case expressions/transfers

Henry Cejtin henry@sourcelight.com
Wed, 9 Jan 2002 17:32:42 -0600


Yes, just as you say: the stupid range check is there even though it can
never possibly be required.  In the byte-code interpreter case it turns out
that just that extra overhead is noticeble, especially if your byte-code
instructions are very light-weight (think stack machine for example).  This
despite the fact that the branch prediction is going to be perfect.