Re: [OF miscellaneous] Hilbert curve: recursion faster than loop?

[Juan Pablo Carbajal]

>> Octave 4.0.1
>>
>> 7.34 ms [0.11 ms]
>> 4.73 ms [0.06 ms]
>> 1.27 ms [0.03 ms]
>>
>> Octave 4.3.0+
>>
>> 7.61 ms [0.16 ms]
>> 4.82 ms [0.05 ms]
>> 1.42 ms [0.03 ms]
>>
>> Matlab R2016b
>>
>> 8.65 ms [0.58 ms]
>> 2.98 ms [0.10 ms]
>> 1.30 ms [0.08 ms]
>
> Same figures here except for nuances using Octave 4.3.0+ and Matlab R2017a
> as Juan and jbect got.

Thanks jbect and Kai.

The lesson learned seems to be that recursion can beat loops and pre-allocation.
It seems matlab is even worse than Octave at slicing and asignation
(assuming second line is the iterations without allocation). Does
anybody know how is this possible? Why is re-writting an array's
values more expensive than extending the array?

I guess we will stick to the recursion then, and maybe to move it to
C++, hopping this will further reduce the cost.