On Nov 4, 2007 6:18 PM, Jarrod Millman wrote:

On Nov 4, 2007 11:53 AM, Emanuel Woiski 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 1.0rc2and 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.... Thanks woiski

Thanks. I will try them later and let you know. Any chance to have those for 2.4 as well?:) If it works, I will be able to upgrade to matplotlib 0.90.1.... Regards woiski On Nov 6, 2007 5:25 AM, David Cournapeau wrote:

Emanuel Woiski wrote:

...

On Nov 4, 2007 6:18 PM, Jarrod Millman mailto:millman@berkeley.edu> wrote:

On Nov 4, 2007 11:53 AM, Emanuel Woiski 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 _______________________________________________ Numpy-discussion mailing list Numpy-discussion@scipy.org http://projects.scipy.org/mailman/listinfo/numpy-discussion

yep I have no choice - those are a bunch of lab machines....:) thanks anyway regards woiski On Nov 6, 2007 9:59 AM, David Cournapeau wrote:

...

Thanks. I will try them later and let you know. Any chance to have those for 2.4 as well?:) If it works, I will be able to upgrade to matplotlib 0.90.1.... Well, you will have to find someone else. Using windows is already

Emanuel Woiski wrote: painful enough: I don't have the motivation to build for every supported python versions :)

cheers,

David _______________________________________________ Numpy-discussion mailing list Numpy-discussion@scipy.org http://projects.scipy.org/mailman/listinfo/numpy-discussion

On Nov 6, 2007 10:57 AM, David Cournapeau wrote:

Emanuel Woiski wrote:

...

yep I have no choice - those are a bunch of lab machines....:) thanks anyway regards woiski Can't you build it yourself ? You just have to install mingw, and compiling blas/lapack is not too difficult.

cheers,

Well, actually I did it and all the tests came out OK for 'my' numpy 1.0.4whereas crash (something related to msvcrt71.dll) for 'my' scipy 0.6.0. However, all my numpy/scipy codes seem to be working perfectly now :).... You can find them here (look down the page for numpy and scipy BUILT files):

http://200.145.244.188/Disciplinas/teamspaces/material-de-python-comum-as-di... Portuguese...) BTW I have used ATLAS binaries (PII) from the Scipy site and Enthought mingw installation... regards, woiski