Dear all Recently I use finite difference method code with regular symmetric sparse matrix. I'm going to perform calculations with big amount of matrix meshes. It needs a lot of cputime, and I'm going to use vector supercomputer. Could you tell me information about matrix solver specialize in vector machine ?

Probably there are many such matrix solvers that would vectorize. You could start looking at what is available in LAPACK.

The essential ingredient will be for you to have a compiler for your platform that can recognize loops for vectorization.

Sometimes the platform comes supplied with hand-coded assembly versions of libraries that are faster than what you can obtain from optimized FORTRAN or C. Particularly, you should look for BLAS library implementations to help speed up your calculations.

Finally, you should look into parallelizing your code if memory and CPU time are limiting your calculations.

Good luck.