help-octave
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Very slow mesh on MacOS X


From: Dmitri A. Sergatskov
Subject: Re: Very slow mesh on MacOS X
Date: Fri, 6 Oct 2006 11:12:13 -0600

On 10/6/06, Marcus Vinicius Eiffle Duarte <address@hidden> wrote:
I guess I was a little confusing in my previous post.

I was originally using mesh. When I read your suggestion of modifying the
"__gnuplot_raw__" lines, I read mesh.m and didn't fid such lines in it -
mesh.m uses __gnuplot_set__. So I tried modifying mesh.m to use
__gnuplot_set__ nohidden3d, but the resulting plot was  also very slow (like
the plot done with the unmodified mesh).

I guess mesh.m is slightly different between 2.1.73 and 2.9.9 version (that
is the version I am using).
In any case here is a simple benchmark I did: I created the file tmesh

address@hidden octave]$ cat tmesh
#!/usr/bin/octave -q
sombrero(300);
quit

Then I run it with modified and unmodified mesh.m (I also added
tic/toc timer into
mesh.m to see how much time octave spends preparing data for gnuplot). I tried
both x11 and wxt terminal on 2.2GHz Athlon64 w/ 1M cache and 1 GB of RAM
running 64-bit Fedora Core 5 (will try Aquaterm when I get my hands on Mac).

x11, no hidden lines removed:

address@hidden octave]$ time ./tmesh
Elapsed time is 0.193909 seconds.

real    0m3.743s
user    0m3.024s
sys     0m0.344s

x11, hidden lines removed:

address@hidden octave]$ time ./tmesh
Elapsed time is 0.080994 seconds.

real    1m43.943s
user    1m39.346s
sys     0m1.864s

-------

wxt, NO hidden lines removed:

address@hidden octave]$ time ./tmesh
Elapsed time is 0.107468 seconds.

real    0m7.189s
user    0m6.616s
sys     0m0.408s

wxt, hidden lines removed:

address@hidden octave]$ time ./tmesh
Elapsed time is 0.084267 seconds.

real    1m42.784s
user    1m41.374s
sys     0m1.012s

===================

This is with 4.1 (development) version of gnuplot.
I did similar test ~2 years ago comparing gnuplot to plplot, and plplot
was _much_ faster. Unfortunately it was not as stable as gnuplot.

That's why I guess the problem lies in AquaTerm, and not in gnuplot.

Perhaps Aquaterm has its own problems...


Thanks,

Marcus Vinicius



Sincerely,

Dmitri.
--


reply via email to

[Prev in Thread] Current Thread [Next in Thread]