[llvm] [MachinePipeliner] Add an abstract layer to manipulate Data Dependenc… (PR #109918)

[Michael Marjieh via llvm-commits]

================
@@ -114,10 +115,123 @@ class MachinePipeliner : public MachineFunctionPass {
   bool useWindowScheduler(bool Changed);
 };
 
+/// Represents a dependence between two instruction.
+class SwingSchedulerDDGEdge {
+  SUnit *Dst = nullptr;
+  SDep Pred;
+  unsigned Distance = 0;
+
+public:
+  /// Creates an edge corresponding to an edge represented by \p PredOrSucc and
+  /// \p Dep in the original DAG. This pair has no information about the
+  /// direction of the edge, so we need to pass an additional argument \p
+  /// IsSucc.
+  SwingSchedulerDDGEdge(SUnit *PredOrSucc, const SDep &Dep, bool IsSucc)
+      : Dst(PredOrSucc), Pred(Dep), Distance(0u) {
+    SUnit *Src = Dep.getSUnit();
+
+    if (IsSucc) {
+      std::swap(Src, Dst);
+      Pred.setSUnit(Src);
+    }
+
+    // An anti-dependence to PHI means loop-carried dependence.
+    if (Pred.getKind() == SDep::Anti && Src->getInstr()->isPHI()) {
+      Distance = 1;
+      std::swap(Src, Dst);
+      auto Reg = Pred.getReg();
+      Pred = SDep(Src, SDep::Kind::Data, Reg);
----------------
mmarjieh wrote:

I found the issue I think.
I had a bug (compiler crash on assertion) on an offline target (not in LLVM community).
I investigated and got to a conclusion that this is the issue.

You changed the phi anti-dep handling.
So when originally having,

some_instruction -> phi; anti dep

You have in your new DAG:
some_instruction -> phi, data dep

In the original consideration, some_instruction is a successor of the phi with anit-dep.
and it was missed out in succ_l.
See my fix, does it make sense to you?
![image](https://github.com/user-attachments/assets/e55ce3c3-9c3d-444b-98ea-0af859d67454)


https://github.com/llvm/llvm-project/pull/109918