Well, this is actually an interesting story. In my first test the code was 
about 100 times slower for which reason I did not investigate Julia further 
and concentrated on a Numpy like multidimensional array package called 
numcpp: https://github.com/tknopp/numcpp. And with C++11 I got actually 
quite far.

But at some point I reevaluated Julia and thanks to 
https://github.com/JuliaLang/julia/issues/4832 got the performance issues 
sorted out.

In the end I was about a factor 1.3 slower than the C version. At least for 
1D and 2D. ND currently only has a fallback implementation. The cartesian 
macros could make this a lot faster.

Am Freitag, 16. Mai 2014 13:18:39 UTC+2 schrieb Ariel Keselman:
>
> How close did you get to c speed?

Reply via email to