[PATCH] D14361: Let SelectionDAG start to use probability-based interface to add successors.

[Xinliang David Li via llvm-commits]

On Fri, Nov 20, 2015 at 4:54 PM, Cong Hou <congh at google.com> wrote:

> congh added inline comments.
>
> ================
> Comment at: lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp:1312
> @@ -1311,3 +1311,3 @@
>      UnwindDest.first->setIsEHPad();
> -    addSuccessorWithWeight(FuncInfo.MBB, UnwindDest.first,
> UnwindDest.second);
> +    addSuccessorWithProb(FuncInfo.MBB, UnwindDest.first,
> UnwindDest.second);
>    }
> ----------------
> davidxl wrote:
> > Let's add a TODO here for a follow up. In follow up, find a way to
> assert the violations and examine the root cause
> OK. But why here? It is a specific place and the follow up is fixing a
> more general problem. After the whole patch is checked in, I will work on
> how to validate the successors' probabilities.
>

SGTM.


>
> It this patch ready? I will update the test cases.
>
>
Please update test cases and we can take a look.

David


>
> http://reviews.llvm.org/D14361
>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20151120/b28f7631/attachment.html>