This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH, rs6000] Switch the rs6000 port over to LRA
- From: David Edelsohn <dje dot gcc at gmail dot com>
- To: Peter Bergner <bergner at vnet dot ibm dot com>
- Cc: GCC Patches <gcc-patches at gcc dot gnu dot org>, Segher Boessenkool <segher at kernel dot crashing dot org>, Michael Meissner <meissner at linux dot vnet dot ibm dot com>, Bill Schmidt <wschmidt at linux dot vnet dot ibm dot com>
- Date: Wed, 3 Aug 2016 19:03:57 -0400
- Subject: Re: [PATCH, rs6000] Switch the rs6000 port over to LRA
- Authentication-results: sourceware.org; auth=none
- References: <d4245e94-6a1e-93ad-0d13-8eb016f04029@bergner.org> <e4752d7a-341b-ace3-8366-0c30558923f1@vnet.ibm.com>
On Wed, Aug 3, 2016 at 6:59 PM, Peter Bergner <bergner@vnet.ibm.com> wrote:
> On 8/2/16 3:17 PM, Peter Bergner wrote:
>>
>> Now that Vlad has fixed PR69847, which was the last problem holding the
>> rs6000 port from switching from reload to LRA, we are ready to flip the
>> switch.
>>
>> Is the following ok once bootstrap/regtesting on both LE and BE
>> (32 & 64 regtesting) comes out clean?
>
>
> So we have two "regressions":
>
> +FAIL: gcc.target/powerpc/bool3-p7.c scan-assembler-not [ \\t]xxlnor
> +FAIL: gcc.target/powerpc/bool3-p8.c scan-assembler-not [ \\t]xxlnor
>
>
> Looking into these "failures", they show up because when we enable
> LRA, we also implicitly enable -mvsx-timode and these failures are
> due to -mvsx-timode. The same test cases fail when we use -mvsx-timode
> with reload.
>
> I'll note that these failures are not code correctness bugs, but
> performance bugs. I plan to open a bugzilla to track the fixing
> of these failures.
>
> My question, is since these failures are not due to LRA, do we
> want to consider the switch to LRA ok to commit or do we want to
> wait until the -mvsx-timode performance bug is fixed?
Peter,
Please go ahead and commit the LRA switch patch.
Please open a Bugzilla for the rs6000 backend about the vsx-timode
performance regression. The vsx-timode regression needs to be fixed
for GCC 7.
Thanks, David