Abstract | ||
---|---|---|
Cache parameters for CPU architectures are typically defined for a best overall match regarding the targeted field of applications. However, such may hinder high-performance execution of single applications and also does not account for cache access phases as occurring in long-running applications, e.g. from field of high-performance computing. It has been shown that matching the cache parameters to a running application results in both application speed-up and increased energy efficiency. The aim of the presented work is to create a versatile, reconfigurable cache hardware infrastructure for cache performance analysis. Such an infrastructure enables real-time monitoring of running applications and therefore is able to better trace a running application's behaviour compared to off-line analysis of a more or less reduced trace. In this paper, we will address the problems and side-effects of a run-time reconfigurable cache architecture. to which we present appropriate solutions. We will also give,in outline of the upcoming hardware prototype. |
Year | Venue | Field |
---|---|---|
2007 | PARALLEL COMPUTING: ARCHITECTURES, ALGORITHMS AND APPLICATIONS | Computer science,Parallel computing,Cache-only memory architecture |
DocType | Volume | ISSN |
Conference | 15 | 0927-5452 |
Citations | PageRank | References |
5 | 0.52 | 8 |
Authors | ||
3 |
Name | Order | Citations | PageRank |
---|---|---|---|
Fabian Nowak | 1 | 39 | 6.52 |
Rainer Buchty | 2 | 143 | 18.44 |
Wolfgang Karl | 3 | 372 | 34.84 |