Re: [eigen] problem with MKL bindings for dense Eigen

[Gael Guennebaud <gael.guennebaud@xxxxxxxxx>]

On Mon, Apr 9, 2018 at 3:09 PM, Peter <list@xxxxxxxxxxxxxxxxx> wrote:

Dear Gael,

Am 09.04.2018 um 13:31 schrieb Gael Guennebaud:
> I see.
>
> Problem fixed: https://bitbucket.org/eigen/eigen/commits/3a08a093adac (in 3.3 too).

indeed, thanks.

BTW, the MKL version is now consistently more accurate than Eigen without MKL.
Could be a hint on the usage of FMA, but I'm just guessing.

It's more likely more accuracy in the lowest eigenvalues as we assume that eigenvalues lower than machine_epsilon * largest_eigenvalues are zero. 

gael

 

Best regards,
Peter