> You need to benchmark, test and tweak in order to figure out what works best. Benchmarking is particularly important because, unless you're dealing with large data or very hot code, there's a good chance that effort won't yield measurable benefits.

How to benchmark something like this sounds like it could become a pretty good article in its own right.

Usually, you can run something like perf or callgrind, with instruction level profiling, and you will get a good idea. Its not benchmarking in the traditional sense, but it has a similar result.

SIMD can make your code faster, or slower - profiling is a great way to tell exactly how much faster or slower.