Benchmarks for IBM Risc/6000

[Ed Clarke/10240000]

>From article <25834 at ut-emx.UUCP>, by nghiem at ut-emx.UUCP (Alex Nghiem):
> and Macintosh II A/UX 1.01 (~2900 drystones). The RISC/6000 is
> advertised to run ~45K drystones.

Ok, I tried a copy that Dan Ehrlich sent to me from shire.cs.psu.edu 
using the xlc compiler with "-DTIMES -DHZ=100 -O" and no loader flags.
I got 39920 dhrystones and it didn't make any significant difference
whether or not I used register variables.  The benchmark says "C/2.0"
in the source, while my OS build level is 8951(O).  This is a floor
standing machine, but not the real fast one ( strangely enough, all
the faster machines went to the people who put them together ... ).
My code wanted a number input through stdio - I used 1000000 to be 
sure things evened out.

Open mouth insert foot time ... not exactly about the RS-6000 -

I'd be happier if this benchmark were slightly modified.  Take a
small driver program that runs twenty copies of Dhrystone 2.0 in
parallel.  Give me the real elapsed time between the start of the
first copy and the death of the last ... use a standard loop count
of 1000000.  Benchmark should be run on a freshly re-booted system
with all the normal ( sendmail, named, NFS, gettys, X11 ) tasks running
but only one user ( not the same as 'single user maint mode' ).

I'm an "End User", and I don't really care how fast you can calculate
floating point numbers and move strings.  Gime a real number - how
long am I going to sit here waiting for my job to finish?  The numbers
right now just give user process time.  That's ok, but what about
paging and task switching and spooling and and and ... 
-- 
Ed Clarke      | Happiness, n.  An agreeable sensation arising from
acheron!clarke | contemplating the misery of another. - Ambrose Bierce