLAGS})
-    endif()
-
     if (onnxruntime_USE_NCCL)
       target_include_directories(${target} PRIVATE ${NCCL_INCLUDE_DIRS})
       target_link_libraries(${target} PRIVATE ${NCCL_LIBRARIES})

cmake/onnxruntime_providers_rocm.cmake

@@ -158,9 +158,6 @@
 
   if (onnxruntime_ENABLE_TRAINING)
     target_include_directories(onnxruntime_providers_rocm PRIVATE ${ORTTRAINING_ROOT} ${CMAKE_CURRENT_BINARY_DIR}/amdgpu/orttraining ${MPI_CXX_INCLUDE_DIRS})
-    if(onnxruntime_USE_MPI)
-      target_link_libraries(onnxruntime_providers_rocm PRIVATE ${MPI_LIBRARIES} ${MPI_CXX_LINK_FLAGS})
-    endif()
 
     # RCCL is enabled by default for ROCM builds
     #if (onnxruntime_USE_NCCL)

cmake/patches/dawn/dawn.patch

@@ -36,3 +36,24 @@ index 5bfac41dcc..71a153daaa 100644
  void WGPUBufferImpl::Destroy() {
    emwgpuBufferDestroy(this);
    AbortPendingMap("Buffer was destroyed before mapping was resolved.");
+diff --git a/src/tint/utils/memory/aligned_storage.h b/src/tint/utils/memory/aligned_storage.h
+index c532c4fc38..19c950af4c 100644
+--- a/src/tint/utils/memory/aligned_storage.h
++++ b/src/tint/utils/memory/aligned_storage.h
+@@ -31,6 +31,9 @@
+ #include <cstddef>
+
+ #include "src/tint/utils/memory/bitcast.h"
++#include "src/tint/utils/macros/compiler.h"
++
++TINT_BEGIN_DISABLE_WARNING(UNSAFE_BUFFER_USAGE);
+
+ namespace tint {
+
+@@ -50,4 +53,6 @@ struct alignas(alignof(T)) AlignedStorage {
+
+ }  // namespace tint
+
++TINT_END_DISABLE_WARNING(UNSAFE_BUFFER_USAGE);
++
+ #endif  // SRC_TINT_UTILS_MEMORY_ALIGNED_STORAGE_H_

java/build.gradle

@@ -211,6 +211,7 @@ test {
 		'USE_TENSORRT',
 		'USE_QNN',
 		'USE_XNNPACK',
+		'USE_WEBGPU',
 		])
 	testLogging {
 		events "passed", "skipped", "failed"

java/src/main/java/ai/onnxruntime/OrtSession.java

@@ -1339,6 +1339,18 @@ public void addCoreML(Map<String, String> providerOptions) throws OrtException {
       addExecutionProvider(CoreMLProviderName, providerOptions);
     }
 
+    /**
+     * Adds WebGPU as an execution backend.
+     *
+     * @param providerOptions Configuration options for the WebGPU backend. Refer to the WebGPU
+     *     execution provider's documentation.
+     * @throws OrtException If there was an error in native code.
+     */
+    public void addWebGPU(Map<String, String> providerOptions) throws OrtException {
+      String webGpuProviderName = "WebGPU";
+      addExecutionProvider(webGpuProviderName, providerOptions);
+    }
+
     private native void setExecutionMode(long apiHandle, long nativeHandle, int mode)
         throws OrtException;
 

java/src/test/android/app/src/androidTest/java/ai/onnxruntime/example/javavalidator/SimpleTest.kt

@@ -52,6 +52,11 @@ class SimpleTest {
         runSigmoidModelTestImpl(1, OrtProvider.QNN)
     }
 
+    @Test
+    fun runSigmoidModelTestWEBGPU() {
+        runSigmoidModelTestImpl(1, OrtProvider.WEBGPU)
+    }
+
     @Throws(IOException::class)
     private fun readModel(fileName: String): ByteArray {
         return InstrumentationRegistry.getInstrumentation().context.assets.open(fileName)
@@ -90,6 +95,16 @@ class SimpleTest {
                     }
                 }
 
+                OrtProvider.WEBGPU -> {
+                    if (OrtEnvironment.getAvailableProviders().contains(OrtProvider.WEBGPU)) {
+                        val providerOptions = Collections.emptyMap<String, String>()
+                        opts.addWebGPU(providerOptions)
+                    } else {
+                        Log.println(Log.INFO, TAG, "NO WEBGPU EP available, skip the test")
+                        return
+                    }
+                }
+
                 OrtProvider.CPU -> {
                     // No additional configuration is needed for CPU
                 }

java/src/test/java/ai/onnxruntime/InferenceTest.java

@@ -749,6 +749,12 @@ public void testQNN() throws OrtException {
     runProvider(OrtProvider.QNN);
   }
 
