Pixman Has Improved Gradients, Is Much Faster

[phoronix]

Phoronix: Pixman Has Improved Gradients, Is Much Faster

While Cairo is frequently mentioned on Phoronix, mentioned less but used by Cairo (as well as the X.Org Server) for pixel manipulation is the Pixman library. Soeren Sandmann announced a new release candidate of Pixman 0.19.6 this afternoon and it has a few interesting changes worth noting...

http://www.phoronix.com/vr.php?view=ODY5Ng

[markg85]

Any intention of doing benchmarks for that gradient part? I wonder how much faster it got compared to the previous release.

[curaga]

This reminds me, does pixman have a regression suite/tests for graphical corruptions?

If there's none, I suppose it would be useful if Phoronix could help there.

I recall once an update to a stable pixman version caused text corruption in gtk, that was luckily gone when the next 0.x major stable branch came out.
Another time a pixman update stopped WM icons from showing.

[sandmann]

Any intention of doing benchmarks for that gradient part? I wonder how much faster it got compared to the previous release.

The gradient improvements are mainly for correctness: The radial ones are now drawn with PDF semantics; the linear ones uses more accurately drawn, which is important for large images.

We do hope to get some significant performance improvements at some point.

[sandmann]

This reminds me, does pixman have a regression suite/tests for graphical corruptions?


If there's none, I suppose it would be useful if Phoronix could help there.

I recall once an update to a stable pixman version caused text corruption in gtk, that was luckily gone when the next 0.x major stable branch came out.
Another time a pixman update stopped WM icons from showing.

Yes, we do have a test suite that tests for graphical corruption among other things. If you notice bugs after a pixman update, please file bugs or send mail to pixman@lists.freedesktop.org. We tend to respond pretty quickly.