Leveraging the Potential of Control-Flow Error Resilient Techniques in Multithreaded Programs

TL;DR

This paper introduces a compile-time software-based method for recovering control-flow errors in multithreaded programs, utilizing dependency graphs to detect and correct faults with high accuracy and manageable overhead.

Contribution

It proposes a novel technique that leverages dependency graphs to recover control-flow errors in multithreaded programs, demonstrating high fault coverage and efficiency.

Findings

Detects and corrects 91.9% to 93.8% of injected faults

Effective in multithreaded benchmarks like quick sort and matrix multiplication

Maintains acceptable performance and memory overheads

Abstract

This paper presents a software-based technique to recover control-flow errors in multithreaded programs. Control-flow error recovery is achieved through inserting additional instructions into multithreaded program at compile time regarding to two dependency graphs. These graphs are extracted to model control-flow and data dependencies among basic blocks and thread interactions between different threads of a program. In order to evaluate the proposed technique, three multithreaded benchmarks quick sort, matrix multiplication and linked list utilized to run on a multi-core processor, and a total of 5000 transient faults has been injected into several executable points of each program. The results show that this technique detects and corrects between 91.9% and 93.8% of the injected faults with acceptable performance and memory overheads.