Bill Allombert on Wed, 23 Sep 2009 19:44:29 +0200

[Date Prev] [Date Next] [Thread Prev] [Thread Next] [Date Index] [Thread Index]

Re: matkerint versus matsnf

On Wed, Sep 23, 2009 at 06:55:07PM +0200, Jeroen Demeyer wrote:
> Hello pari-dev,
> If I want to compute a Z-basis of the kernel of a matrix, the most  
> obvious GP function would be matkerint().  However, that does a  
> LLL-reduction which is not needed for my application.

> The kernel of an integral matrix can also be found using matsnf().  In  
> many cases, matsnf is faster if the matrix has more columns than rows  
> (more unknowns than relations).

This is almost exactly what matkerint(,2) does, apart for the final
reduction... However this is still very surprising. What kind of
matrix is that ?

> Considering this, I think that it makes sense to add a flag to  
> matkerint() indicating that matsnf should be used.

Why not instead add a flag to matkerint so that it does not perform the
final LLL reduction ?