Re: [eigen] discrepancy in (triangular view + transpose) and self-adjointness

[Christoph Hertzberg <chtz@xxxxxxxxxxxxxxxxxxxxxxxx>]

Christoph Hertzberg wrote:
> Manoj Rajagopalan wrote:
> > But, the following code produces a discrepancy:
> >
> >     MatrixXi B(N,N);
> >     B.setRandom();
> >     C1 = A.selfadjointView<Lower>() * B;
> >     C2 = (A_l * B)   +   (A_zl.transpose() * B);
> >     assert(C1 == C2); // <--------  FALSE!
>
> seems like A_zl.transpose() * B ignores the ZeroDiag flag.
> replacing (A_zl.transpose() * B) by
> (A_zl.transpose().toDenseMatrix() * B)
> makes it working ...
> I can't tell where exactly the error happens

damn, thats hard to debug.
It seems that the Problem is that in 
/Eigen/src/Core/products/TriangularMatrixMatrix.h:90

> // implements col-major += alpha * op(triangular) * op(general)
> template <typename Scalar, typename Index, int Mode,
>           int LhsStorageOrder, bool ConjugateLhs,
>           int RhsStorageOrder, bool ConjugateRhs>
> struct ei_product_triangular_matrix_matrix<Scalar,Index,Mode,true,
>                                            LhsStorageOrder,ConjugateLhs,
>                                            RhsStorageOrder,ConjugateRhs,ColMajor>;

does check for UnitDiag, but never checks for ZeroDiag.

Can't help with fixing, though ...


--
----------------------------------------------
Dipl.-Inf. Christoph Hertzberg
Cartesium 0.051
Universität Bremen
Enrique-Schmidt-Straße 5
28359 Bremen

Tel: (+49) 421-218-64252
----------------------------------------------