GPU-accelerated time-domain circuit simulation
Agilent Technol., Santa Rosa, CA, USA
IEEE Custom Integrated Circuits Conference, 2009. CICC ’09
@inproceedings{poore2009gpu,
title={GPU-accelerated time-domain circuit simulation},
author={Poore, RE},
booktitle={Custom Integrated Circuits Conference, 2009. CICC’09. IEEE},
pages={629–632},
year={2009},
organization={IEEE}
}
Time-domain circuit simulation is often dominated by the transistor model evaluation time. An analysis of a test suite of 27 circuits shows 66% of the transient runtime is spent evaluating the core BSIM4 transistor model code. A modern graphics processing unit (GPU) is a highly paralled, high performance computer suitable for non-graphics tasks. Circuit simulation is sped up by a factor of 3 to 6, with no loss in accuracy, by moving the transistor evaluation to the NVIDIA GTX 280 GPU. Maximum GPU performance is obtained with single precision arithmetic, so transistor model code should be written with this in mind.
September 2, 2011 by hgpu
