Block operations could be faster (ever so slightly)

[Pouya-moh]

Hi,

One way to speed up, is to use fixed size matrix block operations which according to Eigen documentation is slightly faster. However overall it could mount to something considerable.

// Block of size (p,q), starting at (i,j) 
matrix.block(i,j,p,q); // dynamic
matrix.block<p,q>(i,j); // fixed

This holds for segment, head, and tail operations for vectors (down below the same documentation page) and corner operations for matrices. This of course cannot be a single fix. It has be done slowly over the course time.

Cheers,
Pouya.

[EnricoMingo]

Ok, maybe I can prepare a branch and work on it.

[Pouya-moh]

I can also contribute to that branch...