[Benchmarking] Java scalability issue slide

[Andrea Aime]

I already tried it, and Pisces is a tragedy :-)
It indeed scales up pretty nicely but it starts so slow that it's not
worth using, it reaches the speed of Ductus somewhere around 32
threads....
However, I tried it only in synthetic tests that were hitting only
java2d, so it might be worth looking into it again using a real
testing setup.
I also looked into the attempt to use Cairo rendering, which makes for
nicer output actually, but again does not scale up well (and it's
slower
than Ductus at lower threads):
http://linuxhippy.blogspot.com/2010/06/xrender-pipeline-now-in-jdk7-master.html

I'd be interested in hearing what your findings are if you try that out

Cheers
Andrea