Good to know about the first and last part. And as for the second part... Meh. I don't see any optimizations that hard-coding calling something named "memcpy"

Good to know about the first and last part. And as for the second part... Meh. I don't see any optimizations that hard-coding calling something named "memcpy" (or whatever) does that cannot be enabled by looking at the actual code that gets linked. Albeit with more difficulty.