help-octave
[Top][All Lists]
Advanced

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

planning to try 64 bit octave with Atlas 3.8


From: John W. Eaton
Subject: planning to try 64 bit octave with Atlas 3.8
Date: Sat, 20 Oct 2007 10:13:41 -0400

On 19-Oct-2007, Michael Creel wrote:

| > If you compile with --enable-64, you will see the following message at
| > the end of the configure run:
| > 
| > configure: WARNING: You used the EXPERIMENTAL --enable-64 option.
| > configure: WARNING: Are you sure that is what you want to do?
| > configure: WARNING: 
| > configure: WARNING: You must ensure that the Fortran compiler generates
| > configure: WARNING: code with 8 byte signed INTEGER values, and that your
| > configure: WARNING: BLAS and LAPACK libraries are compiled to use 8 byte
| > configure: WARNING: signed integers for array indexing.
| > 
| > In addition to the library/compiler issues, there are a number of
| > other things that are not comletely implemented.  People who are
| > interested in seeing a fully functional 64-bit aware version of Octave
| > are encouraged to contribute.
| > 
| > jwe
| > _______________________________________________
| > Help-octave mailing list
| > address@hidden
| > https://www.cae.wisc.edu/mailman/listinfo/help-octave
| > 
| > 
| 
| Anticipating some good performance improvements due to threading on multiple
| cores,  I'm planning on trying to make a 64 bit Octave using Atlas 3.8, so
| I'd appreciate any pointers about how to do so before wading in. In addition
| to atlas, what else might I need to self-compile? I'm using Debian. Thanks,

You'll need to compile the Fortran bits in a way that INTEGERs are
8-bytes, which probably means using gfortran -fdefault-integer-8.

Octave doesn't really know about large files, so loading
multi-gigabyte files will probably fail.

The binary file formats are 32-bit, so saving large arrays will
probably cause trouble.

There are probably other things I can't think of at the moment (or
haven't thought of at all; that's why it is still experimental).

jwe


reply via email to

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