High-performance computing under availability constraints to solve dense triangular system

Abstract

In this paper, we focus on parallel planning applied to a 2-step graph with a constant task cost which is the precedence graph of the algorithm solving a triangular system. We sort the tasks of 2-steps graph using critical path scheduling and we present a new schedule without and with some availability constraints. Some processors may not be available for some time interval. For each described scheduling, we determine the theoretical value of its makespan. Finally, we expose some experimental results.