java aliasing rules

Thu Apr 4 12:57:00 GMT 2002

Jeff Sturm writes:
 > On 2002年3月31日, Bryce McKinlay wrote:
 > > >>lhz r9, [ps1.f1]
 > > >>addi r9,r9,1
 > > >>sth r9, [ps1.f1]
 > > >>
 > > >
 > > >Wouldn't accessing r9 immediately after the load cause a pipeline stall?
 > > >Assuming an in-order processor? That would be a significant performance
 > > >hit.
 > > 
 > > Right, its not optimal scheduling, but there's no way to avoid that and 
 > > still have the correct behaviour for NullPointers. And as you suggest, a 
 > > modern processor may be speculativly executing the following loads, so 
 > > it probibly doesn't matter too much.
If it doesn't annul such speculation on SEGV I don't think it's
correct Java semantics.
 > Am I correct in thinking this is only an issue for -fnon-call-exceptions?
Yes.
 > It might be useful to turn this "correctness" off with a compiler option,
 > as we do with -fno-bounds-check. I habitually check for null in my code,
 > and don't do anything useful with a NullPointerException besides aborting.
 > I suspect that's true of a great deal of Java code.
I know what you mean, but turning this off is a pretty major departure
from the Java language spec.
Andrew.