I think the author misses the point: you can't just look at code and expect to understand performance, in any language. That's what profilers are for.
I will agree that micro-optimization of known hotspots is mechanical, but finding the right areas can be deliciously difficult.
How can the author be "missing" his own point? This isn't a reply to anything.