Initial support for DRM async page flip v2 - #1948
Open
cmeissl wants to merge 3 commits into
Open
Conversation
cmeissl
force-pushed
the
feature/async_pageflip
branch
4 times, most recently
from
February 28, 2026 21:43
34c75fb to
5d6e9d7
Compare
Member
|
Works as expected on my integrated AMD Radeon cards 👌 |
|
Works as expected in Niri with very minor changes to it on AMD :) Thank you for the good work |
|
Using #1946 + #1949 on NVIDIA Turing (with the Worth noting that DRM debugger reports the following message: |
|
I've been waiting for this for so long... 🥹 |
cmeissl
force-pushed
the
feature/async_pageflip
branch
from
April 9, 2026 18:49
5d6e9d7 to
80c793c
Compare
Co-Authored-By: Christian Meissl <meissl.christian@gmail.com>
cmeissl
force-pushed
the
feature/async_pageflip
branch
from
April 9, 2026 18:55
80c793c to
f8a8f0f
Compare
cmeissl
marked this pull request as ready for review
April 9, 2026 18:56
Collaborator
Author
Hm, |
|
tested in niri with rtx 4060 mobile (proprietary 610.43.02) via nvidia-drm. kernel 6.18 (arch). async flips engage as needed. Thank you! |
cushycush
added a commit
to cushycush/smithay
that referenced
this pull request
Jul 13, 2026
Adds a sticky PresentationMode { Vsync, Async } to DrmCompositor (and
a DrmOutput passthrough), consulted at flip time. An async flip is the
ordinary atomic request plus PAGE_FLIP_ASYNC (PageFlipFlags::ASYNC on
the legacy backend), gated on DRM_CAP_ATOMIC_ASYNC_PAGE_FLIP /
DRM_CAP_ASYNC_PAGE_FLIP queried once per surface.
The request is otherwise unchanged: IN_FENCE_FD and FB_DAMAGE_CLIPS are
kept (recent kernels exempt both from the async-flip property check, and
keeping the fence means the kernel still waits for the buffer to be
ready, so a tearing flip never scans out an unfinished buffer). A commit
the kernel refuses on the async path is retried synchronously with the
same request inside the surface page_flip; the actual per-frame outcome
is recorded on the pending frame and exposed via pending_presentation_mode
so the caller can classify the completed flip by what really happened.
Carried on top of 298ebc9 for Drift DRIFT-984 until upstream async-flip
support (PR Smithay#1948) lands.
cushycush
added a commit
to cushycush/smithay
that referenced
this pull request
Jul 13, 2026
Adds a sticky PresentationMode { Vsync, Async } to DrmCompositor (and
a DrmOutput passthrough), consulted at flip time. An async flip is the
ordinary atomic request plus PAGE_FLIP_ASYNC (PageFlipFlags::ASYNC on
the legacy backend), gated on DRM_CAP_ATOMIC_ASYNC_PAGE_FLIP /
DRM_CAP_ASYNC_PAGE_FLIP queried once per surface.
The request is otherwise unchanged: IN_FENCE_FD and FB_DAMAGE_CLIPS are
kept (recent kernels exempt both from the async-flip property check, and
keeping the fence means the kernel still waits for the buffer to be
ready, so a tearing flip never scans out an unfinished buffer). A commit
the kernel refuses on the async path is retried synchronously with the
same request inside the surface page_flip; the actual per-frame outcome
is recorded on the pending frame and exposed via pending_presentation_mode
so the caller can classify the completed flip by what really happened.
Carried on top of 298ebc9 for Drift DRIFT-984 until upstream async-flip
support (PR Smithay#1948) lands.
|
Any expectations when this will get merged? So excited for this! |
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
7 participants
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.
Description
This is based on #1335, but takes a slightly different route.
Instead of calculating the
PresentationModeinsideDrmCompositor::render_frameit expects to receivethe desired mode.
This should allow the following things:
the implicit handling might not have enough information about user-preference and calculating it
per frame could result in flip-flip between sync and async page flip which should be avoided imo
this should increase the chance that we can fulfill the async request
Note: This won't work as is on some intel hardware, because not all formats support async page flips.
I will try to come up with a solution, but that can wait for another day.
There is also a new environment variable in anvil to test this, running with
ANVIL_FORCE_TEARING=1shouldenable it.
@PolyMeilex I opened a separate PR because I did not like to overwrite your wonderful work, but I kept you
as the main author. I hope you don't mind.
Checklist