[tree-ssa] Removal of gotos from cfg based ir

[Andrew Haley]

Jason Merrill writes:
 > On Fri, 14 Nov 2003 17:27:09 +0100, Jan Hubicka <hubicka@ucw.cz> wrote:
 > 
 > >> Surely we dont have a cfg edge from every single stmt that can trap when
 > >> any arbitary stmt in a  block can cause an exception? thats ugly. ewww
 > >> :-)
 > 
 > > That is what we do currently at RTL.  Better alternatives?
 > 
 > I believe Ada has rules something like an extrapolation of the C sequence
 > point rules to specify which side-effects must be done before a throw, in
 > order to give the compiler more latitude in reordering across a potential
 > trap like this.  Does anyone have the details handy?
 > 
 > AFAIK Java doesn't give the compiler this sort of latitude, so our hands
 > are tied.

The Java front end already knows quite accurately which exprs may
trap.  I'd like to attach this as an attribute and pass it to the
middle end.  This would perhaps be rather better than what we do at
present.

Andrew.