The conditions for creating optimizing parallelizing compilers for computing systems with distributed memory are described. Target computing systems are microcircuits of the “supercomputer on a chip” type. Both optimizing program transformations specific to systems with distributed memory and those transformations that are needed both for computing systems with distributed memory and for computing systems with shared memory are presented. The issues of minimizing interprocessor transfers when parallelizing a recursive function are discussed. The main approach to creating such compilers is block-affine data placement in distributed memory with minimization of inter-processor transfers. It is shown that parallelizing compilers for computing systems with distributed memory should be created on the basis of a high-level internal representation and a high-level output language.

This paper is concerned with development of parallelizing compiler onto computer system with distributed memory. Industrial parallelizing compilers create programs for shared memory systems. Transformation of sequential programs onto systems with distributed memory requires development of new functions. This is becoming topical for future computer systems with hundreds and more cores. Conditions for program loop parallelization onto computer system with distributed memory is formulated in terms of information dependence graph.