Trying to understand allocations while creating vectors

On my laptop it appears that creating a Vector{Float64} with >2048 elements triggers an extra allocation

julia> @btime Vector{Float64}(undef, 2048);
  379.093 ns (1 allocation: 16.13 KiB)

julia> @btime Vector{Float64}(undef, 2049);
  454.149 ns (2 allocations: 16.14 KiB)

julia> sizeof(Float64)*2048

Something similar is seen with eltype ComplexF64

julia> @btime Vector{ComplexF64}(undef, 1024);
  403.270 ns (1 allocation: 16.13 KiB)

julia> @btime Vector{ComplexF64}(undef, 1025);
  478.369 ns (2 allocations: 16.14 KiB)

I’m not familiar with how allocations occur, could someone please explain why this extra allocation takes place once we overshoot a certain memory limit (16.13 KiB in this case)?

It’s just an internal implementation detail.

1 Like