This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [ARM][PATCH, PR 68223] arm_[su]min_cmp pattern fails
- From: Ramana Radhakrishnan <ramana dot radhakrishnan at foss dot arm dot com>
- To: Michael Collison <michael dot collison at linaro dot org>, GCC Patches <gcc-patches at gcc dot gnu dot org>
- Date: Fri, 6 Nov 2015 07:49:39 +0000
- Subject: Re: [ARM][PATCH, PR 68223] arm_[su]min_cmp pattern fails
- Authentication-results: sourceware.org; auth=none
- References: <563C4CB3 dot 9050703 at linaro dot org>
On 06/11/15 06:46, Michael Collison wrote:
> The patterns arm_smin_cmp and arm_umin_cmp patterns fail if operand 0 and operand 2 are equal and both are less than operand 1. The solution is to remove the two patterns.
Agreed. There is no way this can be made to work without additional branches and that means the original pattern is probably more efficient.
>
> 2015-11-06 Michael Collison <michael.collison@linaro.org
> Ramana Radhakrishnan <ramana.radhakrishnan@linaro.org>
>
> PR target/68223
Instead just say .. (watching out formatting in the Changelog).
Revert:
2015-08-01 Michael Collison <michael.collison@linaro.org
Ramana Radhakrishnan <ramana.radhakrishnan@linaro.org>
* config/arm/arm.md (*arm_smin_cmp): New pattern.
(*arm_umin_cmp): Likewise
I would like to have a simple execution test added as a follow-up and not rely on a test failing in BE testing to show this failure up.
You do not need approval to revert your own patch unless you see that it's become enmeshed with other people's work - So, this is OK assuming no regressions.
regards
Ramana