Re: [eigen] patch for tridiagonal matrices in eigen

[ Thread Index | Date Index | More Archives ]

OK ok,

first of all, I'm taking care of the Upper--->UpperTriangular change
tomorrow, so, Mauro, don't worry about it.

But then I thought about implementing the change "from & to ==" as
described in the previous email below.

A bit of thinking shows that it's really an intrusive change! So I
propose to postpone that too to after we branch 2.0.

All what matters is that we keep API stability after 2.0. The
Upper--->UpperTriangular change affects the API so it must be done
now, but the rest, intrusive as it is, doesn't affect the API.


2008/12/19 Benoit Jacob <jacob.benoit.1@xxxxxxxxx>:
> 2008/12/19 Mauro Iazzi <mauro.iazzi@xxxxxxxxx>:
>> I do not understand properly the meaning of these flags. For example
>> part needs assumes some to be exclusive but not others (i.e.
>> UpperTriangular is not UpperHessenberg, but StrictlyUpper is
>> UpperTridiagonal...). If part() checked using == instead of & I think
>> it would be safe to use the Flags to mark the type of the matrix (it
>> already does in some points).
> Hm, good remark.
> It seems that the current design, whcih worked quite well do far, does
> not scale very well.
> Maybe indeed it's time to start using exclusive flags and == instead of &.
> Cheers,
> Benoit


Mail converted by MHonArc 2.6.19+