Risc System/6000 (use of 1.1 Dhrystones)

[Dick Dunn]

khb at chiba.Eng.Sun.COM (Keith Bierman - SPD Advanced Languages) writes:
 [I wrote]
>    Then there's the business of reporting Dhrystone 1.1 numbers.  That's just
>    a disgrace.
...
> Far be it from me to defend IBM; but dhrystone 1.1 is what many of the
> chip houses still use for quoting. Thus many marketing departments
> feel obligated to use it...

"But Mom!  Everybody else is doin' it!"

I won't retract my flame about this.  It's still wrong.  It's still either
dishonest (if they know better) or mind-bogglingly ignorant (if they
don't).  What I should have pointed out is that IBM is not alone in this
sort of travesty...which does not make it any less a travesty.

As long as marketing departments feel they can get away with reporting an
over-optimized 1.1 Dhrystone, they're likely to do it.  That's why I'm
making noise.  I don't think we should let them get away with it.  Also,
there are folks who use 2.1 Dhrystone, and they should not be penalized for
doing so.

Also remember that the big headline-generating MIPS numbers came directly
from this old benchmark-of-ill-repute.

Anybody got some 2.1 Dhrystone numbers on the 6000 boxes?  Anybody looked
at the generated code to see if they're playing by the rules?
-- 
Dick Dunn     rcd at ico.isc.com    uucp: {ncar,nbires}!ico!rcd     (303)449-2870
   ...Don't lend your hand to raise no flag atop no ship of fools.