So i've been trying to use clapack for a finite differences Quantum Mechanics problem, this boils down to eigenvaluing a very sparse non-symmetric matrix. I've been using the dgherd_ and dhseqr_ routines and have been getting good results, however when the order of my matrices (doubles) gets above 27^2 i get a segmentation fault which i simply cannot track down (if it's 26^2 it works out fine...).
The code is available at: http://www.relain.co.uk/stuff/physics/sqhfix.c
I'm compiling the source with the Mac Os 10.3.9 version of gcc and using the clapack and cblas libraries that are included within the power-pc optimised framework called veclib.
I'm running this on a G4 (32bit) with 768mb of ram, am i simply running out of memory? Surely not.
I'm more of a physicist than a computer scientist sooo maybe it's something very basic with my coding? Do i need to be dynamically allocating memory for the arrays?
Thanks for your help,
also if anyone out there is actually an author of lapack i think it's great!