[Mono-list] Quick question reguarding local allocation
Miguel de Icaza
25 Feb 2002 20:21:05 -0500
> No I think we had the same thing in mind, what I forgot to say at the bottom of that last post was, if the mcs compiler does squeeze the locals into as few as possible, then a compiler that takes the IL produced as input then has the knowledge that all the locals must interfere with each other, ie it doesn't have to do liveness analysis on the locals since it's already been done. Of course this is a bit naughty anyway since such a compiler should really accept IL produced from any source_language->IL compiler so it should do it's own liveness analysis anyway.
Ok. I know we are reusing some variables sometimes, and the mechanism
is in place, it is just not as wildly deployed as it should be.
So I would not assume that a slot is not reused ;-)