30-bit displaying with Krita?

[Boudewijn Rempt]

On Wed, 23 Dec 2009, Kai-Uwe Behrmann wrote:

> In Krita 2.0.2 on xf86_64 Linux image displaying appears to be limited to 
> 8-bit per channel even with OpenGL switched on. The source material 
> is a 16-bit/3-channels gray gradient Tiff file. But the display output 
> looks like 8-bit. Kritas colour picker shows fine differences on a pixel 
> level. So loading appears to be correct. However the Krita display showes 
> lesser differences. Does Krita display its data in OpenGL as 
> GL_RGB16/GL_UNSIGNED_SHORT?
> 
> Any otheridea why Krita gives only 8-bit output?
> 

Adrian Page is the person who can answer this most authoritatively, but I think
that Krita always uses 8 bits rgb textures, unless the colorspace is RGBAF16HALF,
and I think that that colorspace has been removed in favour of the openctl-based
colorspaces, so the following lines in Krita are in error:

./krita/ui/opengl/kis_opengl_image_textures.cpp:                    if (m_image->colorSpace()->id() == "RGBAF16HALF") {
./krita/ui/opengl/kis_opengl_image_textures.cpp:    if (colorSpaceId == "RGBAF16HALF") {
./krita/ui/opengl/kis_opengl_image_textures.cpp:        if (colorSpaceId == "RGBAF16HALF") {
./krita/plugins/formats/tiff/kis_tiff_writer_visitor.cpp:    if (KoID(cs->id()) == KoID("RGBAF16HALF") || KoID(cs->id()) == KoID("RGBAF32")) 

16 bits integer colorspaces are downsampled to 8 bits for display in opengl as well.
I never realized this issue, maybe because I only have laptops and those have very 
crappy displays :-(.

> In my tests a small test OpenGL viewer shows the rendering in 30-bit 
> display mode ("RGB weight 101010") different from Krita and very smooth as 
> expected.

The relevant code is in krita/ui/opengl and definitely needs a bit of maintenance.

Boudewijn