https://bugs.kde.org/show_bug.cgi?id=511269

--- Comment #2 from Shengyu Qu <[email protected]> ---
(In reply to Zamundaaa from comment #1)
> Linear doesn't follow human perception at all either, but gamma 2.2 is quite
> close to what people expect a percentage level of opacity to look like.
> Using linear caused a lot of complaints about things looking bad, in one
> extreme case even completely unreadable, and the opacity slider in Konsole
> not behaving like expected.
> 
> Consistency between Windows and Plasma is a non-goal, we need consistency
> with our own applications and themes. Blending in some gamma space is the
> only way to achieve that.
> 
> If we get a Vulkan compute renderer done at some point, we can do custom
> blending functions in there, if there's Windows applications that depend on
> linear blending. Until we have that and apps can explicitly request a
> specific blending space.
> 
> > Also, with the incoming KMS color pipeline API, color blending using linear 
> > space won't cost as much performance as before.
> It only changes when we can use overlay planes, but nothing else about the
> performance impact when compositing with the GPU.
> 
> (In reply to Shengyu Qu from comment #0)
> > (Off-topic)Also, I'm thinking about whether we should add more detailed
> > options than just a "performance/precision" settings, for example, dedicated
> > option to enable KMS offload and direct scanout.
> We have environment variables for these things, which should be enough for
> debugging and working around driver issues.
> 
> About the actual option to do linear blending, I don't think it would be a
> good idea. There's basically no reason for anyone to want linear blending,
> it only has downsides and would make the code more complicated for no good
> reason.

Thanks for reply. I have another question: Switching between linear and
non-linear would only affect alpha blending? Or other color space operation are
also affected?

-- 
You are receiving this mail because:
You are watching all bug changes.

Reply via email to