[PATCH] D61105: [MS] Emit S_HEAPALLOCSITE debug info in SelectionDAG

[Eli Friedman via Phabricator via llvm-commits]

efriedma added a comment.

> Is this SDNode map a good idea?

It's not ideal. I mean, if possible, it would be more clear to store the information as a property of the call node itself.  And it's generally not setting a good example to use a map with SDNode pointers rather than some sort of handle.  Could you use an MDNodeSDNode here as an operand to the CALL?  (You can create one with SelectionDAG::getMDNode().)

The address of an X86ISD::CALL is likely stable, but there isn't really any guarantee; it's just incidental due to the way call nodes are constructed.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D61105/new/

https://reviews.llvm.org/D61105