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
They present portability problems since D3D11/12 can't sample from them, and their use cases are very niche. We could add these back in the future if someone has a real-world need for them.
The text was updated successfully, but these errors were encountered:
We got some feedback where this is format is useful outside of sampling - if these formats can be used (within reason) as render targets and shader read resources, it might be worth keeping. Probably needs a giant asterisk in the docs, but we can figure that out after beta.
I'd prefer to keep the UINT formats if we can: I'm using SDL_GPU_TEXTUREFORMAT_R8_UINT quite a bit for emulating paletted surfaces, and while I can't use them with samplers any more, it works fine as a shader read resource instead.
That being said, I (a) haven't tested this with anything other than the Vulkan backend yet, and (b) can relatively easily work around it if they're missing. But logically I'm dealing with a texture with UINT properties, and either emulating that by un-normalising a UNORM version, or using something else (putting the data in a buffer) isn't quite as clean.
That's pretty similar to the other use case we had - alright, we'll try to make sure to document this as one of those formats that's useful outside of just sampling, probably as part of #217.
They present portability problems since D3D11/12 can't sample from them, and their use cases are very niche. We could add these back in the future if someone has a real-world need for them.
The text was updated successfully, but these errors were encountered: