[Help-glpk] Re: glpk and ATLAS/GSL.

[Andrew Makhorin]

>  i have profiled the netlib DFL001 problem on different processors
>  (athlon, p4, alpha and ppc) and found out that the most time
>  is spent on LU decomposition and backsubstitution.
>  Is it possible to use CLAPACK routines from 
>  GSL or better ATLAS instead of the "native glpk" implementation ?

LAPACK (and therefore CLAPACK) does not have routines for general sparse
matrices. Which routines do you mean?

>  Is there any clean way to hide the ``for(i=1;i<=n)'' constructs in 
>  lpx_* API  for gsl_* API ?

The question is unclear.


Andrew Makhorin