[racket] LAPACK FFI

From: Neil Toronto (neil.toronto at gmail.com)
Date: Sat Oct 20 23:40:40 EDT 2012

On 10/20/2012 07:58 PM, Richard Cleis wrote:
> Is anybody working on the LAPACK FFI listed on the Hackathon page?
> How does "BSD Licensed Matrix Library" apply to a Racket FFI?
> I'd like to know which makes more sense:
> - write/find the few Gaussian Elimination solvers that I need
> - Figure out how to do it with FFI and LAPACK

You have another option:

  - Wait for the math library.

That might not be quick enough, though. The initial check-in is a week 
or two off, and there will still be polishing and debugging left to do 
afterward. Oh, and math/matrix currently doesn't compile because Sam has 
to fix something. Also, you'd have to use a nightly build.

Okay, maybe not such a great option yet. :D

I'd first try an FFI to LAPACK. Here's my experience. I've written my 
own arbitrary-precision floating-point library in Racket before. For 
math/bigfloat, though, I made an FFI to MPFR. The latter was faster to 
produce, it has more functions, and I'm much more certain that 
everything in it works.

But now I have to reveal my bias: I might personally like to have an FFI 
to LAPACK, but I don't have time to write it. ;)

Neil ⊥

Posted on the users mailing list.