Optimizing Speedup on Multicore Platform with OpenMP Schedule Clause and Chunk Size

Abstract

Despite the description of schedule type which can be easily found at any OpenMP reference material, little is known about the effect of different schedule type and chunk size on the parallel performance of shared memory multicore processor. Literature shows that performance analysis on different multicore platform overlooked the effect of different schedule type and chunk size, where often it was not explicitly specified. Hence, default assignment of the loop iterations among threads is assumed. By default, static schedule is used and size of chunk which is the ratio of total number of iteration to the number of threads is implemented. This research analyses the effect of different schedule type and chunk size on speedup achieved of different shared memory multicore platform under regular workload. Apart from that, the performance gain obtained after turning on/off certain multicore technologies and after turning on/off selected number of active cores per processor is also analysed. Results shows that different multicore technology exhibit different speedup value under different combination of schedule type and chunk size. Apart from that, it also observe that different multicore platform is better than the other in terms of speedup as the number of cores are increased.