[llvm] [MachinePipeliner] Make Recurrence MII More Accurate (PR #105475)

[Ryotaro Kasuga via llvm-commits]

================
@@ -1680,13 +1680,19 @@ void SwingSchedulerDAG::Circuits::createAdjacencyStructure(
         Added.set(N);
       }
     }
-    // A chain edge between a store and a load is treated as a back-edge in the
-    // adjacency matrix.
+    // Order edges of the following:
+    // 1. Load -> Store
+    // 2. Store -> Load
+    // are treated as a back-edge in the adjacency matrix.
+    // Store after store was handled above.
     for (auto &PI : SUnits[i].Preds) {
-      if (!SUnits[i].getInstr()->mayStore() ||
+      if (PI.getKind() != SDep::Order ||
           !DAG->isLoopCarriedDep(&SUnits[i], PI, false))
         continue;
-      if (PI.getKind() == SDep::Order && PI.getSUnit()->getInstr()->mayLoad()) {
+      if ((SUnits[i].getInstr()->mayLoad() &&
+           PI.getSUnit()->getInstr()->mayStore()) ||
+          (SUnits[i].getInstr()->mayStore() &&
+           PI.getSUnit()->getInstr()->mayLoad())) {
----------------
kasuga-fj wrote:

Oh sorry, I made a mistake in the previous comment https://github.com/llvm/llvm-project/pull/105475#discussion_r1728130892.  I meant to say store-to-store, not store-to-load.  My intention was to remove conditions based on the type of instruction.  In this case, I think no conditional branching is needed here.

I'm not sure what your comment "Store after store was handled above" means.  If you mean output-dependencies, they are not relevant: Output-dependence is a write-after-write dependence for physical registers. 

Also, could you please fix the following comment?
https://github.com/llvm/llvm-project/blob/17726a8bdbb952b30957aa462b9b18d0c724464c/llvm/include/llvm/CodeGen/MachinePipeliner.h#L376

https://github.com/llvm/llvm-project/pull/105475