Richards spends a fair amount of time in not-equal comparisons, much of it the immediate case, which could be inlined.
Created attachment 23773 [details]
Comment on attachment 23773 [details]
It seems really unfortunate that you don't inline all immediate cases :-/
I could probably inline more of them, but there are various weird special cases, such as true == 1, false == 0. We have a special opcode for comparing to null, so booleans and undefined are the only immediates missing fast cases.