Newsgroups: comp.parallel.pvm
From: wrankin@ee.duke.edu (William T. Rankin)
Subject: Re: Numerical Recipes (was Re: Wanted: fortran source for matrix operations using PVM)
Organization: Duke University EE Dept.; Durham, NC
Date: 21 Oct 1994 12:15:22 GMT
Message-ID: <388bcq$9e7@news.duke.edu>

In article <TIM.94Oct16140122@isis.usi.utah.edu>, tim@isis.usi.utah.edu (Timothy E. Burns) writes:
|> 
|>    Or you may try the book " Numerical Receipt for fortran ".
|>    [snip]
|>    Brenda
|> 
|> Dear Brenda,
|> 
|> I know you mean well, but this is poor advice.  I can think of no
|> algorithm in numerical recipes that treats important issues in 
|> optimization.  Even from a pedagogical point of view, you are
|> much better off learning about linear algebra algorithms from Golub
|> and Van Loan's _Matrix_Computations_.

If you are at a university, you are *much* better off to 
learn about linear algebra algorithms by sitting in on a grad level 
class on the subject :-)

Ictually, I found _Recipes_ to be an invaluable aid while taking 
that linear algebra class.  most of the explanations of the algorithms
are well written and concise.  as an engineer, it was nice to be able
to get a concise example for the principles discussed in our main text, 
in this case Kincaid and Cheney's book.

|> Books like _Numerical_Recipes_ have done a lot of damage to scientific
|> computing, and I know that once I was victim.

I don't think anyone claims that _Recipes_ is the end all source for
highly optimized numerical algorithms.  But it is a good start to
try to get a handle on what algorithm you need to be looking for.
True, a more efficient and optimized algorithm may be available on
some of the archive servers, but you have to know what you are looking
for first.

For example, for our research involving multipole methods to solve
particle systems, we needed algorithms to generate Legendre and
Gegenbauer polynomials.  Recipes_ (in C) provided us with what we
needed and allowed us to continue our research without the necessity of
reinventing the wheel.

If the bottleneck for our program was generating Legendre polynomials
then we would have taken a long hard look at the most efficient algorithm
for doing this.  But to waste time carefully optimizing a section of
code that is called fairly infrequently would have been a rather silly waste
of our time.  _Recipes_ provided us with working, tested code that allowed us
to get on with more important things.

I know, this is wandering a bit from the original topic, but to dismiss 
"Recipes" as useless is doing a great disservice.  It is an extreamly
useful book for scientific programmers.

|> Sincerely,
|> Tim Burns
|> 
|> --
|> Tim Burns                               tim@osiris.usi.utah.edu
|> Utah Supercomputing Institute           (Day) 801-581-5172
|> 85 SSB, Univ. of Utah, SLC, UT 84112    http://usi.utah.edu/burns/tim.html

-- 
----                                /       __/    /    /
bill rankin                        /              /    /
wrankin@ee.duke.edu               ___  /    /    /    /
philosopher/coffee-drinker       /    /    /    /    /
molecules-R-us                  /    /    /    /    /
                             _______/  __/  __/  __/

"A distributed system is one in which I cannot get something done
because a machine I've never heard of is down"   --Leslie Lamport

