ARPACK vs. LAPACK

mkrems · 05-14-2008, 03:40 PM

Hi there,

So I am working on a code to solve an eigenvalue problem, and I am confronted with using ARPACK or LAPACK for the eigensolver. I know that ARPACK uses some of the routines from LAPACK, but what are the main differences between using one over the other? Which one is preferred (I am guessing it depends on the type of problem)? Also, am I correct in assuming that the same comparisons between ARPACK and LAPACK will apply to PARPACK and scalaPACK (the parallel versions of these codes)?

Thanks!

mosca · 05-14-2008, 10:41 PM

ARPACK is for large, sparse matrices. LAPACK is for dense matrices or sparse matrices with trivial structure.

archtoad6 · 05-15-2008, 08:38 AM

This & #3 as well as #4 are (currently) in
LinuxQuestions.org > Forums > Non-*NIX Forums > Programming > ARPACK vs. LAPACK,
please move all 3 to a new thread in an appropriate forum.

jarick, when you make a mistake like this, it is recommended that you report your own thread to the moderators asking them to move it to the place you should have posted it in the 1st place.

mkrems · 05-15-2008, 01:00 PM

Quote:

Originally Posted by mosca

ARPACK is for large, sparse matrices. LAPACK is for dense matrices or sparse matrices with trivial structure.

Is there any general criteria for using one or the other? As a guess LAPACK or ARPACK would be equally good for finding the eigenvectors of a tridiagonal matrix for example, but ARPACK would be much better at this for a large sparse random matrix? Does one only use LAPACK for small matrices?