[llvm] [LoopUnroll] Rotate loop before unrolling inside of UnrollRuntimeLoopRemainder (PR #148243)

[Marek Sedláček via llvm-commits]

================
@@ -587,21 +536,116 @@ llvm::UnrollLoop(Loop *L, UnrollLoopOptions ULO, LoopInfo *LI,
       UnrollRuntimeEpilog.getNumOccurrences() ? UnrollRuntimeEpilog
                                               : isEpilogProfitable(L);
 
-  if (ULO.Runtime &&
-      !UnrollRuntimeLoopRemainder(L, ULO.Count, ULO.AllowExpensiveTripCount,
-                                  EpilogProfitability, ULO.UnrollRemainder,
-                                  ULO.ForgetAllSCEV, LI, SE, DT, AC, TTI,
-                                  PreserveLCSSA, ULO.SCEVExpansionBudget,
-                                  ULO.RuntimeUnrollMultiExit, RemainderLoop)) {
+  bool LoopRotated = false;
+  bool ReminderUnrolled = false;
+  if (ULO.Runtime) {
+    // Call unroll with disabled rotation, to see if it is possible without it.
+    ReminderUnrolled = UnrollRuntimeLoopRemainder(
+        L, ULO.Count, ULO.AllowExpensiveTripCount, EpilogProfitability,
+        ULO.UnrollRemainder, ULO.ForgetAllSCEV, LI, SE, DT, AC, TTI,
+        PreserveLCSSA, ULO.SCEVExpansionBudget, ULO.RuntimeUnrollMultiExit,
+        RemainderLoop);
+
+    // If unroll is not possible, then try with loop rotation.
+    if (!ReminderUnrolled) {
+      BasicBlock *OrigHeader = L->getHeader();
+      BranchInst *BI = dyn_cast<BranchInst>(OrigHeader->getTerminator());
+      if (BI && !BI->isUnconditional() &&
+          isa<SCEVCouldNotCompute>(SE->getExitCount(L, L->getLoopLatch())) &&
+          !isa<SCEVCouldNotCompute>(SE->getExitCount(L, OrigHeader))) {
+        LLVM_DEBUG(
+            dbgs() << "  Rotating loop to make the exit count computable.\n");
+        SimplifyQuery SQ{OrigHeader->getDataLayout()};
+        SQ.TLI = nullptr;
+        SQ.DT = DT;
+        SQ.AC = AC;
+        LoopRotated =
+            llvm::LoopRotation(L, LI, TTI, AC, DT, SE,
+                               /*MemorySSAUpdater*/ nullptr, SQ,
+                               /*RotationOnly*/ false, /*Threshold*/ 16,
+                               /*IsUtilMode*/ false, /*PrepareForLTO*/ false,
+                               [](Loop *, ScalarEvolution *) { return true; });
----------------
mark-sed wrote:

UnrollRuntimeLoopReminder could do it, but it will still need to do all the checks it does after it does the rotation, this means UnrollRuntimeLoopReminder would have to take a flag if rotation is allowed (so that UnrollAndJam does not trigger rotation) and then it would do all the unroll checks, if they failed it will rotate and then do the checks again (then unrolling).
So doing it inside of reminder unroll, will copy the checks, which is similar to current stare where the call is copied.
But I see your point and it will make the reminder unroll more powerful and could be used later on in other parts, so I can change that.

https://github.com/llvm/llvm-project/pull/148243