[SciPy-dev] Re: Linalg2

Pearu Peterson pearu at cens.ioc.ee
Mon Mar 25 02:41:11 CST 2002


On Sun, 24 Mar 2002, Travis Oliphant wrote:

> Nothing should be lost.  I just renamed the files (e.g. from fblas to fblas2) 
> so that they could be installed along with linalg.  Otherwise distutils 
> writes linalg2/fblas over linalg/fblas  or vice-versa depending on which one 
> was installed first.
> 
> It's possible I forgot to commit a file.   For example, do the fblas.pyf (now 
> fblas2.pyf) files need to be committed or are they automatically generated 
> from the generic_fblas.pyf files?
> 
> Can you be more specific about what you think is "lost."

See
http://scipy.net/cgi-bin/viewcvs.cgi/scipy/linalg2/generic_fblas2.pyf.diff?r1=1.2&r2=1.3&sortby=date

Previously there were files

generic_fblas.pyf         <- this includes what follows
generic_fblas1.pyf        <- signatures for blas 1 routines
generic_fblas2.pyf        <- signatures for blas 2 routines

But then, I suspect, you did

mv generic_fblas.pyf generic_fblas2.pyf

so that you overwrite an existing file when renaming and now we have

generic_fblas1.pyf
generic_fblas2.pyf

As a result one file is lost that contained wrappers to BLAS 2 routines
gemv,hemv,symv,trmv.

I think there was no need to rename all these files.
Use the following

def process_all():
    # process the standard files.
    for name in ['fblas2','cblas2','clapack2','flapack2']:
        generate_interface(name,'generic_%s.pyf'%(name[:-1]),name+'.pyf')

in interface_gen.py.

Regards,
	Pearu




More information about the Scipy-dev mailing list