#
#asahi-gpu · 2023-04-16 00:40
alyssauber 4K: 5fps
#asahi-gpu · 2023-04-16 00:40
alyssanative uber: 105fps
#asahi-gpu · 2023-04-16 00:39
alyssaspec 4K: 61fps
#asahi-gpu · 2023-04-16 00:39
alyssaopt sink+move hack:
#asahi-gpu · 2023-04-16 00:25
alyssa(given that we're off at 37% of where we should be... if it were 90% there I'd say yeah all the little stuff matters)
#asahi-gpu · 2023-04-16 00:25
alyssaidk. if it's "death by a thousand cuts" i guess we can deal with that, but usually there's something more glaring
#asahi-gpu · 2023-04-16 00:24
alyssaregister cache hinting might help if that's our bottleneck
#asahi-gpu · 2023-04-16 00:23
alyssafirst ~40 instructions benefit from some parallelism
#asahi-gpu · 2023-04-16 00:22
alyssaTellowKrinkle: re ld_tile placement, Metal isn't sinking it past the first block but it is sinking it to the bottom of the first block
#asahi-gpu · 2023-04-16 00:20
alyssano mystery ops, that's a good sign
#asahi-gpu · 2023-04-16 00:20
alyssaeyes
#asahi-gpu · 2023-04-16 00:14
TellowKrinkleBTW if you're curious, here's a Dolphin ubershader and its disassembly as of about 4 months ago: https://gist.github.com/TellowKrinkle/ced8ea4781f4a2bcfa292ac50d93dc25
#asahi-gpu · 2023-04-16 00:12
TellowKrinkleOkay can confirm, if you add MTLTextureUsagePixelFormatView to all textures, Metal performs within 1fps of MoltenVK at 4K
#asahi-gpu · 2023-04-16 00:08
alyssathreads helped: dolphin-ubers/231.shader_test MESA_SHADER_FRAGMENT: 448 -> 512 (14.29%)
#asahi-gpu · 2023-04-16 00:08
alyssathreads helped: dolphin-ubers/221.shader_test MESA_SHADER_FRAGMENT: 512 -> 576 (12.50%)
#asahi-gpu · 2023-04-16 00:08
alyssaeg
#asahi-gpu · 2023-04-16 00:08
alyssathis is despite shader-db showing thread count improvements in a pile of ubershaders
#asahi-gpu · 2023-04-16 00:07
alyssascheduler doesn't break anything but doesn't seem to be much help
#asahi-gpu · 2023-04-15 23:51
TellowKrinkleLet me quick run a no-PFV test of melty molten galaxy and make sure that accounts for all the performance difference between MVK and Metal
#asahi-gpu · 2023-04-15 23:47
alyssaWas worth a try ;P
#asahi-gpu · 2023-04-15 23:47
alyssaAww :p
#asahi-gpu · 2023-04-15 23:45
TellowKrinkleI don't think Dolphin has too much Linux specific stuff. I'm not too familiar with the more uncommon stuff though, I only joined the project a year ago.
#asahi-gpu · 2023-04-15 23:43
TellowKrinkleI've used Apple's equivalent of that for debugging almost all the GPU-related optimizations I've done for either PCSX2 or Dolphin
#asahi-gpu · 2023-04-15 23:42
alyssaTellowKrinkle: also is there any cool stuff Dolphin can do on Linux but not macOS that I can use to sell this if perf doesn't pan out? :~P
#asahi-gpu · 2023-04-15 23:41
alyssa's a question for lina I think
#asahi-gpu · 2023-04-15 23:40
alyssaoh that
#asahi-gpu · 2023-04-15 23:40
alyssalooks up what gpuvis is
#asahi-gpu · 2023-04-15 23:40
alyssaprobably not
#asahi-gpu · 2023-04-15 23:40
TellowKrinkleBTW does gpuvis work on Asahi yet? That might be a good way to see if we're running into scheduling issues.
#asahi-gpu · 2023-04-15 23:39
alyssahere I'm going to try copypasting some code from panfrost and see if that makes things go brr
#asahi-gpu · 2023-04-15 23:36
alyssaas I mentioned above loading gl_FragColor at the top of the shader is going to force the shaders to be totally serialized, which is also going to magnify the issue (maybe)
#asahi-gpu · 2023-04-15 23:36
alyssaright. if specialized shaders don't usually gl_FragDepth write, but ubershaders do, that'll be a gap
#asahi-gpu · 2023-04-15 23:36
TellowKrinkle(Unless you have accurate depth or whatever that flag is called enabled)
#asahi-gpu · 2023-04-15 23:35
TellowKrinkleWe shouldn't do that in specialized shaders though
#asahi-gpu · 2023-04-15 23:34
alyssaHSR is going to be hurt by the gl_FragDepth write, I think
#asahi-gpu · 2023-04-15 23:34
TellowKrinkleHybrid in particular sees much worse slowdowns when small amounts of ubershader are inserted into an otherwise non-ubershader render than I see on AMD
#asahi-gpu · 2023-04-15 23:33
TellowKrinkleUhh do you know if things like HSR and stuff are getting enabled for non-ubershaders? I've noticed Apple GPUs having a much larger gap between uber and non-uber shaders than e.g. AMD, and have been wondering if that's due to Apple GPUs getting a good bit of their performance from optimizations that get blocked by ubershaders.
#asahi-gpu · 2023-04-15 23:32
alyssain terms of both pressure and latency, given how big these are
#asahi-gpu · 2023-04-15 23:32
alyssascheduling might be significant for the ubershaders though
#asahi-gpu · 2023-04-15 23:31
alyssaseeing as I've literally never seen an ALU saving optimization matter for fps
#asahi-gpu · 2023-04-15 23:30
alyssathere's a long tail of ALU cycle shaving we could do for the ubershaders, but I'm not convinced any of it would move the needle
#asahi-gpu · 2023-04-15 23:30
alyssahopefully some of that is something dumb
#asahi-gpu · 2023-04-15 23:29
alyssa(Granted, 37% the performance of Metal would still be totally playable. And there's piles of content we'll have available that's just not possible on macOS but still.)
#asahi-gpu · 2023-04-15 23:29
alyssadoesn't want to be 37% the performance of Metal :-V
#asahi-gpu · 2023-04-15 23:28
alyssashould maybe look at one of those disassembled Metal ubershaders
#asahi-gpu · 2023-04-15 23:27
alyssaWell. That's not great but not our problem then, heh
#asahi-gpu · 2023-04-15 23:27
alyssaOh
#asahi-gpu · 2023-04-15 23:27
TellowKrinkleLoad happens before the main ubershader loop
#asahi-gpu · 2023-04-15 23:27
alyssasink the load or not?
#asahi-gpu · 2023-04-15 23:27
TellowKrinkleIf you look at the decompilation of ubershaders, Metal does the same

plus ›