A problem with LU (culaSgetrf)

General CULA Dense (LAPACK & BLAS) support and troubleshooting. Use this forum if you are having a general problem or have encountered a bug.

A problem with LU (culaSgetrf)

Postby LaChupacabra » Tue Feb 02, 2010 11:52 am

Hi!
Can someone explain to me why culaSgetrf doesn't factorize some matrices, for example a square 5x5 matrix filled with numbers 0..24, I get data error. Matlab, which uses LAPACK can handle such matrices. Is it a bug in CULA? If not, it should be pointed out in the manuals.
LaChupacabra
 
Posts: 2
Joined: Tue Jan 12, 2010 1:23 am

Re:A problem with LU (culaSgetrf)

Postby john » Tue Feb 02, 2010 2:53 pm

We report the error exactly as LAPACK/MKL does - it in defined as an error condition for a matrix to be singular to the getrf routine, and we report it accordingly as a culaDataError. Matlab actually receives this error and disregards it. (Note that the Matlab inv() operator correctly reports it as singular.)

It's worth noting that the L*U*P can still be used to reconstruct A, but you can't solve a system because you will be forced to divide by zero.

I see your point though. LAPACK still provides the factorization, even if singular. CULA 1.2 will include this, as per your finding.

Thanks for the report,
John
john
Administrator
 
Posts: 587
Joined: Thu Jul 23, 2009 2:31 pm


Return to CULA Dense Support

Who is online

Users browsing this forum: No registered users and 2 guests

cron