#asahi-gpu · 2021-03-27 00:46
bloomand fp16 is so much cheaper in hw than fp32 that it again works out#asahi-gpu · 2021-03-27 00:46
bloom^ integer#asahi-gpu · 2021-03-27 00:45
bloomIt follows naturally from 32-bit arithmetic. I.e. packed 2x16-bit add is the same as 32-bit add up to handling of carry bits.#asahi-gpu · 2021-03-27 00:45
bloomFor Mali, even have 4 channel 8-bit vectors targeting ML workloads#asahi-gpu · 2021-03-27 00:44
bloomSome AMD and Mali GPUs are scalar for 32-bit instructions, but allow 2 channel 16-bit vectors (what AMD markets as rapid packed math or sth)#asahi-gpu · 2021-03-27 00:43
bloomIt's.. cute#asahi-gpu · 2021-03-27 00:42
bloomOlder GPU arches were genuinely vector, some newer ones (GCN, Bifrost) have a fp16vec2 thing going on, special oddball handling goes to Adreno which has a special "repeat N times" modifier :-p#asahi-gpu · 2021-03-27 00:42
bloomDifferent vendors cope with this in different ways.#asahi-gpu · 2021-03-27 00:42
bloom(Since everything is 4x worse over in graphics land, since you repeat the same instruction a bunch of times.)#asahi-gpu · 2021-03-27 00:41
bloom're not careful about code density, you'll blow through your i-cache budget#asahi-gpu · 2021-03-27 00:41
bloomIf you#asahi-gpu · 2021-03-27 00:41
bloomthat especially matters since G13 is a pure scalar arch, yet it's designed for graphics (vector) workloads#asahi-gpu · 2021-03-26 22:42
bloomOh, dunno then#asahi-gpu · 2021-03-26 20:46
bloom(software defined, not hw)#asahi-gpu · 2021-03-26 20:45
bloomused to pass the base address of the buffer from the driver#asahi-gpu · 2021-03-26 20:45
bloomuniforms, I assume#asahi-gpu · 2021-03-25 21:45
bloomDarkShadow44: looks like you're having fun :)#asahi-gpu · 2021-03-21 11:29
bloomMesa doesn't even ave asin/acos/atan acceleration wired up, it's lowered to a (iirc polynomial..?) approximation unconditionally :-P#asahi-gpu · 2021-03-21 11:29
bloomMesa doesn't even ave asin/acos/atan acceleration wired up, it's lowered to a (iirc polynomial..?) approximation unconditionally :-P#asahi-gpu · 2021-03-21 11:24
bloomasin/acos/atan even more so#asahi-gpu · 2021-03-21 11:24
bloomasin/acos/atan even more so#asahi-gpu · 2021-03-21 11:24
bloom(nowadays)#asahi-gpu · 2021-03-21 11:24
bloom(nowadays)#asahi-gpu · 2021-03-21 11:24
bloomdougall: not that I'm discouraging more r/e, but fast sin/cos is surprisingly uninteresting on GPUs#asahi-gpu · 2021-03-21 11:24
bloomdougall: not that I'm discouraging more r/e, but fast sin/cos is surprisingly uninteresting on GPUs#asahi-gpu · 2021-03-20 23:22
bloomsnooping GPU memory#asahi-gpu · 2021-03-20 23:22
bloomsnooping GPU memory#asahi-gpu · 2021-03-18 20:09
bloomCoco for Coco Puff!#asahi-gpu · 2021-03-18 16:54
bloomdefinitely not, I'm just like this#asahi-gpu · 2021-03-18 16:48
bloomthat's normal i just say say things#asahi-gpu · 2021-03-18 16:47
bloomthey don't I was making a joke#asahi-gpu · 2021-03-18 16:46
bloomhttps://en.wikipedia.org/wiki/Coco_(2017_film)#Accolades#asahi-gpu · 2021-03-18 16:46
bloomdid it not win any awards it should've :>#asahi-gpu · 2021-03-18 16:44
bloom"-P#asahi-gpu · 2021-03-18 16:44
bloomused for accelerating video decode of the award-winning Pixar film#asahi-gpu · 2021-03-18 16:44
bloom;P#asahi-gpu · 2021-03-18 16:44
bloomcocoprocessor#asahi-gpu · 2021-03-18 16:41
bloomdaniels: lol, true#asahi-gpu · 2021-03-18 16:20
bloomAlas#asahi-gpu · 2021-03-18 16:19
bloom4k on mobile is rather painful withot compression, so#asahi-gpu · 2021-03-18 16:17
bloom(IIRC Arm has this too, also not shipping afaik)#asahi-gpu · 2021-03-18 16:17
bloomLikewise, can the video decoder recompress to the native format on the fly?#asahi-gpu · 2021-03-18 16:17
bloomI know Arm has camera stuff feeding into AFBC but I don't know if it (or anything like it) ever shipped.#asahi-gpu · 2021-03-18 16:16
bloom(at least on the iPhones with fancy 4k@60..)#asahi-gpu · 2021-03-18 16:16
bloomBetter question is if the cameras support the compression scheme too#asahi-gpu · 2021-03-18 16:14
bloombut again Apple is far too competent to not have their display controller ingest their compression scheme as well#asahi-gpu · 2021-03-18 16:13
bloomAdmittedly I don't nknow what the display controller ingests, I just saw this in flight from app<-->compositor#asahi-gpu · 2021-03-18 16:12
bloomApple has managed to excise every mention of a compression scheme from public docs/source.#asahi-gpu · 2021-03-18 16:12
bloomGotta hand it to Apple, even Arm and Qualcomm brag about AFBC/UBWC despite keeping the internals proprietary.#asahi-gpu · 2021-03-18 16:11
bloomM1 has full support for tiled compressed framebuffers, documented absolutely nowhere