| HAL-Inria | Publications, software ... of Inria's scientists |
Journal articles
Safe Overclocking for CNN Accelerators through Algorithm-Level Error Detection
Abstract : In this article, we propose a technique for improving the efficiency of convolutional neural network hardware accelerators based on timing speculation (overclocking) and fault tolerance. We augment the accelerator with a lightweight error detection mechanism to protect against timing errors in convolution layers, enabling aggressive timing speculation. The error detection mechanism we have developed works at the algorithm-level, utilizing algebraic properties of the computation, allowing the full implementation to be realized using high-level synthesis tools. Our prototype on ZC706 demonstrated up to 60% higher throughput with negligible area overhead for various wordlength implementations.
https://hal.inria.fr/hal-03094811
Contributor : Steven Derrien Connect in order to contact the contributor
Submitted on : Wednesday, June 23, 2021 - 4:05:06 PM
Last modification on : Thursday, November 4, 2021 - 10:54:02 AM
Long-term archiving on: : Friday, September 24, 2021 - 6:51:57 PM
Contributor : Steven Derrien Connect in order to contact the contributor
Submitted on : Wednesday, June 23, 2021 - 4:05:06 PM
Last modification on : Thursday, November 4, 2021 - 10:54:02 AM
Long-term archiving on: : Friday, September 24, 2021 - 6:51:57 PM
File
FINAL VERSION.pdf
Files produced by the author(s)