Re: [eigen] Recursion and block matrices

[Norman Goldstein <normvcr@xxxxxxxxx>]

Thanks for explaining Stride<0,0>.

A few weeks, ago, when I read Gael's email, I was already sufficiently 
overwhelmed with Eigen, and did not want to get involved in the Devel 
branch.  Today I took a look, and the Ref certainly seems related, but I 
can't say that I understand it well enough to use it.  It seems that the 
inner stride must be 1?

What I am aiming for is conceptually simple, and it is allowing me to 
learn the Eigen class hierarchy well enough to implement sub-matrices 
that will be less efficient as expressions, but will not have infinite 
recursion, and will share the same memory as the original matrix.  In a 
nutshell, a sub-matrix is a Map of a rectangular subset of a PlainObject.

Will let you guys know when I am done.  I'm sure I will not have done 
this in the best way, and any suggestions you have for improvement will 
be taken to heart.  I hope will not have too many question to bother you 
along the way.  If you have any other ideas for defining submatrices 
with the qualities I described above, please shout.  Although I enjoy 
the challenge, I would not classify this as fun, and am looking forward 
to using this functionality.

Thanks, again,
Norm



On 09/18/2012 08:59 AM, Christoph Hertzberg wrote:
> On 18.09.2012 17:51, Norman Goldstein wrote:
> > I figured this one out by decomposing the C++ statements until the
> > compiler message was more specific.
> > It turns out that the compiler was being asked to match Stride<0,0> with
> > Stride<-1,-1>.  The former is
> > defaulted in some declarations, while the latter is for dynamic stride.
> > When would one want to use
> > Stride<0,0>?
>
> Stride<0,0> means "natural" stride, so colstride is number of rows, 
> rowstride is 1 (assuming colmajor matrices).
> Have you yet tried to use the Ref<MatrixXd> class, as Gael suggested 
> some mails ago? Or are you implementing this more for the fun of it?
>
> Christoph
>
>
> > > > On 08/28/2012 01:01 AM, Gael Guennebaud wrote:
> > > > > With the devel branch you might try:
> > > > >
> > > > > void incr(Ref<MatrixXd> mat) {
> > > > >
> > > > > // no cast needed
> > > > > ...
> > > > >
> > > > > }