Jan

An idea for the API design. Add a function to `wgpu.extras` that can set instance flags so global `_the_instance` keeps hold of these. Maybe `set_instance_extras()` with a python friendly kwargs...

as far as I can tell the instance extras work, like you can request debug mode from fxc and vulkan for sure (shows up in renderdoc). While dxc works for...

I think I have finally figured out what caused the crash and can finally continue to get it working. for reference (mostly myself if this happens again next update). Creating...

I do still have some issues with dx12 remaining regarding various features:  Vulkan and even OpenGL seem to work. No crashes but incorrect values for push constants and indirect...

> Is this ready to go from your end? yeah, perhaps a changelog entry and some kind of notice that Dx12 with native features can produce incorrect numbers in compute...

@almarklein should be good to go now! for the remaining errors( https://github.com/pygfx/wgpu-py/pull/718#issuecomment-3006617272), the stencil warnings are due to how the test is setup (hijacking the logger) and Dx12 outputs additional...

Would it make sense to include a pre commit config that has the ruff steps included? Since ruff changes quite frequently, not pinning a version and using latest does sneak...

there was a release upstream with one Bugfix that sounds similar https://github.com/gfx-rs/wgpu/releases/tag/v24.0.3

consider compiling prebuilt wheels with AOT enabled, especially the windows native variant. It should be s much better experience, and you only need to support Arc dGPU. As PVC is...

> [...] building from source with AOT as described [here](https://github.com/intel/intel-extension-for-pytorch/blob/xpu-master/docs/tutorials/AOT.md). found an error in the linked doc. [first sentence](https://github.com/intel/intel-extension-for-pytorch/blob/5a4f4050e48a7a1ca7517a35afb6e0049c6b2719/docs/tutorials/AOT.md?plain=1#L20) claims the prebuilt wheels have AOT enabled or both GPU devices....