On Fri, 2 Jun 2006, Tim Jenness wrote:
> I'll try rerunning the profiler on the entire system rather than just
> clinplot itself.
Nothing much changed except that it turns out that 15% of the runtime is
spent in all the forked ls calls that GRP does (I think it's GRP).
The profiler still only accounts for about half of the elapsed time
though (the timing chart shows gaps where no samples were taken at all -
it looks likle those are the gaps where the line is actually drawn on the
plot in which case, drawing fewer lines on the plot would help a lot.
The other problem with the profiler is that it doesn't go into system
calls so disk I/O also turns up as gaps in the profile.
--
Tim Jenness
JAC software
http://www.jach.hawaii.edu/~timj
|