[llvm] [VPlan] Move tail folding logic out of addMiddleCheck. NFC (PR #203475)

[via llvm-commits]

llvmorg-github-actions[bot] wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-llvm-transforms

Author: Luke Lau (lukel97)

<details>
<summary>Changes</summary>

We simplify the TripCount == VectorTrip count condition with tail
folding, but we can just do that in foldTailByMasking and keep the
logic in one place instead.


---
Full diff: https://github.com/llvm/llvm-project/pull/203475.diff


3 Files Affected:

- (modified) llvm/lib/Transforms/Vectorize/LoopVectorize.cpp (+1-2) 
- (modified) llvm/lib/Transforms/Vectorize/VPlanConstruction.cpp (+14-7) 
- (modified) llvm/lib/Transforms/Vectorize/VPlanTransforms.h (+1-1) 


``````````diff
diff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
index f9b325a5233e6..c3dbf0a1ee3fa 100644
--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
@@ -6563,8 +6563,7 @@ VPlanPtr LoopVectorizationPlanner::tryToBuildVPlan1() {
   // loop with an in-loop mask, then the middle check has already been
   // created to compare against the actual number of lanes executed.
   if (EEStyle != UncountableExitStyle::MaskedHandleExitInScalarLoop)
-    RUN_VPLAN_PASS(VPlanTransforms::addMiddleCheck, *VPlan0,
-                   CM.foldTailByMasking());
+    RUN_VPLAN_PASS(VPlanTransforms::addMiddleCheck, *VPlan0);
   RUN_VPLAN_PASS(VPlanTransforms::createLoopRegions, *VPlan0,
                  getDebugLocFromInstOrOperands(Legal->getPrimaryInduction()));
   if (CM.foldTailByMasking())
diff --git a/llvm/lib/Transforms/Vectorize/VPlanConstruction.cpp b/llvm/lib/Transforms/Vectorize/VPlanConstruction.cpp
index 3e41133ff7463..dd3c8bebd1cd1 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanConstruction.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanConstruction.cpp
@@ -1301,7 +1301,7 @@ bool VPlanTransforms::handleEarlyExits(VPlan &Plan, UncountableExitStyle Style,
   return true;
 }
 
-void VPlanTransforms::addMiddleCheck(VPlan &Plan, bool TailFolded) {
+void VPlanTransforms::addMiddleCheck(VPlan &Plan) {
   auto *MiddleVPBB = VPBlockUtils::getPlainCFGMiddleBlock(Plan);
   // If MiddleVPBB has a single successor then the original loop does not exit
   // via the latch and the single successor must be the scalar preheader.
@@ -1332,12 +1332,9 @@ void VPlanTransforms::addMiddleCheck(VPlan &Plan, bool TailFolded) {
   auto *LatchVPBB = cast<VPBasicBlock>(MiddleVPBB->getSinglePredecessor());
   DebugLoc LatchDL = LatchVPBB->getTerminator()->getDebugLoc();
   VPBuilder Builder(MiddleVPBB);
-  VPValue *Cmp;
-  if (TailFolded)
-    Cmp = Plan.getTrue();
-  else
-    Cmp = Builder.createICmp(CmpInst::ICMP_EQ, Plan.getTripCount(),
-                             &Plan.getVectorTripCount(), LatchDL, "cmp.n");
+  VPValue *Cmp =
+      Builder.createICmp(CmpInst::ICMP_EQ, Plan.getTripCount(),
+                         &Plan.getVectorTripCount(), LatchDL, "cmp.n");
   Builder.createNaryOp(VPInstruction::BranchOnCond, {Cmp}, LatchDL);
 }
 
@@ -1449,6 +1446,16 @@ void VPlanTransforms::foldTailByMasking(VPlan &Plan) {
         Builder.createNaryOp(VPInstruction::ExtractLane, {LastActiveLane, Op});
     R.getVPSingleValue()->replaceAllUsesWith(Ext);
   }
+
+  // VectorTripCount now equals TripCount so simplify the MiddleVPBB branch.
+  assert(match(Plan.getMiddleBlock()->getTerminator(),
+               m_BranchOnCond(m_OneUse(m_SpecificICmp(
+                   CmpInst::ICMP_EQ, m_Specific(Plan.getTripCount()),
+                   m_Specific(&Plan.getVectorTripCount()))))) &&
+         "Unexpected MiddleVPBB branch");
+  VPValue *OldCond = Plan.getMiddleBlock()->getTerminator()->getOperand(0);
+  Plan.getMiddleBlock()->getTerminator()->setOperand(0, Plan.getTrue());
+  OldCond->getDefiningRecipe()->eraseFromParent();
 }
 
 /// Insert \p CheckBlockVPBB on the edge leading to the vector preheader,
diff --git a/llvm/lib/Transforms/Vectorize/VPlanTransforms.h b/llvm/lib/Transforms/Vectorize/VPlanTransforms.h
index b186594cd5b1b..51984b803ab43 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanTransforms.h
+++ b/llvm/lib/Transforms/Vectorize/VPlanTransforms.h
@@ -172,7 +172,7 @@ struct VPlanTransforms {
 
   /// If a check is needed to guard executing the scalar epilogue loop, it will
   /// be added to the middle block.
-  LLVM_ABI_FOR_TEST static void addMiddleCheck(VPlan &Plan, bool TailFolded);
+  LLVM_ABI_FOR_TEST static void addMiddleCheck(VPlan &Plan);
 
   // Create a check in \p CheckBlock to see if the vector loop should be
   // executed. May create VPExpandSCEV recipes in the plan's entry block.

``````````

</details>


https://github.com/llvm/llvm-project/pull/203475