ChanServ changed the topic of #zink to: official development channel for the mesa3d zink driver || https://docs.mesa3d.org/drivers/zink.html
ced117 has quit [Ping timeout: 480 seconds]
ced117 has joined #zink
ced117 has quit [Ping timeout: 480 seconds]
ced117 has joined #zink
pendingchaos_ has joined #zink
pendingchaos has quit [Ping timeout: 480 seconds]
pendingchaos_ has quit [Read error: Connection reset by peer]
pendingchaos has joined #zink
<zmike> baryluk: that antichamber game is trippy
<baryluk> :) I got all benchmarks finish, and the analysis script fixed, as I broken it somehow. But all benchmarks looks green compared to previous run.
<zmike> hooray
<baryluk> +500%
<zmike> #bigfps
<baryluk> I do not see results for mesh exploder thingy. Checking now why.
<baryluk> Also +500% cumulative improvement
<baryluk> heaven: about 5% up
<baryluk> there are some other weird results. but they do have very high variance. Probably my GPU is acting up again, and not maintaining gpu clocks properly.
<baryluk> actually now. according to my logs, gpu clocks were 1050MHz nonstop entire time every second.
<baryluk> I see. some vram issue still.
<baryluk> i.e. glmark2 after running for 60 minutes (i.e. each of 22 test 11 times), was reporting 1800MB VRAM. But in runs from June was peaking at 277MB.
<baryluk> also valley extreme, and superposition extreme, crashed. the non-extreme settings did finish and good performance.
<baryluk> will rerun manually and see what is with that.
<zmike> hmmm
<zmike> probably post on the 21.3 ticket about the vram leak with a repro method
<zmike> the slow ones are the worst
<baryluk> malloc(): unsorted double linked list corrupted
<baryluk> ok. time for adress sanitizer / valgrind.
<baryluk> ;D
<zmike> :/
<baryluk> malloc(): unaligned tcache chunk detected
<baryluk> in some other
<baryluk> geex mesh shader exploder, doesn't crash my gpu anymore on kernel 5.14. Yay.
<baryluk> But it does not render correctly compared to radeonsi.
<zmike> progress!
<zmike> 🤔
<baryluk> running both at the same time, shows that the pink quads are correct, but the yellow lines are out of phase, by about 90 degree. some minus sign, or some corrdinate axis flipped somwhere somehow