#
…

2022-12-20

00:07MajorBiscuit has quit [Ping timeout: 480 seconds]
02:07alyssa has joined #asahi-gpu
02:07alyssa 8: 0509000e00c0f200 device_load 0, i32, xyzw, r1_r2_r3_r4, u0_u1, r0, unsigned
02:07alyssa 10: 0529040e00c0f200 device_load 0, i32, xyzw, r5_r6_r7_r8, u2_u3, r0, unsigned
02:07alyssa 18: 0541080e00c0f200 device_load 0, i32, xyzw, r8_r9_r10_r11, u4_u5, r0, unsigned
02:07alyssa 20: 3800 wait 0
02:07alyssaI wonder why this code sequence doesn't want to work :V
02:10alyssaI just want multiple loads in flight :d
02:10alyssa2 loads seems to work, but 3 no?
02:15alyssawhich is strange because Apple's compiler is happy to generate sequences like that
02:15alyssawhy does my wait not work but there's does? :P
02:17alyssaoh, hold on
02:17alyssanow when I make it load vec4 it instead changes to
02:17alyssa 1e: 0501000a00c8f200 device_load.TODO 0, i32, xyzw, r0_r1_r2_r3, u0_u1, r0, unsigned, lsl 2, 0, 0, 4, 0
02:17alyssa 26: 0529a00e00c8f200 device_load 0, i32, xyzw, r5_r6_r7_r8, u0_u1, r5, unsigned, lsl 2
02:17alyssa 2e: 0551800e00c8f200 device_load 0, i32, xyzw, r10_r11_r12_r13, u0_u1, r4, unsigned, lsl 2
02:17alyssa 36: 3800 wait 0
02:17alyssaain't this what we saw this morning?
02:18alyssamuch ado about a single bit.
02:19alyssadoes it for vec3
02:20alyssaand vec2
02:20alyssabut not scalars
02:20alyssaso, so many questions
02:20alyssawhat in the world could this single celebrity bit do
02:22bluetail0 has joined #asahi-gpu
02:25bluetail8 has joined #asahi-gpu
02:26bluetail8 has quit []
02:29bluetail has quit [Ping timeout: 480 seconds]
02:33bluetail0 has quit [Ping timeout: 480 seconds]
02:35TellowKrinkleWeird, I'm getting this on my M1 (no .TODO):
02:35TellowKrinkle 24: 0529a40e00c8f200 device_load 0, i32, quad, r5_r6_r7_r8, u2_u3, r5, unsigned, lsl 2
02:35TellowKrinkle 34: 3800 wait 0
02:35TellowKrinkle 1c: 0501040e00c8f200 device_load 0, i32, quad, r0_r1_r2_r3, u2_u3, r0, unsigned, lsl 2
02:35TellowKrinkle 2c: 0551240e01c8f200 device_load 0, i32, quad, r10_r11_r12_r13, u2_u3, r9, unsigned, lsl 2
02:35TellowKrinkle(Compute shader that loads 3 float4s, does an fma, and stores the result)
02:42alyssaTellowKrinkle: and the plot thickens
02:42alyssathis is in a fragment shader. I wonder if that makes a difference.
02:44alyssaindeed I don't get it in a vertex shader
02:44alyssaso if I can get the non-TODO one in a VS
02:45alyssawhy isn't this work for my compiler
02:48alyssaso... what gives?!
02:50alyssai get it, it's only okay when apple's compiler does it..
02:52alyssasame index, different base, maybe that's part
02:55alyssa 14: 3e9108e02500 convert f_to_u32, r4, r15, rtz
02:55alyssa 1a: 0501800e00c8f200 device_load 0, i32, xyzw, r0_r1_r2_r3, u0_u1, r4, unsigned, lsl 2
02:55alyssa 22: 0531840e00c8f200 device_load 0, i32, xyzw, r6_r7_r8_r9, u2_u3, r4, unsigned, lsl 2
02:55alyssa 2a: 0559880e00c8f200 device_load 0, i32, xyzw, r11_r12_r13_r14, u4_u5, r4, unsigned, lsl 2
02:55alyssa 32: 3800 wait 0
02:55alyssanope no difference
02:55alyssaso apply just... generates asm that looks like mine... but it doesn't work for me...
