This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [fix] PR39300: make PRE not disturb vectorizer
On Wed, Jul 22, 2009 at 8:34 AM, Michael Matz<matz@suse.de> wrote:
> Hi,
>
> On Tue, 21 Jul 2009, Richard Guenther wrote:
>
>> IMHO the performance numbers speak for themselves. ?Can you re-check
>> with the above change (and also throw it onto the C++ tester once) and
>> add a vectorization testcase as HJ suggests?
>
> I did the requested changes, regstrapped without regression. ?The c++
> tester and spec2000 are unimpressed by the patch, except applu, which
> regressed on the testrun, but I can't reproduce any regression, so it
> hopefully was just a hickup on the machine.
>
>> Ok with that changes.
>
> Checked in a r149942. ?For reference the patch is below. ?(I included
> -ftree-pre explicitely in the dg-options in case anybody would like to
> "fix" a problem by removing PRE from -O3 ).
>
>
> Ciao,
> Michael.
> --
> ? ? ? ?PR tree-optimization/35229
> ? ? ? ?PR tree-optimization/39300
>
> ? ? ? ?* tree-ssa-pre.c (includes): Include tree-scalar-evolution.h.
> ? ? ? ?(inhibit_phi_insertion): New function.
> ? ? ? ?(insert_into_preds_of_block): Call it for REFERENCEs.
> ? ? ? ?(init_pre): Initialize and finalize scalar evolutions.
> ? ? ? ?* Makefile.in (tree-ssa-pre.o): Depend on tree-scalar-evolution.h .
>
> testsuite/
> ? ? ? ?* gcc.dg/vect/vect-pre-interact.c: New test.
The new testcase fails on Linux/ia32:
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40830
H.J.