+  @Test
+  @EnabledIfSystemProperty(named = "USE_WEBGPU", matches = "1")
+  public void testWEBGPU() throws OrtException {
+    runProvider(OrtProvider.WEBGPU);
+  }
+
   private void runProvider(OrtProvider provider) throws OrtException {
     EnumSet<OrtProvider> providers = OrtEnvironment.getAvailableProviders();
     assertTrue(providers.size() > 1);

js/common/lib/env.ts

@@ -41,16 +41,22 @@ export declare namespace Env {
     numThreads?: number;
 
     /**
-     * set or get a boolean value indicating whether to enable SIMD. If set to false, SIMD will be forcely disabled.
+     * set a value indicating whether to enable SIMD.
+     *
+     * ONNX Runtime will perform feature detection based on the value of this property. Specifically, when the value is
+     * set to:
+     * - `undefined`, `true` or `"fixed"`: will check availability of Fixed-width SIMD.
+     * - `"relaxed"`: will check availability of Relaxed SIMD.
+     * - `false`: will not perform SIMD feature checking.
+     *
+     * Setting this property does not make ONNX Runtime to switch to the corresponding runtime automatically. User need
+     * to set `wasmPaths` or `wasmBinary` property to load the corresponding runtime.
      *
      * This setting is available only when WebAssembly SIMD feature is available in current context.
      *
      * @defaultValue `true`
-     *
-     * @deprecated This property is deprecated. Since SIMD is supported by all major JavaScript engines, non-SIMD
-     * build is no longer provided. This property will be removed in future release.
      */
-    simd?: boolean;
+    simd?: boolean | 'fixed' | 'relaxed';
 
     /**
      * set or get a boolean value indicating whether to enable trace.

js/web/docs/webnn-operators.md

@@ -49,6 +49,7 @@ platforms. Check the [WebNN status](https://webmachinelearning.github.io/webnn-s
 | GlobalLpPool| ai.onnx(7+) | l2Pool2d | Only supports 4-D input, 'p' value is 2 |
 | Greater | ai.onnx(7-8, 9-12, 13+) | greater | |
 | GreaterOrEqual | ai.onnx(12-15, 16+) | greaterOrEqual | |
+| GroupQueryAttention | com.microsoft(1+) | add, cast, concat, constant, cumulativeSum, div, expand, lesser, matmul, reshape, scatterND, softmax, transpose, where | Only supports input total_sequence_length is constant and past_sequence_length of past kv equals to present_sequence_length of present kv. Does not support cos_cache and sin_cache inputs |
 | GRU | ai.onnx(7-13, 14-21, 22+) | gru | Only supports 'layout' == 0. 'clip' is not supported. The activation functions in 'activations' must be one of 'Relu', 'Tanh', 'Sigmoid'. Forward and backward activations must be the same if bidirectional. 'sequence_lens' if present should be constant with values equal to the first dimension length of input 'X' |
 | HardSigmoid | ai.onnx(7+) | hardSigmoid | |
 | HardSwish | ai.onnx(14+) | hardSwish | |
@@ -63,6 +64,7 @@ platforms. Check the [WebNN status](https://webmachinelearning.github.io/webnn-s
 | LRN | ai.onnx(7-12, 13+) | pad, averagePool2d, transpose, add, mul, pow, div | |
 | LSTM | ai.onnx(7-13, 14-21, 22+) | lstm | Only supports 'layout' == 0, 'input_forget' == 0. 'clip' is not supported. The activation functions in 'activations' must be one of 'Relu', 'Tanh', 'Sigmoid'. Forward and backward activations must be the same if bidirectional. 'sequence_lens' if present should be constant with values equal to the first dimension length of input 'X' |
 | MatMul | ai.onnx(7-8, 9-12, 13+) | matmul | |
+| MatMulNBits | com.microsoft(1+) | add, dequantizeLinear, matmul, reshape, transpose | Inputs 'B' and 'zero_points' (if present) should be constants, input 'g_idx' is not supported, only bits=4 is supported |
 | Max | ai.onnx(7, 8-11, 12, 13+) | max | |
 | MaxPool | ai.onnx(7, 8-9, 10, 11, 12+) | maxPool2d | Only supports 4-D input, 2-D 'kernel_shape', 'storage_order' != 1, one output |
 | Min | ai.onnx(7, 8-11, 12, 13+) | min | |

js/web/lib/backend-wasm.ts

@@ -17,12 +17,13 @@ export const initializeFlags = (): void => {
     env.wasm.initTimeout = 0;
   }
 
-  if (env.wasm.simd === false) {
+  const simd = env.wasm.simd;
+  if (typeof simd !== 'boolean' && simd !== undefined && simd !== 'fixed' && simd !== 'relaxed') {
     // eslint-disable-next-line no-console
     console.warn(
-      'Deprecated property "env.wasm.simd" is set to false. ' +
-        'non-SIMD build is no longer provided, and this setting will be ignored.',
+      `Property "env.wasm.simd" is set to unknown value "${simd}". Reset it to \`false\` and ignore SIMD feature checking.`,
     );
+    env.wasm.simd = false;
   }
 
   if (typeof env.wasm.proxy !== 'boolean') {

js/web/lib/wasm/wasm-factory.ts

@@ -64,6 +64,34 @@ const isSimdSupported = (): boolean => {
   }
 };
 
