[cairo] [cairo-announce] [ANNOUNCE] pixman release 0.15.20 (Release Candidate) now available

[Siarhei Siamashka]

On Wednesday 12 August 2009, Soeren Sandmann wrote:
> Koen Kooi <k.kooi at student.utwente.nl> writes:
> > Siarhei has some more improvements for ARM and you have the
> > glyph-fast-path, fast-path-cache, faster-fast-path, validate and
> > pixel-fetch branches, would those go into the 0.16.x series, or is all
> > that 0.17.x material?
>
> The glyph-fast-path, validate, and pixel-fetch branches are part of
> 0.15.20. The others, along with additional ARM improvements, are
> 0.17.x material.
>
> Basically, it has to be a performance regression, and a quite serious
> one at this point, to get into 0.16.0.

It's understandable. Any extras can exist as separate patchsets, as long
as pixman core is stable.

> Btw., I would appreciate if you could try blitter-test and see if it
> passes on ARM.

It does not. Function 'neon_composite_over_n_8_0565' is not completely
identical to C path, we discussed it on irc. This problem was not
detected by an earlier variant of blitters-test, that's why I did not
notice it before. Probably it is because the test was increasing the
size of test images too slow (progressively increasing sizes of images
while testing is done in order to have higher probability of spotting
problems on simpler testcases involving small sizes, which are easier to
debug).

But just disabling this function results in quite a measurable performance
regression on text operations.

-- 
Best regards,
Siarhei Siamashka