Quoting Maarten Lankhorst (2018-12-03 14:00:31)
> The high precision gradient path was falling back to the 8-bit path,
> then expanding it to float. This removed any advantage of the high
> precision path, so make sure our gradients are computed as floats
> by adding a separate path.
>
> Signed-of
The high precision gradient path was falling back to the 8-bit path,
then expanding it to float. This removed any advantage of the high
precision path, so make sure our gradients are computed as floats
by adding a separate path.
Signed-off-by: Maarten Lankhorst
---
pixman/pixman-conical-gradient