[Ocaml-biz] Let's choose a market

[Brian Hurt]

On Sat, 11 Sep 2004, Brandon J. Van Every wrote:

> > Rendering farms, crypto- probably not.  These are two places
> > where even a 10% performance hit isn't acceptable.
> 
> I'm not believing you here.  If what you said was true, rendering farms
> would get written in C rather than C++.  I think this is a C FFI /
> Bigarray issue, not a "can't use a HLL" issue.  ILM uses Python to
> control its rendering farms.  Of course, that's just the network glue,
> not the rendering code.

The glue code, yes.  The rendering itself?

As for C/C++, the "conventional wisdom" is that C++ is as fast as C.  I've 
never actually seen a benchmark proving this, and I've seen a number of 
benchmarks disproving it, but the nice thing about CW is that it doesn't 
need to be true to be CW.

Most crypto is done in C, when it isn't done in assembler.  As they 
actually do measure the performance of different implementations.

> 
> > I'm really starting to think something Ocaml needs is a kick
> > ass, take no prisoners, best of breed linear algebra library.
> 
> Several exist at http://caml.inria.fr/humps/caml_Mathematics.html
> Do any kick sufficient ass per your definitions?

No.  There are a couple of BLAS/LAPACK wrappers.  I want something better 
than BLAS/LAPACK.

-- 
"Usenet is like a herd of performing elephants with diarrhea -- massive,
difficult to redirect, awe-inspiring, entertaining, and a source of
mind-boggling amounts of excrement when you least expect it."
                                - Gene Spafford 
Brian