Re: [eigen] 2.0.16 and 3.0.1 for the end of the week.

[Anton Gladky <gladky.anton@xxxxxxxxx>]

I have checked that, and there is really a regression since 2.0.10.
But, unfortunately, 2.0.16 will not solve your problem...

I would recommend you to use 3.0.0 or coming 3.0.1.
It takes only 7.5ms to execute the code and changes are minimal:
just get rid of "#include <Eigen/Array>".

Anton




On Thu, May 26, 2011 at 9:28 PM, Robert Lupton the Good
<rhl@xxxxxxxxxxxxxxxxxxx> wrote:
> Compiled -O3 n a 64-bit Linux box the following code takes 9ms with 2.0.10 and 285ms with 2.0.15
>
> I don't know the hg root to clone the release candidate for 2.0.16 so I haven't checked.
>
>                                R
>
>
> #include <Eigen/Core>
> #include <Eigen/Array>
>
> int main(){
>  int n = 50;
>  Eigen::MatrixXd m1(n,n);
>  Eigen::MatrixXd m2(n,n);
>  Eigen::MatrixXd m3(n,n);
>  m1.setRandom();
>  m2.setRandom();
>  m3.setRandom();
>
>  Eigen::VectorXd v1(n);
>  Eigen::VectorXd v2(n);
>  Eigen::VectorXd v3(n);
>  v1.setRandom();
>  v2.setRandom();
>  v3.setRandom();
>
>  for (int i = 0; i != 100; ++i) {
>     v3 = m1*(m2*v1+m3.transpose()*m2*v2);
>  }
>
>  return 0;
> }
>
>
>
>