[llvm] [VPlan] Unroll VPReplicateRecipe by VF. (PR #142433)

[via llvm-commits]

github-actions[bot] wrote:

<!--LLVM CODE FORMAT COMMENT: {clang-format}-->


:warning: C/C++ code formatter, clang-format found issues in your code. :warning:

<details>
<summary>
You can test this locally with the following command:
</summary>

``````````bash
git-clang-format --diff HEAD~1 HEAD --extensions h,cpp -- llvm/lib/Transforms/Vectorize/LoopVectorize.cpp llvm/lib/Transforms/Vectorize/VPlan.cpp llvm/lib/Transforms/Vectorize/VPlan.h llvm/lib/Transforms/Vectorize/VPlanAnalysis.cpp llvm/lib/Transforms/Vectorize/VPlanPatternMatch.h llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp llvm/lib/Transforms/Vectorize/VPlanTransforms.h llvm/lib/Transforms/Vectorize/VPlanUnroll.cpp
``````````

</details>

<details>
<summary>
View the diff from clang-format here.
</summary>

``````````diff
diff --git a/llvm/lib/Transforms/Vectorize/VPlan.h b/llvm/lib/Transforms/Vectorize/VPlan.h
index 454accba7..8764454a7 100644
--- a/llvm/lib/Transforms/Vectorize/VPlan.h
+++ b/llvm/lib/Transforms/Vectorize/VPlan.h
@@ -907,10 +907,12 @@ public:
     BranchOnCount,
     BranchOnCond,
     Broadcast,
-    /// Creates a struct of fixed-width vectors containing all operands. The number of operands
-/// matches the number of fields in the struct.
+    /// Creates a struct of fixed-width vectors containing all operands. The
+    /// number of operands
+    /// matches the number of fields in the struct.
     BuildStructVector,
-    /// Creates a fixed-width vector containing all operands. The number of operands matches the vector element count.
+    /// Creates a fixed-width vector containing all operands. The number of
+    /// operands matches the vector element count.
     BuildVector,
     ComputeAnyOfResult,
     ComputeFindLastIVResult,
diff --git a/llvm/lib/Transforms/Vectorize/VPlanPatternMatch.h b/llvm/lib/Transforms/Vectorize/VPlanPatternMatch.h
index 7c427ac18..48a96e010 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanPatternMatch.h
+++ b/llvm/lib/Transforms/Vectorize/VPlanPatternMatch.h
@@ -276,8 +276,7 @@ using UnaryVPInstruction_match =
     UnaryRecipe_match<Op0_t, Opcode, VPInstruction>;
 
 template <unsigned Opcode>
-using ZeroOpVPInstruction_match =
-    ZeroOpRecipe_match<Opcode, VPInstruction>;
+using ZeroOpVPInstruction_match = ZeroOpRecipe_match<Opcode, VPInstruction>;
 
 template <typename Op0_t, unsigned Opcode>
 using AllUnaryRecipe_match =
@@ -310,8 +309,7 @@ using AllBinaryRecipe_match =
     BinaryRecipe_match<Op0_t, Op1_t, Opcode, Commutative, VPWidenRecipe,
                        VPReplicateRecipe, VPWidenCastRecipe, VPInstruction>;
 
-inline ZeroOpVPInstruction_match<VPInstruction::BuildVector>
-m_BuildVector() {
+inline ZeroOpVPInstruction_match<VPInstruction::BuildVector> m_BuildVector() {
   return ZeroOpVPInstruction_match<VPInstruction::BuildVector>();
 }
 
diff --git a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
index 2eb03ab64..4a1d7fd85 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
@@ -493,10 +493,9 @@ Value *VPInstruction::generate(VPTransformState &State) {
   }
   case Instruction::ExtractElement: {
     assert(State.VF.isVector() && "Only extract elements from vectors");
-    unsigned IdxToExtract = cast<ConstantInt>(getOperand(1)->getLiveInIRValue())
-                                ->getZExtValue();
-    return State.get(getOperand(0),
-                     VPLane(IdxToExtract));
+    unsigned IdxToExtract =
+        cast<ConstantInt>(getOperand(1)->getLiveInIRValue())->getZExtValue();
+    return State.get(getOperand(0), VPLane(IdxToExtract));
   }
   case Instruction::Freeze: {
     Value *Op = State.get(getOperand(0), vputils::onlyFirstLaneUsed(this));
@@ -611,11 +610,11 @@ Value *VPInstruction::generate(VPTransformState &State) {
     auto *StructTy =
         cast<StructType>(State.TypeAnalysis.inferScalarType(getOperand(0)));
     auto NumOfElements = ElementCount::getFixed(getNumOperands());
-    Value *Res = PoisonValue::get(
-        toVectorizedTy(StructTy, NumOfElements));
