[Pixman] [PATCH/RFC] Use pixman_fixed_16_16_t type in _pixman_gradient_walker_{reset/pixel}

[Luca Barbato]

On 02/19/2010 12:31 PM, Siarhei Siamashka wrote:
> From: Siarhei Siamashka <siarhei.siamashka at nokia.com>
> 
> This argument is explicitly cast to 'int' or 'int32_t' anyway when it
> gets processed, except for 'if (pos < stops[n].x)' construct.
> 
> And at least the value coming from radial gradients processing code is
> expected to be in [0.0, 1.0], which should fit into pixman_fixed_16_16_t
> data type nicely.

The stddev variation seems to say that's not always true or at least the
rounding changes are present.

I wonder if that difference is worth having a configure switch to change
the types so who is more concerned to precision (and has an output
target large enough to have such difference perceptible) could use the
larger formats and who would enjoy the speedup without concerns could
use the smaller one.

lu

-- 

Luca Barbato
Gentoo/linux
http://dev.gentoo.org/~lu_zero