Re: [eigen] Slow matrix-matrix multiply

[Sameer Agarwal <sameeragarwal@xxxxxxxxxx>]

Gael,

Unfortunately the original code and data used to produce this is not mine. Let me ask the original authors if its okay to share. Otherwise I will construct a new example and send it to you.  Give me a few,

BTW at this point Ceres uses eigen for matrix-matrix operations where all the dimensions are fixed.  When one of the dimensions is dynamic we drop down to using a naive loop.

Sameer

On Mon, Apr 8, 2013 at 4:21 AM, Gael Guennebaud <gael.guennebaud@xxxxxxxxx> wrote:

Hi,

On Thu, Apr 4, 2013 at 1:13 AM, Sameer Agarwal <sameeragarwal@xxxxxxxxxx> wrote:
> The thing which is common to all the problems above is that the matrices in
> question are all statically sized. Another problem we are interested in
> involves semi-statically sized matrices, and there the performance
> improvements are much more dramatic.
>
> 8. 8 Cameras 1 Shared calibration 2190 points
>           old-eigen new-eigen new-blas
>   gcc        0.17     0.15      0.04
> clang        0.16     0.13      0.03
>
> In summary, for fixed sized matrices Clang and Eigen have some talking to
> do. For dynamic/semi-static matrices it seems (based on one example) for
> both GCC and Clang custom routines beat Eigen.

This last example is more interesting, and something we expected. How
can I reproduce it?

gael