Characterizing the Memory Behavior of Compiler-Parallelized Applications

Compiler-parallelized applications are increasing in importance as moderate-scale multiproces-sors become common. This paper evaluates how features of advanced memory systems (e.g., longer cache lines) impact memory system behavior for applications amenable to compiler par-allelization. Using full-sized input data sets and applications taken from standard benchmark suites, we measure statistics such as speedups, synchronization and load imbalance, causes of cache misses, cache line utilization, data traac and memory costs. This exploration allows us to draw several conclusions. First, we nd that larger granularity parallelism often correlates with good memory system behavior, good overall performance, and high speedup in these applications. Second, we show that when long (512 byte) cache lines are used, many of these applications suuer from false sharing and low cache line utilization. Third, we identify some of the common artifacts in compiler-parallelized codes that can lead to false sharing or other types of poor memory system performance, and we suggest methods for improving them. Overall, this study ooers both an important snapshot of the behavior of applications compiled by current parallelizing compilers, as well as an increased understanding of the interplay between cache line size, program granularity, and memory performance in moderate-scale multiprocessors.