> Yup, I've fully populated my filesystem with NetBSD/sparc binaries.
> Here's the output of bonnie:  (yes, I'm a zealot)

Cool :-)

>               -------Sequential Output-------- ---Sequential Input-- --Random--
>               -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
> Machine    MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU  /sec %CPU
>           100    87 98.4   110 -3418377.6    60 7646566.9    82 -7882796.3   134 34.9  11.1 99.0
> 
> Ack!  Well, you can still see the number.. :)

A problem that I found when Chuck and I first started on the 4/100 
code is that floating point doesn't work right (under NetBSD) on some 
versios of the 4/100 CPU..  I have 2 4/110's, one which has the "right"
fp version, and one with the "wrong" fp version. 

You'll notice that some programs give weird numbers, and others will
cause fp exceptions.  If you compile the same code with a SunOS compiler
and linker it works fine...  

The fp fix has to do with the C compiler/linker and libraries, SunOS 
does some magic to fix the fp code for binaries that run on a 4/100.
At one point we thought that it was a simple code fix applied by 'ld' at
link time, but I was never able to reproduce the effect by hand. 

Anyone have any new ideas on this?

    John Stone
    johns@cs.umr.edu

-- 
*       John Stone - Computer Science Graduate Student / Unix Admin       *
*       University of Missouri - Rolla    Office: 208 MCS, 341-6434       *
*                           Unix Is Good For You                          *