[llvm] [LoopVectorize] Enable vectorisation of early exit loops with live-outs (PR #120567)

[David Sherwood via llvm-commits]

================
@@ -9086,19 +9092,38 @@ addUsersInExitBlocks(VPlan &Plan,
       if (Op->isLiveIn())
         continue;
 
-      // Currently only live-ins can be used by exit values from blocks not
-      // exiting via the vector latch through to the middle block.
-      if (ExitIRI->getParent()->getSinglePredecessor() != MiddleVPBB)
-        return false;
-
       LLVMContext &Ctx = ExitIRI->getInstruction().getContext();
-      VPValue *Ext = B.createNaryOp(VPInstruction::ExtractFromEnd,
-                                    {Op, Plan.getOrAddLiveIn(ConstantInt::get(
-                                             IntegerType::get(Ctx, 32), 1))});
+      VPValue *Ext;
+      VPBasicBlock *PredVPBB =
+          cast<VPBasicBlock>(ExitIRI->getParent()->getPredecessors()[Idx]);
----------------
david-arm wrote:

In one of my early implementations last year I actually kept the users on separate lists, a list for "normal" exits and a list of early exits. But then I think you restructured the code in collectUsersInExitBlocks and addUsersInExitBlocks to handle arbitrary numbers of exits (https://github.com/llvm/llvm-project/pull/115066), which is the direction it seemed you wanted this to go. If we go down the route of doing this in `handleUncountableEarlyExit` then we can should also clean up `collectUsersInExitBlocks` as it doesn't need to be so general anymore.

https://github.com/llvm/llvm-project/pull/120567