Exploiting Resource Parallelism for Integrated Fault Monitoring and Recovery
Abstract
Most emerging high-performance processors, such as superscalar, and VLIW processors, possess substantial fine-grain parallelism in the form of multiple pipeline stages, multiple functional units, and multiple buses. The objective of the proposed research is to develop a new method, called integrated monitoring, which effectively exploits the dynamic availability of parallel resources, e.g. delay slots in the pipe stages or idling functional units, for concurrently detecting program execution errors. Such an approach alleviates the need of a separate and dedicated hardware monitor, and minimizes or eliminates the performance penalty incurred for fault monitoring.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 20, 1992
- Accession Number
- ADA248843
Entities
People
- John P. Shen
Organizations
- Carnegie Mellon University