Re: [ublas] About SuperLU and other stuff

From: Paul C. Leopardi (leopardi_at_[hidden])
Date: 2005-05-22 20:11:34

Dear Luca,
Reply below. Best regards
On Sun, 22 May 2005 02:46 am, luca regini wrote:
> I am writing a scientific application. For the moment my vectors are just
> dinamically allocated array of doubles.
> To use UBLAS should i rewrite the code to use UBLAS' vector or is there an
> adapter??
> After performing some sparse matrix computations i must solve a system
> using SuperLU. What's the state of
> the driver?? How does the driver behave? Does it copy stuff and allocate
> new memory??

Have you tried using uBLAS lu_factorize and lu_substitute, defined in lu.h ?
Do you know how it compares to SuperLU ? Has anyone made this comparison ?
Also, how large are your sparse matrices?

GluCat ( http://glucat.sf.net ) uses lu_factorize and lu_substitute to 
implement division of multivectors. It works well AFAIK, but I have not 
compared it to SuperLU.