In article <5907@calmasd.Prime.COM> gdh@calmasd.Prime.COM (Gerald Hall) writes:
>In article <13306@lhdsy1.lahabra.chevron.com> shc@lhdsy1.lahabra.chevron.com (S.H. Couturie) writes:
>>In article <23qh9i$e6@info.epfl.ch> brossard@sic.epfl.ch writes:
>>> S10-512 (5.2):
>>>real 2.0
>>>user 1.9
>>>sys 0.0
>>>
>>OK, can't resist. I copied the /usr/share/man/man1/csh.1 from our
>>SparcCenter 1000 to my R4000 Indigo, and:
>> /bin/time nroff -man csh.1 >/dev/null
>> real 2.1
>> user 1.6
>> sys 0.0
>>Gee, and my Indigo is a year old.....
>
>Ok, now try this on your SC10000 and your Indigo (preferably from a
>script so they all start about the same time) the average the times:
>
> /bin/time nroff -man csh.1 >/dev/null &
> /bin/time nroff -man csh.1 >/dev/null &
> /bin/time nroff -man csh.1 >/dev/null &
> /bin/time nroff -man csh.1 >/dev/null &
>
>Hint: the SC1000 will do four of these (or more depending on how
>many CPUs you have installed) in essentially the SAME amount of time
>your Indigo does one, while your Indigo will take more than four
>times as long to do each one with four running. You may do only one
>thing at a time, but for a busy system multiprocessing is a big win.
>
So?
SGI has been shipping working SMP systems for almost 5 years now.
We have an old one (4D/240GTX), with 4 25MHz R3000 processors, and the
process-level multiprocessing is a big win.
But, the time won't *really* be the same for each process, given
finite disk bandwidth, finite memory, etc....
BTW, if I had a 36 Processor Challenge, I could do 36 copies of this stupid
test in the same time.....
(We also have a Sun 690, 4 Processor, junker, still running 4.1.2, because of
the idiotically ridiculously stupid Sun upgrade procedure, when extra kernel
products are involved.... and NO you DO NOT get good multiprocessing with
that! No SMP, etc...)
Steve <DISCLAIMER: I speak for myself only...>
-- Steve Couturie Voice/Fax: (310) 694-9332/7709 La Habra Unix Systems Internet: shc@chevron.com Chevron Information Technology Co. UUCP: ...!uunet!lhdsy1!shc P.O. Box 446, La Habra, CA 90633-0446
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:08:08 CDT