Eduardo H. M. Cruz, Matthias Diener, Philippe O. A. Navaux

Research output: Chapter in Book/Report/Conference proceedingChapter


Locality of memory accesses is one of the most important aspects to be considered when designing an architecture or developing software. With the introduction of multicore architectures, the memory hierarchy had to evolve to able to provide the necessary bandwidth to several cores operating in parallel. With this evolution, memory hierarchies started to present several caches in the same level, some levels shared by multiple cores, and other private to a core. Another important step was the incorporation of a memory controller inside the processor, in which multiprocessor systems presented NUMA characteristics. Due to the introduction of such technologies, the performance of memory hierarchies and the systems as a whole were even more dependent on memory locality. In this context, techniques such as sharing-aware thread and data mapping are able to increase memory locality and thereby performance. Our experiments indicate performance improvements of up to 200% in a scientific application.

Original languageEnglish (US)
Title of host publicationSpringerBriefs in Computer Science
PublisherSpringer Nature
Number of pages1
StatePublished - 2018

Publication series

NameSpringerBriefs in Computer Science
ISSN (Print)2191-5768
ISSN (Electronic)2191-5776

ASJC Scopus subject areas

  • General Computer Science


Dive into the research topics of 'Conclusions'. Together they form a unique fingerprint.

Cite this