Bug 227226 – BabylonJS Under water demo is slower than it should be on Intel

• Component: WebGL
• Status: RESOLVED FIXED

[Kimmo Kinnunen]

BabylonJS Under water demo is slower than it should be on Intel
https://playground.babylonjs.com/#LPTLZM

MacBookPro19,1 Intel / AMD 

WebGL on Metal on 10fps
WebGL on Metal off 30fps

Should be 60fps.

[Kimmo Kinnunen]

Related to rdar://79216506 but not strictly the same

[Radar WebKit Bug Importer]

<rdar://problem/79872484>

[Kyle Piddington]

Created attachment 432552 [details]
Patch

[EWS Watchlist]

Note that there are important steps to take when updating ANGLE. See https://trac.webkit.org/wiki/UpdatingANGLE

[Kenneth Russell]

Comment on attachment 432552 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=432552&action=review

> Source/ThirdParty/ANGLE/ChangeLog:6
> +	Remove fastMath restriction on Intel, restore explicit [[invariant]] tag to position, fragcoord 

Please fix indentation in ChangeLog.

> Source/ThirdParty/ANGLE/src/compiler/translator/TranslatorMetalDirect/EmitMetal.cpp:977
> +    (decl.isField() ? mInvariants.contains(decl.field()) : mInvariants.contains(decl.variable())) || (qualifier == TQualifier::EvqPosition || qualifier == TQualifier::EvqFragCoord);

Making gl_Position and gl_FragCoord implicitly invariant seems like a far-reaching change. If fastMath is re-enabled on Intel, do more applications than just this one need this implicit invariant qualifier on these built-in variables?

[Kyle Piddington]

Created attachment 432563 [details]
Patch

[Kenneth Russell]

Comment on attachment 432563 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=432563&action=review

Looks good to me. A few minor comments about the commit message. r+

> Source/ThirdParty/ANGLE/ChangeLog:7
> +        Removing fastmath in all scenarios leads to unnacceptable performance on integrated graphics

Consider adding periods at end of sentences.

> Source/ThirdParty/ANGLE/ChangeLog:8
> +        Currently, webgl conformance tests and the safrai tests don't have any invariance tests that show issue. 

safrai -> safari

show issue -> "show issues" or "show the issue"

[Kyle Piddington]

Created attachment 432704 [details]
Patch for landing

[EWS]

Committed r279466 (239322@main): <https://commits.webkit.org/239322@main>

All reviewed patches have been landed. Closing bug and clearing flags on attachment 432704 [details].

[Truitt Savell]

It looks like the changes in https://trac.webkit.org/changeset/279466/webkit

broke 4 Webgl tests

tracking in https://bugs.webkit.org/show_bug.cgi?id=227596

[Kyle Piddington]

*** Bug 226841 has been marked as a duplicate of this bug. ***