Publication Details

TFApprox: Towards a Fast Emulation of DNN Approximate Hardware Accelerators on GPU

VAVERKA Filip, MRÁZEK Vojtěch, VAŠÍČEK Zdeněk and SEKANINA Lukáš. TFApprox: Towards a Fast Emulation of DNN Approximate Hardware Accelerators on GPU. In: 2020 Design, Automation & Test in Europe Conference & Exhibition (DATE). Grenoble: Institute of Electrical and Electronics Engineers, 2020, pp. 294-297. ISBN 978-3-9819263-4-7.
Czech title
TFApprox: Efektivní emulace hardwarových akcelerátorů DNN obsahujících přibližně počítající násobičky na GPU
Type
conference paper
Language
english
Authors
Keywords

approximate circuits, approximate multipliers, deep neural networks, DNN accelerator

Abstract

Energy efficiency of hardware accelerators of deep neural networks (DNN) can be improved by introducing approximate arithmetic circuits. In order to quantify the error introduced by using these circuits and avoid the expensive hardware prototyping, a software emulator of the DNN accelerator is usually executed on CPU or GPU. However, this emulation is typically two or three orders of magnitude slower than a software DNN implementation running on CPU or GPU and operating with standard floating point arithmetic instructions and common DNN libraries. The reason is that there is no hardware support for approximate arithmetic operations on common CPUs and GPUs and these operations have to be expensively emulated. In order to address this issue, we propose an efficient emulation method for approximate circuits utilized in a given DNN accelerator which is emulated on GPU. All relevant approximate circuits are implemented as look-up tables and accessed through a texture memory mechanism of CUDA capable GPUs. We exploit the fact that the texture memory is optimized for irregular read-only access and in some GPU architectures is even implemented as a dedicated cache. This technique allowed us to reduce the inference time of the emulated DNN accelerator approximately 200 times with respect to an optimized CPU version on complex DNNs such as ResNet. The proposed approach extends the TensorFlow library and is available online at https://github.com/ehw-fit/tf-approximate.

Published
2020
Pages
294-297
Proceedings
2020 Design, Automation & Test in Europe Conference & Exhibition (DATE)
Conference
Design, Automation and Test in Europe Conference, Grenoble, FR
ISBN
978-3-9819263-4-7
Publisher
Institute of Electrical and Electronics Engineers
Place
Grenoble, FR
DOI
UT WoS
000610549200053
EID Scopus
BibTeX
@INPROCEEDINGS{FITPUB12072,
   author = "Filip Vaverka and Vojt\v{e}ch Mr\'{a}zek and Zden\v{e}k Va\v{s}\'{i}\v{c}ek and Luk\'{a}\v{s} Sekanina",
   title = "TFApprox: Towards a Fast Emulation of DNN Approximate Hardware Accelerators on GPU",
   pages = "294--297",
   booktitle = "2020 Design, Automation \& Test in Europe Conference \& Exhibition (DATE)",
   year = 2020,
   location = "Grenoble, FR",
   publisher = "Institute of Electrical and Electronics Engineers",
   ISBN = "978-3-9819263-4-7",
   doi = "10.23919/DATE48585.2020.9116299",
   language = "english",
   url = "https://www.fit.vut.cz/research/publication/12072"
}
Back to top