Font Size: a A A

Increasing memory performance in multi-core processors

Posted on:2011-12-19Degree:M.SType:Thesis
University:Oklahoma State UniversityCandidate:Magee, Marion CFull Text:PDF
GTID:2448390002957338Subject:Computer Science
Abstract/Summary:
To continue to develop performance increases in modern processors, an old paradigm in computer science must be re-explored. Flat memory architecture, provided with modern processor construction, may provide those performance increases in the future.;Modern processors utilize a hierarchical memory cache systems to hide latency that is inherent in off-chip memories. Traditional RAM has always lagged processor clock speeds, giving rise to latency in transferring the required data and instructions. By moving all external RAM onto the chip using current cache memory technologies, latencies produced by cache misses can be eliminated.;Simulations show that if the size of a given program or program set is known, only half of that amount will be needed to produce a gain in performance over traditional, hierarchical cache memory systems.
Keywords/Search Tags:Memory, Performance, Cache
Related items