To:	djgpp AT delorie DOT com
Subject:	Re: Interesting benchmark results
Message-ID:	<19970310.174935.10167.11.chambersb@juno.com>
References:	<m0w45NG-000S1nC AT natacha DOT inti DOT edu DOT ar>
From:	chambersb AT juno DOT com (Benjamin D Chambers)
Date:	Mon, 10 Mar 1997 20:46:27 EST

On Mon, 10 Mar 1997 16:57:56 +0000 "Salvador Eduardo Tropea (SET)"
<salvador AT natacha DOT inti DOT edu DOT ar> writes:
>About the post of Robert and the benchmarks results:
>He forgot to say what machine is using (486DX2 66 if he didn't changed 
>it) so 
>people using a Pentium perhaps can be benefited by the -O3 option.
OTOH, people using a Pentium might be _hurt_ a lot more than he is by
using -O3.  AFAIK, the Pentium will only go superscalar if the code
_is_already_in_the_cache_and_has_been_executed.  This imposes quite a
limit on things; ie you have to keep code size small (which is the
opposite of -O3) and looping (which is kind of the opposite of -O3 :)  Of
course, branches can wreak havoc too... The general rule is, there _is_
no general rule.  Try everything out for each individual program to find
out for sure.

...Chambers

- Raw text -