Bug 239377 – [WebGPU] Implement hardware limits

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

[Myles C. Maxfield]

[WebGPU] Implement hardware limits

[Myles C. Maxfield]

Created attachment 457684 [details]
Patch

[Myles C. Maxfield]

Created attachment 457686 [details]
Patch

[Myles C. Maxfield]

Created attachment 457720 [details]
Patch

[Myles C. Maxfield]

Created attachment 457722 [details]
Patch

[Darin Adler]

Comment on attachment 457722 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=457722&action=review

> Source/WebGPU/WebGPU/Adapter.h:45
> +    static Ref<Adapter> create(id<MTLDevice> device, Instance& instance, WGPULimits&& limits)

There is no benefit to passing a structure full of scalars with an rvalue reference and calling WTFMove. Just const& is fine for this kind of object, unless this is generic code where it might be something with non-scalars in it, or if we think we’ll be evolving it some day.

> Source/WebGPU/WebGPU/HardwareLimits.mm:41
> +static WGPULimits apple3()

constexpr

> Source/WebGPU/WebGPU/HardwareLimits.mm:75
> +static WGPULimits apple4()

constexpr

> Source/WebGPU/WebGPU/HardwareLimits.mm:109
> +static WGPULimits apple5()

constexpr

> Source/WebGPU/WebGPU/HardwareLimits.mm:144
> +static WGPULimits apple6()

constexpr

> Source/WebGPU/WebGPU/HardwareLimits.mm:178
> +static WGPULimits apple7()

constexpr

> Source/WebGPU/WebGPU/HardwareLimits.mm:338
> +static bool checkLimits(const WGPULimits& limits)

This could be constexpr

> Source/WebGPU/WebGPU/HardwareLimits.mm:386
> +bool isValid(const WGPULimits& limits)

This could be constexpr

> Source/WebGPU/WebGPU/HardwareLimits.mm:391
> +bool anyLimitIsBetterThan(const WGPULimits& target, const WGPULimits& reference)

This could be constexpr

[Myles C. Maxfield]

Committed r292940 (249705@trunk): <https://commits.webkit.org/249705@trunk>

[Radar WebKit Bug Importer]

<rdar://problem/91842461>

[Ryosuke Niwa]

This patch broke builds:
https://build.webkit.org/#/builders/12/builds/11560

/Volumes/Data/worker/bigsur-debug/build/Source/WebGPU/WebGPU/HardwareLimits.mm:338:23: error: constexpr function never produces a constant expression [-Winvalid-constexpr]

[Myles C. Maxfield]

(In reply to Ryosuke Niwa from comment #8)
> This patch broke builds:
> https://build.webkit.org/#/builders/12/builds/11560
> 
> /Volumes/Data/worker/bigsur-debug/build/Source/WebGPU/WebGPU/HardwareLimits.
> mm:338:23: error: constexpr function never produces a constant expression
> [-Winvalid-constexpr]

Trying to reproduce now

[Myles C. Maxfield]

(In reply to Myles C. Maxfield from comment #9)
> (In reply to Ryosuke Niwa from comment #8)
> > This patch broke builds:
> > https://build.webkit.org/#/builders/12/builds/11560
> > 
> > /Volumes/Data/worker/bigsur-debug/build/Source/WebGPU/WebGPU/HardwareLimits.
> > mm:338:23: error: constexpr function never produces a constant expression
> > [-Winvalid-constexpr]
> 
> Trying to reproduce now

I can't reproduce this (possibly because I have a different SDK installed than the one bot uses), but I can see by inspection that the error is correct - that function is marked as constexpr but will never produce a constant expression. I'll commit a prospective fix.

[Myles C. Maxfield]

(In reply to Myles C. Maxfield from comment #10)
> (In reply to Myles C. Maxfield from comment #9)
> > (In reply to Ryosuke Niwa from comment #8)
> > > This patch broke builds:
> > > https://build.webkit.org/#/builders/12/builds/11560
> > > 
> > > /Volumes/Data/worker/bigsur-debug/build/Source/WebGPU/WebGPU/HardwareLimits.
> > > mm:338:23: error: constexpr function never produces a constant expression
> > > [-Winvalid-constexpr]
> > 
> > Trying to reproduce now
> 
> I can't reproduce this (possibly because I have a different SDK installed
> than the one bot uses), but I can see by inspection that the error is
> correct - that function is marked as constexpr but will never produce a
> constant expression. I'll commit a prospective fix.

https://commits.webkit.org/r292941