Mittal, Sparsh and Wang, H and Jog, A and Vetter, J S
(2017)
Design and Analysis of Soft-Error Resilience Mechanisms for GPU Register File.
In: International Conference on VLSI Design (VLSID), Hyderabad, India.
Abstract
Modern graphics processing units (GPUs) are
using increasingly larger register file (RF) which occupies
a
large fraction of GPU core area and is very frequently access
ed.
This makes RF vulnerable to soft-errors (SE). In this paper,
we
present two techniques for improving SE resilience of GPU RF
.
First, we propose compressing the RF values for reducing the
number of vulnerable bits. We leverage value similarity and
the presence of narrow-width values to perform compression
at
warp or thread-level, respectively. Second, we propose sel
ective
hardening to design a portion of register entry with SE immun
e
circuits. By collectively using these techniques, higher r
esilience
can be provided with lower overhead. Without hardening, our
warp and thread-level compression techniques bring 47.0%
and 40.8% reduction in SE vulnerability, respectively.
Actions (login required)
 |
View Item |