Let me ask an honest question: does performance really matter that
much for small/medium problems? If so, isn't it more cost effective just
to buy better hardware than to mess with esoteric C++ optimization
techniques? For small to medium problems, I would think that what
matters more is programmer time, indicating that a language designed
with software engineering in mind rather than performance would be
more appropriate. As one slides up the scale to very, very large problems
where the last flop is important for producing management-pleasing
bullets on slides, performance becomes relatively more important, perhaps
passing programmer time enough that dropping to fast languages or
jumping through hoops to get other language to act fast may be worth it...
Perhaps I am biased by being in DOE and seeing problems in the hundreds
of millions, but a few thousand equations seems like something that should
be done on a calculator these days. I'm truly interested in being educated
on the practical aspects of small/medium problems...
Cheers,
Andy
At 05:22 PM 3/28/00 -0900, you wrote:
>On Tue, 28 Mar 2000, Win-Tsuang Hwang at vector-space.com wrote:
>
>> Hi all:
>>
>
>> >It seems to be a fact that numerically intensive code spends a lot of
>> time performing matrix maths.
>>
>> One more example (besides DSP) that this statement is not always correct
>> is in finite element method. For small to medium size problems (less
>> than a few thousands d.o.f.), more time (at least many folds) is spend
>> on computing components of element stiffness matrix, not on solving
>> matrix problem of the global stiffness matrix.
>>
>
>I agree with that. Also, for certain nonlinear finite element problems,
>lots of iterations are done at the gauss points which can take lots of
>time relative to solving the global equations.
>
>Jonah
>
>> Thanks,
>>
>> Win T. Hwang
>> winhwang@vector-space.com
>> VectorSpace Programming, http://www.vector-space.com
>>
>> --------------------- Object Oriented Numerics List
>--------------------------
>> * To subscribe/unsubscribe: use the handy web form at
>> http://oonumerics.org/oon/
>> * If this doesn't work, please send a note to owner-oon-list@oonumerics.org
>>
>
>--------------------- Object Oriented Numerics List --------------------------
>* To subscribe/unsubscribe: use the handy web form at
>http://oonumerics.org/oon/
>* If this doesn't work, please send a note to owner-oon-list@oonumerics.org
Dr. Andrew J. Cleary
Computational Scientist
Center for Applied Scientific Computing
Lawrence Livermore National Lab
L-561, PO Box 808, Livermore, CA 94551
925-422-1939
--------------------- Object Oriented Numerics List --------------------------
* To subscribe/unsubscribe: use the handy web form at
http://oonumerics.org/oon/
* If this doesn't work, please send a note to owner-oon-list@oonumerics.org
This archive was generated by hypermail 2b29 : Wed Feb 20 2002 - 03:20:11 EST