Newsgroups: comp.parallel.mpi
From: "Hans Leichtl" <hans@ntplx.net>
Subject: Re: MPI_Wtick and MPI_Wtime
Organization: NETPLEX Internet Access (860) 233-1111
Date: Mon, 23 Mar 1998 21:04:32 -0500
Message-ID: <6f74bj$bbd$1@news.ntplx.net>

that would explain why the resolution returned by MPI_Wtick fluctuations.
i guess it's up to the implementers of MPI to tell us how they did it.

sun has a function gethrtime(). digital unix also has a us counter. (The
real time stuff has to be installed for that to be available)
all of them have cycle counters, and that gets pretty accurate (2ns on alpha
with 500 mhz).
nt on alpha uses the cycle counter, on intel they have a sub us resolution
hardware timer.

please let me know if you find some more stuff.

hans