00:44 karolherbst: okay.. this a lot better now
00:56 karolherbst: ... iris flushes more often now :'(
00:58 karolherbst: ... the mi stuff is like super expensive in iris
00:58 karolherbst: 3.7% is just spending time in the mi part
01:00 karolherbst: mhhh are queries reusable? mhhh
01:04 karolherbst: maybe I jsut got unlucky.. seems other runs are better
09:13 mareko: can somebody please review this NIR MR? https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39743
17:39 robclark: karolherbst: https://gitlab.freedesktop.org/robclark/mesa/-/commits/rusticl/profiling .. re-worked to add PIPE_QUERY_TIMESTAMP_RAW (but not rebased on your latest)
17:41 karolherbst: robclark: yeah.. you probably want to rebase, because I simplified quite a bit of the code
17:44 robclark: ok.. but not sure I'll have time today
17:44 robclark: anyways, at least that gives you an idea what I had in mind.. I guess it shouldn't be too hard to add iris
17:45 karolherbst: yeah.. I have some thoughts how dealing with different query types can be simplified