Re: Plotting with 3.2.0 on Windows is SLOOOOOOWWWW

[John W. Eaton]

On 29-Jul-2009, Matthias Brennwald wrote:

| On Jul 29, 2009, at 2:17 AM, Tatsuro MATSUOKA wrote:
| 
| > Hello
| >
| > Hmmmmm!
| >
| > The slowness seem to strongly depends on the system.
| >
| > In the test previous post performed on the computer in my home  
| > (Windows XP home Cereron M 1.2Ghz 512Mb
| > Memory).
| > I have tested the same tests by Matthias Brennwald on the computer  
| > the university (Windows XP
| > professionan HT Pentium 3.4GHz 1GB Memory)
| >
| > The speed of the tests on the octave 3.2 (3.2.0 MinGW distributed on  
| > the OctaveForge and Octave 3.2.2
| > built myself)is comparable to octave 3.0.5 MinGW distributed on the  
| > OctaveForge).
| > Of course the result for 3.0.5 is faster than that ion the computer  
| > in mu home.  However it is
| > reasonable to the different in CPU performance.  However the  
| > difference octave 3.2 on MinGW is too big
| > to attribute the cpu performance.
| >
| >
| > Note that I do not use John's patch for this test.
| >
| > For the computer in my university, the John's patch seem to be not  
| > needed.
| > I will test again when I return to my home.
| >
| > Hello Matthias Brennwald
| >
| > Can you perform the test with John's patch by yourself?
| >
| > Regards
| >
| > Tatsuro
| 
| I don't think I am able to add this patch and recompile Octave myself  
| on Windows. I am an absolute idiot when it comes to Windows, and I  
| have never compiled anything on this platform. I don't even know what  
| compilers and IDEs there are. So, I'd be very grateful if someone else  
| with more experience could try this.

The change I sent was a single character change to one .m file.  You
don't need to recompile Octave to test it.  Just edit the file
gnuplot_drawnow.m and change the second occurrence of

  plot_stream = __gnuplot_open_stream__ (2, h);

to

  plot_stream = __gnuplot_open_stream__ (1, h);

then restart Octave and try your plotting command again.  Is it
faster?  I'd be surprised if it is, but some obscure difference in the
way that popen2 vs. popen is connecting to gnuplot is the only thing I
can think of at the moment that might cause a difference in the speed
of plotting.  But this is just a guess.

jwe