Poster is on display and will be presented at the poster pitch session.
This poster describes an approach to predict the performance of HPC workloads. Based on data of runtime hardware counters from a set of HPC applications and benchmarks, we train an artificial intelligence model that is able to predict the performance of other HPC applications. This work differs from current research on the granularity of training and prediction, as the model is built using individual computation bursts of the whole training set. Here, we prove that a prediction of the instructions per cycle (IPC) metric of unseen applications is possible based on architectural performance counters that can be obtained easily with already used and convenient performance tools.
Contributors: 
