SISAL versus FORTRAN: a comparison using the Livermore loops
D.C. Cann, John Feo · Conference on High Performance Computing (Supercomputing) · 1990
The acquisition of parallel processors in the scientific community is increasing, but the difficulties of programming parallel machines persist. Two approaches have emerged: automatic parallelizing compilers for extant languages, and new languages professed to provide simpler and cleaner parallel programming models. Unfortunately, because of their semantics most new languages have acquired a reputation for inefficiency. This paper compares the performance of SISAL, an applicative language for parallel numerical computations, and FORTRAN using the Livermore Loops. Our intent is to show that applicative programs, when compiled using a set of powerful yet simple optimization techniques, can achieve sequential execution speeds comparable to FORTRAN, and automatically utilize conventional shared memory multiprocessors.