This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: alias time explosion
On Tue, 2006-03-21 at 22:16 -0500, Andrew MacLeod wrote:
> On Tue, 2006-03-21 at 17:42 -0500, Daniel Berlin wrote:
> > On Tue, 2006-03-21 at 17:30 -0500, Andrew MacLeod wrote:
> > > On T
> > > I seem to have narrowed it down to this patch:
> > >
> > > http://gcc.gnu.org/ml/gcc-patches/2006-01/msg00908.html
> > >
> >
> > That's quite a while ago :).
> >
>
> It was, I am suprised no one took note of it befor enow. I was suprised
> I had to go back that far to find it.
This is probably a pathological testcase for something.
> >
> > i will note the patch is pretty much required for correctness. We were
> > getting seriously wrong answers before in some cases.
> >
>
> yeah, well wrong pessimistic, or wrong incorrect code. If its the
> latter, there must be a better way to do it.
Well, both, actually.
>
> hopefully its just some check ordering issue thats easily resolved. Its
> seems like an insane amount of time.
Looking at the testcase, it's probably the removal of is_global_var from
is_call_clobbered, or something close that used to not check the bitmap
and now does. This would actually generate wrong code in some cases
where we got out of whack between the real call clobbered bitmap and
what is_call_clobbered would return.
I will check if this is the case, and if so, it is trivial to fix.
--Dan