Asynchronous Communication for Finite-Difference Simulations on GPU Clusters using CUDA and MPI
Computer Science, Massey University, Albany, North Shore 102-904, Auckland, New Zealand
Computational Science Technical Note CSTN-125, 2011
@article{playne2011asynchronous,
title={Asynchronous Communication for Finite-Difference Simulations on GPU Clusters using CUDA and MPI},
author={Playne, DP and Hawick, KA and Albany, N.S.},
year={2011}
}
Graphical processing Units (GPUs) are finding widespread use as accelerators in computer clusters. It is not yet trivial to program applications that use multiple GPU-enabled cluster nodes efficiently. A key aspect of this is managing effective communication between GPU memory on separate devices on separate nodes. We develop an algorithmic framework for Finite-Difference numerical simulations that would normally require highly synchronous data-parallelism so they can effectively use loosely coupled GPU-enabled cluster nodes. We employ asynchronous communications and appropriate memory overlay of computations and communications to hide latency.
October 16, 2011 by hgpu
