[PATCH] D65593: [Attributor] Deduction based on path exploration

[Hideto Ueno via Phabricator via llvm-commits]

uenoku marked an inline comment as done.
uenoku added a comment.

> I have not convinced myself that this covers the recursive case properly, maybe that was never the intend though. Can we have a test like this:

Yes, recursive branches are not handled in this patch to simplify. I'll add on a later patch.



================
Comment at: llvm/lib/Transforms/IPO/Attributor.cpp:1032
+    // ParentState_m = ChildState_{m, 1} /\ ChildState_{2, 2} /\ ... /\
+    // ChildState_{1, n_m}
+    //
----------------
jdoerfert wrote:
> I don't get the comment. What are the two indices of ChildState and why is the second ChildState "out-of-sync"?
It was broken. Sorry for the confusion.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D65593/new/

https://reviews.llvm.org/D65593