Great – thanks for reporting back. It's nice that you could get that kind of good performance here without too much shenanigans.
On Sat, Dec 6, 2014 at 5:50 PM, Andre Bieler <andre.biele...@gmail.com> wrote: > for completeness: > > with the inner loops now going through the first index as suggested by > Jeff, > there was another increase in speed. So now I stand at *16.8 s* on average > with julia. > > The same thing in python/numpy takes roughly *6800 s* to run > (however not vectorized in numpy, using for loops as in the examples > above) >
