Update

[ghstack-poisoned]

Author: JCNTH

backends/webgpu/runtime/WebGPUCompat.h

@@ -10,8 +10,6 @@
 
 #include <webgpu/webgpu.h>
 
-#include <cstdint>
-
 #if defined(__EMSCRIPTEN__)
 #include <emscripten/emscripten.h>
 #else
@@ -21,14 +19,9 @@
 
 namespace executorch::backends::webgpu {
 
-// Caller's instance must enable TimedWaitAny; returns the WaitAny status.
-inline WGPUWaitStatus webgpu_wait(WGPUInstance instance, WGPUFuture future) {
-  WGPUFutureWaitInfo info = {};
-  info.future = future;
-  return wgpuInstanceWaitAny(instance, 1, &info, UINT64_MAX);
-}
-
-// Blocking event-loop pump for the output-map readback (browser: JS yield).
+// Make progress on pending WebGPU callbacks; callers loop until their done flag
+// is set. Native (Dawn): pump the event queue + brief yield (no busy-spin).
+// Browser: yield to the JS event loop.
 inline void webgpu_poll(WGPUInstance instance) {
 #if defined(__EMSCRIPTEN__)
   (void)instance;

backends/webgpu/runtime/WebGPUDevice.cpp

@@ -22,10 +22,12 @@ namespace {
 
 struct AdapterResult {
   WGPUAdapter adapter = nullptr;
+  bool done = false;
 };
 
 struct DeviceResult {
   WGPUDevice device = nullptr;
+  bool done = false;
 };
 
 void on_adapter_request(
@@ -45,6 +47,7 @@ void on_adapter_request(
         static_cast<int>(message.length),
         message.data);
   }
+  result->done = true;
 }
 
 void on_device_request(
@@ -64,6 +67,7 @@ void on_device_request(
         static_cast<int>(message.length),
         message.data);
   }
+  result->done = true;
 }
 
 void on_device_error(
@@ -85,36 +89,34 @@ void on_device_error(
 WebGPUContext create_webgpu_context() {
   WebGPUContext ctx;
 
-  // TimedWaitAny lets webgpu_wait() block on futures via wgpuInstanceWaitAny.
-  WGPUInstanceDescriptor instance_desc = {};
-#if defined(__EMSCRIPTEN__)
-  instance_desc.capabilities.timedWaitAnyEnable = true;
-  instance_desc.capabilities.timedWaitAnyMaxCount = 1;
-#else
-  WGPUInstanceFeatureName features[1] = {WGPUInstanceFeatureName_TimedWaitAny};
-  instance_desc.requiredFeatureCount = 1;
-  instance_desc.requiredFeatures = features;
-#endif
-  ctx.instance = wgpuCreateInstance(&instance_desc);
+  ctx.instance = wgpuCreateInstance(nullptr);
   if (!ctx.instance) {
     throw std::runtime_error("Failed to create WebGPU instance");
   }
 
+  // Request adapter using AllowSpontaneous mode (fires during
+  // wgpuInstanceProcessEvents or any other API call).
   AdapterResult adapter_result;
   WGPURequestAdapterCallbackInfo adapter_cb = {};
-  adapter_cb.mode = WGPUCallbackMode_WaitAnyOnly;
+  adapter_cb.mode = WGPUCallbackMode_AllowSpontaneous;
   adapter_cb.callback = on_adapter_request;
   adapter_cb.userdata1 = &adapter_result;
 
-  // No backend pin or forced fallback; Dawn auto-selects the adapter.
+  // Release Dawn has no bundled fallback adapter; pick the platform backend
+  // (Metal on Apple, Vulkan elsewhere -- SwiftShader via VK_ICD_FILENAMES).
   WGPURequestAdapterOptions adapter_opts = {};
-  adapter_opts.powerPreference = WGPUPowerPreference_HighPerformance;
+#if defined(__APPLE__)
+  adapter_opts.backendType = WGPUBackendType_Metal;
+#else
+  adapter_opts.backendType = WGPUBackendType_Vulkan;
+#endif
   adapter_opts.forceFallbackAdapter = false;
-  WGPUWaitStatus adapter_wait = webgpu_wait(
-      ctx.instance,
-      wgpuInstanceRequestAdapter(ctx.instance, &adapter_opts, adapter_cb));
+  wgpuInstanceRequestAdapter(ctx.instance, &adapter_opts, adapter_cb);
+  while (!adapter_result.done) {
+    webgpu_poll(ctx.instance);
+  }
 
-  if (adapter_wait != WGPUWaitStatus_Success || !adapter_result.adapter) {
+  if (!adapter_result.adapter) {
     wgpuInstanceRelease(ctx.instance);
     ctx.instance = nullptr;
     throw std::runtime_error(
@@ -126,7 +128,7 @@ WebGPUContext create_webgpu_context() {
   // Request device
   DeviceResult device_result;
   WGPURequestDeviceCallbackInfo device_cb = {};
-  device_cb.mode = WGPUCallbackMode_WaitAnyOnly;
+  device_cb.mode = WGPUCallbackMode_AllowSpontaneous;
   device_cb.callback = on_device_request;
   device_cb.userdata1 = &device_result;
 
@@ -139,11 +141,12 @@ WebGPUContext create_webgpu_context() {
   }
   device_desc.uncapturedErrorCallbackInfo.callback = on_device_error;
 
-  WGPUWaitStatus device_wait = webgpu_wait(
-      ctx.instance,
-      wgpuAdapterRequestDevice(ctx.adapter, &device_desc, device_cb));
+  wgpuAdapterRequestDevice(ctx.adapter, &device_desc, device_cb);
+  while (!device_result.done) {
+    webgpu_poll(ctx.instance);
+  }
 
-  if (device_wait != WGPUWaitStatus_Success || !device_result.device) {
+  if (!device_result.device) {
     wgpuAdapterRelease(ctx.adapter);
     wgpuInstanceRelease(ctx.instance);
     ctx.adapter = nullptr;

backends/webgpu/runtime/WebGPUGraph.cpp

@@ -619,7 +619,6 @@ void WebGPUGraph::copy_outputs(std::vector<std::pair<void*, size_t>>& outputs) {
         cb_info);
   }
 
-  // WaitAny returns pre-queue on CI Dawn+SwiftShader here; pump until mapped.
   bool all_mapped = false;
   while (!all_mapped) {
     webgpu_poll(instance_);