This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: PR c/8068
> On Fri, Feb 14, 2003 at 11:36:04PM +0100, Arend Bayer wrote:
> > Two general questions on const-fold.c:
> > 1. I was very much surprised to see this happen at -O0. Shouldn't
> > extract_muldiv(), or at least it's less obvious simplications such as
> > (op + const1) * const2 -> op*const2 + const1*const2 etc., only be executed at
> > -O1 and higher?
>
> *shrug* We've always done it this way.
>
> > 2. I find it very hard to verify that similar exponential behaviour
> > will not get triggered with other testcases. Maybe extract_muldiv() should
> > have a bound how many recursive calls it tries while simplifying one
> > expression?
>
> Indeed, based on the fact that this can happen with / as well,
> this does seem the best solution. I'm testing this.
Hi,
the patch you commited into 3.3 branch causes GCC to misscopile gnatmake
so make gnatlib_and_tools fails on x86-64. Any idea on this?
Honza