base_name + '_WebGPUCanvas2DDrawImage',
Peter McNeeleyyou forgot to rename!!!
Done
Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
Code-Review | +1 |
Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
Code-Review | +1 |
// This is either 'rgba8unorm' (android) or 'bgra8unorm' on mac
nit: mac and most or all other platforms
// Only if we neeed bgra do we need to request the feature
I didn't think this feature would be available on all bgra8unorm platforms we test, but I guess it is?
Test could just use rgba8unorm to be safe but up to you since apparently this works in practice.
(however for CTS reftests we would want to always use rgba8unorm, or have coverage for both)
Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
// This is either 'rgba8unorm' (android) or 'bgra8unorm' on mac
nit: mac and most or all other platforms
I am about to change linux 😊
So i just left it as the the things that we know for sure.
I am also not certain where d3d really falls here.
Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
// Only if we neeed bgra do we need to request the feature
I didn't think this feature would be available on all bgra8unorm platforms we test, but I guess it is?
Test could just use rgba8unorm to be safe but up to you since apparently this works in practice.
(however for CTS reftests we would want to always use rgba8unorm, or have coverage for both)
The test asks for the preferred canvas format. This should be RGBA for Android and in the future linux. Everywhere else we need this feature for this test to work.
// This is either 'rgba8unorm' (android) or 'bgra8unorm' on mac
Peter McNeeleynit: mac and most or all other platforms
I am about to change linux 😊
So i just left it as the the things that we know for sure.
I am also not certain where d3d really falls here.
Acknowledged
// Only if we neeed bgra do we need to request the feature
Peter McNeeleyI didn't think this feature would be available on all bgra8unorm platforms we test, but I guess it is?
Test could just use rgba8unorm to be safe but up to you since apparently this works in practice.
(however for CTS reftests we would want to always use rgba8unorm, or have coverage for both)
The test asks for the preferred canvas format. This should be RGBA for Android and in the future linux. Everywhere else we need this feature for this test to work.
Right, I'm suggesting that the test could just use `rgba8unorm` instead of the preferred format. That's always allowed (and always supports storage), just maybe less efficient. It's the simplest way to use storage to write to a canvas on all platforms without any conditionals.
Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
// Only if we neeed bgra do we need to request the feature
Peter McNeeleyI didn't think this feature would be available on all bgra8unorm platforms we test, but I guess it is?
Test could just use rgba8unorm to be safe but up to you since apparently this works in practice.
(however for CTS reftests we would want to always use rgba8unorm, or have coverage for both)
Kai NinomiyaThe test asks for the preferred canvas format. This should be RGBA for Android and in the future linux. Everywhere else we need this feature for this test to work.
Right, I'm suggesting that the test could just use `rgba8unorm` instead of the preferred format. That's always allowed (and always supports storage), just maybe less efficient. It's the simplest way to use storage to write to a canvas on all platforms without any conditionals.
Ah yes I didnt think about that? I think you are correct that this could be better.
Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
// Only if we neeed bgra do we need to request the feature
Peter McNeeleyI didn't think this feature would be available on all bgra8unorm platforms we test, but I guess it is?
Test could just use rgba8unorm to be safe but up to you since apparently this works in practice.
(however for CTS reftests we would want to always use rgba8unorm, or have coverage for both)
Kai NinomiyaThe test asks for the preferred canvas format. This should be RGBA for Android and in the future linux. Everywhere else we need this feature for this test to work.
Peter McNeeleyRight, I'm suggesting that the test could just use `rgba8unorm` instead of the preferred format. That's always allowed (and always supports storage), just maybe less efficient. It's the simplest way to use storage to write to a canvas on all platforms without any conditionals.
Ah yes I didnt think about that? I think you are correct that this could be better.
Done
Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
New webgpu pixel test that uses pixel storage in compute
This pixel test acts as a 'smoke test' for WebGPU compute
Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |