Yeah, exactly. I feel like the author is missing the point of the optimizer. It's not something that attempts to fix bad code, it just expresses the programmers

Yeah, exactly. I feel like the author is missing the point of the optimizer. It's not something that attempts to fix bad code, it just expresses the programmers intent efficiently. For example, if I have a common subexpression, I would expect the compiler to notice it and take care of it automatically instead of me having to tediously code it explicitly to do the right thing. Given how large open source libraries are today, all of those little inefficiencies add up (in size and space) if we just let the compiler be stupid. But the easiest counterargument is to just to tell someone to compile their program at -O0 (which I've done inadvertently on an number of occasions). The program is noticeably bigger and slower.