In this webinar members of the ARCHER2 CSE team will show how the LIKWID toolsuite can be used to quantify application performance by measuring hardware performance counters and using metrics such as FLOPS, memory bandwidth, cache misses, etc.
We will show how to perform measurements of MPI and/or thread parallel applications on ARCHER2 and relate these to realistic peak hardware capabilities obtained by running LIKWID microbenchmarks.
This webinar should be of interest to anyone wanting to assess how well an application exploits ARCHER2 node-level hardware, including through construction of a roofline plot.