Libjava failures status

Tue Mar 12 03:20:00 GMT 2002

Jan Hubicka writes:
 > > On Sun, Mar 10, 2002 at 01:06:36PM +0100, Jan Hubicka wrote:
 > > > I plan to add symetrical check verifying that we do not discover
 > > > trapping instruciton in the middle of basic block...
 > > 
 > > If you do this, you'll need another representation for 
 > > instructions that, examined individually, may trap, but
 > > global information has determined that it cannot.
On a somewhat related subject:
I remember a while ago I wrote a patch to the Java front end that
marked MEMs as either potentially trapping or not, and as far as I
could tell this information was preserved throughout all optimization
stages. This was pretty easy to do, and eliminated about of half the
instructions from being treated as potentially trapping.
Andrew.