[llvm] [LoopPeel] Implement initial peeling off the last loop iteration. (PR #139551)

[Florian Hahn via llvm-commits]

================
@@ -325,19 +326,71 @@ static unsigned peelToTurnInvariantLoadsDerefencebale(Loop &L,
   return 0;
 }
 
-// Return the number of iterations to peel off that make conditions in the
-// body true/false. For example, if we peel 2 iterations off the loop below,
-// the condition i < 2 can be evaluated at compile time.
+bool llvm::canPeelLastIteration(const Loop &L, ScalarEvolution &SE) {
+  const SCEV *BTC = SE.getBackedgeTakenCount(&L);
+  Value *Inc;
+  CmpPredicate Pred;
+  BasicBlock *Succ1;
+  BasicBlock *Succ2;
+  // The loop must execute at least 2 iterations to guarantee that peeled
+  // iteration executes.
+  // TODO: Add checks during codegen.
+  if (isa<SCEVCouldNotCompute>(BTC) ||
+      !SE.isKnownPredicate(CmpInst::ICMP_UGT, BTC, SE.getOne(BTC->getType())))
+    return false;
+
+  // Check if the exit condition of the loop can be adjusted by the peeling
+  // codegen. For now, it must
+  // * exit via the latch,
+  // * the exit condition must be a NE/EQ compare of an induction with step
+  // of 1.
+  BasicBlock *Latch = L.getLoopLatch();
+  return Latch == L.getExitingBlock() &&
----------------
fhahn wrote:

Ah yes thanks! Added a test with multiple latches & multiple exiting, but couldn't get one to crash because I think at the moment we don't compute BTCs for loops with multiple latches.

https://github.com/llvm/llvm-project/pull/139551