Re: Order of matrix multiplication

[Nicholas Jankowski]

On Mon, Sep 12, 2016 at 2:59 PM, Yu Liu <address@hidden> wrote:

FYI, numpy has a special function np.linalg.multi_dot to do this. https://github.com/numpy/numpy/blob/b91e8d8f164731bb710cc1e5173cc8ec3f8fadf5/numpy/linalg/linalg.py

Their code has very detailed explanation of the well-known algorithm.

definitely looks like something that could be adapted.  That said, I did just recall that there is an existing bug waiting to have a patch tested and committed for mtimes related to improperly flattening an ND array. So if anyone starts looking to make changes it should take the current offered patch into account: https://savannah.gnu.org/bugs/?47173

from that report it seems fixing mtimes was not the most trivial of tasks, so 'tread with care'