#asahi-gpu · 2022-01-16 18:03
alyssabad dirty tracking maybe#asahi-gpu · 2022-01-16 18:03
alyssaoh but only when run in sequence, ugh#asahi-gpu · 2022-01-16 18:01
alyssaalso -bshading:shading=gouraud is broken was it always like that#asahi-gpu · 2022-01-16 18:01
alyssanote to self -btexture:texture-filter=mipmap is broken#asahi-gpu · 2022-01-16 18:01
alyssaI'm really proud of this wacky thing#asahi-gpu · 2022-01-16 17:36
alyssaoh that wants MRT. fine#asahi-gpu · 2022-01-16 17:32
alyssaam I going to nerdsnipe myself into implementing compute kernels? *sweats*#asahi-gpu · 2022-01-16 17:32
alyssaam I going to start running GLES3.1 tests? yes.#asahi-gpu · 2022-01-16 17:25
alyssaMaybe it doesn't handle it correctly? :|#asahi-gpu · 2022-01-16 17:16
alyssalet's see what moltenvk does#asahi-gpu · 2022-01-16 17:16
alyssaGL doesn't want that, wonder if we can force it#asahi-gpu · 2022-01-16 17:16
alyssa(DirectX behaviour, I think)#asahi-gpu · 2022-01-16 17:16
alyssaDiscarded threads are still active as helper invocations#asahi-gpu · 2022-01-16 17:15
alyssaProblem -- Metal's discard is in fact demote, not discard#asahi-gpu · 2022-01-16 17:10
alyssarebases#asahi-gpu · 2022-01-16 17:08
alyssaTruly#asahi-gpu · 2022-01-16 17:08
jnbetter Broadway than Broadcom#asahi-gpu · 2022-01-16 17:07
alyssaBroadway is being piped into my brain. Let's go!#asahi-gpu · 2022-01-16 17:07
alyssainserts headphones#asahi-gpu · 2022-01-16 17:06
alyssaIf marcan is being productive, so should I :>#asahi-gpu · 2022-01-16 17:06
alyssaOkay. No point laying about for hours feeling tired, let's write some Asahi GPU code#asahi-gpu · 2022-01-16 15:07
alyssadoesn't know what M1 GPU stuff to do#asahi-gpu · 2022-01-16 15:07
alyssaI should do M1 GPU stuff#asahi-gpu · 2021-12-15 13:28
alyssayeah, probably#asahi-gpu · 2021-12-15 04:35
chadmedunless that was when you tested GLES3 for funsies#asahi-gpu · 2021-12-15 04:35
chadmedoh damn i thought there were way more failures as of your last stream lmao#asahi-gpu · 2021-12-15 03:51
alyssathe list was that + discard fails over the summer I've been doing not-M1-Mesa#asahi-gpu · 2021-12-15 03:51
alyssaunless your exams started a year ago?#asahi-gpu · 2021-12-15 03:11
chadmedthats insane progress compared to the state we were in when i went AWOL for exams#asahi-gpu · 2021-12-15 03:11
chadmedare these the only GLES2 tests that fail now? #asahi-gpu · 2021-12-14 21:51
alyssaMostly broken depth/stencil attachments and missing cube map / mipmap support#asahi-gpu · 2021-12-14 21:50
alyssa^ current fails list with my patches for discard#asahi-gpu · 2021-12-14 21:50
alyssahttp://backtick.town/~bloom/failures.csv#asahi-gpu · 2021-12-14 21:23
alyssaunfortunately something in -bideas is triggering gpu faults so it's still not really working#asahi-gpu · 2021-12-14 21:22
alyssaglmark2 -bideas now discards correctly anyhow#asahi-gpu · 2021-12-14 21:20
alyssaNot at all convinced the impl is right but hey#asahi-gpu · 2021-12-14 21:20
alyssapassing the gles2 tests#asahi-gpu · 2021-12-14 21:20
alyssaOkay, I have discard implemented in mesa#asahi-gpu · 2021-12-14 20:37
alyssaIt's a large instruction though .. wonder what else it's doing besides sample mask writes.#asahi-gpu · 2021-12-14 20:32
alyssaThis is notably not an ALU source -- copying a uniform to the samepl mask requires an intermediate move#asahi-gpu · 2021-12-14 20:18
alyssa0x100 can't be done with an immediate. so it's bits [16, 22] and [26, 27]#asahi-gpu · 2021-12-14 20:17
alyssaat least 0xff, with bits in [16, 22] but also up top [26, ...]#asahi-gpu · 2021-12-14 20:16
alyssaactually the immediate is larger and discontiguous, as is AGX tradition#asahi-gpu · 2021-12-14 20:14
alyssanot obvious where the reg is, since r0l is all zero anyway#asahi-gpu · 2021-12-14 20:13
alyssabit...23?#asahi-gpu · 2021-12-14 20:13
alyssasorry no#asahi-gpu · 2021-12-14 20:13
alyssaso bit 15 is the immediate? flag#asahi-gpu · 2021-12-14 20:13
alyssaand bit 15 is cleared and the immediate is zeroed#asahi-gpu · 2021-12-14 20:13
alyssathis turns into `mov r0l, 65535`#asahi-gpu · 2021-12-14 20:13
alyssanext up is trying an all-1's sample mask