Maximum Variable Size in Octave

[John W. Eaton]

On  4-Sep-2008, Przemek Klosowski wrote:

| 	I am planning memory-intensive calculations in Octave and/or Matlab.
|    The Mathwork has published the following document describing the
|    maximum variable size in Matlab on various platforms:
|    http://www.mathworks.com/support/tech-notes/1100/1110.html
| 	Is there an equivalent document for Octave?
| 
| For what it is worth, I compiled octave on a 64-bit itanium system with 32 GB
| of memory and allocated over 10 GB arrays. 

If you use --enable-64 (with care to ensure that the libraries you
link with are also compiled properly for 64-bit indexing; it's still
an experimental feature) then you should be able to allocate arrays up
to roughly 2^63 bytes if your operating system will allow it.  But you
can't read or write arrays larger than approximately 2^31 bytes since
no one has yet done the work to fix the I/O parts of Octave to be
64-bit aware.

On 32-bit systems, you can allocate arrays up to approximately 2^31
bytes, so that is 2^23 double precision elements.  If you have enough
memory, you can allocate many arrays of this size, but any single
array can't be larger than 2^31 bytes.

jwe