Machine Learning for CUDA+MPI Design Rules

Carl Pearson, Aurya Javeed, Karen Devine

We present a new strategy for automatically exploring the design space of key CUDA+MPI programs and providing design rules that discriminate slow from fast implementations. In such…