[llvm] [BOLT] Gadget scanner: reformulate the state for data-flow analysis (PR #131898)

[Anatoly Trosinenko via llvm-commits]

================
@@ -169,16 +186,26 @@ struct State {
   std::vector<SmallPtrSet<const MCInst *, 4>> LastInstWritingReg;
   State() {}
   State(unsigned NumRegs, unsigned NumRegsToTrack)
-      : NonAutClobRegs(NumRegs), LastInstWritingReg(NumRegsToTrack) {}
-  State &operator|=(const State &StateIn) {
-    NonAutClobRegs |= StateIn.NonAutClobRegs;
+      : SafeToDerefRegs(NumRegs), LastInstWritingReg(NumRegsToTrack) {}
+
+  /// Returns S, so that S.merge(S1) == S1.merge(S) == S1.
+  static State getMergeNeutralElement(unsigned NumRegs,
----------------
atrosinenko wrote:

My idea was that this static function returns a neutral element w.r.t. `merge` operation (in an algebraic sense), this is the reason for keeping it right in the `State` class, as opposed to `getStartingStateAt*` methods of `PacRetAnalysis`, as this function should be aligned with `merge`. Maybe this is a bit "overengineered" name for a function returning an initial state, but after all, "starting state" returned by `PacRetAnalysis::getStartingStateAt*` is "initial" as well, but for a different purpose.

On the other hand, this function definitely deserves an explanation that this "initial" state is initial in the same sense as 0 being an initial value for accumulator when computing a sum of a set of integers. I will add a comment explaining what is the expected usage of this function, thank you for pointing this out!

https://github.com/llvm/llvm-project/pull/131898