Execution of regular DO loops on asynchronous multiprocessors

Abstract

Studies the execution of a Fortran DO loop, modeled as an iteration space and several dependence vectors, on an asynchronous shared-memory multiprocessor, where each iteration is the basic unit to be executed by a single processing element. During the execution of a DO loop, an iteration will pass through four states, namely, idle, pending, ready, and finished states. By describing an execution scheme, the paper studies (1) the number of iterations not depending on any other iterations, (2) the maximum number of ready iterations at any instance during the execution, (3) the maximum number of pending iterations at any instance during the execution, (4) a hash function to disperse different pending iterations, and (5) the parallel execution time.<>