Bill Allombert on Mon, 13 Mar 2017 17:00:28 +0100

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

Re: Linear algebra via CUP decomposition and reduction to matrix multiplication

On Mon, Mar 13, 2017 at 08:26:11AM +0100, Peter Bruin wrote:
> Thanks!  Here are two follow-up patches to simplify a few things:
> - The CUP_gauss functions were unnecessarily moving certain matrix
>   entries up, thereby destroying known zeroes which then had to be put
>   back by other functions; the first patch avoids this unnecessary work.
> - The matrix U and the permutation P were always computed, even when
>   only the echelon form was needed; the second patch allows U = P = NULL
>   in the CUP_gauss functions.  Also, this patch removes a useless
>   gerepile in Flm_CUP_gauss and moves the gerepile in FlxqM_CUP_gauss
>   into the outermost loop (potentially uses less memory with the same
>   time complexity).