[llvm] r268539 - [AArch64] Use the reciprocal estimation machinery

[Evandro Menezes via llvm-commits]

Hi, Renato.

Indeed, combining some operations into a series could be improved by a 
cost model and evaluation of the BB context, since the series adds 
multiplications which could end up regressing the performance of an 
already multiplication intensive context.

But improvements to existing implementations should not imply their 
reversal.

Thank you,

-- 
Evandro Menezes

On 09/20/16 05:11, Renato Golin wrote:
> On 20 September 2016 at 01:49, Eric Christopher via llvm-commits
> <llvm-commits at lists.llvm.org> wrote:
>> I'm going to ask you to revert this patch please. There are a few reasons:
> I agree.
>
>
>> a) I don't see a review thread here? This changes the port quite a bit.
> There were some discussions around this topic, and from what
> everyone's saying, it's a complex topic.
>
> https://llvm.org/bugs/show_bug.cgi?id=27107
> http://lists.llvm.org/pipermail/llvm-dev/2016-September/104676.html
>
> We can't just decide this based on flags (*especially not* isExynosM1()!).
>
> Please revert this and join the discussion upstream.
>
> cheers,
> --renato
>
>
>