#
…

2024-01-01

00:00steven has quit [Quit: ZNC 1.8.2 - https://znc.in]
00:01steven has joined #asahi-gpu
00:10al3xtjames2 has joined #asahi-gpu
00:10al3xtjames has quit [Read error: Connection reset by peer]
00:10al3xtjames2 is now known as al3xtjames
01:14enick_397 has quit [Ping timeout: 480 seconds]
01:14rhysmdnz has quit [Ping timeout: 480 seconds]
01:25enick_397 has joined #asahi-gpu
01:25rhysmdnz has joined #asahi-gpu
01:30crabbedhaloablut has quit []
02:38jnn is now known as jn
03:02jeisom has quit [Ping timeout: 480 seconds]
03:02kidplayer666 has quit [Quit: Connection closed for inactivity]
03:43rhysmdnz has quit [Quit: Bridge terminating on SIGTERM]
03:43enick_397 has quit [Quit: Bridge terminating on SIGTERM]
03:47rhysmdnz has joined #asahi-gpu
03:48Guest12483 has joined #asahi-gpu
04:14Compassion has quit [Ping timeout: 480 seconds]
04:17Compassion has joined #asahi-gpu
04:20marvin24 has joined #asahi-gpu
04:23marvin24_ has quit [Ping timeout: 480 seconds]
04:38jnn has joined #asahi-gpu
04:41jn has quit [Ping timeout: 480 seconds]
05:34i509vcbI wonder if this info would be helpful for figuring out how to prevent the unrecoverable timeouts I get. I noticed with deqp-runner if I set the maximum number of jobs to 1 or 2 (although I think the unrecoverable timeouts will occur with less frequency) I don't get unrecoverable timeouts. By default deqp-runner will use all my cores (8 on the M2 air)
05:35i509vcbI think this might hint at some bug with the gpu recovery?
05:42i509vcbUpon further thought for vk-meta-copy to properly do 3d copies without doing something horrible for performance like submitting 8192 draw calls for a full 8192x8192x8192 copy for an image, compute shaders with imageLoad and imageStore might be best?
05:43i509vcbUnless I've missed something, we'd need to write in the worst case 8192 layers of a 2d copy for each layer of the 3d image
05:45i509vcbAlthough I assume there is a reason why vk-meta is currently using a fragment shader for that?
05:54possiblemeatball has quit [Quit: Quit]
07:09rhysmdnz has quit [Quit: Bridge terminating on SIGTERM]
07:09Guest12483 has quit [Quit: Bridge terminating on SIGTERM]
07:13rhysmdnz has joined #asahi-gpu
07:14Guest12499 has joined #asahi-gpu
07:57chadmed has quit [Remote host closed the connection]
08:10chadmed has joined #asahi-gpu
08:47kidplayer666 has joined #asahi-gpu
09:01crabbedhaloablut has joined #asahi-gpu
09:07yuyichao_ has joined #asahi-gpu
09:11yuyichao has quit [Ping timeout: 480 seconds]
11:37kidplayer666 has quit [Quit: Connection closed for inactivity]
12:41WindowPa- has quit [Ping timeout: 480 seconds]
12:45maximbaz has quit [Quit: bye]
12:46WindowPain has joined #asahi-gpu
12:46maximbaz has joined #asahi-gpu
12:56kidplayer666 has joined #asahi-gpu
13:02WindowPain has quit [Ping timeout: 480 seconds]
13:15WindowPain has joined #asahi-gpu
14:01alyssa has joined #asahi-gpu
14:01alyssai509vcb: "sfloat" is indeed just float
14:02alyssapossibly this is the issue of "copies are supposed to preserve nans/denorms but they get flushed when copying as floats"
14:03alyssacasting to a size-equivalent safe format (e.g. r32_uint for r32_float) should deal with that
14:03alyssathis means we can't compress rgb9e5 which is annoying but meh
14:04alyssai509vcb: 3d copies can be done in one draw call with a layered framebuffer, the vertex shader can write the layer on our hardware
14:04alyssa(1 draw call, 8192 rectangles)
14:05alyssai509vcb: as for why fragment, it was originally thought that we needed to use the 3d pipe (fragment) in order to get apple's lossless compression
14:05alyssathis is the case as exposed in metal (imageStore disables compression)
14:05alyssaand is how e.g. Mali works
14:06alyssaI have since learned that imageStore works on compressed images in limited cases, but it's unclear if it's enough to build a blitter.
14:06alyssa(without a 3d pipe fallback at least
14:06alyssaagxv isn't doing compression yet but it will
14:07alyssaat some point we need to see how metal does copies
14:07alyssai don't think there's a straight-up hw DMA engine but maybe there's something helpful
14:10crabbedhaloablut has quit [Read error: Connection reset by peer]
14:12crabbedhaloablut has joined #asahi-gpu
14:57kidplayer666 has quit [Quit: Connection closed for inactivity]
15:07possiblemeatball has joined #asahi-gpu
15:23crabbedhaloablut has quit [Read error: Connection reset by peer]
15:24crabbedhaloablut has joined #asahi-gpu
17:05kidplayer666 has joined #asahi-gpu
19:15kidplayer666 has quit [Quit: Connection closed for inactivity]
19:16jeisom has joined #asahi-gpu
19:31i509vcbhmm okay thanks for the info
19:31i509vcbI guess I am forced to figure out layered rendering for 3d copies
19:32i509vcb(I know there is stuff you pointed me to in the past, just need to try a few things)
19:48alyssaor eat the draws for now, it's not the end of the world
19:48alyssai509vcb: meta->cmd_draw_volume
19:53alyssamm, I think that should be straightforward to wire up, you don't need GS or anything
19:56i509vcbyeah I guess I'll try that first
19:56i509vcbI did kind of vanish for a month lol
19:57possiblemeatball has quit [Quit: Quit]
20:00alyssaall good :-)
20:01anders__2 has joined #asahi-gpu
20:08anders_2 has quit [Ping timeout: 480 seconds]
20:13Mrmaxmeier has quit [Quit: The Lounge - https://thelounge.chat]
20:13Mrmaxmeier has joined #asahi-gpu
20:37zalyx has quit [Quit: later alligator]
20:47possiblemeatball has joined #asahi-gpu
20:47zalyx has joined #asahi-gpu
21:22kidplayer666 has joined #asahi-gpu
21:40i509vcbI'm not 100% on how I'd do that casting. nir_bitcast_vector seems wrong since it wants the bitsize (probably for f32 to f64 casts or something like that) rather than a transmute of sorts
22:50seb91nihel has quit [Quit: ZNC 1.8.2 - https://znc.in]
22:50seb4nihel has joined #asahi-gpu
23:14linuxgemini8 has quit []
23:15linuxgemini has joined #asahi-gpu
23:57ella-0alyssa: as far as i could tell metal always just uses a compute pass for copies
…