On Saturday, 12 October 2013 at 05:20:11 UTC, SomeDude wrote:
On Friday, 11 October 2013 at 16:10:21 UTC, FreeSlave wrote:
There is "Matrices and linear algebra" module in wish list.
Let's discuss its design. D is complicated language so it's
difficult to choose the right way here. We need to find
compromise between efficiency and convenient interface. I'm
going to make some suggestions how this module should look
like.
Please, make your suggestions too.
Firstly, my opinion is, it shouldn't be in the std package.
Secondly, if it's not super highly performant, noone will use
it. The highest performing linalg libraries implement the low
level BLAS API or are vendor specific, like the Intel MKL.
These are super optimized libraries where the asm code is hand
optimized by specialists (GotoBLAS and the derived OpenBLAS for
instance), and these will be *very* hard to beat. So much so
that if BLAS isn't used as the engine, it's very unlikely that
your library will be used for serious number crunching.
Built above BLAS, you can find a number of libraries that make
using linear algebra more pleasant. One such library, currently
the best IMHO is Armadillo C++. It has the very big advantage
that it mimics Matlab functions. This feature is extremely
useful, because scientists very often model their algorithms in
Matlab before porting them to C++. This is what is done in the
company I work for.
So the best I could suggest is to study the Armadillo C++
library and port it to D.
For these cases we may let users to choose low-level backend if
they need. High-level interface and default implementation are
needed anyway.
I called it std.linalg because there is website
http://www.linalg.org/ about C++ library for exact computational
linear algebra. Also SciD has module scid.linalg. We can use
std.linearalgebra or something else. Names are not really
important now.
Ok, things are more clear now. Today I look what I can do.
