Nicolas Silva

Results 152 comments of Nicolas Silva

An possible approach is to handle it outside of wgpu-core, near the code that deals with error scopes since it's a central place where errors are processed. That means all...

My plan is now to detect device lost errors when converting the hal device error type to the wgpu-core one. It will require a bit of plumbing but has the...

> My plan is now to detect device lost errors when converting the hal device error type to the wgpu-core one. Scratch that, it's way too invasive.

This is accurate. We need to find a good way to inspect every error, detect when it's a device loss and react accordingly (mark the device as lost so that...

Another solution that was discussed is to do something similar to what I described in my previous comment, but in the backends instead of in wgpu-core. The backends have simpler...

This blocks a few of the official webgpu samples, see https://bugzilla.mozilla.org/show_bug.cgi?id=1806659#c4

I like the general idea. How would a user of the API know when a task is done?

Sounds good to me. Should the new constraint be (for each `Thing`) `max[Thing]PerShaderStage

Alright, I preferred your earlier suggestion but in the short term Firefox will restrict `max[Thing]PerShaderStage` to small values, since it's simple and should not impact users too much. > It...

> > (Firefox): Stop providing the IDs during resource creation and do ID mapping in the gecko-side bindings. > > How exactly would that look like? Would that be something...