12:26 karolherbst: some "leak" https://gist.githubusercontent.com/karolherbst/f758cd896334c86d897c9c8c90901966/raw/4d5103c06d460db568e9562ec55a2cbda21221af/gistfile1.txt
13:24 sravn: For the ignorant - what is "dEQP-VK.compute.basic.ssbo_rw_single_invocation" - in terms of user visible functionality?
13:25 sravn: From karolherbst leak it is obvious an important milestone - so well-done is in place here.
13:27 mupuf: karolherbst: congrats for getting basic compute shaders working \o/
13:27 mupuf: NVK is a good name :)
13:47 karolherbst: mupuf: wasn't me though, I just wrote the code for parsing the push buffers :D
13:47 karolherbst: currently trying to figure how the trap handler works
13:47 mupuf: karolherbst: oh, who else is working on nvk?
13:47 karolherbst: last "CTS" run "new: Pass: 18927, Fail: 9366, Crash: 27836, Warn: 2, Skip: 1490189, Flake: 541, Duration: 50:28, Remaining: 0"
13:47 karolherbst: mupuf: dave and jason
13:48 mupuf: hehe, the usual suspects :D
13:48 karolherbst: :D
13:49 karolherbst: anyway.. debugging OpenGL bugs sucks in nouveau, so I am focusing on make those things easier
13:49 karolherbst: the other two are working out shader stuff
13:49 karolherbst: we already got images working as well
13:51 mupuf: well, good luck to you, guys!
13:52 karolherbst: thanks
14:31 karolherbst: mwk: by any chance, do you know (or was that even you) who wrote this one document on shader debugging/trap handler/breakpoint stuff?
14:31 karolherbst: trying to set it all up, but _something_ I am missing
14:39 mwk: I think I wrote something like this, yes
14:39 mwk: what document exactly?
14:42 karolherbst: I already found the URL: https://people.freedesktop.org/~imirkin/mwk-singlestep-nvc0.txt
14:42 karolherbst: but I am on turing and it seems like regs are different, but nvgpu does point to the correct ones
14:42 karolherbst: just something isn't working out alright and I am sure that's due to lack of understanding
14:46 mwk: ... well
14:46 mwk: I have very little memory of writing that thing
14:46 karolherbst: yeah... that's not an issue, I know I got it to work at some point, not quite sure what I am missing. But maybe it doesn't help if the kernel kills the channel the moment it gets a trap
14:54 karolherbst: maybe I'll try to use bpt trap first