>Hmmmm..... I think that the extra if test in the timing routine
>is just a sanity check to make sure that the elapsed time is positive.
Actually, it's checking for wrap-around of the microsecond units of the
timer, and trying to adjust for it. Except that the adjustment is not
needed and causes it to add an extra second to the times for some
runs (I looked at the timing of individual runs to verify the problem).
>I did not write those routines, so I don't know what the author had
>in mind --- did you try the other C version (cstream.c)?
I was thinking of mailing the author directly, but I thought I'd let
you know first. The cstream.c uses a different timing routine and
works fine (although it gives very different performance on our SGI
Indy than stream_d for two of the tests).
-- Ben Gamsa firstname.lastname@example.org - http://www.eecg.toronto.edu/~ben Dept. of C.S., University of Toronto, Toronto, Ont. CANADA, M5S 1A4
This archive was generated by hypermail 2b29 : Tue Apr 18 2000 - 05:23:04 CDT