Please use this identifier to cite or link to this item: http://dx.doi.org/10.14279/depositonce-6341
Main Title: An automatic input-sensitive approach for heterogeneous task partitioning
Author(s): Kofler, Kofler
Grasso, Ivan
Cosenza, Biagio
Fahringer, Thomas
Type: Conference Object
Language Code: en
Abstract: Unleashing the full potential of heterogeneous systems, consisting of multi-core CPUs and GPUs, is a challenging task due to the difference in processing capabilities, memory availability, and communication latencies of different computational resources. In this paper we propose a novel approach that automatically optimizes task partitioning for different (input) problem sizes and different heterogeneous multi-core architectures. We use the Insieme source-to-source compiler to translate a single-device OpenCL program into a multi-device OpenCL program. The Insieme Runtime System then performs dynamic task partitioning based on an offline-generated prediction model. In order to derive the prediction model, we use a machine learning approach based on Artificial Neural Networks (ANN) that incorporates static program features as well as dynamic, input sensitive features. Principal component analysis have been used to further improve the task partitioning. Our approach has been evaluated over a suite of 23 programs and respectively achieves a performance improvement of 22% and 25% compared to an execution of the benchmarks on a single CPU and a single GPU which is equal to 87.5% of the optimal performance.
URI: https://depositonce.tu-berlin.de//handle/11303/7020
http://dx.doi.org/10.14279/depositonce-6341
Issue Date: 2013
Date Available: 26-Oct-2017
DDC Class: 004 Datenverarbeitung; Informatik
Subject(s): code analysis
compilers
gpu
heterogeneous computing
machine learning
runtime system
task partitioning
Usage rights: Terms of German Copyright Law
Proceedings Title: Proceedings of the 27th International ACM Conference on International Conference on Supercomputing
Publisher: Association for Computing Machinery (ACM)
Publisher Place: New York, NY
Publisher DOI: 10.1145/2464996.2465007
Page Start: 149
Page End: 160
ISBN: 978-1-4503-2130-3
Appears in Collections:Fachgebiet Architektur eingebetteter Systeme » Publications

Files in This Item:
File SizeFormat 
An_automatic.pdf1.36 MBAdobe PDFView/Open


Items in DepositOnce are protected by copyright, with all rights reserved, unless otherwise indicated.