It's not language agnostic in the sense that some languages don't have the necessary functionality to allocate everything up front and avoid allocations in the hot loop.
For example, if you were to benchmark this in Java, the HashMap<> class allocates (twice!) on every insertion. Allocations are a bit cheaper with the GC than they would be via malloc/friends, but still we'd expect to see significant allocator and hence GC overhead in this benchmark
For example, if you were to benchmark this in Java, the HashMap<> class allocates (twice!) on every insertion. Allocations are a bit cheaper with the GC than they would be via malloc/friends, but still we'd expect to see significant allocator and hence GC overhead in this benchmark