Re: Expm giving NaN

[David Bateman]

Marco Caliari wrote:
> Dear maintainers,
>
> the patch against CMatrix.cc in 2.9.14 fixing the "expm giving NaN" bug.
> Here the result:
>
> octave:1> version                        
> ans = 2.9.14
> octave:2> expm(1000*i-10)
> ans = 2.5532e-05 + 3.7540e-05i
> octave:3> expm([1000*i-10 0;0 1000*i-10])
> ans =
>
>   2.5532e-05 + 3.7540e-05i  0.0000e+00 + 0.0000e+00i
>   0.0000e+00 + 0.0000e+00i  2.5532e-05 + 3.7540e-05i
>
>
> I also rewrote the computation of the rational approximation avoiding a 
> matrix-scalar product and including a for loop. I'm not sure if it is 
> better. 
>
> Best regards,
>
> Marco
>   

Marco,

John should probably comment on this patch before it is committed, but
can you resupply to patch as a diff with context. The type of patch you
sent is difficult to apply if the code has changed (and in fact it has).
To generate a diff with context, do something like

diff -u liboctave/CMatrix.cc.orig liboctave/CMatrix.cc

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