-    assert(NumOfElements.getKnownMinValue() == StructTy->getNumElements() && "number of operands must match number of elements in StructTy");
+    Value *Res = PoisonValue::get(toVectorizedTy(StructTy, NumOfElements));
+    assert(NumOfElements.getKnownMinValue() == StructTy->getNumElements() &&
+           "number of operands must match number of elements in StructTy");
     for (const auto &[Idx, Op] : enumerate(operands())) {
-      for (unsigned I = 0 ; I != NumOfElements .getKnownMinValue(); I++) {
+      for (unsigned I = 0; I != NumOfElements.getKnownMinValue(); I++) {
         Value *ScalarValue = Builder.CreateExtractValue(State.get(Op, true), I);
         Value *VectorValue = Builder.CreateExtractValue(Res, I);
         VectorValue =
@@ -628,8 +627,7 @@ Value *VPInstruction::generate(VPTransformState &State) {
   case VPInstruction::BuildVector: {
     auto *ScalarTy = State.TypeAnalysis.inferScalarType(getOperand(0));
     auto NumOfElements = ElementCount::getFixed(getNumOperands());
-    Value *Res = PoisonValue::get(
-        toVectorizedTy(ScalarTy, NumOfElements));
+    Value *Res = PoisonValue::get(toVectorizedTy(ScalarTy, NumOfElements));
     for (const auto &[Idx, Op] : enumerate(operands()))
       Res = State.Builder.CreateInsertElement(Res, State.get(Op, true),
                                               State.Builder.getInt32(Idx));
@@ -2800,7 +2798,8 @@ void VPReplicateRecipe::execute(VPTransformState &State) {
   Instruction *UI = getUnderlyingInstr();
 
   if (!State.Lane) {
-    assert(IsSingleScalar && "VPReplicateRecipes outside replicate regions must be unrolled");
+    assert(IsSingleScalar &&
+           "VPReplicateRecipes outside replicate regions must be unrolled");
     scalarizeInstruction(UI, this, VPLane(0), State);
     return;
   }
diff --git a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
index df65874f6..906763e6d 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
@@ -1142,18 +1142,20 @@ static void simplifyRecipe(VPRecipeBase &R, VPTypeAnalysis &TypeInfo) {
 
   // Look through ExtractLastElement (BuildVector ....).
   if (match(&R, m_VPInstruction<VPInstruction::ExtractLastElement>(
-            m_BuildVector()))) {
+                    m_BuildVector()))) {
     auto *BuildVector = cast<VPInstruction>(R.getOperand(0));
-    Def->replaceAllUsesWith(BuildVector->getOperand(BuildVector->getNumOperands() - 1));
+    Def->replaceAllUsesWith(
+        BuildVector->getOperand(BuildVector->getNumOperands() - 1));
     return;
   }
   // Look through ExtractPenultimateElement (BuildVector ....).
   if (match(&R, m_VPInstruction<VPInstruction::ExtractPenultimateElement>(
-            m_BuildVector()))) {
+                    m_BuildVector()))) {
     auto *BuildVector = cast<VPInstruction>(R.getOperand(0));
-    Def->replaceAllUsesWith(BuildVector->getOperand(BuildVector->getNumOperands() - 2));
+    Def->replaceAllUsesWith(
+        BuildVector->getOperand(BuildVector->getNumOperands() - 2));
     return;
-}
+  }
 
   // Some simplifications can only be applied after unrolling. Perform them
   // below.
diff --git a/llvm/lib/Transforms/Vectorize/VPlanUnroll.cpp b/llvm/lib/Transforms/Vectorize/VPlanUnroll.cpp
index 9b0476114..2b15e9ca0 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanUnroll.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanUnroll.cpp
@@ -459,12 +459,14 @@ static VPReplicateRecipe *cloneForLane(VPlan &Plan, VPBuilder &Builder,
       continue;
     }
     if (Lane.getKind() == VPLane::Kind::ScalableLast) {
-      NewOps.push_back(Builder.createNaryOp(VPInstruction::ExtractLastElement, {Op}));
+      NewOps.push_back(
+          Builder.createNaryOp(VPInstruction::ExtractLastElement, {Op}));
       continue;
     }
     // Look through buildvector to avoid unnecessary extracts.
     if (match(Op, m_BuildVector())) {
-      NewOps.push_back(cast<VPInstruction>(Op)->getOperand(Lane.getKnownLane()));
+      NewOps.push_back(
+          cast<VPInstruction>(Op)->getOperand(Lane.getKnownLane()));
       continue;
     }
     VPValue *Idx =

``````````

</details>


https://github.com/llvm/llvm-project/pull/142433