Potential macOS Metal & vsync improvements #309
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
I have zero clue at the moment why this works. MoltenVK sets this property during initialization of the swapchain, and logging confirms it's set, but setting the proprty ahead of time seems to actually disable vsync and result in the expected rendering smoothness. At worst, this change does nothing that wouldn't be done by MoltenVK anyway. At best, it seems to provide much smoother render on a number of tested macOS setups (Intel 2015 MBP/Mojave, Intel 2015 MBP/Catalina, 2020 M1 MBP/Big Sur).
The performance difference for Playback isn't so significant that this makes sense to check for Rosetta 2 and change accordingly.
NikhilNarayana
force-pushed
the
slippi
branch
from
b0868c6
to
aaad1a2
Compare
|
This has been updated to Slippi 2.3.3. |
vladfi1
pushed a commit
to vladfi1/slippi-Ishiiruka
that referenced
this pull request
Jan 1, 2022
* Force CAMetalLayer to disable vsync. I have zero clue at the moment why this works. MoltenVK sets this property during initialization of the swapchain, and logging confirms it's set, but setting the proprty ahead of time seems to actually disable vsync and result in the expected rendering smoothness. At worst, this change does nothing that wouldn't be done by MoltenVK anyway. At best, it seems to provide much smoother render on a number of tested macOS setups (Intel 2015 MBP/Mojave, Intel 2015 MBP/Catalina, 2020 M1 MBP/Big Sur). * Inject a custom CALayer rendering target to avoid wxWidgets drawRect * ENV feature flag this * Formatting * Flag MoltenVK ENV variables on Playback The performance difference for Playback isn't so significant that this makes sense to check for Rosetta 2 and change accordingly. * Pin ffmpeg in CI due to upstream changes. * Try 2.8 I guess.
jlambert360
added a commit
to jlambert360/Ishiiruka
that referenced
this pull request
Jan 29, 2022
jlambert360
added a commit
to jlambert360/Ishiiruka
that referenced
this pull request
Jan 30, 2022
jordan-zilch
pushed a commit
to jordan-zilch/Ishiiruka
that referenced
this pull request
May 17, 2022
* Force CAMetalLayer to disable vsync. I have zero clue at the moment why this works. MoltenVK sets this property during initialization of the swapchain, and logging confirms it's set, but setting the proprty ahead of time seems to actually disable vsync and result in the expected rendering smoothness. At worst, this change does nothing that wouldn't be done by MoltenVK anyway. At best, it seems to provide much smoother render on a number of tested macOS setups (Intel 2015 MBP/Mojave, Intel 2015 MBP/Catalina, 2020 M1 MBP/Big Sur). * Inject a custom CALayer rendering target to avoid wxWidgets drawRect * ENV feature flag this * Formatting * Flag MoltenVK ENV variables on Playback The performance difference for Playback isn't so significant that this makes sense to check for Rosetta 2 and change accordingly. * Pin ffmpeg in CI due to upstream changes. * Try 2.8 I guess.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Reopening this PR given that issues with the Launcher being the culprit were identified and have potential fixes in the pipeline.
Definitely open to more testing, but wanted to open for discussion if nothing else.
This does a few things, detailed below. Tested on the following devices:
2020 M1 MBP, macOS 11.3
2015 Intel i5 MBP, macOS 10.15.7
2020 M1 MBP (8GB), macOS 11.4
2015 MBP, macOS 10.14.6
27inch 5k Retina iMac running 10.14.6, didn't grab year from user
2019 MBP, macOS 10.15.7, didn't grab year from user
On these devices, these patches generally deliver smoother performance with less screen tearing when rendering with Metal. This doesn't fix the odd stuttering for every 8GB M1 Pro, but it has resulted in improved performance for some of them - no idea past this on whether it's hardware or what.
SLP_METAL_DOUBLE_BUFFER) which can be used to tell Metal to use double-buffer instead of triple-buffer behind the scenes. This is only really interesting for iMacs at the moment (Macbooks choke on it), hence why I didn't bother making a UI option for it - mostly want the ability to document it on the wiki and see how it fares for users. This would be possible on a mainline port as well, so I don't think it's going too far into left field either.MVK_CONFIG_SYNCHRONOUS_QUEUE_SUBMITS, which we've documented on the wiki and most users who've tried it report it leading to less screen tearing under Vulkan. This setting helps here, but shouldn't be necessary in a mainline port.Note that the only difference from the original PR is that
MVK_CONFIG_SYNCHRONOUS_QUEUE_SUBMITSis ifdef'd out for Playback, as it exhibits a race condition on M1 Macs running under Rosetta 2.