[llvm] Introducing a new ISD::POISON SDNode to represent the poison value in the IR. (PR #125883)

[zhijian lin via llvm-commits]

================
@@ -690,8 +690,17 @@ END_TWO_BYTE_PACK()
   /// \<target\>ISD namespace).
   bool isTargetOpcode() const { return NodeType >= ISD::BUILTIN_OP_END; }
 
-  /// Return true if the type of the node type undefined.
-  bool isUndef() const { return NodeType == ISD::UNDEF; }
+  /// Returns true if the node type is UNDEF or, when DoNotIncludeExplicitPoison
+  /// is false, POISON.
+  /// @param DoNotIncludeExplicitPoison Determines whether to check only for
+  /// UNDEF.
+  bool isUndef(bool DoNotIncludeExplicitPoison = false) const {
+    return NodeType == ISD::UNDEF ||
+           (!DoNotIncludeExplicitPoison && NodeType == ISD::POISON);
----------------
diggerlin wrote:

If I understand your comment correct. 
I think we have a discussion on the https://github.com/llvm/llvm-project/pull/125883#discussion_r1951504462

`Don't we usually want to keep treating poison as undef to avoid updating a bunch of places? In IR, PoisonValue inherits from UndefValue so isa<UndefValue> returns true for poison or undef. I believe SelectionDAGBuilder was previously creating ISD::UNDEF for poison. So continuing to treat ISD::POISON like ISD::UNDEF in many cases would be consistent.`


https://github.com/llvm/llvm-project/pull/125883