Which version on which OS is fastest and stable?

[Reiner.Suikat]

Hello,

I have two questions:
Which version on what OS should I use to get good performance and which
version does not crash when doing repeated operations on very large
data-sets?

Current situation:
I am using the old 2.1.50 windows stand-alone binary. This works fairly
well and is the fastest I have tested so far. But it repeatedly crashes
when working with large amount of data, mostly it does manage to put out
an "out of memory" error-message before disappearing. My assumption is
that memory gets fragmented and then it fails to allocate the contigous
memory chunks required for the data arrays. I tried to do a clear in
between to reclaim memory but it doesn't really help.

I have tested both the 2.169 and the 2.92 versions running in a
cygwin-environment. When compiled with gcc as delivered with cygwin both
are extremely slow and I learned from this mailing list that this is due
to sjlj exception handling. So I recompiled gcc not to use
sjlj-exceptions (took almost 2 hours just to compile) and then
recompiled octave.

Result:
It appears to be more stable than the 2.1.50 windows binary, loop
execution has sped up significantly (at least a factor of 10 compared to
the sjlj exception version), but overall execution including file-i/o is
still significantly slower than the 2.1.50 version. 

What about native linux versions? I wouldn't mind going through the
trouble of installing linux as a second os if I knew in advance that
then octave would be faster and more stable. Can anyone share some
insights with me regarding these items? I do need the fastest octave I
can get (needed: binary file-i/o, fft/stft/resample, loop execution) and
it must be able to cope with repeated operations on very large amounts
of data (currently it makes it through one, may be two datasets, then it
crashes on working with the third one).

Any help is greatly appreciated

Thanks
Reiner



-------------------------------------------------------------
Octave is freely available under the terms of the GNU GPL.

Octave's home on the web:  http://www.octave.org
How to fund new projects:  http://www.octave.org/funding.html
Subscription information:  http://www.octave.org/archive.html
-------------------------------------------------------------