ITS: An ILP-based combined instruction/task static scheduling algorithm

Abstract

Our combined task and instruction static scheduling algorithm implemented in the COINS compiler uses an Integer Linear Programming model to find a schedule for a program on a symmetric multiprocessor system-on-chip. We compare our work to state of the art approaches and on average we find a speedup as high as 1.49 compared to a static task scheduling approach without instruction scheduling. Depending on the computation to communication ratio of the application we estimate an average speedup of 2.55 to 2.63 in application execution time compared to sequential code.