Clarification questions on design

[goretkin]

I am trying to read into the comment: https://github.com/JuliaCI/BenchmarkTools.jl/blob/45052b1953029a0bfd0829ab735dd055d01d9676/src/execution.jl#L268-L270

Is this to allow tune!ing benchmarks as a separate step from running them?

At the risk of being further off-topic, I'm trying to understand what is happening with allocations here: https://github.com/KristofferC/NearestNeighbors.jl/blob/73eeb606efb504cd0638fd3f4eb5dd2cd025296d/benchmarks/runbenchmarks.jl#L24

If I understand correctly, for each loop iteration, there is a new reference to a different tree "instances", and there is memory allocated for each of those. If that's the case, then it seems it might be beneficial to interleave this with running the benchmarks. But I'm not sure how to organize that.