Add bc4-r-unorm and astc-12x12-unorm brain images

[beaufortfrancois]

This PR adds the possibility for users to pick bc4-r-unorm and astc-12x12-unorm brain images in the "Volume Rendering - Texture 3D" sample.

[beaufortfrancois]

FYI @mehmetoguzderin

[beaufortfrancois]

@mwyrzykowski I've noticed using the bc4-r-unorm texture format in this sample is causing the canvas camera to be stuck somehow... Is this a known issue with Safari Tech Preview or am I holding it wrong?

bc4-safari.mov

[mwyrzykowski]

Ah looks like we were rejecting 3D compressed texture creation even when the extension was enabled, thanks for catching @beaufortfrancois

I suppose the sample could detect the validation error we generated but I addressed it with https://bugs.webkit.org/show_bug.cgi?id=291690:

Screen.Recording.2025-04-17.at.10.55.29.AM.mov

[kainino0x]

Ah looks like we were rejecting 3D compressed texture creation even when the extension was enabled, thanks for catching

@mwyrzykowski Is it firing the GPUDevice uncapturederror event? The samples are supposed to pop up a dialog if that happens.

[mwyrzykowski]

Ah looks like we were rejecting 3D compressed texture creation even when the extension was enabled, thanks for catching

@mwyrzykowski Is it firing the GPUDevice uncapturederror event? The samples are supposed to pop up a dialog if that happens.

@kainino0x it appears the related CTS tests https://gpuweb.github.io/cts/standalone/?q=webgpu:api,validation,error_scope:* and https://gpuweb.github.io/cts/standalone/?q=webgpu:api,operation,uncapturederror:* are passing in WebKit as the pattern from https://www.w3.org/TR/webgpu/#example-bf3af559 works in WebKit

But the method used in the samples:

  device.onuncapturederror = (ev) => {
    fail(`Uncaptured error:\n${ev.error.message}`);
  };

does not. Replacing that with:

    device.addEventListener('uncapturederror', (ev) => {
        fail(`Uncaptured error:\n${ev.error.message}`);
    });

results in:

[beaufortfrancois]

@mwyrzykowski My understanding is that both device.onuncapturederror = ... and device.addEventListener('uncapturederror', ... should work. Is this an STP issue?

[kainino0x]

Thank you for investigating, Mike! Filed CTS bug gpuweb/cts#4361

[beaufortfrancois]

@kainino0x I've added astc-12x12-unorm brain image as well in this PR.

[beaufortfrancois]

To illustrate the value of both formats, here's a Chromium DevTools network screenshot:

[mwyrzykowski]

@mwyrzykowski My understanding is that both device.onuncapturederror = ... and device.addEventListener('uncapturederror', ... should work. Is this an STP issue?

Yes, presumably so

[beaufortfrancois]

@mwyrzykowski My understanding is that both device.onuncapturederror = ... and device.addEventListener('uncapturederror', ... should work. Is this an STP issue?

Yes, presumably so

@mwyrzykowski I've tried https://red-triangle-webgpu.glitch.me/uncapturederror.html in Chrome, Firefox Nightly, and Safari Tech Preview.
Both Chrome and Firefox Nightly log "onuncapturederror" and "addEventListener("uncapturederror")" while Safari Tech Preview only logs "addEventListener("uncapturederror")"

[beaufortfrancois]

FYI whatwg/compression#34 shows that chrome won't have brotli support soon sadly.

…

On Tue, Apr 22, 2025, 03:00 Kai Ninomiya ***@***.***> wrote: ***@***.**** commented on this pull request. ------------------------------ In sample/volumeRenderingTexture3D/main.ts <#501 (comment)> : > const gui = new GUI(); +const brainImages = { + r8unorm: { + bytesPerBlock: 1, + blockLength: 1, + feature: undefined, + dataPath: + '../../assets/img/volume/t1_icbm_normal_1mm_pn0_rf0_180x216x180_uint8_1x1.bin.gz', update: Since we realized we need to use DecompressionStream, we are stuck with gzip (or deflate). — Reply to this email directly, view it on GitHub <#501 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAE243QRKTJLJPUYT4DO2VT22WIDDAVCNFSM6AAAAAB3KSX44WVHI2DSMVQWIX3LMV43YUDVNRWFEZLROVSXG5CSMV3GSZLXHMZDOOBSGQ4TEOBSG4> . You are receiving this because you modified the open/close state.Message ID: ***@***.***>

[kainino0x]

The concerns there seem to be about brotli compression, but we only need decompression.

[beaufortfrancois]

According to whatwg/compression#34 (comment), Safari has Brotli support to Compression Streams by using Apple's Compression Framework. See WebKit/WebKit@0d3451c

[mdanowar3]

This is the most important part of the best interest way