This is better because you don't need to allocate ALL the benchmarks
before benchmarking them!