Bug 199012 – REGRESSION: ( r246394 ) webgpu/whlsl-buffer-fragment.html and webgpu/whlsl-buffer-vertex.html are failing

• Status: RESOLVED
• Resolution: FIXED
• Component: WebGPU

[Truitt Savell]

The changes in r246394 introduced two broken webgpu/ tests.

webgpu/whlsl-buffer-fragment.html 
webgpu/whlsl-buffer-vertex.html

History:
https://webkit-test-results.webkit.org/dashboards/flakiness_dashboard.html#showAllRuns=true&tests=webgpu%2Fwhlsl-buffer-fragment.html%20webgpu%2Fwhlsl-buffer-vertex.html

Diff:
https://build.webkit.org/results/Apple%20High%20Sierra%20Release%20WK2%20(Tests)/r246586%20(11821)/webgpu/whlsl-buffer-fragment-diffs.html
https://build.webkit.org/results/Apple%20High%20Sierra%20Release%20WK2%20(Tests)/r246586%20(11821)/webgpu/whlsl-buffer-vertex-diffs.html

[Truitt Savell]

Marked tests as failing in https://trac.webkit.org/changeset/246591/webkit

[Radar WebKit Bug Importer]

<rdar://problem/51900620>

[Saam Barati]

Why not mark this as flaky? They're only failing on high Sierra AFAIK

[Ryan Haddad]

(In reply to Saam Barati from comment #3)
> Why not mark this as flaky? They're only failing on high Sierra AFAIK
Instead of flaky, we should mark them as failing specifically for High Sierra. Took care of that in https://trac.webkit.org/changeset/246619/webkit

[Alexey Proskuryakov]

Is it High Sierra specific or GPU specific? I believe we use older hardware there.

[Myles C. Maxfield]

I’m in the process of checking now.

[Myles C. Maxfield]

This bug is both High-Sierra AND GPU-specific. The test passes on this GPU on Mojave, and the test passes on High Sierra on other hardware.

[Myles C. Maxfield]

Created attachment 372611 [details]
WIP

[Myles C. Maxfield]

Assigning this to Justin to bring this across the finish line.

[Myles C. Maxfield]

Created attachment 372613 [details]
WIP

[Myles C. Maxfield]

Created attachment 372618 [details]
Needs test harness update

[Justin Fan]

Comment on attachment 372618 [details]
Needs test harness update

View in context: https://bugs.webkit.org/attachment.cgi?id=372618&action=review

> Source/WebCore/platform/graphics/gpu/cocoa/GPUDeviceMetal.mm:44
> +    if (deviceName.startsWith("Intel(R) Iris(TM) Graphics "))

I'm sure we have more than one generation of MBP with Intel Iris graphics (Iris 5100, Iris 6100, Iris 540). Will this only affect the one(s)we're concerned about?

> Source/WebCore/platform/graphics/gpu/cocoa/GPUDeviceMetal.mm:48
> +    return false;

Is this line ever reached?

> LayoutTests/webgpu/bind-groups.html:10
> +    getBasicDevice().then(function(device) {

promise_test callback should return the promise created by .then().

> LayoutTests/webgpu/pipeline-layouts.html:25
> +    getBasicDevice().then(async function(device) {

ditto.

[Myles C. Maxfield]

(In reply to Justin Fan from comment #12)
> Comment on attachment 372618 [details]
> Needs test harness update
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=372618&action=review
> 
> > Source/WebCore/platform/graphics/gpu/cocoa/GPUDeviceMetal.mm:44
> > +    if (deviceName.startsWith("Intel(R) Iris(TM) Graphics "))
> 
> I'm sure we have more than one generation of MBP with Intel Iris graphics
> (Iris 5100, Iris 6100, Iris 540). Will this only affect the one(s)we're
> concerned about?

AFAIK this is a driver problem, and all those GPUs share a driver. At least 5000 and 6100 share a driver. So I think we want to disallow the whole line.

> 
> > Source/WebCore/platform/graphics/gpu/cocoa/GPUDeviceMetal.mm:48
> > +    return false;
> 
> Is this line ever reached?
> 
> > LayoutTests/webgpu/bind-groups.html:10
> > +    getBasicDevice().then(function(device) {
> 
> promise_test callback should return the promise created by .then().
> 
> > LayoutTests/webgpu/pipeline-layouts.html:25
> > +    getBasicDevice().then(async function(device) {
> 
> ditto.

[Myles C. Maxfield]

Created attachment 372652 [details]
Patch

[Myles C. Maxfield]

Created attachment 372816 [details]
fails tests

[Myles C. Maxfield]

Created attachment 372877 [details]
Patch

[Saam Barati]

Comment on attachment 372877 [details]
Patch

r=me

[Myles C. Maxfield]

Created attachment 372889 [details]
Patch for committing

[EWS Watchlist]

Comment on attachment 372889 [details]
Patch for committing

Attachment 372889 [details] did not pass mac-wk2-ews (mac-wk2):
Output: https://webkit-queues.webkit.org/results/12576755

New failing tests:
webgpu/whlsl-read-modify-write.html
webgpu/whlsl-use-undefined-variable.html
webgpu/whlsl-use-undefined-variable-2.html

[EWS Watchlist]

Created attachment 372896 [details]
Archive of layout-test-results from ews107 for mac-highsierra-wk2

The attached test failures were seen while running run-webkit-tests on the mac-wk2-ews.
Bot: ews107  Port: mac-highsierra-wk2  Platform: Mac OS X 10.13.6

[Myles C. Maxfield]

Created attachment 372897 [details]
Patch for committing

[Myles C. Maxfield]

Created attachment 372898 [details]
Patch for committing

[Myles C. Maxfield]

Committed r246846: <https://trac.webkit.org/changeset/246846>