Are several memory copies and context switches free for you? They are most expensive operations on modern hardware (yes, I know about splice syscall in Linux, but it is Linux-specific and don't avoid context switches). I think, on very high-end modern storage subsystem (like, stripe built out of several enterprise-grade NMVe SSDs) impact will be well-measurable in case of simple "wc".

I can imagine if you are running 70s vintage DEC hardware, there might be appreciable performance benefits to not spawning that extra process.