HPC Blog The HPC Advisory Council exists to spread the word about high-performance computing tech, provide a network of expertise, equipment for members, and generally educate all-comers about the wonders of HPC.
Thrillingly, it also has a bunch of equipment to play with in-house, including clusters ranging from four to 38 nodes, and a couple of GPU-enabled, four-node clusters. Members can use these for testing, development, or just kick the tires on something new.
In line with its educational mission, HPCAC has published more than 160 profiles for popular HPC applications. Everything from Abaqus to WRF, with a bunch of GADGET, GROMACS, and OpenFOAM in between.
When profiling applications, they particularly concentrate on communication patterns and activity, since that is often where you see the most opportunity for optimisation.
At the recent HPCAC workshop in Lugano, Switzerland, the tsar of the HPCAC profiling efforts, Pak Lui, gave a talk about how they profile applications, their capabilities, and the trends he's seeing. You get the slides here (PDF) or watch the video below.