Automatic Fusions of CUDA-GPU Kernels for Parallel Map

Warning

This publication doesn't include Faculty of Economics and Administration. It includes Faculty of Informatics. Official publication website can be found on muni.cz.
Authors

FOUSEK Jan FILIPOVIČ Jiří MADZIN Matúš

Year of publication 2011
Type Article in Proceedings
Conference Second International workshop on highly-efficient accelerators and reconfigurable technologies (HEART)
MU Faculty or unit

Faculty of Informatics

Citation
Web http://www.fi.muni.cz/~xfilipov/s4-1.pdf
Field Informatics
Keywords GPU, CUDA, kernels fusion, map, mapped function
Description When implementing a function mapping on the contemporary GPU, several contradictory performance factors affecting distribution of computation into GPU kernels have to be balanced. A decomposition-fusion scheme suggest to decompose computational problem to be solved by several simple functions implemented as standalone kernels and some of these functions later fuse into more complex kernels to improve memory locality. In this paper, a prototype of source-to-source compiler automating the fusion phase is presented and the impact of fusions generated by the compiler as well as compiler efficiency is experimentally evaluated.
Related projects:

You are running an old browser version. We recommend updating your browser to its latest version.