🚀 go-pugleaf

   In <1989Nov4.004529.10049@ico.isc.com> Dick Dunn wrote:
   >                                     Second, I would expect better locality
   > for code reference than for data reference, hence the I cache ought to do
   > more good than the D cache.  Aren't the pathological cache-busting programs
   > generally ones which spray data accesses all over the place?
   
No.  That is typical of pathological demand-paged-memory-busting
techniques.  The reason is that typically caches consist of a large
set of realatively short entries (typically a few-to-several bytes) as
compared to demand-paged-memory which consists of a set of LONG
entry-measured in kilobytes.  A truly random cache access pattern is
pretty good because the lookup scheme is usually a hash function for
which it is easy to generate bad access patterns (typically (addr >>
m)).  Furthermore, the amount of memory moved (therefore the time
involved with moving) for a single cache-update is not very great, so
if the extra stuff is never used, very little effort was wasted.  In
comparison a single page-in takes so much time, and moves so much
data, that random addressing patterns will cause pathologically bad behavior.

Caches, however are only useful if the same data is used
repetitively before it is flushed.  This seems to be more an issue of
algorithm design than programming style.
 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
				Eric Freudenthal
				NYU Ultracompter Lab
				715 Broadway, 10th floor
				New York, NY  10012
				
				Phone:(212) 998-3345 work
				      (718) 789-4486 home
				Email:freudent@ultra.nyu.edu
--
 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
				Eric Freudenthal
				NYU Ultracompter Lab
				715 Broadway, 10th floor
				New York, NY  10012
				
				Phone:(212) 998-3345 work
				      (718) 789-4486 home
				Email:freudent@ultra.nyu.edu