05470718

j

05470718 Lefevre, Vincent Optimized storage management for automatically parallelized programs. (Gestion optimis\?ee des donn\?ees dans les programmes automatiquement parall\?elis\?es.) RAIRO, Tech. Sci. Inf. 16, No. 9, 1111-1139 (1997). 1997 \'Editions HERMES, Paris EN C.1.2 D.1.3 F.2.2 parallel processing parallel programming automatic parallelization storage management array dataflow analysis scheduling Summary: There are essentially two methods of parallel programming. The first one consists in specifying all the parallelism in the source program, the second one leaves the compiler detect and exploit automatically the parallelism. Automatic parallelization techniques usually lead to an inflation of data structures memory size in the parallelized program. This article presents a method which reduces this inflation, with the help of elementary algebra techniques, providing the source program is limited to DO loops and arrays with affine subscripts.