[PATCH] D122375: [CoroSplit] Handle argument being the frame pointer

[John McCall via Phabricator via llvm-commits]

rjmccall accepted this revision.
rjmccall added a comment.

LGTM



================
Comment at: llvm/lib/Transforms/Coroutines/CoroSplit.cpp:878
+    DummyArgs.push_back(
+        new BitCastInst(UndefValue::get(A.getType()), A.getType()));
+    VMap[&A] = DummyArgs.back();
----------------
nikic wrote:
> rjmccall wrote:
> > Is it really okay to add an unparented instruction to the value map?
> At least I'm not aware of any issues that would cause. Parenting it would be a bit awkward here, as the function is completely empty at this point (doesn't have any basic blocks to insert into).
Ah, yes, somehow I missed that this was in clone-specific code.  Alright, if it's not causing problems, I guess we can do this.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D122375/new/

https://reviews.llvm.org/D122375