Performance vs PyCall + Python's dcor #2
Description
EnergyStatistics.jl seems to have considerable worse performance when compared with PyCall + Python's dcor, so I'm opening this issue and if I have time in the future I might look into it.
using BenchmarkTools, EnergyStatistics, PyCall
pydcor = pyimport("dcor")
distance_correlation = pydcor.distance_correlation
a = rand(10000)
b = rand(10000)@btime distance_correlation(a,b)
42.030 ms (13 allocations: 720 bytes)
0.015306081174822574@btime dcor(a,b)
1.546 s (19 allocations: 763.47 MiB)
0.015306081175246091