2018/12/19 by Mahesh Lakshminarasimhan, Lakshminarasimhan, Mahesh, Catherine Olschanowsky +1
Computer Science · #FOS: Computer and information sciences #Performance (cs.PF) #cs.PF
paper · pdf · doi:10.48550/arxiv.1812.07778
arxiv created 2018/12/19 · arxiv updated 2018/12/20
Optimizing scientific applications to take full advan-tage of modern memory subsystems is a continual challenge forapplication and compiler developers. Factors beyond working setsize affect performance. A benchmark framework that exploresthe performance in an application-specific manner is essential tocharacterize memory performance and at the same time informmemory-efficient coding practices. We present AdaptMemBench,a configurable benchmark framework that measures achievedmemory performance by emulating application-specific accesspatterns with a set of kernel-independent driver templates. Thisframework can explore the performance characteristics of a widerange of access patterns and can be used as a testbed for potentialoptimizations due to the flexibility of polyhedral code generation.We demonstrate the effectiveness of AdaptMemBench with casestudies on commonly used computational kernels such as triadand multidimensional stencil patterns.