Date   : Tue, 07 Dec 1982 10:56:00-PST
From   : chesley.tsca@Sri-Tsc
Subject: BDS-C Benchmarks

       One thing to watch out for when doing BDS-C benchmarks is the
huge difference between the code generated with no switches, and that
generated with -o and -e (and making any variables that can be globals).
(The person doing the Aztec benchmark didn't mention which way he did it.)
       I recently did a benchmark of BDS-C versus C/80, using the Byte
prime number program.  Without -o and -e, BDS-C took 41 seconds (this is
timed from after the program loads to the next "A>" prompt (i.e., after
the reboot), and thus may not compare to other numbers times from different
points).  With -o and -e, it took 17 seconds!  C/80 took (if I remember
correctly) 28 seconds.
       --Harry...