Re: goals for 3.1

[David Bateman]

David Bateman wrote:
> I can't generate these failures, but I've moved on since and have
> eliminated all but splu of the dispatched sparse functions. So perhaps I
> fixed the issue during that process. Can you wait till I also get rid of
> splu and I send an additional patch and see if that fixes it?
>
>   
Ok here is the patch that eliminates the rest of the dispatched sparse
functions. It also should fix the bug you saw John. It also modifies the
lu and chol functions to be compatible with matlabR2007b adding the
"vector" argument, eliminating initial column permutations for the
sparse matrices, and adding the possibility of scaling factors for
sparse matrices. The initial Q value can be obtained by replacing
existing lu(A,Q) calls with lu(A*Q), so this is in fact no lose which is
probably why mathworks removed it.

Another notable change is the single argument return value of LU.. This
now return the matrix returned by xGETRF in matlab2007b, or a simulated
version of it for sparse matrices, that includes the L and U matrices.
Essentially this is Y = L + U - eye(size(U)). However a single argument
return from the lu function is essentially useless even after this
change as the permutation information is lost.

Regards
David

-- 
David Bateman                                address@hidden
Motorola Labs - Paris                        +33 1 69 35 48 04 (Ph) 
Parc Les Algorithmes, Commune de St Aubin    +33 6 72 01 06 33 (Mob) 
91193 Gif-Sur-Yvette FRANCE                  +33 1 69 35 77 01 (Fax) 

The information contained in this communication has been classified as: 

[x] General Business Information 
[ ] Motorola Internal Use Only 
[ ] Motorola Confidential Proprietary

patch7578.bz2
Description: BZip2 compressed data