The implementation of dojox.math.matrix.inverse() is inefficient

[enzo]

Here are the results of a comparison between dojox.math.matrix.inverse() and the inverse() method of the Matrix object in ​Sylvester :

  Size   Sylvester      Dojo
 
 20x 20:     17 ms        753 ms
 30x 30:     57 ms      5,216 ms
 40x 40:    123 ms     25,126 ms
100x100:  1,778 ms       ???
200x200: 15,566 ms       ???

I would recommend to adopt the same method as Sylvester (Gauss-Jordan) rather than the less efficient ​adjugate matrix method (which, by the way, in dojox.math is ​considered incorrectly referenced as "adjoint"). Sylvester is licensed under a BSD-style license, and it may be possible to use directly its code.

[enzo]

Test case (requires Dojo and Sylvester installed in the same directory wher this file lives)

[dylan]

Given that no one has shown interest in creating a patch in the past 5+ years, I'm closing this as patchwelcome.