This PR replaces our old method of reducing the benchmark results by using an array to using atomics instead. This should help us implement single threaded benchmarks.
This PR replaces our old method of reducing the benchmark results by using an array to using atomics instead. This should help us implement single threaded benchmarks.