21:58 <cphealy> alyssa: did you ever figure out what the issue was with those glmark2-es2 tests that got way slower and you said it was doing CPU operations instead of GPU?

22:16 <HayashiEsme[m]> o/ I may or may not have made a rookie mistake which meant I ended up talking to myself for a day (Hi I'm Esme, if the name's recognisable I asked a question to Alyssa at ELC/OSS about contributing so here I am trying to be more helpful than effort -- hopefully)

22:17 <robclark> cphealy: alyssa probably meant CPU limited instead of GPU limited? Maybe `perf record` gives a hint?

22:18 <HayashiEsme[m]> I was working on #3679 (the issue re: refactoring and eliminating the use of magic numbers for gpu_ids) -- but I'll clarify that another time (there seems to be different implementations for gpu_id between the kernel and mesa) -- when I stumbled on a typo in the panfrost docs I made a patch for. I figured I'd bring it up here https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13296

23:00 <alyssa> cphealy: no, I did not. would be curious about the bisect

23:34 <cphealy> ack I'll give it a try then.