GATE
- 2 June 2019
- conference paper
- conference paper
- Published by Association for Computing Machinery (ACM) in Proceedings of the 56th Annual Design Automation Conference 2019
Abstract
Although approximate computing is widely used, it requires substantial programming effort to find appropriate approximation patterns among multiple pre-defined patterns to achieve a high performance. Therefore, we propose an automatic approximation framework called GATE to uncover hidden opportunities from any data-parallel program regardless of the code pattern or application characteristics using two compiler techniques, namely subgraph-level approximation (SGLA) and approximate thread merge(ATM). GATE also features conservative/aggressive tuning and dynamic calibration to maximize the performance while maintaining the TOQ level during runtime. Our framework achieves an average performance gain of 2.54x over the baseline with minimum accuracy loss.Keywords
Funding Information
- Ministry of Science ICT and Future Planning (2017-0-00142)
- National Research Foundation of Korea (2017R1A4A1015498)
- Ministry of Trade, Industry and Energy (10077609)
This publication has 8 references indexed in Scilit:
- Input responsiveness: using canary inputs to dynamically steer approximationPublished by Association for Computing Machinery (ACM) ,2016
- A Survey of Techniques for Approximate ComputingACM Computing Surveys, 2016
- ImageNet Large Scale Visual Recognition ChallengeInternational Journal of Computer Vision, 2015
- ParaproxPublished by Association for Computing Machinery (ACM) ,2014
- SAGEPublished by Association for Computing Machinery (ACM) ,2013
- SIMD defragmenterACM SIGARCH Computer Architecture News, 2012
- GreenPublished by Association for Computing Machinery (ACM) ,2010
- Rodinia: A benchmark suite for heterogeneous computingPublished by Institute of Electrical and Electronics Engineers (IEEE) ,2009