| Summary: | When Night Light is active Plasma is unusable on Intel GPU | ||
|---|---|---|---|
| Product: | [Plasma] kwin | Reporter: | Martin <martin.tk> |
| Component: | night color | Assignee: | KWin default assignee <kwin-bugs-null> |
| Status: | RESOLVED NOT A BUG | ||
| Severity: | normal | CC: | nate, xaver.hugl |
| Priority: | NOR | ||
| Version First Reported In: | 6.5.4 | ||
| Target Milestone: | --- | ||
| Platform: | Fedora RPMs | ||
| OS: | Linux | ||
| Latest Commit: | Version Fixed/Implemented In: | ||
| Sentry Crash Report: | |||
|
Description
Martin
2025-12-28 16:19:50 UTC
> Graphics Processor: llvmpipe
You don't have working hardware acceleration, likely because the GPU is just too old to run shaders with more than a few instructions.
Color management features like night light require a bit more complex shader, and that affects llvmpipe performance badly. It's only faster on X11 and older versions of Plasma because night light isn't fully implemented there.
With a core duo, even slightly complex shaders in llvmpipe being very slow isn't really unexpected. To be honest, I'm surprised it's even usable without night light.
(In reply to Zamundaaa from comment #1) > > Graphics Processor: llvmpipe > You don't have working hardware acceleration, likely because the GPU is just > too old to run shaders with more than a few instructions. > > Color management features like night light require a bit more complex > shader, and that affects llvmpipe performance badly. It's only faster on X11 > and older versions of Plasma because night light isn't fully implemented > there. > > With a core duo, even slightly complex shaders in llvmpipe being very slow > isn't really unexpected. To be honest, I'm surprised it's even usable > without night light. That's disappointing and surprising. It used to work fine on Wayland before 6.2 (or 6.3, I can't remember when exactly this started). And there are no such issues when I use X11 session. Night light engages and there's no slow down. So the hardware can manage it. It seems like the implementation changed at some point and is highly inefficient. It really only has a few more instructions in the shader vs. the normal one. It doesn't seem surprising to me that a 19 year old laptop CPU struggles with that though. Like I said already, it's astonishing it's even fast enough for direct copying with the simpler shaders. |