Hadoop+Aparapi: Making heterogenous MapReduce programming easier
Computer Science Department, University of Illinois at Urbana Champaign
University of Illinois at Urbana Champaign, 2012
@article{lin2012hadoop,
title={Hadoop+Aparapi: Making heterogenous MapReduce programming easier},
author={Lin, Y. and Okur, S. and Radoi, C.},
year={2012}
}
Lately, programmers have started to take advantage of GPU capabilities of cloud-based machines. Using the GPUs can decrease the number of nodes required to perform the computation by increasing the productivity per node. We combine Hadoop, a widely-used MapReduce framework, with Aparapi, a new Java-to-OpenCL conversion tool from AMD. We propose an easy-to-use API which allows easy implementation of MapReduce algorithms that make use of the GPU. Our API improves upon Hadoop by further hiding the complexity of GPU programming, thus allowing the programmer to concentrate on her algorithm. We also propose an accompanying refactoring that allows the programmer to specify the GPU part of their map computation by using very lightweight annotation.
April 10, 2012 by hgpu
