Empirical models of miss rates

Kangwoo Lee, Michel Dubois

Research output: Contribution to journalArticlepeer-review

Abstract

In this paper, we introduce an empirical model for the cache miss rate of parallel applications in shared-memory multiprocessors. A few samples are collected by simulating an application with small input data sets and a statistical extrapolation technique is applied to produce analytical models of the miss rate. The analytical prediction of the various miss rates for very large input data obtained from empirical models is extremely accurate for most applications.

Original languageEnglish
Pages (from-to)205-219
Number of pages15
JournalParallel Computing
Volume24
Issue number2
DOIs
StatePublished - Feb 1998

Keywords

  • Cache misses
  • Data access
  • Prediction
  • Shared-memory multiprocessor

Fingerprint

Dive into the research topics of 'Empirical models of miss rates'. Together they form a unique fingerprint.

Cite this