gnuplot development

I'm using gnuplot 4.2 patchlevel 3.

I have a very large dataset with small differences in numbers, but many
points get plotted as the same point in the .eps file, seriously increasing
render time vs. when you strip out the duplicate points by hand:

~/hpc/proj1/bug$ time epstopdf chiba_final3.eps --outfile=/dev/null

real 0m12.419s
user 0m9.909s
sys 0m1.944s

~/hpc/proj1/bug$ time epstopdf chiba_final3.min.eps --outfile=/dev/null

real 0m1.820s
user 0m1.364s
sys 0m0.312s

My attached script eps_reducer.sh is modified exactly to work on my plots
generated by the attached plot.sh; it does a `sort -u` on all the cross
data of the plots, and the rest of the file is untouched.

I wanted to include my original test data (250MB -> 35MB bzipped), but it
doesn't look like SF will let me upload such a large file, so I'll see
later about uploading that... In the meantime, I'm attaching an original
.eps file and its "minified" version - I'll see if I can upload some
original data later if more space is available somewhere.