XRender pipeline - updated benchmark analysis

[Clemens Eisserer]

Hello,

The benchmark results submitted as part of the project documentation
for the XRender pipeline project showed not that impressive results
for swing benchmarks (the non-remote cases) and small operations (e.g.
20x20 image tests).
I did extensive profiling to find the causes and one major problem
seems to be very high per-primitive costs in Xorg-server-1.5 (caused
by an expensive implementation of dixLookupPrivate).
A fix is already in the works.

A more detailed analysis can be found in the updated benchmarks.html attached.

Thanks, Clemens
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.openjdk.java.net/pipermail/challenge-discuss/attachments/20080814/383a19c2/attachment.html