Time-predictable chip-multiprocessor design

Abstract

Real-time systems need time-predictable platforms to enable static worst-case execution time (WCET) analysis. Improving the processor performance with superscalar techniques makes static WCET analysis practically impossible. However, most real-time systems are multi-threaded applications and performance can be improved by using several processor cores on a single chip. In this paper we present a time-predictable chip-multiprocessor system that aims to improve system performance while still enabling WCET analysis. The proposed chip-multiprocessor (CMP) uses a shared memory with a time-division multiple access (TDMA) based memory access scheduling. The static TDMA schedule can be integrated into the WCET analysis. Experiments with a JOP based CMP showed that the memory access starts to dominate the execution time when using more than 4 processor cores. To provide a better scalability, more local memories have to be used. We add a processor local scratchpad memory and split data caches, which are still time-predictable, to the processor cores.