"No benchmarks or any other numbers."

And yet plenty of logical argument. The author assumes a level of optimization experience on the part of the reader, for example that the reader appreciates the physical cost of branch mispredictions.