#
…

2021-01-12

00:03TheLink has joined #asahi-gpu
00:04aratuk has joined #asahi-gpu
00:09aratuk has quit [Ping timeout: 264 seconds]
00:25ransom has joined #asahi-gpu
00:45ransom has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
00:48bloomAlright, let's see what I can get done in an hour :)
00:53bloom(nothing is if I just read random github issues fo an hour..)
01:01bloomI guess wiring in the dwetile
01:02icecream95"Oddly NEON is making things slower". Because GCC's handling of ARM vector intrinsics sucks?
01:02bloomicecream95: Yeah, probably :p
01:03icecream95Memory access patterns are also important. Ideally you want to write entire cache lines at once, but I think Aarch64 dropped that instruction
01:04bloomDetiling should write linearly but read scattered, just as scalar code.
01:04jn__bloom: not sure if it's intended, but your deleted text comes through to the other side of IRC, e.g. "dw␡etile"
01:04bloomjn__: ?
01:05hthh_i just see "I guess wiring in the dwetile"
01:05jn__dw<character that represents backspace>tile
01:05jn__U+2421 ␡ GRAPHIC FOR DELETE <- this character
01:06jn__ah wait, i mean "dw<backspace>etile"
01:07jn__no offence meant, it just seems like an obscure misconfiguration of something
01:07icecream95The matrix coprocessor probably has an instruction for tiling anyway
01:09bloomjn__: Contrary to what you might think based on my code, I am not good at using computers. ^^
01:09bloomicecream95: That's cheating :<
01:10bloomBut you might be right :)
01:10jn__bloom: no worries :)
01:26bloomOk, got the detiling wired in.
01:27bloommarcan: Turns out the framebuffer needs to be 128 byte aligned. I was previously aligning to 64 bytes only, and that's where the missing 64 bytes came from.
02:09klaus_ has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
02:48brentr123[m]So is metal going to be the graphics driver or will it be based off of moltenvk/molten gl?
02:51icecream95brentr123[m]: AFAIK the plan is to write a native OpenGL and/or Vulkan driver with no Metal involved
02:57bloom^^ that
02:57bloomI am still undecided about GL or Vulkan first.
03:00brentr123[m]Anyway to do both?
03:00bloomWith 2x the work..
03:00brentr123[m]Get the best of both worlds
03:00bloomGL-on-VK is a thing but meh
03:17ransom_ has joined #asahi-gpu
03:18taowa has joined #asahi-gpu
03:23ransom_ has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
03:30aratuk has joined #asahi-gpu
03:35aratuk has quit [Ping timeout: 265 seconds]
04:12aratuk has joined #asahi-gpu
04:21ransom has joined #asahi-gpu
04:25aratuk has quit []
04:53khronokernel[m] has joined #asahi-gpu
05:01aratuk has joined #asahi-gpu
05:05davidrysk[m]the device probably maps better to VK
05:08ConeOfAttack[m]Quick question, I'm out of my league here but don't we already know Apple is running some custom instructions?
05:08davidrysk[m]it's a GPU, it's entirely custom instructions
05:11ConeOfAttack[m]... I figured they'd reuse as much between CPU and GPU as possible. So I was referring to ARM instructions.
05:13ConeOfAttack[m]Sorry, I was looking at an earlier discussion.
05:15aratuk_ has joined #asahi-gpu
05:18aratuk has quit [Ping timeout: 246 seconds]
05:25wiloon has joined #asahi-gpu
05:36aratuk_ has quit [Remote host closed the connection]
05:39ransom has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
05:41Tokamak_ has joined #asahi-gpu
05:41artemistLooking at benchmarks, Zink performans way better than I would expect
05:43Tokamak has quit [Ping timeout: 256 seconds]
05:45Tokamak_ has quit [Ping timeout: 265 seconds]
05:55Tokamak has joined #asahi-gpu
06:01Tokamak has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
06:04ransom has joined #asahi-gpu
06:13ransom has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
06:27Tokamak_ has joined #asahi-gpu
07:05austriancoderbloom: Vulkan first :)
07:36aratuk has joined #asahi-gpu
07:41aratuk has quit [Ping timeout: 256 seconds]
07:47Tokamak_ has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
07:51marcanbloom: that makes sense
07:52marcanthat reminds me of the wii NAND ECC calculation hardware... it ostensibly writes the calculated ECC buffer after the ECC spare buffer, but it actually writes it to the address XOR 0x40
07:52marcanso if you don't align it ends up *before* the buffer
07:53marcan(I figured out this behavior via experimentation, then many years later when the official docs leaked, it turns out they don't document it properly... my docs are better than the official ones :))
08:41klaus_ has joined #asahi-gpu
