The performance of most processors today is improved through appropriate use of caches, which
learn and predict various pieces of information. The effectiveness of a cache generally improves
as its size increases because larger caches are able to contain the frequently used information of a
larger number of applications. However, large caches consume a lot of power, and much of this
power is wasted when some region of an application is able to perform almost as well with a
considerably smaller cache. In this paper, we analyze the behavior of a typical cache and show
that applications have widely varying behavior, with phases in which the cache is heavily
utilized often interleaved with phases in which large parts of the cache remain idle. We use these
observations to propose a new type of cache called the quenched cache, which periodically
quenches the cache by turning it off completely. The effect of quenching is a cache that adapts its
power consumption based on actual activity, doing so without elaborate learning or feedback
mechanisms in hardware. The paper presents results of experiments to show the effectiveness of
quenched caches.
By: Ravi Nair
Published in: RC22543 in 2002
LIMITED DISTRIBUTION NOTICE:
This Research Report is available. This report has been submitted for publication outside of IBM and will probably be copyrighted if accepted for publication. It has been issued as a Research Report for early dissemination of its contents. In view of the transfer of copyright to the outside publisher, its distribution outside of IBM prior to publication should be limited to peer communications and specific requests. After outside publication, requests should be filled only by reprints or legally obtained copies of the article (e.g., payment of royalties). I have read and understand this notice and am a member of the scientific community outside or inside of IBM seeking a single copy only.
Questions about this service can be mailed to reports@us.ibm.com .