v148 · WebGPU · WGSL
Compute Benchmark
Run the same sum-of-squares compute shader two ways: the old manual 3D-to-linear conversion, and the Chrome 148 built-in global_invocation_index. Compare GPU execution time, shader line count, and correctness.
Checking WebGPU support…
Larger = longer GPU execution, clearer timing difference.
Manual index pre-148
Computes flat index by hand: x + y*W + z*W*H
—
GPU time (ms)
— extra WGSL lines
global_invocation_index Chrome 148
Built-in linear index; compiler handles the math
—
GPU time (ms)
— WGSL lines needed
Benchmark log
Click "Run benchmark" to start…
Shader comparison
// Before Chrome 148 — manual 3D→1D conversion
@compute @workgroup_size(64, 1, 1)
fn main(@builtin(global_invocation_id) id: vec3<u32>,
@builtin(num_workgroups) nwg: vec3<u32>) {
let W = nwg.x * 64u;
let H = nwg.y;
let flat = id.x + id.y * W + id.z * W * H; // ← boilerplate every shader
if (flat >= arrayLength(&data)) { return; }
result[flat] = data[flat] * data[flat];
}
// Chrome 148 — global_invocation_index built-in
@compute @workgroup_size(64, 1, 1)
fn main(@builtin(global_invocation_index) flat: u32) { // ← one line replaces four
if (flat >= arrayLength(&data)) { return; }
result[flat] = data[flat] * data[flat];
}references
implementation reference
Need the exact API surface, compatibility boundaries, errors, lifecycle, and source links? Read the matching gendn reference ↗