new SSA IL

Jagannathan, Suresh
Tue, 7 Aug 2001 23:57:09 -0400

It's interesting that you guys are reconsidering SSA.  I remember we
had a discussion on this a while ago with Richard.  My recollection
at the time was that we concluded there'd be little difference in
the optimizations that would result.  Given the fairly natural translation
between CPS and SSA, I'm surprised you think there'd be a lot to gain
(except for handling exceptions which I agree with).  Do you have a sense
(based on the size of jump bodies currently) how much would be gained by a
more accurate dominator graph?  In particular, do you have good (i.e.,
realistic) examples in mind where the liveness properties of variables
defined in the current IL would be significantly different than that
computed by SSA?

On the sourceforge issue -- I think it's a good idea.  Whatever gets
decided, I think it's important to *not* make MLton@sourcelight become
public since there's too much traffic dedicated to implementation that's not
of general interest.  Any other mailing list or web-based distribution forum
is fine with me.

-- Suresh