forum.alglib.net

Hi, I am using smatrixevd to solve eigenvalues for a large matrix (10000 * 10000). But it reported aborted core dump. I wonder how large this algorithm can handle. Thank you very much.

Several thousands by several thousands is possible, but 10000x10000 is too large. It is 100 000 000 elements, 800 megabytes, and function allocates several copies of array as temporaries. I suppose that you used it on 32-bit system, so it was unable to allocate enough memory.

So how much memory do you expect to use for 10000 * 10000 matrix?

If you don't need eigenvectors - then 1600 Mbytes (2x800, one for a original matrix, one for a copy). 3200 Mbytes - if you need eigenvectors.