Uh oh!
There was an error while loading. Please reload this page.
difftest: cleanup wgpu runner - #587
Conversation
3b37201 to
08f6d09Compare603efd7 to
408a0bbCompare39dd493 to
376f532Compare408a0bb to
c0cb9e0Compare81640a0 to
f1250b9Compare
eddyb
left a comment
There was a problem hiding this comment.
LGTM modulo a few small things, anything else blocking this?
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
| let (sender, receiver) = futures::channel::oneshot::channel(); | ||
| buffer_slice.map_async(wgpu::MapMode::Read, move |res| { | ||
| let _ = sender.send(res); | ||
| }); | ||
| device.poll(wgpu::PollType::wait_indefinitely())?; | ||
| block_on(receiver) | ||
| .context("mapping canceled")? | ||
| .context("mapping failed")?; |
There was a problem hiding this comment.
Does the wgpu example runner do it this way? Or does it just unwrap the Result in the closure passed to map_async?
There was a problem hiding this comment.
It actually does, that said, I'm not a huge fan of the download code AI wrote so let me refactor it a little. encoder.map_buffer_on_submit() is way cleaner as a map_async() replacement.
Also noticed it was only ever writing out the first writable buffer, if you had two the second one would just be ignored... I'll just append all the buffers. But I also don't want to spend too much time on this and rather completely refactor the difftest framework yet again, will propose my idea on zulip today.
eddyb
left a comment
There was a problem hiding this comment.
We should've landed this already - feel free to bump wgpu to 30 if it's straightforward.
… shader passthrough for rust-gpu
6c892fd to
dbbdea8CompareFirestar99
commented
Jul 13, 2026
Updated it to wgpu 30, will merge when ready. I also found the branch |
Uh oh!
There was an error while loading. Please reload this page.
Extracted from #334