Performance of Dictionaries.jl vs Base.Dict

Internals & Design
, , ,
25

I apologize for bringing up these old messages, but for me, and perhaps for others too, they are of current interest.
I tried to use the instructions of the update function!() but I get no improvements.
Maybe I apply it wrong

julia> function solve(stones, n)
           d = Dict{Int64,Int64}()
           for s ∈ stones
               d[s] = 1
           end
           for _ in 1:n
               d_n = Dict{Int64,Int64}()
               for (s, n) ∈ d
                   for s2 ∈ blink(s)
                       s2 == NO_STONE && continue
                       index = Base.ht_keyindex2!(d_n, s2)
           if index > 0
               @inbounds d_n.vals[index] += n # = func(dict.vals[index])
           else
               @inbounds Base._setindex!(d_n, n, s2, -index)
           end
                       # if haskey(d_n, s2)
                       #     d_n[s2] += n
                       # else
                       #     d_n[s2] = n
                       # end
                   end
               end
               d = d_n
           end
           sum(values(d))
       end
solve (generic function with 1 method)

julia> solve(stones, 25)
197357

julia> solve(stones, 75)
234568186890978

julia> @btime solve(stones, 25)  # 260 us
  635.100 μs (9800 allocations: 496.95 KiB)
197357

julia> @btime solve(stones, 75)  # 12ms
  29.340 ms (590514 allocations: 23.42 MiB)
234568186890978