[PATCH] Make sure branch edge weights are not dropped in TailDuplicatePass

[Manman Ren]

Sorry I was thinking about static MBPI. Yes, using branch weight metadata
means we can change it. Your original number looks good then :]

LGTM.

Thanks,
Manman



On Tue, Feb 11, 2014 at 12:26 PM, Akira Hatanaka <ahatanak at gmail.com> wrote:

> Yes, I changed both the weights in the metadata and the string that checks
> the weights in the machine IR dump.
>
> Did that answer your question?
>
>
> On Tue, Feb 11, 2014 at 12:20 PM, Manman Ren <manman.ren at gmail.com> wrote:
>
>>
>>
>>
>> On Tue, Feb 11, 2014 at 12:15 PM, Akira Hatanaka <ahatanak at gmail.com>wrote:
>>
>>> Updated patch attached. I added FileCheck to the command line and
>>> changed the weight to something that looks more normal.
>>>
>>
>> Are the weights fixed with the testing case? I don't get why you can just
>> change the weights to something.
>> LGTM otherwise,
>>
>> Manman
>>
>>
>>>
>>>
>>> Please review. Thank you.
>>>
>>>
>>> On Tue, Feb 11, 2014 at 11:20 AM, Manman Ren <manman.ren at gmail.com>wrote:
>>>
>>>>
>>>> The testing case seems to miss FileCheck.
>>>> +; RUN: llc -march=arm -print-machineinstrs=tailduplication
>>>> -tail-dup-size=100 -enable-tail-merge=false -disable-cgp < %s
>>>>
>>>> +; CHECK: Successors according to CFG: BB#1(1234) BB#2(5678)
>>>> +; CHECK: Successors according to CFG: BB#1(4321) BB#2(8765)
>>>>
>>>> The above numbers look weird 1234 5678 :]
>>>>
>>>> Thanks for working on this.
>>>> Manman
>>>>
>>>>
>>>>
>>>> On Tue, Feb 11, 2014 at 10:36 AM, Akira Hatanaka <ahatanak at gmail.com>wrote:
>>>>
>>>>> There are two lines in TailDuplication.cpp where edge weights are not
>>>>> passed to MachineBasicBlock::addSuccessor, which results in the weights
>>>>> being dropped from MachineBasicBlock. The attached patch fixes this bug.
>>>>>
>>>>> _______________________________________________
>>>>> llvm-commits mailing list
>>>>> llvm-commits at cs.uiuc.edu
>>>>> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>>>>>
>>>>>
>>>>
>>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20140211/10de8a9a/attachment.html>