[MMTK] MMTK and NumPy

[Konrad Hinsen]

On 30.05.2007, at 13:26, Gisle Sælensminde wrote:

> In a thread with the name "Segmentation fault, normal modes  
> calculation", it is said (by Konrad) that you need to use MMTK with  
> Python 2.5 + NumPy to use arrays bigger than 4GB, which I interpret  
> the way that you _can_ use MMTK with NumPy instead of Numeric,  
> which you could not do earlier. Is this right?

Yes. I am using this combination on one of my daily work machines, so  
it does get some testing. Nevertheless, there may still be problems;  
I am currently at about one NumPy-related fix per week.

> I now try to configure and install MMTK with numpy instead of  
> Numeric, but can't find any instructions on how to do so. It may be  
> that I have been looking in the wrong place. How can I do that?

You need to use Scientific Python 2.7.x and MMTK 2.15.x. Ideally,  
take the current code base from the Mercurial archives (http:// 
dirac.cnrs-orleans.fr/hg/) because I have made a number of NumPy- 
related fixes recently and didn't yet find the time to make new  
releases.

To install ScientificPython with NumPy, do

	python setup.py build --numpy
	python setup.py install --numpy

Then install MMTK as usually; it will use the same array package as  
ScientificPython.

MMTK comes with a growing number of unit tests (directory Tests); if  
your installation passes all of them, it is probably as good as mine.

Konrad.
--
---------------------------------------------------------------------
Konrad Hinsen
Centre de Biophysique Moléculaire, CNRS Orléans
Synchrotron Soleil - Division Expériences
Saint Aubin - BP 48
91192 Gif sur Yvette Cedex, France
Tel. +33-1 69 35 97 15
E-Mail: hinsen at cnrs-orleans.fr
---------------------------------------------------------------------