02:57TellowKrinkleBTW on that fragment shader one, were you doing anything with side effects (e.g. writing to device memory)? I can only get the .TODO if I do that and I don't also add "[[early_fragment_tests]]"
02:57alyssano side effects..
02:58alyssanow i'm not getting the .TODO in the frag shader either
02:59alyssaright, wasn't for scalar loads, only vector
02:59alyssauh why don't i get it for this vector.
02:59alyssasigned index, 3 different buffers, same idx
03:00alyssa3 different buffers, same idx, but an UNSIGNED index and I get the .TODO on the first
03:00alyssathere now i got it with a signed index
03:00alyssai honestly think the logic is just
03:01alyssabool APPLE_PROPRIETARY_COMPILER_should_set_random_bit(void) {
03:01alyssa if (setting_the_bit_would_fuck_with_alyssa_more_than_not_setting_it()) {
03:01alyssa return true;
03:01alyssa } else {
03:01alyssa return false;
03:03alyssamaybe it's some sort of cache hint, idk
03:04alyssain terms of data structure changes
03:04swaggie has quit [Read error: Connection reset by peer]
03:04swaggie has joined #asahi-gpu
03:04alyssafor the VS, unk2 is 1 on the no buffer load case and 2 on the 3 load case
03:04alyssauses way more regs obviously
03:04alyssadifferent uniforms obviosuly
03:05alyssafor the FS, unk 2 is set in the FS when loads and unset when no loads
03:05alyssathat seems potentially more interesting
03:05alyssabut other than nothing
03:05alyssaso we have 2 random magic bits with no discernible pattern
03:06alyssaand inexplicable deqp fails on linux
03:06alyssaand ... yeah, this seems like a normal state of affairs, let's play supertuxkart instead
03:11alyssaunless i have some unrelated bug where the VS is now faster causing me to lose some other race condition
03:11alyssathat's.. unlikely but i'm running out of ideas
03:39cylm_ has quit [Ping timeout: 480 seconds]
03:44balrog has quit [Quit: Bye]
03:54alyssa__builtin_ffs();
03:54alyssaI found my bug
03:54alyssaUnrelated to any of the magic, just a pure academic error
03:55alyssathis is going on Spot The Bug
03:55TellowKrinkleOkay the ".TODO" seems to only be added if the input comes from a flat interpolated input?
03:56alyssaTellowKrinkle: I was getting it with uint(smooth input)
03:59TellowKrinklealyssa: Good to know
04:01TellowKrinkleOkay I was able to get the TODO with smooth input
04:02TellowKrinklebuffer[uint(smooth.x)] * buffer[uint(smooth.y)] => "device_load.TODO"
04:03TellowKrinklebuffer[uint(smooth.x) * 8 + 0] * buffer[uint(smooth.x) * 8 + 2] => "device_load"
04:04balrog has joined #asahi-gpu
04:04TellowKrinklebuffer[uint(smooth.x) * 8 + 0] * buffer[uint(smooth.y) * 8 + 2] => device_load
04:13balrog has quit [Quit: Bye]
04:14alyssatruly an RNG
04:22TellowKrinkleSo far, it seems that if the input is smooth, doing anything other than just converting it to a uint removes the .TODO. But if the input is flat, I can do all sorts of math on it and it stays .TODO. Can't possibly imagine what such a flag would be useful for...
