Clarification of Question by
arintel-ga
on
29 Dec 2004 18:30 PST
Hello Pythagoras,
I have read through chapter 6 of Optimizations.pdf. I could not find
discussion of cache locking in there. It only discusses about
prefetch, which is different from cache locking.
In prefetch, "Prefetching too far ahead may cause eviction of cached
data from the caches prior to actually being used in execution; not
prefetching far enough ahead can reduce the ability to overlap memory
and execution latencies." (page 42)
Whereas, in cache locking, "when an application has a number of
important variables that are frequently accessed, it may improve
performance when those variables are locked into the cache. This keeps
them from being evicted when the cache becomes full, resulting in
fewer accesses to main memory."
(http://www.intel.com/design/iio/papers/27387202.pdf, page 11)
Chapter 6 in Optimizations.pdf also discusses cache blocking. However,
as I understand it, cache blocking is just a technique that uses
prefetch to deal with one- or two-dimensional arrays, which is, again,
not cache locking.
Regarding the document Instructions.pdf, I'm sorry I don't understand
its language. Do you have its english version?
Thanks a lot for your replies.
Regards,
ARINTEL