Subject: Re: Strange libc shared vs. unshared performance
To: None <mrg@mame.mu.OZ.AU, port-sparc@NetBSD.ORG>
From: Chris Torek <torek@BSDI.COM>
List: port-sparc
Date: 11/12/1994 22:17:09 
I would bet you are seeing cache effects.  Try a dynamically
linked dhrystone that has `padding' code inserted at the front,
to see if the addresses of the inner dhrystone loops are conflicting
with the addresses of the (dynamic) libc functions.

Chris