In this slidecast, Mark O’Connor from Allinea Software describes how the company’s Performance Reports tool can help HPC users increase the energy efficiency of their computations.
A crazy idea was born at ISC14 while answering questions about the new energy metrics in Allinea Performance Reports and Allinea MAP – could the information in these reports show us an easy way to increase energy efficiency without having to change the program? The idea was to use CPU frequency scaling to run memory-bound jobs at a lower clock frequency. In Lazy Energy Efficiency Challenge One, I found that on a synthetic benchmark I could increase the iterations per watt by 19% on a memory-bound code. A large proportion of real-world workloads are memory-bound. The only code to achieve 100% CPU FPU utilization on most clusters is the Linpack benchmark. Real codes have to spend time waiting for memory accesses, for MPI communications and for filesystem I/O. Could some of them benefit from reducing the CPU frequency when indicated by a performance report?