[Pixman] [PATCH 5/8] pixman-access.c: Add floating point accessor functions

[Alexander Larsson]

On lör, 2012-08-25 at 21:45 -0400, Søren Sandmann Pedersen wrote:
> The other direction is currently done with 4096-entry table which
> works fine for 16 bit integers, but not so great for floating
> point. So instead this patch uses a binary search in the sRGB->linear
> table. The existing 32 bit accessors for the sRGB format are also
> converted to use this method.

I implemented a custom x^2.4 and x^(1/2.4) for doubles in babl:
http://git.gnome.org/browse/babl/commit/?id=373cb5af0d511ded2005a6ee9a9330c5319b8acd

If you convert it to float and limit the number of chebychev terms it
might be fast enough for pixman.