Home | english  | Impressum | Sitemap | KIT

Application-independent Autotuning for GPUs

Application-independent Autotuning for GPUs
Tagung:

Konferenzartikel 

Jahr:

2014 

Autoren:

Martin Tillmann
Thomas Karcher
Carsten Dachsbecher
Walter F. Tichy

Links:PDFPDF

Summary

Autotuning is an established technique for adjusting performance-critical parameters of applications to their specific run-time environment. In this paper, we investigate the potential of online autotuning for general purpose computation on GPUs. Our application-independent autotuner AtuneRT optimizes GPU-specific parameters such as block size and loop-unrolling degree. We also discuss the peculiarities of autotuning on GPUs. We demonstrate tuning potential using CUDA and by instrumenting the parallel algorithms library Thrust. We evaluate our online autotuning approach with various GPUs and sample applications.

Beteiligte Mitarbeiter (zufällige Reihenfolge)
Titel Vorname Nachname

Bibtex

@inproceedings{,
author={Martin Tillmann, Thomas Karcher, Carsten Dachsbacher, Walter F. Tichy},
title={Application-independent Autotuning for GPUs},
year=2014,
booktitle={Parallel Computing: Accelerating Computational Science and Engineering (CSE)},
publisher={IOS Press},
editor={Michael Bader, Arndt Bode, Hans-Joachim Bungartz, Michael Gerndt, Gerhard R. Joubert, Frans Peters},
series={Advances in Parallel Computing},
url={https://ps.ipd.kit.edu/downloads/},
isbn={978-1-61499-380-3},
doi={10.3233/978-1-61499-381-0-626},
abstract={Autotuning is an established technique for adjusting performance-critical parameters of applications to their specific run-time environment. In this paper, we investigate the potential of online autotuning for general purpose computation on GPUs. Our application-independent autotuner AtuneRT optimizes GPU-specific parameters such as block size and loop-unrolling degree. We also discuss the peculiarities of autotuning on GPUs. We demonstrate tuning potential using CUDA and by instrumenting the parallel algorithms library Thrust. We evaluate our online autotuning approach with various GPUs and sample applications.},
number={Volume 25},
pages={626-635},
pptUrl={https://ps.ipd.kit.edu/downloads/},