This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Fix PR32540 exponential time behavior in PRE
On 11/4/07, Richard Guenther <richard.guenther@gmail.com> wrote:
> On 11/4/07, Sebastian Pop <sebpop@gmail.com> wrote:
> > Hi,
> >
> > The attached patch fixes the PR32540 and PR33922. The patch is based
> > on the patch that Nick Clifton sent earlier, but adapted to stop the
> > computation before the phi translation. I also changed the default
> > cutoff to 100 anticipable values, but I'm not sure about the best
> > value here. Daniel what is your opinion on this cutoff value?
>
> I think 100 is reasonable. Can you change the pr32540-1.c testcase
> to omit the 0x1 << 0 case? Otherwise due to phiopt this one is not
> different from pr32540-2.c.
>
> > The patch regstrapped on amd64-linux. Okay for trunk?
>
> Let Danny some days to comment, otherwise this is ok.
100 is fine as long as we are satisfied these are not problems with
value numbering deciding two values are different when they aren't.
(I haven't looked back at these bugs in a while).