A parallel scheme for accelerating parameter sweep applications on a GPU

摘要

This paper proposes a parallel scheme for accelerating parameter sweep applications on a graphics processingrnunit . By using hundreds of cores on the graphics processing unit, we found that our schemernsimultaneously processes multiple parameters rather than a single parameter. The simultaneous sweepsrnexploit the similarity of computing behaviors shared by different parameters, thus allowing memory accessesrnto be coalesced into a single access if similar irregularities appear among the parameters’ computationalrntasks. In addition, our scheme reduces the amount of off-chip memory access by unifying the data that arerncommonly referenced by multiple parameters and by placing the unified data in the fast on-chip memory. Inrnseveral experiments, we applied our scheme to practical applications and found that our scheme can performrnup to 8.5 times faster than a naive scheme that processes a single parameter at a time. We also include arndiscussion on application characteristics that are required for our scheme to outperform the naive scheme.