#asahi-gpu · 2021-01-15 02:49
davidrysk[m]com.apple.audio.IOKitten -- part of the audio stack I assume#asahi-gpu · 2021-01-15 02:48
bloom-----What is IOKitten? :p#asahi-gpu · 2021-01-15 02:47
davidrysk[m]Apple person: "Your KEXT?"#asahi-gpu · 2021-01-15 02:47
davidrysk[m](that whole forum thread)#asahi-gpu · 2021-01-15 02:46
davidrysk[m]bloom: HAHAH#asahi-gpu · 2021-01-15 02:44
bloomkIOReturnBadArgument apparently#asahi-gpu · 2021-01-15 02:43
bloomhttps://developer.apple.com/forums/thread/113747 Blames code signing something#asahi-gpu · 2021-01-15 02:43
bloom-EDONTTRYTOWRITEYOUROWNDRIVERS#asahi-gpu · 2021-01-15 02:42
bloom^^ Seems awfully specific return code :|#asahi-gpu · 2021-01-15 02:39
bloomret E00002C2#asahi-gpu · 2021-01-15 01:20
bloom(I mean.. modifying the shaders is certianly one way to find the purpose hah)#asahi-gpu · 2021-01-15 01:19
bloomAlthough right now I'm a lot more concerned by a half dozen driver internal shaders with no clear purpose.#asahi-gpu · 2021-01-15 01:19
bloomUnclear.#asahi-gpu · 2021-01-15 01:13
modwizcodeWhy would they vectorize only half of it?#asahi-gpu · 2021-01-15 01:03
bloomActually more like 6..#asahi-gpu · 2021-01-15 00:48
bloomMaybe a coordinate shader?#asahi-gpu · 2021-01-15 00:48
bloomMeanwhile, there is at least one 'hidden' shader being pulled in :|#asahi-gpu · 2021-01-15 00:27
bloom(Frustratingly asymmetric ^^)#asahi-gpu · 2021-01-15 00:27
bloomthe load is vectored.#asahi-gpu · 2021-01-15 00:27
bloom* store vary#asahi-gpu · 2021-01-15 00:27
bloomAlso interesting is that I/O is scalarized.#asahi-gpu · 2021-01-15 00:26
bloomThen the shader explicitly does address calculations with imad and iadd, and loads the resulting (64-bit) address with a generic load instruction.#asahi-gpu · 2021-01-15 00:25
bloomWhat should be attributes is just mapped as a generic memory, and the vertex ID is accessible via a special register.#asahi-gpu · 2021-01-15 00:24
bloomOh, this is an interesting detail: there is no "load attribute" instruction on the M1 GPU#asahi-gpu · 2021-01-14 22:12
bloomhthh_: I have seen stranger things :p#asahi-gpu · 2021-01-14 22:08
davidrysk[m]and just assume it works#asahi-gpu · 2021-01-14 22:08
davidrysk[m]almost like they're not bothering to check the return value at all#asahi-gpu · 2021-01-14 22:06
hthh_haha yep, definitely a weird one#asahi-gpu · 2021-01-14 22:06
davidrysk[m]hthh_: I figured it was a bug, but that's really funny#asahi-gpu · 2021-01-14 22:05
hthh_davidrysk[m]: heh, i can't be sure, but i don't think it's intentional that they only return KERN_SUCCESS on failure#asahi-gpu · 2021-01-14 21:51
j`eyfill-in-the-blanks! :P#asahi-gpu · 2021-01-14 21:50
davidrysk[m]ok!#asahi-gpu · 2021-01-14 21:50
bloomit will be#asahi-gpu · 2021-01-14 21:50
davidrysk[m]that's fine, just would be nice if everything were to be compilable#asahi-gpu · 2021-01-14 21:50
bloomon my todo list to cleanup and fix and make things buildable#asahi-gpu · 2021-01-14 21:50
bloomdavidrysk[m]: in order, 1) not yet but very soon, 2) clang *.c -framework IOKit 3) no but the code that's not is being held for good reason#asahi-gpu · 2021-01-14 21:39
davidrysk[m]wrap.c:145:44: error: no member named 'is_cmdbuf' in 'struct agx_allocation'#asahi-gpu · 2021-01-14 21:39
davidrysk[m]also are all your files checked in?#asahi-gpu · 2021-01-14 21:37
davidrysk[m]also what's your compile line for wrap?#asahi-gpu · 2021-01-14 21:35
davidrysk[m]bloom: are there any examples that you can check into the repo that wrap can be used with?#asahi-gpu · 2021-01-14 21:25
bloomdhewg: Ack.#asahi-gpu · 2021-01-14 21:24
opticronheh#asahi-gpu · 2021-01-14 21:24
dhewgi expect 8 penguins on simplefb soon, so go for at least 9 triangles#asahi-gpu · 2021-01-14 21:24
sven:-)#asahi-gpu · 2021-01-14 21:24
bloom("Something tells me you won't be sleeping tonight")#asahi-gpu · 2021-01-14 21:23
bloomcube spinning go brrgh? ;P#asahi-gpu · 2021-01-14 21:23
bloom12 triangles#asahi-gpu · 2021-01-14 21:23
bloomopticron: why not like