Parallelizing Compilers for Recursive Functions

Parallelizing compilers is one of the most important topics in parallel programming. It is developed to automatically parallelize sequential programs to reduce the burden of programming on parallel computers. At start it concentrates on parallelizing loops then it opens its scope to other aspects such as recursive functions. Consequently, new parallelizing compilers are proposed to automatically parallelize recursive functions in order to be executed efficiently on different parallel systems such as distributed memory massively parallel multicomputers and multiprocessors. A special concern has been for applications of divide-and-conquer algorithms. New algorithms for dependence analysis, partitioning, scheduling, and generating SPMD message passing program are introduced. Also new virtual topologies are proposed. Different types of recursive functions have been examined using variety of parallel machines. Ultimate performance measurements are used aiming at reaching optimality.