Tweet
Originally posted by indepe
View Post
-
There is a backport patch available for plasma 6.0.x, provided from Xaver if distributions want to include it. -
from the mailing list
i am so confused by this.All Vulkan drivers have gained support for explicit synchronisation on
Wayland and X11.
My understanding is that x11 and wayland explicit sync support is related to "vulkan window system integration", though i don't know what does that even mean.
i don't understand then what is the point of this https://gitlab.freedesktop.org/mesa/...requests/18032.Comment
-
I think the bit done is to allow desktop compositors to use explicit sync while you are running vulkan apps on top of them. Not that the vulkan app itself is, which would be that MR you linked to.Originally posted by Ray_o View PostComment
-
Well, no performance improvement was expected. Though I was surprised to find explicit sync can be ~10% slower than implicit sync with light apps: mesa!29199Originally posted by RejectModernity View PostJust tested mesa-git with kwin with explicit sync patches on AMD. Zero improvements. ZERO.
They wrote "kwin with explicit sync patches", which presumably means with those patches applied manually.Originally posted by indepe View Post
Don't feel bad, you're in good company here. The fact that "explicit sync" isn't a single thing, but many things in different contexts, is mighty confusing to many people.Originally posted by Ray_o View Post
It's about synchronization of GPU access to buffers shared between the client and display server via the Wayland protocol and X11 DRI3/Present extensions, to ensure correct presentation of client frames.My understanding is that x11 and wayland explicit sync support is related to "vulkan window system integration", though i don't know what does that even mean.
It's about driver-internal synchronization of GPU access. In contrast to explicit sync in the display protocols (which is independent and not required for this), this actually improves performance with some real-world apps.i don't understand then what is the point of this https://gitlab.freedesktop.org/mesa/...requests/18032.Comment
-
-
Originally posted by MrCooper View Post
Still, exactly because it isn't a "single thing", I wouldn't expect such patches to resolve all the (unspecified) sync problems they might expect. I'd expect that Kwin will need some adjustments before the release in mid-June, and even then perhaps not result in all possible sync improvements right away.Originally posted by MrCooper View Post
Also, on AMD generally things are already working quite well (at least in my own very limited experience), so I wouldn't expect everyone will easily see improvements just from some general "support". (I probably won't.)Last edited by indepe; 23 May 2024, 10:03 AM.Comment
-
Frame pacing will be better when the system is stressed, especially with older hardware. Not to mention this speeds up adaptation of Linux now that NVIDIA works (better)Originally posted by RejectModernity View PostJust tested mesa-git with kwin with explicit sync patches on AMD. Zero improvements. ZERO.
There's also more improvements on the way
Micro optimizations are awesome!
(PS.. Some parts are still not finalized (like wlroots) or pending in git (such as kwin))
Last edited by Kjell; 23 May 2024, 03:50 PM.Comment
-
Yeah, that's my MR. I wouldn't characterize narrowing (not closing) a ~10% performance gap to implicit sync as "more improvements" though, more like "damage control".Originally posted by Kjell View PostThere's also more improvements on the way
Micro optimizations are awesome!
Comment
-
Originally posted by MrCooper View Post
Yeah, that's my MR. I wouldn't characterize narrowing (not closing) a ~10% performance gap to implicit sync as "more improvements" though, more like "damage control".
However eventually the gap will be closed (if not turned around), or is there something standing in the way of that?Comment
Comment