PVM implementation of the symmetric-Galerkin method

摘要

We report on initial progress towards a parallel virtual machine (PVM) implementation of the symmetric-Galerkin boundary integral method. We take advantage of software packages specifically designed to solve linear algebra problems on distributed memory parallel computers. In particular we use linear algebra routines from the ScaLAPACK, PBLAS and BLACS libraries. These routines assume a block cyclic decomposition of the matrix operands. The decomposition of the operands and its impact on the construction of the coefficient matrix are described.