Call the program <input file for a> <rank> <input file for b> <method> <iterations> Ex. ./serial _Input/sparse_rank900_fdlaplacian 900 _Input/rhs_fdlaplacian_900x1 1 1000 ./parallel _Input/sparse_rank900_fdlaplacian 900 _Input/rhs_fdlaplacian_900x1 1 1000 You will probably want to pipe the stdout to a file using the '>' operator. NOTES: <method> is 1 - Gauss Siedel; 2 - Jacobi When using the parallel version, don't forget to set OMP_NUM_THREADS to the desired number of threads. Thread counts tested: 1, 2, 4, 8, 16, 24 When using serial cblas compilation, make sure OMP_NUM_THREADS=1. GOTO_NUM_THREADS has not be tested thoroughly, do not use. MAKEFILE: "make" will compile serial.c into serial and parallel.c into parallel FOR ATLAS "make parallel" will just compile parallel FOR ATLAS "make serial" will just compile serial FOR ATLAS "make goto" will compile serial.c into serial and parallel.c into parallel FOR GotoBLAS2 "make clean" will clean directory
mattm401/pDAMGES
A parallel (OpenMP) Diagonally-dominant Algebraic Multigrid Equation Solver by Benjamin G. Jimenez, Virginia Forstall, and Matthew Louis Mauriello. This project was created as part of the requirements for completion of CMSC714: High Performance Computing at the University of Maryland. Feel free to use with credit given to the authors in any derived work. To give credit where credit is due, these programs must be configured to work with ATLAS, GotoBLAS2, or another LAPACK/CBLAS package which are the property of their respective authors. If you would like to use another linear algebra package, you should just have to reconfigure the MAKEFILE for them - as long as it includes standard CBLAS/LAPACK you should be fine.
C