Re: [eigen] geometry module

[ Thread Index | Date Index | More Archives ]

On Wed, Sep 8, 2010 at 2:04 PM, Gael Guennebaud
<gael.guennebaud@xxxxxxxxx> wrote:
> The problem is not the copy which are well optimized away by gcc, but
> the extra 5 shuffling. Maybe some shuffling can be removed by directly
> vectorizing the quaternion * vector product (we currently vectorize
> quat*quat only).

sorry, there are not 5 extra shuffling only, but 7 (!), and gcc
already optimized out 2 of them. I'm wondering how this vectorized
cross3 function can be faster than a scalar cross product....


Mail converted by MHonArc 2.6.19+