Simulation Symposium, Annual
Download PDF

Abstract

The performance of modern computer systems is increasingly limited by long latencies of accesses to their memory systems. Instruction-level multithreading is a technique to tolerate long latencies of memory accesses by switching from one instruction thread to another. The paper shows that the simulation-based performance evaluation of distributed-memory multithreaded multiprocessor systems can be significantly simplified by using approximate models, composed of only a few processors, but with some parameters adjusted to represent the behavior of the original system.
Like what you’re reading?
Already a member?
Get this article FREE with a new membership!