I noticed that fps drops heavily when dumping images to file, but that
processor usage isn't often
100%
Initially I thought that this was because ImageIO.write flushes its output, and
that perhaps it was
spending a while blocking while the OS waits on the disk. On further
inspection, it seems like most
of the time went into actually encoding the image rather than writing it out.
In any case, a fairly straightforward thread pool implementation has given ~40%
speedup in my
tests. On more multiprocessor machines I'd expect an even more substantial
improvement. I've
attached a patch with my code.
More info here:
http://github.com/rictic/code_swarm/commit/ea68f4dbc5a88605379225263a1853838c3a0
581
Original issue reported on code.google.com by ric...@gmail.com on 24 Aug 2008 at 2:17
Original issue reported on code.google.com by
ric...@gmail.com
on 24 Aug 2008 at 2:17Attachments: