A partition method for solving block pentadiagonal linear system on intel hypercube iPSC/860
We present a parallel partition algorithm for the solution of block pentadiagonal linear systems suitable for computation on computers with distributed memory. The method belongs to the direct methods and is based on the partition method derived in  for non-block banded matrices. The parallelization is achieved by dividing of the original block matrix into large blocks that can be processed almost independently. The time measurement results achieved by implementation of this method in the message passing Fortran on Intel hypercube iPSC/860 have shown a dependence of the effectivity of the implementation on the size of the large blocks.