A Generic Approach for Developing Highly Scalable Particle-Mesh Codes for GPUs
Forschungszentrum DresdenRossendorf e.V., D-01328 Dresden, Germany
Symposium on Application Accelerators in High Performance Computing, 2010
@article{honig2010generic,
title={A Generic Approach for Developing Highly Scalable Particle-Mesh Codes for GPUs},
author={H{\”o}nig, W. and Schmitt, F. and Widera, R. and Burau, H. and Juckeland, G. and M{\”u}ller, MS and Bussmann, M.},
booktitle={Application Accelerators in High Performance Computing, 2010 Symposium, Papers},
year={2010}
}
We present a general framework for GPU-based low-latency data transfer schemes that can be used for a variety of particle-mesh algorithms [8]. This framework allows to hide the latency of the data transfer between GPU-accelerated computing nodes by interleaving it with the kernel execution on the GPU. We discuss as an example the fully relativistic particle-in-cell (PiC) code PIConGPU [5] currently used to simulate particle acceleration by extremely short high-energy laser pulses. The PiC algorithm is a versatile algorithm used frequently in plasma physics-especially for large-scale simulations of fusion plasmas [13]-, in astrophysics [7], [9], or for the simulation of particle accelerators [11]. A special Cell processor version is used as a benchmark code for the Roadrunner system at Los Alamos National Lab [4].
February 19, 2011 by hgpu