From: lusk@donner.mcs.anl.gov (Rusty Lusk)
Newsgroups: comp.parallel.mpi
Subject: Re: Timing calls
Date: 9 Oct 1998 14:31:33 GMT
Organization: Argonne National Laboratory
Message-Id: <6vl6o5$h2v@milo.mcs.anl.gov>
References: <6vk4k6$hs$1@nnrp1.dejanews.com>


In article <6vk4k6$hs$1@nnrp1.dejanews.com>, chetan@cdac.ernet.in writes:
|> Hi all,
|> In one of my programs (FORTRAN) I have noticed that the time
|> reported by MPI_WTIME does not match with the time reported
|> by other calls like dtime (I am using Solaris 2.6 on a SPARC).
|> The ratio of the two time values is too big to be taken lightly
|> (like 4 times in one case). MPI_WTIME reports larger time value.

You did not show in your example exacly how you called dtime and how
you used the results.  There are a number of ways in with dtime can
be accidentally misused.  Can you post the details of the calls to
dtime and how you calculated the elapsed time with its results?

Also note from the man page on Solaris....

     NOTE: The very first call to etime or dtime may  be  inaccu-
     rate.

Regards,
Rusty