[llvm] r280364 - [SimplifyCFG] Handle tail-sinking of more than 2 incoming branches

[James Molloy via llvm-commits]

Hi Joerg,

I’ll take a look. I hate this code - it’s produced so many more problems than it’s fixed; I really need to get around to polishing up GVNSink so it can be replaced.

Did you raise a bugzilla ticket for this? if not, could you? And to be clear, this is a functional regression rather than a performance regression?

Cheers,

James
> On 18 Feb 2017, at 02:33, Joerg Sonnenberger <joerg at bec.de> wrote:
>
> On Thu, Sep 01, 2016 at 12:58:14PM -0000, James Molloy via llvm-commits wrote:
>> Author: jamesm
>> Date: Thu Sep  1 07:58:13 2016
>> New Revision: 280364
>>
>> URL: http://llvm.org/viewvc/llvm-project?rev=280364&view=rev
>> Log:
>> [SimplifyCFG] Handle tail-sinking of more than 2 incoming branches
>
> Hello James,
> this change introduces a regression from 3.9->4.0 breaking Xen kernel builds.
> Test case is attached. The original code has an always_inline attribute,
> but that's not relevant here. The difference is that before, the default
> branch would have been eliminated and now it remains.
>
> Joerg
> <vpmu.c>

IMPORTANT NOTICE: The contents of this email and any attachments are confidential and may also be privileged. If you are not the intended recipient, please notify the sender immediately and do not disclose the contents to any other person, use it for any purpose, or store or copy the information in any medium. Thank you.