Emanuel Woiski wrote:
On Nov 4, 2007 6:18 PM, Jarrod Millman <millman@berkeley.edu <mailto:millman@berkeley.edu>> wrote:
On Nov 4, 2007 11:53 AM, Emanuel Woiski <woiski@gmail.com <mailto:woiski@gmail.com>> wrote: > I am havig a lot of trouble for quite some time wih numpy - scipy Windows > binaries for old AMD (pre-XP) processors. Since I teach in a lab with lots > of machines using that kind of processor, I got stuck with numpy 1.0rc2 and > scipy 0.5.1, which is, believe me, the only combination that woks...Well not > only tests fail but Python crashes! > FYI there are many people over the world still using not-so-recent > computers... > Yes I have tried s0.6 and n1.0.3.1 without success.
Are you building from source? If so, have you tried the most recent development code?
Nope, just WinXP (and WIn2000) binaries....
The problem is a distribution problem (that is how the binaries were built). Do you need optimized BLAS/LAPACK (for fast linear algebra) ? If not, would you mind trying those ? http://www.ar.media.kyoto-u.ac.jp/members/david/archives/numpy-1.0.3.1.win32... http://www.ar.media.kyoto-u.ac.jp/members/david/archives/scipy-0.6.0.win32-p... Note that those do not use any optimized libraries (no ATLAS, no fftw, etc...). This is the bare minimal to make it work on any x86 machine (no need for SSE/SSE2). cheers, David