JacobiSVD Returns Incorrect Matrices

[nriaziat]

Hi,

I'm attempting to use the JacbiSVD algorithm on some 3x3 matrices and consistently get wildly incorrect SVD results (comparing to MATLAB). Both the singular values and the resulting U,V matrices are off by several orders of magnitude. Any ideas why this could be? I'm using stlport.h to compile for an Arduino Mega 2560.

Thanks!

[rpavlik]

Not sure though I haven't used this in forever. You could check upstream Eigen to see if they've fixed any bugs in that code there. (The only easy guess I have is possibly memory order, see if the result on the transpose looks better...)

[nriaziat]

I'll check the transpose but the singular values should still be invariant under transposition so I'd expect those values to be correct.