regression: web gpu working incorrectly

[sigmaSd]

deno run  --import-map https://raw.githubusercontent.com/denoland/webgpu-examples/main/deno.json https://
raw.githubusercontent.com/denoland/webgpu-examples/main/hello-compute/mod.ts

outputs [ "1", "4", "3", "295" ] instead of ["0", "2", "7", "55"]

This works with version 1.43.0 but not 1.44.0 @crowlKats suspects #23684

meta:
Linux 1E-55-11-07-05-2E 6.9.11-200.fc40.x86_64 #1 SMP PREEMPT_DYNAMIC Thu Jul 25 18:17:34 UTC 2024 x86_64 GNU/Linux
Brand Raw: Intel(R) Core(TM) i3-3110M CPU @ 2.40GHz

[sigmaSd]

Here is a self contained example

async function addWithWebGPU(a: number, b: number) {
  if (!navigator.gpu) {
    throw new Error("WebGPU not supported on this browser.");
  }

  console.log("Requesting adapter...");
  const adapter = await navigator.gpu.requestAdapter();
  if (!adapter) {
    throw new Error("No appropriate GPUAdapter found.");
  }

  console.log("Requesting device...");
  const device = await adapter.requestDevice();

  console.log("Creating buffer...");
  const bufferSize = 3 * Float32Array.BYTES_PER_ELEMENT;
  const buffer = device.createBuffer({
    size: bufferSize,
    usage: GPUBufferUsage.STORAGE | GPUBufferUsage.COPY_SRC |
      GPUBufferUsage.COPY_DST,
  });

  console.log("Writing input data...");
  const input = new Float32Array([a, b, 0]);
  device.queue.writeBuffer(buffer, 0, input);

  console.log("Creating shader module...");
  const shaderModule = device.createShaderModule({
    code: `
            @group(0) @binding(0) var<storage, read_write> data: array<f32>;

            @compute @workgroup_size(1)
            fn main(@builtin(global_invocation_id) global_id : vec3<u32>) {
                data[2] = data[0] + data[1];
            }
        `,
  });

  console.log("Creating compute pipeline...");
  const pipeline = device.createComputePipeline({
    layout: "auto",
    compute: {
      module: shaderModule,
      entryPoint: "main",
    },
  });

  console.log("Creating bind group...");
  const bindGroup = device.createBindGroup({
    layout: pipeline.getBindGroupLayout(0),
    entries: [{ binding: 0, resource: { buffer: buffer } }],
  });

  console.log("Encoding commands...");
  const commandEncoder = device.createCommandEncoder();
  const passEncoder = commandEncoder.beginComputePass();
  passEncoder.setPipeline(pipeline);
  passEncoder.setBindGroup(0, bindGroup);
  passEncoder.dispatchWorkgroups(1);
  passEncoder.end();

  console.log("Creating read buffer...");
  const gpuReadBuffer = device.createBuffer({
    size: bufferSize,
    usage: GPUBufferUsage.COPY_DST | GPUBufferUsage.MAP_READ,
  });

  console.log("Encoding copy command...");
  commandEncoder.copyBufferToBuffer(buffer, 0, gpuReadBuffer, 0, bufferSize);

  console.log("Submitting GPU commands...");
  const gpuCommands = commandEncoder.finish();
  device.queue.submit([gpuCommands]);

  console.log("Mapping read buffer...");
  await gpuReadBuffer.mapAsync(GPUMapMode.READ);
  const resultArrayBuffer = gpuReadBuffer.getMappedRange();
  const result = new Float32Array(resultArrayBuffer);

  console.log("Raw result array:", result);

  // Extract the result before unmapping
  const finalResult = result[2];

  gpuReadBuffer.unmap();

  console.log("Computation complete.");
  return finalResult;
}

// Usage
async function testAddition() {
  try {
    const result = await addWithWebGPU(5, 7);
    console.log(`5 + 7 = ${result}`);
  } catch (error) {
    console.error("An error occurred:", error);
  }
}

// Run the test
testAddition().then(() => console.log("Test completed")).catch((error) =>
  console.error("Test failed:", error)
);

it outputs 0 instead of 12

but this works on firefox nightly 131.0a1

[sigmaSd]

All the other examples work, it seems like the issue is specifically with compute

[sigmaSd]

If I run the rust wgpu compute example, I get

No numbers were provided, defaulting to [1, 2, 3, 4]
MESA-INTEL: warning: Ivy Bridge Vulkan support is incomplete
thread 'main' panicked at /home/mrcool/.cargo/registry/src/index.crates.io-6f17d22bba15001f/wgpu-22.1.0/src/backend/wgpu_core.rs:3411:5:
wgpu error: Validation Error

Caused by:
  In Device::create_compute_pipeline
    Internal error: The selected version doesn't support Features(DYNAMIC_ARRAY_SIZE)

[sigmaSd]

if I run with XDG_RUNTIME_DIR="" deno run -A a.ts (unsetting XDG_RUNTIME_DIR) it actually works

[sigmaSd]

the issue seem to be specifically with XDG_RUNTIME_DIR="/run/user/1000" if I put another value like "/tmp" it works

[sigmaSd]

the same happens in rust webgpu, so its a webgpu bug

[sigmaSd]

I guess the only thing deno can do better here is surface such errors

[sigmaSd]

wgpu now doesn't support incomplete vulkan implementation, that's why it stopped working, it works on xorg because for some reason its picking llvmpipe (vulkan cpu implemtation) as the adapter which workarounds this issue