Fast Interrupt Priority Management in Operating System Kernels
Abstract
This paper describes a new technique, optimistic interrupt protection, that efficiently schedules and handles processor interrupts. While modern processor architectures have led to substantial overall performance improvements, operating systems have received significantly less benefit than application code 1,2,3. One processor function that has not scaled well with processor speed is interrupt management. Operating systems use interrupts to control scheduling and input/output, and use interrupt masking to guarantee integrity of system resources shared across interrupt levels. This approach was efficient in many previous processor architectures (e.g, VAX), where the cost changing interrupt levels was small - generally less than ten instructions 4,5. In modern architectures, however, interrupt masking may be up to an order of magnitude more expensive, contributing to poorer performance of system code.
Document Details
- Document Type
- Technical Report
- Publication Date
- May 01, 1993
- Accession Number
- ADA266638
Entities
People
- Brian N. Bershad
- Daniel Stodolsky
- J. B. Chen
Organizations
- Carnegie Mellon University