Techniques for Efficient Software Checking

TL;DR: A compiler algorithm is developed that finds code patterns that correspond to outcome tolerant branches and removes the unnecessary replicas and can reduce overhead by up to 50% when the most aggressive optimization is applied.

Abstract: Dramatic increases in the number of transistors that can be integrated on a chip make processors more susceptible to radiation-induced transient errors. For commodity chips which are cost- and energy-constrained, we need a flexible and inexpensive technology for fault detection. Software approaches can play a major role for this sector of the market because they need little hardware modifications and can be tailored to fit different requirements of reliability and performance. However, software approaches add a significant overhead. In this paper we propose two novel techniques that reduce the overhead of software error checking approaches. The first technique uses boolean logic to identify code patterns that correspond to outcome tolerant branches. We develop a compiler algorithm that finds those patterns and removes the unnecessary replicas. In the second technique we evaluate the performance benefit obtained by removing address checks before load and stores. In addition, we evaluate the overheads that can be removed when the register file is protected in hardware. Our experimental results show that the first technique improves performance by an average 7% for three of the SPEC benchmarks. The second technique can reduce overhead by up-to 50% when the most aggressive optimization is applied.