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

[Greg Lindahl]

On Tue, Nov 20, 2007 at 12:33:43PM -0600, Nathan Moore wrote:

> After reflection though, I've started to wonder about the wisdom of my
> choice.

I haven't seen a vendor lib that didn't have a Fortran interface (ACML
does, ATLAS does, goto does), and I hear that CUDA may well have
Fortran soonish -- after all, the compiler infrastructure it's built
on does Fortran. But the concrete points you make are probably just
symptoms of your general unease.

For me the biggest concern would be pointers. It's well known that
students new to C struggle with them. You're trying to teach more than
2 languages in a single course. And the C pointer hell is a lot more
than 3 weeks.

> Also, and more relevant, I assume that most employers word-associate
> "Fortran is to backwards as C is to competence".

Unless you're a physicist. And if we're talking about employing
someone as a _scientific_ programmer, I'd expect them to know both.

-- greg