In article <29942.1159082142@splode.eterna.com.au>
mrg@eterna.com.au writes:
>    2.1_STABLE	Pentium 4	doesn't reproduce

One correction; this machine reproduced the problem.
I ran the test program for about two hours and got

% ./a.out
1159086053 580515 < 1159086054 580515
1159086834 550754 < 1159086835 550754
1159090587 291847 < 1159090588 291847
^C
%

> works fine on p4-m, g4/733 and ultrasparc/440.

So, we need several days to say "works fine." :-)

>                                                 i am wondering
> if the above system would be OK with timecounters?

Possibly.

However, the machines are in production and I may not be able
to test on them until the timecounter code will be in a release.

-- 
ITOH Yasufumi