From be4aef056977c123322272fc152ec313c4cef336 Mon Sep 17 00:00:00 2001
From: Jie Chen <jie.a.chen@intel.com>
Date: Thu, 26 Mar 2026 08:59:29 +0800
Subject: [PATCH] Fix WebGPU buffer segment offset alignment

Align maxStorageBufferBindingSize down to the nearest multiple of
minStorageBufferOffsetAlignment after querying device limits. This
ensures that when large buffers are split into segments, each
segment's byte offset satisfies WebGPU's bind group offset alignment
requirement (typically 256 bytes).
---
 onnxruntime/core/providers/webgpu/webgpu_context.cc | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/onnxruntime/core/providers/webgpu/webgpu_context.cc b/onnxruntime/core/providers/webgpu/webgpu_context.cc
index c61d5826cb885..ec20bf2fdbdfb 100644
--- a/onnxruntime/core/providers/webgpu/webgpu_context.cc
+++ b/onnxruntime/core/providers/webgpu/webgpu_context.cc
@@ -124,6 +124,12 @@ void WebGpuContext::Initialize(const WebGpuContextConfig& config) {
     device_queue_ = device_.GetQueue();
     // cache device limits
     ORT_ENFORCE(Device().GetLimits(&device_limits_));
+    // Align maxStorageBufferBindingSize down to minStorageBufferOffsetAlignment so that
+    // buffer segment offsets are always properly aligned for WebGPU bind group creation.
+    if (device_limits_.minStorageBufferOffsetAlignment > 0) {
+      device_limits_.maxStorageBufferBindingSize -=
+          (device_limits_.maxStorageBufferBindingSize % device_limits_.minStorageBufferOffsetAlignment);
+    }
     // cache device features
     wgpu::SupportedFeatures supported_features;
     Device().GetFeatures(&supported_features);