This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH][RFC] Fix P1 PR77498
Hi Richard,
On 30 March 2017 at 09:13, Richard Biener <rguenther@suse.de> wrote:
> On Wed, 29 Mar 2017, Jeff Law wrote:
>
>> On 03/29/2017 04:05 AM, Richard Biener wrote:
>> >
>> > After quite some pondering over this and other related bugs I propose
>> > the following for GCC 7 which tames down PRE a bit (back to levels
>> > of GCC 6). Technically it's the wrong place to fix this, we do
>> > have measures in place during elimination but they are not in effect
>> > at -O2. For GCC 8 I'd like to be more aggressive there but that
>> > would require to enable predictive commoning at -O2 (with some
>> > limits to its unrolling) to not lose optimization opportunities.
>> >
>> > The other option is to ignore this issue and postpone the solution
>> > to GCC 8.
>> >
>> > Bootstrapped / tested on x86_64-unknown-linux-gnu.
>> >
>> > Any preference?
>> >
>> > Thanks,
>> > Richard.
>> >
>> > 2017-03-29 Richard Biener <rguenther@suse.de>
>> >
>> > PR tree-optimization/77498
>> > * tree-ssa-pre.c (phi_translate_1): Do not allow simplifications
>> > to non-constants over backedges.
>> >
>> > * gfortran.dg/pr77498.f: New testcase.
>> I've got a slight preference for this patch.
>>
>> If you had a good start on the real fix then I'd lean more towards postponing.
>
> I wouldn't it yet call "real fix" but just an idea (where I tested the
> PRE side already, with the expected testsuite regressions). I've done
> no benchmarking at all for that. For the proposed patch the situation
> is that we're only going to remove some PRE that was done additionally
> over GCC 6 because of the rev. that caused the regression, so I have
> confidence that it won't make things worse when comparing to GCC 6.
>
> Thus I've now applied the patch.
>
With this patch, the following testcase now fails on arm* targets:
gcc.dg/tree-ssa/pr71347.c scan-tree-dump-not optimized ".* = MEM.*;"
I'll have to rebuild manually if you need the dumps, let me know.
Thanks,
Christophe
> Richard.