Performance

PerFlow is a domain specific framework for performance analysis of parallel applications, which significantly reduces the burden of implementing performance analytical tasks.

Communication traces are increasingly important, both for parallel applications’ performance analysis/optimization, and for designing next-generation HPC systems. Meanwhile, the problem size and the execution scale on supercomputers keep growing, producing prohibitive volume of communication traces.

ScalAna is an automatic tool for scaling loss detection. It uses hybrid static-dynamic analysis to reduce costs and leverages graph algorithms to root cause scaling issues.

Memory monitoring is of critical use in understanding applications and evaluating systems. Due to the dynamic nature in programs’ memory accesses, common practice today leaves large amounts of address examination and data recording at runtime, at the cost of substantial performance overhead (and large storage time/space consumption if memory traces are collected).