05:01evtyz has joined #asahi-gpu
05:03evtyz has quit [Quit: My MacBook Air has gone to sleep. ZZZzzz…]
05:17Gues__________________________ has joined #asahi-gpu
05:22Gues__________________________ has quit [Quit: My MacBook Air has gone to sleep. ZZZzzz…]
05:49balrog has joined #asahi-gpu
06:44WindowPain_ has joined #asahi-gpu
06:49WindowPain has quit [Ping timeout: 480 seconds]
06:59bluetail8 has joined #asahi-gpu
07:22MajorBiscuit has joined #asahi-gpu
07:31Hibyehello_ has joined #asahi-gpu
07:33Major_Biscuit has joined #asahi-gpu
07:36Hibyehello has quit [Ping timeout: 480 seconds]
07:40MajorBiscuit has quit [Ping timeout: 480 seconds]
08:02SSJ_GZ has joined #asahi-gpu
08:43cylm_ has joined #asahi-gpu
09:24swaggie has quit [Remote host closed the connection]
09:24swaggie has joined #asahi-gpu
09:24swaggie has quit [Remote host closed the connection]
09:25swaggie has joined #asahi-gpu
09:31SSJ_GZ has quit [Ping timeout: 480 seconds]
09:41swaggie has quit [Ping timeout: 480 seconds]
09:50bcrumb has joined #asahi-gpu
09:58swaggie has joined #asahi-gpu
10:25swaggie has quit [Ping timeout: 480 seconds]
10:55SSJ_GZ has joined #asahi-gpu
10:59bcrumb has quit [Quit: WeeChat 3.7.1]
11:22bluetail84 has joined #asahi-gpu
11:29bluetail8 has quit [Ping timeout: 480 seconds]
11:29bluetail84 is now known as bluetail8
12:27swaggie has joined #asahi-gpu
12:35swaggie has quit [Ping timeout: 480 seconds]
13:19pjakobsson_ has quit [Remote host closed the connection]
13:40bcrumb has joined #asahi-gpu
13:42bcrumb has quit []
15:05VinDuv has quit [Quit: ZNC 1.8.2+deb2+b1 - https://znc.in]
15:05VinDuv has joined #asahi-gpu
15:27LinuxM1 has joined #asahi-gpu
15:46bcrumb has joined #asahi-gpu
15:47bcrumb has quit []
16:30LinuxM1 has quit [Ping timeout: 480 seconds]
17:06Major_Biscuit has quit [Ping timeout: 480 seconds]
17:24bcrumb has joined #asahi-gpu
17:32snouhaud has joined #asahi-gpu
17:37systwi_ has quit [Ping timeout: 480 seconds]
17:45snouhaud has quit []
17:45snouhaud has joined #asahi-gpu
17:47systwi has joined #asahi-gpu
17:48bcrumb has quit [Quit: WeeChat 3.7.1]
18:14bcrumb has joined #asahi-gpu
18:14bcrumb has quit []
18:17swaggie has joined #asahi-gpu
18:26swaggie has quit [Ping timeout: 480 seconds]
18:35cylm_ has quit [Ping timeout: 480 seconds]
18:49Dcow has quit [Remote host closed the connection]
18:53Dcow has joined #asahi-gpu
19:32Gues__________________________ has joined #asahi-gpu
19:43Gues__________________________ has quit [Ping timeout: 480 seconds]
19:54leftas has joined #asahi-gpu
19:59mkurz has joined #asahi-gpu
20:01mkurzmarcan: Lina updated the GPU tracker issue (https://github.com/AsahiLinux/linux/issues/72), but linked to the same kwin issue twice ("... plus KWin bug, plus core Mesa bug...") She didn't link to the Mesa bug actually.
20:48Dcow has quit [Remote host closed the connection]
20:49Dcow has joined #asahi-gpu
20:49Dcow has quit [Read error: Connection reset by peer]
20:50Dcow has joined #asahi-gpu
20:56Rayyan_ is now known as Rayyan
21:06snouhaud has quit []
21:13alyssaLinear texture: must be of type MTLTextureType2D or linear MTLTextureType2DArray, textureType (MTLTextureType1DArray) disallowed
21:13alyssapray tell, metal, how do I get a linear MTLTextureType2DArray
21:47alyssabloody hell
21:49Dcow has quit [Remote host closed the connection]
21:58Dcow has joined #asahi-gpu
22:08Dcow has quit [Ping timeout: 480 seconds]
23:20SSJ_GZ has quit [Ping timeout: 480 seconds]
23:44Major_Biscuit has joined #asahi-gpu
…