#asahi-gpu · 2023-05-21 00:04
alyssacan't measure a performance hit from forcing all textures bindless#asahi-gpu · 2023-05-15 21:55
alyssa>:)#asahi-gpu · 2023-05-15 21:29
alyssaif we were all writing c++ or rust compilers it *might* be different but I'm not convinced#asahi-gpu · 2023-05-15 21:29
alyssaprobably almost nothing if we're using C#asahi-gpu · 2023-05-15 21:29
alyssait's not clear what, if any, code could be shared between the SSA RAs#asahi-gpu · 2023-05-15 21:28
alyssaand very not great for huge complexity and specificity to the ISA in question#asahi-gpu · 2023-05-15 21:28
alyssawhich is really great for generated code quality + low compile-time#asahi-gpu · 2023-05-15 21:28
alyssathe valve compilers are all SSA-based RA#asahi-gpu · 2023-05-15 21:28
alyssause#asahi-gpu · 2023-05-15 21:28
alyssawe do have a common graph colouring engine (util/ra), which ~~dinky~~ other compilers us#asahi-gpu · 2023-05-15 21:26
alyssaWe've waffled about trying to share RAs in Mesa but... different GPU ISAs are radically different in their RA requirements that it's not clear if that can be done efficiently#asahi-gpu · 2023-05-15 21:26
alyssanot a dumb question at all#asahi-gpu · 2023-05-15 21:26
alyssa(Intel has a source dump somewhere but I have enough self respect not to look at it)#asahi-gpu · 2023-05-15 21:25
alyssaNo idea about LLVM, since only AMD upstreams their LLVM compiler#asahi-gpu · 2023-05-15 21:25
alyssaChaosPrincess: In mesa, yes#asahi-gpu · 2023-05-15 21:20
alyssaDitto for live range splitting in the register allocator.#asahi-gpu · 2023-05-15 20:34
alyssaMaybe we want to schedule a code walkthrough stream so we can do review synchronously and I can fix any issues as they're pointed out in real time?#asahi-gpu · 2023-05-15 20:33
alyssalina: We need a plan for getting MSAA reviewed+merged#asahi-gpu · 2023-05-12 13:09
alyssashrugs#asahi-gpu · 2023-05-12 13:09
alyssashrugs#asahi-gpu · 2023-05-12 12:18
alyssa(Or force the flag on with a 1 line patch on top if that's easier)#asahi-gpu · 2023-05-12 12:17
alyssa(Run your entire desktop with that debug flag. If you do gfx dev and use a login manager, don't. :p)#asahi-gpu · 2023-05-12 12:17
alyssaand also review it#asahi-gpu · 2023-05-12 12:17
alyssamarcan: give https://gitlab.freedesktop.org/asahi/mesa/-/merge_requests/73 a try#asahi-gpu · 2023-05-11 20:55
alyssa(Not urgent by any means. This is just the next item on the "help chip away at the MSAA" mountain. Thank you in advance ^^)#asahi-gpu · 2023-05-11 20:55
alyssa(I think the agx_tilebuffer.c logic is potentially even more busted than it seems on the surface. So maybe lots more tests would be a good thing here...)#asahi-gpu · 2023-05-11 20:54
alyssaYou know the drill ^.^#asahi-gpu · 2023-05-11 20:54
alyssalina: Next time you're looking for something to do, https://gitlab.freedesktop.org/alyssa/mesa/-/commit/8a3acf6eaa3a0abb3dcf7a223714e65842699469#asahi-gpu · 2023-05-07 18:44
alyssaIt all has to get done eventually, heh :p#asahi-gpu · 2023-05-07 18:44
alyssaUp to you :)#asahi-gpu · 2023-05-07 15:22
alyssaMSAA is a big project, hopefully if we both chip away at it we can get through :-D#asahi-gpu · 2023-05-07 15:11
alyssaThese are both needed for MSAA. They seem to work but per the commit message need tests. If you don't have any spicy todos (rust rebase etc) might be nice, I would appreciate those commits getting finished and merged ^_^#asahi-gpu · 2023-05-07 15:10
alyssahttps://gitlab.freedesktop.org/alyssa/mesa/-/commit/6fb2a51b0539fa555112681b0fae46508660755f#asahi-gpu · 2023-05-07 15:10
alyssahttps://gitlab.freedesktop.org/alyssa/mesa/-/commit/23072edf1644a471828c8e1dc6331f940fae37fb#asahi-gpu · 2023-05-07 15:09
alyssalina: IDK if you have plans for this week's streams, but any chance you're interested in an ail sidequest?#asahi-gpu · 2023-05-06 15:13
alyssaI don't see any reason it wouldn't also pass on main#asahi-gpu · 2023-05-06 15:13
alyssalina: Yes, that's in dEQP-GLES31 and all the tests pass on next#asahi-gpu · 2023-04-26 18:03
alyssathat's in GLES3.0 in your CTS, it's in GLES3.1 in mine because I haven't updated in years :p#asahi-gpu · 2023-04-26 18:03
alyssaeh, yeah, right#asahi-gpu · 2023-04-26 17:40
alyssa2. Shrug#asahi-gpu · 2023-04-26 17:39
alyssa1. I don't know exactly how eMRT interacts with partial renders. I'm trying not to think about eMRT until I've finished up everything else in gles3.1#asahi-gpu · 2023-04-26 17:39
alyssa^^#asahi-gpu · 2023-04-26 17:39
alyssaAlso I already had lunch, get some sleep %_%#asahi-gpu · 2023-04-26 17:38
alyssaSounds like a neutral thing ^^#asahi-gpu · 2023-04-26 15:16
alyssa(which is to say, about 37% on a good day)#asahi-gpu · 2023-04-26 15:06
alyssawould not change the level of interesting i am#asahi-gpu · 2023-04-26 15:06
alyssainterested#asahi-gpu · 2023-04-26 15:06
alyssaif you can find a control register bit that you can toggle to make them pass it would make me very interesting#asahi-gpu · 2023-04-26 15:06
alyssathen a bunch of dEQP-GLES3 cases will fail (listed in the commit message)#asahi-gpu · 2023-04-26 15:05
alyssalina: If you revert bf3027c3916 ("mesa/st: Normalize wrap modes for seamless cubes")