#
…

2023-02-18

00:14seeeath has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
00:26capta1nt0ad has joined #asahi-gpu
01:11capta1nt0ad has quit [Remote host closed the connection]
01:21capta1nt0ad has joined #asahi-gpu
01:41kesslerd has quit [Remote host closed the connection]
01:42kesslerd has joined #asahi-gpu
01:44alyssa has joined #asahi-gpu
01:44alyssalina: when you get a chance to look at https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21104 I'd appreciate review on that one
01:44alyssaIt was blocked on !20836 for a minute there but !20836 is now merged so that can go in
01:45alyssaand !20836 is kinda churn so the rebasing around has been really annoying for conflicts
01:45alyssaluckily it's pretty small, since it's been landed piece by piece already (as !20836 and others)
01:46alyssawith !20836, we now finally run most of our NIR passes at CSO create time (shader compile time from the app's perspective)
01:47alyssain theory that means draw time shader variants are cheaper
01:47alyssasince now they're not starting "from scratch", but instead are just applying key-based optimizations to an already "almost ready" NIR shader
01:47alyssa(and then running that NIR through the backend)
01:49alyssa(I have a few other MRs up if you're feeling reviewy, but !21104 is the one that's causing rebase problems)
01:58capta1nt0ad has quit [Ping timeout: 480 seconds]
02:07capta1nt0ad has joined #asahi-gpu
02:16capta1nt0ad has quit [Ping timeout: 480 seconds]
02:17possiblemeatball has joined #asahi-gpu
02:18kesslerd has quit [Remote host closed the connection]
02:20Emantor has quit [Quit: ZNC - http://znc.in]
02:20Emantor has joined #asahi-gpu
02:36kesslerd has joined #asahi-gpu
02:42kesslerd has quit [Remote host closed the connection]
02:43kesslerd has joined #asahi-gpu
04:01i509vcb has quit [Quit: Connection closed for inactivity]
04:04c10l has quit [Read error: Connection reset by peer]
04:04c10l has joined #asahi-gpu
04:12djorz has quit [Ping timeout: 480 seconds]
05:27possiblemeatball has quit [Quit: Quit]
06:33linaalyssa: I'll probably look at it tomorrow, busy today getting Luna packed up for her flight back ^^
07:00stipa has quit [Ping timeout: 480 seconds]
07:14stipa has joined #asahi-gpu
08:09Z750 has quit [Read error: Connection reset by peer]
08:09Z750 has joined #asahi-gpu
09:40ChaosPrincess has quit [Quit: WeeChat 3.8]
09:40ChaosPrincess has joined #asahi-gpu
13:35possiblemeatball has joined #asahi-gpu
14:08md_ has joined #asahi-gpu
14:08md_ has quit []
15:54possiblemeatball has quit [Quit: Quit]
15:55possiblemeatball has joined #asahi-gpu
16:34cylm has quit [Ping timeout: 480 seconds]
16:59nate8 has joined #asahi-gpu
17:46amarioguy has joined #asahi-gpu
17:49alyssawith alpha to cov enabled, enabling writemask causes:
17:49alyssaShader::unk1 to be false
17:49alyssast_tile to be masked, of course
17:50alyssaOcclusion query 2::Unknown from 0x1ffff -> 0x2
17:59seeeath has joined #asahi-gpu
18:11alyssawith no alpha to cov, we still have unk1 false and st_tile masked
18:11alyssaand still 0x2
18:12alyssanotably an... Opaque pass type? which seems broken?
18:30alyssaIt seems like all the difference is from this oq2 unknown
18:30alyssaunless, hmm wait I did a green mask right
18:31alyssaright, yes, blue mask is 0x4
18:31alyssaso it's like a mask of what's enabled
18:31alyssawhere 0x1ffff means "everything is enabled"?
18:32amarioguy has quit [Ping timeout: 480 seconds]
18:33alyssawith blue on 4 render targets, 0x4444
18:36alyssabut I wonder how that works with 8 rt?
18:41alyssawith 8 render targets instead (and switched down to rgba8888) it's switched itself to translucnt and disabled early-z and unk2
18:46alyssaalso fragment_properties::unk3 from 0xf to 0x0 in this case
18:48alyssauhh with 5 render targets (rgba8) and blue
18:48alyssait's still no earlyz/unk2/unk3 and translcuent
18:48alyssabut now the unknown is 0x2492
18:48alyssawhich makes even less sense
18:49alyssaoh, 4 render targets (rgba8) and blue has the first set of changes, just with unknown 0x4444
18:50alyssawith 2 render tagets, 0x44
18:51alyssabut with early-z/etc still enabled
18:51alyssaditto for 3 rts, with 0x444
18:51alyssaonly at 4 does early-z come off
18:52alyssaoh but this was with MSAA 4x
18:52alyssano still no early-z with 4 rts and MSAA
18:55djorz has joined #asahi-gpu
18:56kesslerd has quit [Ping timeout: 480 seconds]
19:39alyssaas for fragment shader side effects
19:39alyssa"set when frag shader spills" is set
19:39alyssaso that's actually a "frag shader side effects" bit
19:39alyssalina: ^^ that one is uapi
19:39alyssaunk 2: 2 -> 3
19:40alyssainterestingly, the frag shader gets a `gl_FragDepth = gl_Position.z` write after the store but before the st_tile
19:41alyssaearly-z is disabled
19:41alyssatri merging is disabled
19:41alyssapunchtrough pass type + conservative depth any thanks to that depth write
19:42alyssawith [[early_fragment_tests]]
19:42alyssastill "spills"
19:42alyssastill unk2->3
19:42alyssano zs_emit, but now we have at the beginning of the shader `gl_SampleMask = 1`
19:42alyssaand still early-z testing is disabled in hardware?!
19:43alyssapasstype of translucent punch through (?!?!?!)
19:46alyssathat's enough hw for my brain rn
19:46alyssazeep zorp
19:58skoobasteeve has joined #asahi-gpu
19:59skoobasteeve has quit []
19:59skoobasteeve has joined #asahi-gpu
20:02alyssamy guess is that the unknown mask is related to HSR
20:03alyssalina: also, I wonder if we have the same ISP_CTL register as powervr
20:03alyssasince that has some line style bits which will be useful later
20:03alyssa(for a vulkan extension)
20:05amarioguy has joined #asahi-gpu
20:10alyssamaybe it's "tile_config"
20:11alyssashrug not too important
20:14alyssa"Another feature that reduces HSR efficiency is write masking."
20:14alyssading ding ding
20:14seeeath has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
20:15seeeath has joined #asahi-gpu
20:15alyssaahaha how did I miss this
20:16alyssa"opaque", "feedback", "translucent"
20:16alyssaso, the passtype enum. nice
20:23skoobasteeve has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
20:27skoobasteeve has joined #asahi-gpu
20:41alyssawell, there's another bug I didn't know I had
21:02possiblemeatball has quit [Quit: Quit]
21:02i509vcb has joined #asahi-gpu
21:35seeeath has quit [Quit: Textual IRC Client: www.textualapp.com]
21:35possiblemeatball has joined #asahi-gpu
21:38seeeath has joined #asahi-gpu
21:42amarioguy has quit [Ping timeout: 480 seconds]
21:56amarioguy has joined #asahi-gpu
22:52possiblemeatball has quit [Quit: Quit]
22:52possiblemeatball has joined #asahi-gpu
22:53possiblemeatball has quit []
23:17possiblemeatball has joined #asahi-gpu
…