[webkit-reviews] review denied: [Bug 238710] [WebGPU] Implement Texture view format compatibility : [Attachment 456504] Patch
bugzilla-daemon at webkit.org
bugzilla-daemon at webkit.org
Sun Apr 3 23:08:26 PDT 2022
Myles C. Maxfield <mmaxfield at apple.com> has denied Myles C. Maxfield
<mmaxfield at apple.com>'s request for review:
Bug 238710: [WebGPU] Implement Texture view format compatibility
https://bugs.webkit.org/show_bug.cgi?id=238710
Attachment 456504: Patch
https://bugs.webkit.org/attachment.cgi?id=456504&action=review
--- Comment #4 from Myles C. Maxfield <mmaxfield at apple.com> ---
Comment on attachment 456504
--> https://bugs.webkit.org/attachment.cgi?id=456504
Patch
View in context: https://bugs.webkit.org/attachment.cgi?id=456504&action=review
>> Source/WebGPU/WebGPU/Texture.mm:1119
>> + return std::nullopt;
>
> There appears to be an inconsistency with the logic in this function. If
`WGPUTextureFormat_ASTC4x4Unorm` returns `WGPUTextureFormat_ASTC4x4Unorm`,
surely these should return themselves? What am I missing?
Yeah, I see what you mean.
Originally, this function was intended to create equivalence classes via
normalization; e.g. RGBA8Unorm and RGBA8UnormSrgb should both be members of the
same equivalence class, and this function returns a representative member of
that class (sort of like how union set works). ASTC4x4Unorm and
ASTC4x4UnormSrgb are in the same equivalence class. I was originally thinking
that formats like R8Unorm wouldn't be part of any equivalence class, so they
would return nullopt here, but I can also understand where you're coming from
about saying R8Unorm should be in an equivalence class of 1. I'll change the
implementation to do that.
>> Source/WebGPU/WebGPU/Texture.mm:1243
>> + if (canonicalizedFormat1 != canonicalizedFormat1)
>
> Comparison to self?!
♂️
More information about the webkit-reviews
mailing list