[Pixman] [PATCH] ARMv6: New fast path implementations that utilise prefetch

[Ben Avison]

On Fri, 11 Jan 2013, Siarhei Siamashka <siarhei.siamashka at gmail.com> wrote:
> The "t-firefox-chalkboard" test seems to have regressed a bit. Is it a
> random noise in the measurements? Or one of the optimizations is poorly
> performing here?

Yes, that was by far the most noticeable regression, although by much less
than most of the improvements in other tests. My best guess is that it was
hitting the same problem as the L1 lowlevel-blt-bench tests for the blit
operations. I've been doing a bit of work on that (and fixing a couple of
bugs that also lead to additional performance improvements) over the last
few days, so I hope that when I next run the cairo benchmarks this may
improve. If so, I will probably post a supplementary patch sooner rather
than later.

Ben