Hi Nishanth! Can you estimate how many nonzero elements your matrices typically contain? Assuming 20000 is the dimension of your vector space and the matrices are "truly" sparse, the size should be well manageable without parallelization using iterative methods. One possible library for C++ is Spectra (https://spectralib.org/index.html). It implements the major algorithms in ARPACK, which is the standard Fortran library for treating sparse matrix problems.

Thank you for your answer Pekko. Yes, my matrix is sparse in nature and I am attempting an iterative method. However after looking at spectra I can confidently say that it should suffice for my problem. However, there is no mention of how to initialize a parallel run?

I don't think Spectra includes a parallelized implementation. But my point was that given the vector dimension of your problem, there probably is no need for parallelization. If your matrices have at most a few million nonzero elements, say, then a non-parallelized iteration should be sufficient. I hope my answer was helpful to you.

Hi Nishanth,

I think you can solve your problem using PETSc (https://www.mcs.anl.gov/petsc/). This is a powerful scientific computation environment that includes SLEPc for working with Eigenvalues.

Best,

Eduardo

Thank you for your Answer Mr. Cesar. I have checked the link, and yes, it is the best solution to my problem.

Regards,

Nishanth