std/num/math: A module for all your math related needs

[Ki11erRabbit]

I was talking with my roommate and realized that linear algebra functions for Koka would be great to have. I did some searching and found that Numpy uses BLAS. I was wondering if it be worthwhile to start work on bindings for Koka or if it would be better to wait for some feature to be merged into Koka that would make writing this module easier.
It would also be great to learn about other fast math libraries to complete this module.

[Ki11erRabbit]

I have just found a library called MAGMA It says it is compatible with BLAS so it might be interesting if Koka could have gpu accelerated linear algebra out of the box.

[TimWhiting]

Cool!

[TimWhiting]

What about having a separate math package. Instead of std/num/math, we can just have math/vector, math/matrix, math/linalg, etc. I don't want import names to get too long, and it would be good not to include too much in the standard libraries, especially if it requires external packages that need installation.

[Ki11erRabbit]

That works for me.

[Ki11erRabbit]

Should I use the naming convention that BLAS uses or should I make it have a "sane" name? I say this because some of the functions have really uninformative names. There is one problem in giving them a good name. This requires that I actually know what a function is doing.

[TimWhiting]

I would suggest that you make the externs follow the BLAS name, but then provide an inline api on top of that renaming things.