+const isRelaxedSimdSupported = (): boolean => {
+  try {
+    // Test for WebAssembly Relaxed SIMD capability (for both browsers and Node.js)
+    // This typed array is a WebAssembly program containing Relaxed SIMD instructions.
+
+    // The binary data is generated from the following code by wat2wasm:
+    // (module
+    //   (func (result v128)
+    //      i32.const 1
+    //      i8x16.splat
+    //      i32.const 2
+    //      i8x16.splat
+    //      i32.const 3
+    //      i8x16.splat
+    //      i32x4.relaxed_dot_i8x16_i7x16_add_s
+    //   )
+    //  )
+    return WebAssembly.validate(
+      new Uint8Array([
+        0, 97, 115, 109, 1, 0, 0, 0, 1, 5, 1, 96, 0, 1, 123, 3, 2, 1, 0, 10, 19, 1, 17, 0, 65, 1, 253, 15, 65, 2, 253,
+        15, 65, 3, 253, 15, 253, 147, 2, 11,
+      ]),
+    );
+  } catch (e) {
+    return false;
+  }
+};
+
 export const initializeWebAssembly = async (flags: Env.WebAssemblyFlags): Promise<void> => {
   if (initialized) {
     return Promise.resolve();
@@ -82,7 +110,14 @@ export const initializeWebAssembly = async (flags: Env.WebAssemblyFlags): Promis
   let numThreads = flags.numThreads!;
 
   // ensure SIMD is supported
-  if (!isSimdSupported()) {
+  if (flags.simd === false) {
+    // skip SIMD feature checking as it is disabled explicitly by user
+  } else if (flags.simd === 'relaxed') {
+    // check if relaxed SIMD is supported
+    if (!isRelaxedSimdSupported()) {
+      throw new Error('Relaxed WebAssembly SIMD is not supported in the current environment.');
+    }
+  } else if (!isSimdSupported()) {
     throw new Error('WebAssembly SIMD is not supported in the current environment.');
   }
 

js/web/package.json

@@ -68,6 +68,7 @@
   "browser": "dist/ort.min.js",
   "exports": {
     ".": {
+      "types": "./types.d.ts",
       "node": {
         "import": "./dist/ort.node.min.mjs",
         "require": "./dist/ort.node.min.js"
@@ -79,24 +80,28 @@
       "require": "./dist/ort.min.js"
     },
     "./all": {
+      "types": "./types.d.ts",
       "import": {
         "onnxruntime-web-use-extern-wasm": "./dist/ort.all.min.mjs",
         "default": "./dist/ort.all.bundle.min.mjs"
       },
       "require": "./dist/ort.all.min.js"
     },
     "./wasm": {
+      "types": "./types.d.ts",
       "import": {
         "onnxruntime-web-use-extern-wasm": "./dist/ort.wasm.min.mjs",
         "default": "./dist/ort.wasm.bundle.min.mjs"
       },
       "require": "./dist/ort.wasm.min.js"
     },
     "./webgl": {
+      "types": "./types.d.ts",
       "import": "./dist/ort.webgl.min.mjs",
       "require": "./dist/ort.webgl.min.js"
     },
     "./webgpu": {
+      "types": "./types.d.ts",
       "import": {
         "onnxruntime-web-use-extern-wasm": "./dist/ort.webgpu.min.mjs",
         "default": "./dist/ort.webgpu.bundle.min.mjs"

js/web/test/e2e/run.js

@@ -83,6 +83,12 @@ async function main() {
     await testExportsMain(PRESERVE, PACKAGES_TO_INSTALL);
   }
 
+  // perform type (typescript related) testing
+  {
+    const testTypeMain = require(path.join(TEST_E2E_RUN_FOLDER, './type/main'));
+    await testTypeMain(PRESERVE, PACKAGES_TO_INSTALL);
+  }
+
   // prepare .wasm files for path override testing
   prepareWasmPathOverrideFiles();
 

js/web/test/e2e/type/main.js

@@ -0,0 +1,23 @@
+// Copyright (c) Microsoft Corporation. All rights reserved.
+// Licensed under the MIT License.
+
+'use strict';
+
+const path = require('path');
+const { installOrtPackages, runShellCmd } = require('./utils');
+
+/**
+ * Entry point for type tests.
+ *
+ * @param {string[]} packagesToInstall
+ */
+module.exports = async function main(PRESERVE, PACKAGES_TO_INSTALL) {
+  console.log('Running type tests...');
+
+  // testcases/module-resolution
+  {
+    await installOrtPackages('module-resolution', PRESERVE, PACKAGES_TO_INSTALL);
+
+    await runShellCmd('npx tsc', { wd: path.join(__dirname, 'testcases', 'module-resolution') });
+  }
+};