04:00 airlied[d]: karolherbst[d]: the perf bench dies for me due to not having workgroup scope support
04:01 airlied[d]: I've pushed out a enw branch to the MR which does a single pass, and drop dim_gran
04:03 orowith2os[d]: chikuwad[d]: It would work under a container using an emulated arch.
04:04 orowith2os[d]: So most of userspace would be using nvk or software rendering.
07:07 karolherbst[d]: airlied[d]: normal coop support
07:08 karolherbst[d]: we do the lowering internally already
07:08 karolherbst[d]: and I wanted to replace it with your pass
07:09 airlied[d]: Ah okay will need your code then
07:48 karolherbst[d]: just advertize 16x32x16 on radv or something
07:48 karolherbst[d]: that's all we do in NVK as well, advertize sizes the hw doesn't support
07:48 karolherbst[d]: (and lower it to smaller ones)
07:55 karolherbst[d]: Mhhh I found a very annoying bug in the rust-analyzer integration with meson... It doesn't detect platform features like `target_has_atomic="32"` and so auto completion in IDEs won't resolve types like AtomicU32...
10:06 airlied[d]: Okay reproduced it, will take a closer look next week
10:09 karolherbst[d]: cool, thanks