Super-scaling

Joe Landman landman at scalableinformatics.com
Wed Feb 19 08:47:21 PST 2003


Super-linearity occurs quite simply when (for a memory bound process)
the resident size of memory for the process is comparable to the size of
the cache, or to the size of the blocks used in the algorithm itself. 
This is effectively a phase transition, from operating mainly in a lower
speed medium (main memory) to operating mainly in a higher speed medium
(re-used cache), and is related to critical scale factors (cache and/or
block size) in the machine and algorithm.

As there are many different critical scale factors in the machines, you
typically get a variety of interesting transitions.  One of the most
important is when the files you read/write can fit entirely into cache
memory.  There are others as well.

Joe

-- 
Joseph Landman, Ph.D
Scalable Informatics LLC,
email: landman at scalableinformatics.com
web  : http://scalableinformatics.com
phone: +1 734 612 4615




More information about the Beowulf mailing list