06:02 Venemo: cheako: waste bandwidth on what?
06:03 Venemo: do you mean the captures?
06:03 Venemo: well, I have not seen this issue in any game that I have tried so far, and neither has other members in the team, so there must be something special about it
06:04 cheako: That's why u can't reproduce it.
06:17 Venemo: what's why I can't reproduce it?
14:55 cheako: If you think you are not doing it correctly, then it's possible to miss an example if only you had tried one more time.
14:58 cheako: I'm putting minimal effort into all my testing, if it is something special it's doing an above average job at acting like it's not.
15:07 cheako: pixelcluster: I heard in vulkan chat that renderpasses are fungible, one can be exchanged for another and even making more than one of each type is pointless. Does that hold in these drivers?
15:13 bnieuwenhuizen: cheako: yes
15:29 MrCooper: fun, just had a GPU reset, and now gnome-shell keeps printing "gnome-shell[10627]: amdgpu: The CS has been rejected (-125). Recreate the context", but somehow it's still working (writing this on it)
15:32 MrCooper: ah, PID 10627 is actually Xwayland (which is also used by hexchat a
15:32 MrCooper: I'm writing this on though)
16:40 Venemo: cheako: I've played hundreds of hours of various games on the Deck and haven't seen this issue yet. but I don't play the same games you play, so I assume there must be something special about the games you played
16:44 cheako: The action I'm doing is restarting the game... though it is remarkable you havn't hit one of the other triggers. I didn't look at the log files, but it ran for two days and I only caught a couple *that were different*... and each test was much faster than what it normally takes to restart the game, because of gfxrc.
16:47 cheako: It's not that a game doesn't trigger this, it's that a game hasn't yet triggered it.
17:06 Venemo: cheako: I'm happy that you found a way to consistently reproduce the problem on hardware that you and we both have, this helps with the investigation a lot, thank you.
17:58 cheako: If it were me, I'd modify gfxreplay to collect some data at some point in the trace... pick a createfence or something to hook a good spot in the "script"... Then when printing out the FPS check to see if it's different from previous tests, in C instead of with perl like I did, and if so print the debug info.
17:59 cheako: This is because the rest of the process is scripted and it takes days to run, so sitting there watching it is not an option.
18:00 Venemo: like I said, I have my hands full with other work right now but I will have some time for your issue in a couple of weeks. maybe pixelcluster can further help if he has time