This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [testcase] optimization/5076: gcc3 regression: yet another strength reduction bug
On Thursday 03 January 2002 22:18, Richard Henderson wrote:
> On Thu, Jan 03, 2002 at 09:47:02PM +0100, Franz Sirl wrote:
> > I'm suggesting the attached bandaid fix for 3.1 and gcc-3_0-branch
>
> [...]
>
> > * unroll.c (unroll_loop): Don't try to unroll loops with a
> > NOTE_INSN_LOOP_CONT.
> > (ujump_to_loop_cont): Delete.
>
> All for loops have a NOTE_INSN_LOOP_CONT. We can do better than this.
All for loops? That's unfortunate, but it would only affect loops with 1
iteration anyway. Do you think the better method I sketched will work out? My
problem here is I'm quite good in analyzing bugs in the loop code in the
meantime, but I usually lack the bigger picture, so it's quite hard for me to
do a correct and full-featured fix.
Franz.