You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Metal documentation specifies that sample locations are not standardized on devices that don't have programmable sample locations: https://developer.apple.com/documentation/metal/mtldevice/2866120-getdefaultsamplepositions#discussion
However, all devices in the current Metal Feature Set Tables support programmable sample locations! According to archived copies of that document, it was only ever unsupported on iOS before 11, and on macOS before 10.13. We require both, so we're in the clear.
Also, my tests pass on M1 on macOS 13.2.1. All other desktop Mac hardware supported D3D11 so would also have the standard sample pattern anyway.
In conclusion, I think we can finally formally decide we are using the standard sample pattern.
The text was updated successfully, but these errors were encountered:
Reopening to track a small editorial change - it's not currently clear that we actually require WebGPU implementations to use the standard sample pattern.
I tightened some of our existing tests to test the exact pattern of multisampling samples: gpuweb/cts#2466
https://learn.microsoft.com/en-us/windows/win32/api/d3d11/ne-d3d11-d3d11_standard_multisample_quality_levels
However, all devices in the current Metal Feature Set Tables support programmable sample locations! According to archived copies of that document, it was only ever unsupported on iOS before 11, and on macOS before 10.13. We require both, so we're in the clear.
Also, my tests pass on M1 on macOS 13.2.1. All other desktop Mac hardware supported D3D11 so would also have the standard sample pattern anyway.
In conclusion, I think we can finally formally decide we are using the standard sample pattern.
The text was updated successfully, but these errors were encountered: