[PATCH] D17652: [CGP] Duplicate addressing computation in cold paths if required to sink addressing mode

[Philip Reames via llvm-commits]

On 03/04/2016 02:36 PM, Philip Reames wrote:
>
>
> On 03/04/2016 02:31 PM, Mehdi Amini wrote:
>>> On Mar 4, 2016, at 2:29 PM, Philip Reames 
>>> <listmail at philipreames.com> wrote:
>>>
>>> reames added inline comments.
>>>
>>> ================
>>> Comment at: test/Transforms/CodeGenPrepare/X86/sink-addrmode.ll:10
>>> @@ +9,3 @@
>>> +; CHECK-LABEL: @test1
>>> +; CHECK: add i64 %sunkaddr, 40
>>> +entry:
>>> ----------------
>>> joker.eph wrote:
>>>> (I know you didn't create, but) having named value should be 
>>>> avoided for optimized build, we should fix CGP. Adding tests that 
>>>> relies on the named value is making it harder to fix this.
>>> Huh?  This is the first I've heard of this advice.  Named values are 
>>> widespread in the optimizer.
>> Any optimization that is using the IRBuilder won't introduce named 
>> value in optimized build.
> Please point to something in the style guide or reference where this 
> has been previously discussed?
>>
>>> If we want to change that, it's definitely a separate change.
>> I don't ask you to change CGP to not emit a named value. I am asking 
>> the test not to rely on them.
> And I'm going to ignore this request since it's irrelevant to the 
> change at hand and should be fixed in a single patch which changes the 
> code if desired.
Let me retract this statement.  My phrasing and the attitude it carried 
wasn't professional.  Sorry.  I'm going to disengage from this until 
Monday and come back at it when I'm in a better headspace.

Philip