Reiner Pope
Hi everyone,
I am pleased to announce hmatrix-static[1], a thin wrapper over Alberto Ruiz's excellent hmatrix library[2] for linear algebra.
The main additions of hmatrix-static over hmatrix are: - vectors and matrices have their length encoded in their types - vectors and matrices may be constructed and destructed using view patterns, affording a clean, safe syntax.
hmatrix-static includes statically-sized versions of hmatrix's linear algebra routines, including: - simple arithmetic (addition, multiplication, of matrices and vectors) - matrix inversion and least squares solutions - determinants / rank / condition number - computing eigensystems - factorisations: svd, qr, cholesky, hessenberg, schur, lu - exponents and sqrts of matrices - norms
See http://code.haskell.org/hmatrix-static/examples/ for example code.
This is quite interesting. I would like to know the performance penalty of such a wrapper. I'll test it by myself when I've got time. But can you give me some idea of how such a wrapper can be optimized by ghc in terms of space and time performance? -- c/* __o/* <\ * (__ */\ <