Bill Spitzak wrote:
Maybe the client should be able to query the color space of the output
and can set the color space of it's buffer to the same one, this would
indicate that no color correction is needed. The compositor could still
color-correct it from one output's color space to another's if it is on
multiple outputs.
This is not as simple to set and as safe as an explicit opt out.
(Apple has had a lot of trouble as a result of doing their
color management opt out this way). It's hard to confirm that you've got
the same profile ("same" in what sense ? Same instance of the profile ?
Same bits ? Same notional colorspace ?)
You would need a special case anyway to implement a null transform,
because things like cLUT profiles have A2B and B2A tables that are not
perfect inverses.
You'd also want to be sure that a null transform is bit perfect, and
that something isn't lost in shaders etc.
If this idea is still attractve, I would suggest supporting device links,
making it reasonably simple to set a null device link.
Ideally a client should be able to do this with just some strings (ie
query the color space and send the same string as the buffer's color
space) so it does not have to open a CMS library that it might otherwise
not be using.
If that's the case, I'd suggest having a string that turns the color management
off. Much simpler and more direct as to the intention.
Graeme Gill.
_______________________________________________
wayland-devel mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/wayland-devel
