Poor time performance on Dict?

[nalimilan]

@y4lu

On a smaller set of values (n=500000), using get!(x,i,i) looked about twice as fast as x[i] = i, and I think dicts may outperform arrays when searching for values, but they do seem to slow down when they get big enough. Maybe indexing related?

I cannot replicate this on master with the following code. The two functions perform very similarly.

function dict_performance()
    n = 100_000_000
    x = Dict{Int,Int}()
    sizehint!(x, n)
    for i in 1:n
        x[i] = i
    end
end

function dict_performance2()
    n = 100_000_000
    x = Dict{Int,Int}()
    sizehint!(x, n)
    for i in 1:n
        get!(x, i, i)
    end
end

using BenchmarkTools
@btime dict_performance()
@btime dict_performance2()

OTOH, it seems that the total time/number of iterations ratio gets worse as the number of unique entries gets larger. More systematic benchmarking would be needed to find out whether that’s a real problem or not.

@phrmoy Does storing 100M unique values correspond to your actual use case? It could be that Julia performs as well as Python for smaller sizes.