john  writes:
>In article ,
>anton@mips.complang.tuwien.ac.at says...
>>
>> john  writes:
>> >To give you another suprising fact - parallel computer design is now
>> >moving away from caring about how many machine cycles are taken.
>> >Spending time worying about that is counter productive.
>> >(Sorry I can't recall the reference for such a bold statement)
>>
>> Outrageous claims are a dime a dozen.  Whoever made it was either
>> bullshitting, or was making a very specific claim (that probably few
>> people care about), and tried to attract attention to his agenda by
>> headlining it with an outrageous statement.  In the latter case the
>> statement is meaningless without the context.  So better forget about
>> that.
>>
>> - anton
>
>Perhaps I should have been clearer anton - I was talking about efficient use
of
>individual CPU cycles in a parallel system that are not so relevant.

You have to be yet clearer.  This claim is still nonsense.  If you
neglect individual CPU cycles such that each individual CPU takes,
say, twice as many cycles the whole application will take twice as
long, on the same parallel machine.  The parallel machine guys buy
these expensive parallel boxes so that their applications run faster,
so they certainly would not appreciate this kind of waste.

>It's not just a random claim - and cetainly not outrageous - it was the result
of
>experimentation or I wouldn't have repeated it.
>(I think it may have been a Sandia document - or possibly something
>Vandierendonk wrote - I will give the link if I come across it again)

I can imagine many valid statements that then become chinese-whispered
into nonsense.  I guess that's what happened here.

>There comes a point where focusing on the fine detail becomes self defeating
is
>the real point I think. Especially where parallel computing is concerned.

Sure, in looking for performance you have to look at where time,
memory, etc. is actually spent, and not where it is not spent, and I
have seen students follow an optimization idea without taking that
into account.  But that's something different from "efficient use of
>individual CPU cycles in a parallel system [...] are not so relevant"

- anton
--
M. Anton Ertl  http://www.complang.tuwien.ac.at/anton/home.html
comp.lang.forth FAQs: http://www.complang.tuwien.ac.at/forth/faq/toc.html
     New standard: http://www.forth200x.org/forth200x.html
   EuroForth 2018: http://www.euroforth.org/ef18/

--- SoupGate-Win32 v1.05