[PATCH] D17781: Codegen: [PPC] Word Rotates are Zero Extending

[Eric Christopher via llvm-commits]

Also why testcases are awesome :)

-eric

On Tue, Mar 1, 2016 at 2:31 PM Eric Christopher <echristo at gmail.com> wrote:

> You're right. We came up with the usual case, but not the normal case.
> Kyle is going to add a check for the arguments as well :)
>
> -eric
>
> On Tue, Mar 1, 2016 at 2:30 PM Hal Finkel <hfinkel at anl.gov> wrote:
>
>> hfinkel added a subscriber: hfinkel.
>> hfinkel requested changes to this revision.
>> hfinkel added a reviewer: hfinkel.
>> hfinkel added a comment.
>> This revision now requires changes to proceed.
>>
>> I don't think this is right. RLWINM starts by taking two copies of the
>> 32-bit amount, one copy in the lower word and one copy in the upper word
>> and rotates the result, Then it composes the mask. The mask is non-zero in
>> the upper bits if MB >= ME.
>>
>>
>> Repository:
>>   rL LLVM
>>
>> http://reviews.llvm.org/D17781
>>
>>
>>
>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20160301/39b7fc3a/attachment.html>