Jump to content

Talk:Comparison of linear algebra libraries

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia

Alglib

[edit]

Alglib supports VB.Net too, just so people are aware of it. (For many users a major strong feature) — Preceding unsigned comment added by 85.164.125.248 (talk) 07:26, 18 December 2011 (UTC)[reply]

Other libraries

[edit]

If I interpret the article correctly, then more libraries that should be covered:

  • AMD Core Math Library (ACML) which has "A full implementation of Level 1, 2 and 3 Basic Linear Algebra Subroutines (BLAS), with key routines optimized for high performance on AMD Opteron™ processors.", as well as customized LAPACK routines, and FFT and random number generation routines.
  • GotoBLAS which is a modern BLAS with good performance

Also, a library for sparse linear algebra (so it technically fits under this page's title) is:

  • OSKI: Optimized Sparse Kernel Interface, from a group at Berkeley including the well known James Demmel.

A library that I don't know as well but might fit into this page:

  • PhiPAC for high-performance BLAS

And also the well known Jack Dongarra has recently (Sept 2011) updated his list of free linear algebra software, so clearly this page should be useful:

Lavaka (talk) 10:09, 20 December 2011 (UTC)[reply]

What about Meschach? Should this be in the table? http://homepage.math.uiowa.edu/~dstewart/meschach/ — Preceding unsigned comment added by 86.152.43.249 (talk) 16:03, 21 November 2015 (UTC)[reply]

I add Newmat in that list if someone ever tries to complete it:

Performance

[edit]

Is there a metric by which these libraries can be compared for performance? --192.31.106.36 (talk) 18:47, 2 December 2013 (UTC)[reply]

Scipy BND matrix support

[edit]

scipy offers a solver for Ax=b with A being a band matrix. However, there is no BND matrix type itself, and the solver works by interpreting the input matrix (which is a 2D array) in a special way, see here: scipy.linalg.solve_banded documentation

I don't know if this counts as having BND matrix support in the second chart. Julainius (talk) 17:18, 25 March 2019 (UTC)[reply]

eigen and boost

[edit]

more of a question: why are eigen and boost not included? I know that boost offers more than just linear elgebra, but I think not including or at least mentioning (maybe with the note that they contain more). This article might be misguiding without any mentioning. — Preceding unsigned comment added by Poritz (talkcontribs) 13:09, 20 April 2019 (UTC)[reply]

Both are added now! Regards, Voorlandt (talk) 20:36, 25 December 2020 (UTC)[reply]