RGB 16 bit as default when starting Krita

[Cyrille Berger]

On Wednesday 20 January 2010, Martin Renold wrote:
> > Third. You must use 16 bit or float for scRGB. Do you think degrading
> > performance twice worth [...]
> 
> If you don't mind a non-Krita developer commenting: I would not be very
> surprised if a high quality operation with all the proper rounding and
> dithering in 8bpc is actually slower than doing the same in 16bpc with
> straight forward math.  Of course much also depends on how carefully
> optimized it is, and my guess might also be wrong.

It goes in the same direction as what pippin said on IRC:
18:47 < pippin> in my testing, 32bit float/component can actually be faster 
than 8bit char/component
18:47 < pippin> since modern cpus actually have to struggle when operating on 
chunks of 8bit memory...
18:47 < pippin> while they are designed to work on wider things

And of course usuage of memcpy balance in favor of 8bits. For the sake of the 
discussion, it would be interesting to have real benchmarks on the subject.


> On another point in this thread, I did write in my blog post that I found
> painting in linear light counter-intuitive.  I should add that I tried it
> only for a few hours, and some people who also tried it (with other
> programs) strongly disagree with me.  I did not take enough time to get
>  used to it first, maybe I would have liked it in the end (or not).

I will write a longer explanation/answer on the two color spaces, but out of 
curiosity, did you do artwork with real-life painting before ? Or did you 
started with digital ? If you did started with digital, it would explain a 
bias toward sRGB painting.

-- 
Cyrille Berger