|Bill Allombert on Wed, 27 May 2009 22:02:08 +0200|
[Date Prev] [Date Next] [Thread Prev] [Thread Next] [Date Index] [Thread Index]
|Re: Another problem with matrix inversion|
On Wed, May 27, 2009 at 12:07:23AM +0200, Bill Allombert wrote: > On Mon, May 25, 2009 at 10:22:39AM +0200, Lorenz Minder wrote: > > Hi, > > > > I would be interested in the numbers, if you still have them at hand. > > What I noticed is that matdet() (when using slow Gaussian elimination) > > is still usable on a slow machine, when matadjoint(,1) wouldn't work for > > me even on a really fast machine. (It was very slow, but the worse > > thing was that it would need way too much RAM.) > > Well, it is a O(n^4) algorithm instead of O(n^3) so it is not > going to be fast, but I think the problem is that RgX_RgM_eval_i is very > inefficient and waste stack space (we should use Brent&Kung). OK, so now I have implemented Brent&Kung for RgX_RgM_eval, so performance are better now. Maybe it also use less memory, because I added garbage collection. Cheers, Bill.