Automatic mapping of system of N-dimensional affine recurrenceequations (SARE) onto distributed memory parallel systems

摘要

The automatic extraction of parallelism from algorithms, and thenconsequent parallel code generation, is a challenging problem. Wenpresent a procedure for automatic parallel code generation in the casenof algorithms described through a SARE (Set of Affine RecurrencenEquations). Starting from the original SARE description in annN-dimensional iteration space, the algorithm is converted into anparallel code for an (eventually virtual) m-dimensional distributednmemory parallel machine (m