Researchers at the Massachusetts Institute of Technology's Computer Science and Artificial Intelligence Laboratory have developed Jenga, a system that reallocates a chip's memory cache access in real time, to create new "cache hierarchies" specifically designed to the needs of particular programs.
The researchers tested the system on a simulation of a chip with 36 cores and found that compared to its best-performing predecessors, the new system increased processing speed by 20% to 30% while reducing energy consumption by 30% to 85%.
Jenga builds on an earlier system, called Jigsaw, which also allocated cache access on the fly but did not build cache hierarchies, making the allocation problem much more complex.
Jenga has to evaluate the trade-off between latency and space for two layers of cache simultaneously, which turns the two-dimensional latency-space curve into a three-dimensional surface.
From MIT News
View Full Article
Abstracts Copyright © 2017 Information Inc., Bethesda, Maryland, USA
No entries found