1873039 - WebGPU: ASTC formats are not supported

Status: RESOLVED FIXED

(Core :: Graphics: WebGPU, defect, P2)

Description

[filip]

User Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/120.0.0.0 Safari/537.36 Edg/120.0.0.0

Steps to reproduce:

Called GPUDevice.createTexture with a format of 'astc-8x8-unorm'

Actual results:

Uncaught TypeError: GPUDevice.createTexture: 'astc-8x8-unorm' (value of 'format' member of GPUTextureDescriptor) is not a valid value for enumeration GPUTextureFormat.

Expected results:

ASTC formats supported if texture-compression-astc was available and enabled.

Pretty sure ASTC being missing from here is the issue: https://github.com/mozilla/gecko-dev/blob/master/dom/webidl/WebGPU.webidl#L368

Comment 1

[filip]

I would consider this a blocker for Mac as well: https://bugzilla.mozilla.org/show_bug.cgi?id=1839761, since Apple Silicon app often rely on ASTC to share code with iOS due to using the same general architecture.

Comment 2

[Erich Gubler [:ErichDonGubler] (he/him)]

filip: RE: iOS: Unless something has changed very recently, the Firefox app on iOS currently uses WebKit for its browser engine, and not Gecko, due to . That specific platform's support is not something we will be able to directly resolve until that does change. 🫤

I do hope it will change soon, though! Mozilla is very interested in using Gecko on iOS.🤞🏻💦

Comment 3

[filip]

Ah right I was so into this that I forgot haha. This does apply to Apple Silicon Macs though!

Comment 4

[filip]

Is there any chance of fixing this so that GPUAdapter.features (https://developer.mozilla.org/en-US/docs/Web/API/GPUAdapter/features) doesn't return texture-compression-astc on Firefox while we're waiting for this to be fixed?

Comment 5

[Erich Gubler [:ErichDonGubler] (he/him)]

Attachment: Bug 1873039 - feat(webgpu): expose ASTC texture support r=#webgpu-reviewers

Comment 6

[Erich Gubler [:ErichDonGubler] (he/him)]

I'm not sure if there's further work beyond exposing the texture format via IDL, but I just made a patch that does so. CI run: try:0743f7bf5083

🤞🏻

Comment 7

[Erich Gubler [:ErichDonGubler] (he/him)]

filip: If you download one of the Firefox builds from ☝🏻 the above try push (filtered view for convenience), is this issue resolved while running it? I don't see any significant CI regressions[^1], so I think we might be able to just land the current patch.

[^1]: Some new CTS tests fail, but that's because some tests that were expecting to fail are now succeeding unexpectedly. 😅 We're using the wrong error mechanism in webgpu:api,validation,capability_checks,features,texture_formats:canvas_configuration_view_formats:* (expected to use a TypeError when the feature isn't enabled, but we currently return a validation error).

Comment 8

[Erich Gubler [:ErichDonGubler] (he/him)]

Attachment: Bug 1873039 - test(webgpu): reset contradictory `expected` outcomes in `backlog` from `try:0743f7bf508344a54d1f415d285ce68bc80a0726` r=#webgpu-reviewers

Comment 9

[Erich Gubler [:ErichDonGubler] (he/him)]

Attachment: Bug 1873039 - test(webgpu): reset contradictory `expected` outcomes in `backlog` from `try:c7c67882a5ee24aa234d26f680c77eb4e6340d9e` r=#webgpu-reviewers

Comment 10

[Phabricator Automation]

Comment on attachment 9423172 [details]
Bug 1873039 - test(webgpu): reset contradictory expected outcomes in backlog from try:c7c67882a5ee24aa234d26f680c77eb4e6340d9e r=#webgpu-reviewers

Revision D221311 was moved to bug 1917059. Setting attachment 9423172 [details] to obsolete.

Comment 11

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/347d2ffd901d
feat(webgpu): expose ASTC texture support r=webgpu-reviewers,webidl,teoxoy,smaug
https://hg.mozilla.org/integration/autoland/rev/845b7034b36b
test(webgpu): reset contradictory `expected` outcomes in `backlog` from `try:0743f7bf508344a54d1f415d285ce68bc80a0726` r=webgpu-reviewers,teoxoy

Comment 12

[tszentpeteri]

https://hg.mozilla.org/mozilla-central/rev/347d2ffd901d
https://hg.mozilla.org/mozilla-central/rev/845b7034b36b

Comment 13

[filip]

(In reply to Erich Gubler [:ErichDonGubler] from comment #7)

filip: If you download one of the Firefox builds from ☝🏻 the above try push (filtered view for convenience), is this issue resolved while running it? I don't see any significant CI regressions[^1], so I think we might be able to just land the current patch.

[^1]: Some new CTS tests fail, but that's because some tests that were expecting to fail are now succeeding unexpectedly. 😅 We're using the wrong error mechanism in webgpu:api,validation,capability_checks,features,texture_formats:canvas_configuration_view_formats:* (expected to use a TypeError when the feature isn't enabled, but we currently return a validation error).

Sorry for the delay, just downloaded the build and can confirm it seems to work!

Comment 14

[Iulian Moraru]

Backed out 4 changesets (bug 1917059, bug 1873039) for causing conflicts when trying to backout Bug 1917102.

• [Push that was backed out(https://treeherder.mozilla.org/jobs?repo=autoland&resultStatus=testfailed%2Cbusted%2Cexception%2Crunnable&revision=6cf53afd32693ef6f9bfff063c88df0580ad299e)

• Backout link

Comment 15

[Iulian Moraru]

Backout merged to central: https://hg.mozilla.org/mozilla-central/rev/8dc2e21d4955

Comment 16

[Erich Gubler [:ErichDonGubler] (he/him)]

I've rebased onto a corrected (soft dep.) patch stack for bug 1917102; hopefully re-landing is uneventful! 🤞🏻

Comment 17

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/1f36edd33503
feat(webgpu): expose ASTC texture support r=webgpu-reviewers,webidl,teoxoy,smaug
https://hg.mozilla.org/integration/autoland/rev/0c3150a96c14
test(webgpu): reset contradictory `expected` outcomes in `backlog` from `try:0743f7bf508344a54d1f415d285ce68bc80a0726` r=webgpu-reviewers,teoxoy

Comment 18

[Cristian Tuns]

https://hg.mozilla.org/mozilla-central/rev/1f36edd33503
https://hg.mozilla.org/mozilla-central/rev/0c3150a96c14