08:47solarkraft has joined #asahi-gpu
09:06Necrosporus has quit [Ping timeout: 240 seconds]
09:15icecream95 has quit [Ping timeout: 265 seconds]
09:17klaus_ has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
09:18klaus has joined #asahi-gpu
09:23DarthCloud has quit [Ping timeout: 240 seconds]
09:26DarthCloud has joined #asahi-gpu
09:30DarthCloud has quit [Remote host closed the connection]
09:31DarthCloud has joined #asahi-gpu
09:37aratuk has joined #asahi-gpu
09:41aratuk has quit [Ping timeout: 240 seconds]
09:49klaus has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
10:15klaus has joined #asahi-gpu
10:23Necrosporus has joined #asahi-gpu
11:57klaus has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
12:01klaus has joined #asahi-gpu
13:14klaus has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
13:26klaus has joined #asahi-gpu
13:33bloomaustriancoder: right, the problem with that is -- afaiu -- Vulkan requires a deep understanding of the hardware, whereas I'm at the point I could throw together a gallium driver passing most of deqp-gles2 by just bitbanging magic numbers
13:33klaus has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
13:33bloom(and more to the point - I am intimately familiar with Gallium, I do not know Vulkan)
13:41austriancoderbloom: passing most of deqp-gles2 .. I think this at least needs one more week for you. but yes I am in the same boat.. i know gallium quite well but never touched vulkan. also in linux userspace is more gles then vulkan atm
13:50bloomaustriancoder: yeah, that's the other thing, there's no point in paying vulkan tax if the entire userspace is going through zink, and that's one more layer of abstraction to debug
13:51bloomespecially on arm
14:00aratuk has joined #asahi-gpu
14:05aratuk has quit [Ping timeout: 246 seconds]
14:09daniels has joined #asahi-gpu
15:24Tokamak has joined #asahi-gpu
15:31bloomaustriancoder: right, the problem with that is -- afaiu -- Vulkan requires a deep understanding of the hardware, whereas I'm at the point I could throw together a gallium driver passing most of deqp-gles2 by just bitbanging magic numbers
15:31* j`eydejavu
15:32brentr123[m]what is gallian? ive never heard of that
15:36klaus has joined #asahi-gpu
15:46modwizcodeIt's the state machine for mesa right?
16:01davidrysk[m]bloom: now I'm curious -- just how *are* vulkan drivers on Linux implemented
16:02davidrysk[m]they seem to be part of mesa
16:07davidrysk[m]> LunarG compared Vulkan to "Gallium3D 2.0." << heh
16:17neobrainI'm actually surprised to hear implementing a Gallium driver requires less understanding of hardware than a Vulkan one, but then again I've never done the former :)
16:18neobrainthough the "bigbanging magic numbers" bit sounds like it wouldn't be a particularly nice or performant driver
16:21rann has joined #asahi-gpu
16:56Tokamak has quit [Ping timeout: 240 seconds]
16:57Tokamak has joined #asahi-gpu
17:12ransom has joined #asahi-gpu
17:17marcanthat's not what that means
17:43DarthCloud has quit [Remote host closed the connection]
17:43DarthCloud has joined #asahi-gpu
17:45robinp has quit [Ping timeout: 260 seconds]
18:00ransom has quit [Quit: Textual IRC Client: www.textualapp.com]
18:02aratuk has joined #asahi-gpu
18:02smist08[m] has joined #asahi-gpu
18:06aratuk has quit [Ping timeout: 260 seconds]
18:37DarthCloud has quit [Remote host closed the connection]
18:38DarthCloud has joined #asahi-gpu
18:52konrad__ has joined #asahi-gpu
18:53austriancoderdavidrysk[m]: they are part of Mesa but the different drivers are not sharing much code as for OpenGL. For OpenGL (ES/CL) there is a common interface called gallium
19:05DarthCloud has quit [Ping timeout: 240 seconds]
19:12DarthCloud has joined #asahi-gpu
19:15bloommarcan: bwah?
19:19marcanbloom: I mean it doesn't mean it won't perform well :p
19:22Tokamak has quit [Ping timeout: 240 seconds]
19:23icecream95 has joined #asahi-gpu
19:27Tokamak has joined #asahi-gpu
20:03aratuk has joined #asahi-gpu
20:07aratuk has quit [Ping timeout: 256 seconds]
20:20modwizcode has quit [Ping timeout: 272 seconds]
21:36ransom has joined #asahi-gpu
22:12ransom has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
22:20ransom has joined #asahi-gpu
22:33TonyRedondo has quit [Quit: Connection closed for inactivity]
22:55Tokamak has quit [Ping timeout: 256 seconds]
22:57Tokamak has joined #asahi-gpu
23:31m42uko has quit [Ping timeout: 260 seconds]
23:36m42uko has joined #asahi-gpu
23:42ransom has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
…