feat(present): offload Present copy compositing to the GPU - #1053
Merged
Conversation
Member
Author
|
For some comparison: Upstream: This PR version: ~10% higher vkmark score, so I consider it is an improvement. Tested with mesa-vulkan-icd-wrapper. |
Member
Author
|
@robertkirkman probably you will like that. Can you please test this one? |
Member
|
the device i currently am able to access cannot run glmark2 because of a closed source android 10 phantom process killer of One UI 2.5, so i tested using luanti and in game fps counter, luanti, adreno 630, 2960x1440
i cannot measure an improvement, but there is no introduced issue. |
robertkirkman
approved these changes
Jul 8, 2026
Member
Author
|
In the case if you did not see the issue probably you did not hit the accelerated path. There were at least two bugs before the force-push. You can check in termux's logcat for updated frame counter, which now includes gpu offload stats (if there was a gpu offload). |
Member
|
I see this and this is this PR's logcat from termux what kind of program should i run that may trigger different behavior? also, do you mean to watch termux logcat or termux:x11 logcat? |
Member
Author
|
I tested it only with mesa-vulkan-icd-wrapper + vkmark. |
Member
|
i will be able to run vkmark later in a different device |
twaik
force-pushed
the
gpu-present-copy
branch
2 times, most recently
from
July 9, 2026 13:19
7442d02 to
f9a2811
Compare
Member
|
Here are Adreno 612 master 151 PR 143 Samsung Galaxy A70 SM-A705FN One UI 8 Android 16
|
Member
Author
|
Well, it is in the margin of error. |
Member
|
Should I test the new version and send logcat result? |
Member
Author
|
I think so. @Pipetto-crypto can you please help us? 
|
Member
|
Member
Author
|
The log does not say if it works fine on your device or not. |
Member
|
For me, vkcube works. Unfortunately, on most of my devices vulkan-wrapper-android did not work very well, particularly with zink, as of March 15 2025, so i do not usually try it, but if I should try it here let me know. |
Member
vkcube works for me and it does not look stretched. sorry i was making an additional message but I posted the log first. |
Member
|
I think that people testing graphics drivers and Termux:X11 should know that I have recently added a new OpenGL test, pkg upgrade
pkg install xscreensaver
export PATH=$PREFIX/libexec/xscreensaver:$PATH
geodesicgearsit is much more intensive than For me it is working fine with this PR. |
Present clients that can't use the flip fast path (windowed, partial damage) now get composited into the root texture via the renderer's GL context instead of a CPU CopyArea. -disable-gpu-present forces the old path; a lost renderer connection also falls back to CPU.
Member
Author
|
Seems like the stretching bug was fixed. |
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
2 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.
Present clients that can't use the flip fast path (windowed, partial damage) now get composited into the root texture via the renderer's GL context instead of a CPU CopyArea.
-disable-gpu-presentforces the old path; a lost renderer connection also falls back to CPU.