#asahi-gpu · 2024-02-20 15:50
rosefromthedead:D#asahi-gpu · 2024-02-20 15:14
rosefromthedeaddamn, didn't know alyssa was secretly my algorithms teacher#asahi-gpu · 2023-12-04 10:39
rosefromthedeadoh neat, ty#asahi-gpu · 2023-12-02 15:27
rosefromthedeadi have a wgpu program wanting 8 maxPerStageDescriptorStorageBuffers but agxv only advertises 4. the metal backend for wgpu hardcodes 31 for this limit, so is 4 a driver limitation or a guess?#asahi-gpu · 2023-12-02 15:26
rosefromthedeadhow were the limits in agxv_limits.h determined?#asahi-gpu · 2023-10-24 23:31
rosefromthedeadalyssa: nice#asahi-gpu · 2023-10-24 21:05
rosefromthedeadwell, my mental models were off in that i thought there existed a non bonkers option :)#asahi-gpu · 2023-10-24 20:56
rosefromthedeadlmk how it goes, will be fun to see how far off my mental models are#asahi-gpu · 2023-10-24 20:56
rosefromthedeadlol nw#asahi-gpu · 2023-09-18 12:38
rosefromthedeadso that's where i am atm#asahi-gpu · 2023-09-18 12:38
rosefromthedeadand i have a feeling it might be easier to wrap my head around if i remove a) by making queries have a list of writers instead of a single writer but my first glance at doing that was a bit scary#asahi-gpu · 2023-09-18 12:36
rosefromthedeadso the challenge isn't in making it work but rather in making the code still be intelligible once it does#asahi-gpu · 2023-09-18 12:35
rosefromthedeadi think the two main confusions for me are a) the current query code assumes only one batch is involved in a query (ref agx_query.c:179) and b) all currently supported query results seem to be written into a bo by the gpu whereas timestamps are written into a struct by the kernel#asahi-gpu · 2023-09-18 12:32
rosefromthedeadi looked once again at timestamp queries and it's doable as soon as i regain the brain power i once had#asahi-gpu · 2023-09-03 09:03
rosefromthedeaddid drm_asahi_get_time land? i can't see it anywhere#asahi-gpu · 2023-08-31 15:01
rosefromthedeadnoted#asahi-gpu · 2023-08-31 14:57
rosefromthedeadgood to know though :eyes:#asahi-gpu · 2023-08-31 14:56
rosefromthedeadmb i meant wrt the timer stuff, because i said i'd do that but i haven't got round to it#asahi-gpu · 2023-08-31 14:49
rosefromthedeadhmm am i needed#asahi-gpu · 2023-08-25 14:22
rosefromthedeadyeah, iirc bevy just invested in Deref abuse and gave up on "doing things right" because it can make apis so much nicer#asahi-gpu · 2023-08-25 13:37
rosefromthedeads/deref/use std::ops::Deref/#asahi-gpu · 2023-08-25 13:37
rosefromthedeadah, the old "you shouldn't deref for polymorphism even though it makes everyone's life easier"#asahi-gpu · 2023-08-17 14:07
rosefromthedead"you're in for a world of pain :3"#asahi-gpu · 2023-08-17 14:07
rosefromthedeadhahahah#asahi-gpu · 2023-08-13 22:30
rosefromthedeadanother key problem is that i think i deleted it#asahi-gpu · 2023-08-13 22:30
rosefromthedeaderm#asahi-gpu · 2023-08-13 22:28
rosefromthedeadwhat i'm aware of that needs fixing in my timestamp impl is 1. the time travelling and 2. it assumes there's only one live query at a time because i'm lazy and didn't want to figure out hash tables or whatever#asahi-gpu · 2023-08-13 22:26
rosefromthedeadgood, it turned out to be a different challenge to what i was expecting but it's fun#asahi-gpu · 2023-08-13 22:25
rosefromthedeadworst case i'll try to get round to publishing what i've got#asahi-gpu · 2023-08-13 22:24
rosefromthedeadmy side project queue rotated recently but gsoc ends in 2 weeks so i might stumble back into mesa#asahi-gpu · 2023-08-13 22:22
rosefromthedeadikr#asahi-gpu · 2023-08-13 22:22
rosefromthedeadhahaha thank you#asahi-gpu · 2023-08-13 22:22
rosefromthedeadyeah i got as far as a hacky local branch that sometimes reports jobs taking negative time#asahi-gpu · 2023-08-13 22:20
rosefromthedeadpraying for another copy paste-based good first issue :)#asahi-gpu · 2023-07-29 14:07
rosefromthedeadoh no#asahi-gpu · 2023-07-23 10:48
rosefromthedeadnot sure how conscious i was when i wrote it#asahi-gpu · 2023-07-23 10:48
rosefromthedeadit's also possible that my code is just mixing up values from different batches or something#asahi-gpu · 2023-07-23 10:47
rosefromthedeadonly on startup of tinywl for some reason, -3150us, -94us, and everything positive from there#asahi-gpu · 2023-07-23 10:42
rosefromthedeadm1 mini#asahi-gpu · 2023-07-23 10:37
rosefromthedeadi'll check for an offset#asahi-gpu · 2023-07-23 10:37
rosefromthedeadlol#asahi-gpu · 2023-07-23 10:37
rosefromthedeadhmm, still getting some negative durations, but i think less now?#asahi-gpu · 2023-07-23 10:36
rosefromthedeadblimey#asahi-gpu · 2023-07-23 10:33
rosefromthedeadwill try with an extra isb, i already have one before the read but ig it would make sense to need ones either side?#asahi-gpu · 2023-07-23 10:33
rosefromthedeadmarcan: not on purpose, just a normal asahi install#asahi-gpu · 2023-07-22 20:16
rosefromthedeadi wonder whether we could (/should) #ifdef the timestamp ioctl behind !ARM64 and use inline asm otherwise#asahi-gpu · 2023-07-22 20:15
rosefromthedeadgonna write that off as "wow, these machines really *are* quick!" for now#asahi-gpu · 2023-07-22 20:13
rosefromthedeadbut only slightly#asahi-gpu · 2023-07-22 20:13
rosefromthedeadhrm, i'm reading timestamps from CNTVCT_EL0 before submitting jobs and the finish timestamp on some of the jobs is ending up earlier than when they were submitted lol#asahi-gpu · 2023-07-21 13:11
rosefromthedeadty, last time i tried to run deqp i gave up for fear of going insane