Open xlucn opened 3 years ago
On another machine I got around 30 fps, which is quite enough. Don't know why.
This could be for a few reasons, but what values do you have for the following variables in your picom.conf?
transition-length, transition-pow-(all of them)
I don't think those parameters are the problem, since I have tried both default and some slow motion parameters (probably length 1000 and pow-* 0.9) on two of my Arch Linux laptops, only one of it has low fps.
I just tested again with no config --config /dev/null
, the one that has the issue now has even lower fps ~ 8. I might restart the laptop later and see if it is still like this.
I see, thanks for confirming that. It may be similar to an issue I see on another machine of mine where the animations will slow over time (over the course of a few days the smoothing gets this slower FPS like motion that you're describing). I don't really know what's the source of it yet since I haven't been able to identify where this leak occurs, and under what scenarios besides uptime. does it happen in both backends? xrender and glx?
does it happen in both backends? xrender and glx?
Both, actually xrender backend is even slower now
Just a thought, I might later try the commit which only this animation patch is added, just to be sure that the issue is brought by that part of the code.
I've heard, but haven't tested or verified, that Ibhagwan's patch actually is the source of the performance leak over time. Just to note, a reboot always solves my problem when it happens to me, what about you?
The reason I thought of that is now just running picom in xrender backend not doing anything causes cpu to go 30% - 70% (of one core), even have core dump if without --experimental-backends
. I don't quite believe a simple (I saw the animation patch was only 200+ lines of code) animation can do all that.
Just to note, a reboot always solves my problem when it happens to me, what about you?
Not rebooted yet, I have some programs running, have to reboot later. But I think this issue should better be solved without rebooting.
Actually the animation patch is indeed causing the other problems (like the core dump, on both my laptops) but I don't have the expertise to debug it. Not sure if the fps issue is related.
51fdb8bcaafc4e1261a644873358031c37e09f83 is the first bad commit
commit 51fdb8bcaafc4e1261a644873358031c37e09f83
Author: blackcapcoder <mailhph@gmail.com>
Date: Sat Oct 7 15:22:37 2017 +0200
If I have more time I might look more into the slow fps issue, although it will be quite hard for me to find any clue
I don't know how to get fps for picom in a proper way, but I recorded desktop in 60fps and set a long transition length. What I see is for each window move, the video has played for 3 to 4 (more towards 4) frames. So my estimation is the fps picom's animation is only around 15. Could it get somewhere about 4x faster?
https://user-images.githubusercontent.com/12032219/103174362-7631b800-489c-11eb-83a2-8941335301f4.mp4