23rd EUROMICRO Conference '97 New Frontiers of Information Technology
Scheduling Parallel Loops for Processing Linear Lists
Budapest, HUNGARY
September 01-September 04
ISBN: 0-8186-8129-2
Imperative languages such as C, FORTRAN 90 or Modula-2 are being increasingly used to program parallel computers. By using these languages, the normal programming environment is disposable, however runtime deteriorates. The problem is the existence of pointers and dynamic data structures, which prevent effective parallelization methods. Here we demonstrate a new technique for shared memory systems. The list approach with pipeline method works on a linear list in parallel. After a theoretical valuation the practical operation on a KSR1 system will be discussed.
Citation:
W. Amme, P. Braun, E. Zehendner, "Scheduling Parallel Loops for Processing Linear Lists," euromicro, pp.426, 23rd EUROMICRO Conference '97 New Frontiers of Information Technology, 1997