Dubious lookup benchmark for unordered sets

[cookiestarfish]

The benchmark report shows ~20ns average lookup time for std::unordered_set, which is less than the time taken by a cache miss, therefore unrealistic for average sized sets. The benchmark uses sets of 100k elements, a std::unordered_set of 100k int32 barely fits in L3 cache and the benchmark makes it look too good in comparison with amc::FlatSet.
I suggest to use bigger set sizes, or repeat the benchmark for different set sizes.

[sjanel]

Hello, thanks for your input!

Actually the comparison with std::unordered_set is supposed to be unfair against FlatSet for lookup operations as it has constant lookup complexity in average. But it's true that benchmarks could be provided for different set sizes. I will add it in my todo list!