[Beowulf] Teaching Scientific Computation (looking for the perfect text)

[Robert G. Brown]

On Tue, 20 Nov 2007, Nathan Moore wrote:

> Our institution has a site license for Mathematica and between that and a
> compiled langauge, I feel guilty telling my students to spend more money on
> something that seems to be of only marginal utility.  Also, (and I'm sure
> I'm wrong on this), Matlab seems like a tool that's permanently in the minor
> leauges.  Sure it has a nice IDE and makes pretty pictures, but so does
> mathematica.  Additionally, (at least when I was using it back in the mid
> 90's), Matlab is an interpreted language.  If you start writing "real" code
> on it that will run for days or months, the compiled (C/fortran) equivalent
> will be significantly faster (I almost said "orders of magnitude faster,"
> but I've never been curious enough to actually make a comparison)

That's correct most of the time.  Even plural.  Sometimes it's not so
bad, though -- it depends on whether you're in a big subroutine call
that is basically looping a block of compiled code on mostly internal
variables or if you're working through the usual list of "objects"
maintained by an interpreted language that does real-time allocation and
garbage collection.  Objects simply don't stream, usually.  So there is
usually a hit per line, a hit for non-sequential memory access, and a
few other hits along the way (e.g. I/O if the routine is busy writing to
your GUI as it goes).

   rgb

-- 
Robert G. Brown
Duke University Dept. of Physics, Box 90305
Durham, N.C. 27708-0305
Phone(cell): 1-919-280-8443
Web: http://www.phy.duke.edu/~rgb
Lulu Bookstore: http://stores.lulu.com/store.php?fAcctID=877977