On Fri, 2 Aug 2002, Melvin Smith wrote: > At 08:50 AM 8/2/2002 -0700, Sean O'Rourke wrote: > >Without performance numbers, this is hard to test, but it can potentially > >turn a single "a = b + c", which is just "add P0, P1, P2" if a, b, and c > >have been referenced, into a hideous five instructions: > > > > fetch_lex P0, 'a' # Because how we store result depends on a's type > > fetch_lex P1, 'b' > > fetch_lex P2, 'c' > > add P0, P1, P2 > > store_lex P0, 'a' > > The last instruction (store_lex) isn't needed since P0 is a pointer to the > lexical.
You are correct, sir. Though interestingly enough, you do need a store_lex (and not a fetch_lex) for "a" in "a = b", at least until we have an assign op. /s
