#
…

2023-02-20

00:48kesslerd has quit [Remote host closed the connection]
00:51pthariensflame has joined #asahi-gpu
00:52kesslerd has joined #asahi-gpu
00:53pthariensflame has quit []
01:17alyssaCoaxing Metal to tell me how bindless textures work seems less effective than just playing with the hardware under Linux and seeing what happens
01:21alyssaso, with flags=3 and everything else as usual, it's looking up "passed uniform + 32-bit register"
01:23alyssathat's enough to emulate bindless but it's not as nice as I was hoping
01:24alyssawhat I really want is a "64-bit register" mode
01:27TellowKrinkleSo does Metal just have a massive array of texture descriptors that it allocates out of and then uses 32 bit offsets into that array as texture pointers?
01:28alyssaI don't know.
01:28alyssaIt's certainly possible.
01:28alyssaThat's basically the "heap" model. It's how NVIDIA works at a hardware level, and how DX12 works at an API level.
01:30alyssawhile heaps are just as expressive as pointer-style bindless, I guess I'm a bit surprised to not see a way to do real pointers
01:30alyssagiven that the hardware is all pointers internally
01:31alyssaLike, my current mental model is that the hardware is internally pointer-style bindless
01:31alyssaand the texture state registers (ts0 etc) and this uniform+reg mode are both conveniences for the driver to emulate binding tables and heaps respectively
01:31alyssabut I'd be surprised if there wasn't a raw pointer mode in here too, if that's what's happening
01:38alyssawhere I'm at is, if this is indeed how bindless works on this hardware (and the only way it works), then I can make this work in the drive
01:39alyssabut it seems more likely that we're missing some encodings than that this is really what Apple intended
01:50alyssaBy the way, does Metal support atomics on images now?
01:51TellowKrinkleStill no image atomics in Metal
01:51alyssak. Maybe that'll come in the next version bump then.
01:59opticron has quit [Ping timeout: 480 seconds]
02:12opticron has joined #asahi-gpu
02:32alyssaTellowKrinkle: I think u4 is the SampleURegDesc
02:32alyssathat said, SampleURegDesc is missing a few bits of extensions
02:33alyssa(Unless you can't use the upper bits for this. Again, while possible that seems less likely than I don't know which bits are the extension.)
02:33TellowKrinkleHow many unknown bits do we have left at this point?
02:35alyssatexture_sample has a known SampleURegDesc but no known extension bits
02:35alyssabut q4 and q6 are both plausibly extension
02:35alyssa(8 bits between them)
02:36alyssawait, q6 is gather stuff at least in part
02:36alyssaand q4 is kill at least in part
02:37alyssaq6 has like 1 free bit
02:37alyssaand q4 has 1 free bit
02:37alyssaand extensions seem to always be 2-bits, so maybe there is no extension
02:37alyssajust seems .. odd
02:38alyssabut regardless, image_write::u4 being a URegDesc (maybe) is a nice take-home message
02:39alyssabecause with that I have enough to implement heaps, even if I'm still unconvinced I should have to
02:39alyssaOTOH, shader code and pipelines and things use a heap-like model too
02:39alyssain that you only get 32-bits of VA but can place it wherever you want
02:39alyssaso, shrug
02:40alyssamaybe it is in line with the hardware
02:40alyssamy brain is very much still in Arm mode
02:41TellowKrinkleAre arm GPUs all pointers? (I know almost nothing about arm GPUs)
02:42alyssanah, mali is ... a mix of models that has changed over time
02:42alyssaI guess it's also like heaps in some ways
02:42alyssathe valhall model i mean
04:21djorz has quit [Ping timeout: 480 seconds]
04:53pbsds has quit [Ping timeout: 480 seconds]
05:06DarkShadow4444 has joined #asahi-gpu
05:11DarkShadow44 has quit [Ping timeout: 480 seconds]
05:22pbsds has joined #asahi-gpu
05:26seeeath has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
05:27seeeath has joined #asahi-gpu
05:43seeeath has quit [Read error: Connection reset by peer]
05:49possiblemeatball has quit [Quit: Quit]
07:52Dementor has quit [Read error: Connection reset by peer]
07:52Dementor has joined #asahi-gpu
08:58pjakobsson has joined #asahi-gpu
09:13i509vcb has quit [Quit: Connection closed for inactivity]
09:35MajorBiscuit has joined #asahi-gpu
09:42kesslerd has quit [Remote host closed the connection]
09:42kesslerd has joined #asahi-gpu
09:48MajorBiscuit has quit [Quit: WeeChat 3.6]
10:37tertu has quit [Ping timeout: 480 seconds]
10:55bisko has joined #asahi-gpu
11:15kesslerd has quit [Quit: Leaving]
11:20kesslerd has joined #asahi-gpu
11:20karolherbst_ has joined #asahi-gpu
11:24karolherbst has quit [Ping timeout: 480 seconds]
11:37karolherbst has joined #asahi-gpu
11:39hightower2 has quit [Ping timeout: 480 seconds]
11:44karolherbst_ has quit [Ping timeout: 480 seconds]
12:06hightower2 has joined #asahi-gpu
13:12cylm has quit [Ping timeout: 480 seconds]
13:13tertu has joined #asahi-gpu
13:24hightower2 has quit [Ping timeout: 480 seconds]
13:43Dementor2 has joined #asahi-gpu
13:46Dementor has quit [Ping timeout: 480 seconds]
13:46Dementor2 is now known as Dementor
14:23bisko has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
14:58bisko has joined #asahi-gpu
15:01bluetail986 has quit []
15:11kesslerd has quit [Remote host closed the connection]
15:19kesslerd has joined #asahi-gpu
15:36zzywysm has quit [Quit: Textual IRC Client: www.textualapp.com]
15:43nela has quit [Ping timeout: 480 seconds]
15:52nela has joined #asahi-gpu
16:07ah- has quit [Read error: Connection reset by peer]
16:07bisko has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
16:11bluetail986 has joined #asahi-gpu
16:13ah- has joined #asahi-gpu
16:24kesslerd has quit [Remote host closed the connection]
16:25kesslerd has joined #asahi-gpu
16:26zzywysm has joined #asahi-gpu
17:15possiblemeatball has joined #asahi-gpu
18:04bluetail986 has quit []
18:04bluetail has joined #asahi-gpu
19:00nyilas has joined #asahi-gpu
19:06alyssalina: thanks for the review, your reward is... more patches to review ;p
20:25possiblemeatball has quit [Quit: Quit]
20:29cylm has joined #asahi-gpu
20:44kesslerd has quit [Quit: Leaving]
20:57djorz has joined #asahi-gpu
21:06yamii_ has joined #asahi-gpu
21:06yamii has quit [Read error: Connection reset by peer]
21:17kesslerd has joined #asahi-gpu
21:19alyssaz32f+s8 msaa4x has, notably
21:19alyssaunk 212=8
21:20alyssasample count=4
21:20alyssasample positions set
21:20alyssatile height=16
21:20nyilas has quit [Quit: Konversation terminated!]
21:20alyssalayout 32x16
21:20alyssasample count 4
21:20alyssabytes per threadgroup 16384
21:20alyssaunk 1 true
21:21alyssaI'm suspicious I'm actually hitting a kernel bug here ..
21:29kesslerd has quit [Remote host closed the connection]
21:43kesslerd has joined #asahi-gpu
22:41hightower2 has joined #asahi-gpu
…