[VPlan] Add opcode to create step for wide inductions.

llvm/lib/Transforms/Vectorize/LoopVectorizationPlanner.h

@@ -172,6 +172,14 @@ class VPBuilder {
           new VPInstruction(Opcode, Operands, *FMFs, DL, Name));
     return createInstruction(Opcode, Operands, DL, Name);
   }
+  VPInstruction *createNaryOp(unsigned Opcode,
+                              std::initializer_list<VPValue *> Operands,
+                              Type *ResultTy,
+                              std::optional<FastMathFlags> FMFs = {},
+                              DebugLoc DL = {}, const Twine &Name = "") {
+    return tryInsertInstruction(new VPInstructionWithType(
+        Opcode, Operands, ResultTy, FMFs.value_or(FastMathFlags()), DL, Name));
+  }
 
   VPInstruction *createOverflowingOp(unsigned Opcode,
                                      std::initializer_list<VPValue *> Operands,

llvm/lib/Transforms/Vectorize/LoopVectorize.cpp

@@ -7949,7 +7949,8 @@ DenseMap<const SCEV *, Value *> LoopVectorizationPlanner::executePlan(
       BestVPlan, BestVF,
       TTI.getRegisterBitWidth(TargetTransformInfo::RGK_FixedWidthVector));
   VPlanTransforms::removeDeadRecipes(BestVPlan);
-  VPlanTransforms::convertToConcreteRecipes(BestVPlan);
+  VPTypeAnalysis TypeInfo(Legal->getWidestInductionType());
+  VPlanTransforms::convertToConcreteRecipes(BestVPlan, TypeInfo);
 
   // Perform the actual loop transformation.
   VPTransformState State(&TTI, BestVF, LI, DT, ILV.Builder, &ILV, &BestVPlan,

llvm/lib/Transforms/Vectorize/VPlan.h

@@ -865,6 +865,7 @@ class VPInstruction : public VPRecipeWithIRFlags,
     CalculateTripCountMinusVF,
     // Increment the canonical IV separately for each unrolled part.
     CanonicalIVIncrementForPart,
+    WideIVStep,
     BranchOnCount,
     BranchOnCond,
     Broadcast,
@@ -1041,11 +1042,18 @@ class VPInstructionWithType : public VPInstruction {
   VPInstructionWithType(unsigned Opcode, ArrayRef<VPValue *> Operands,
                         Type *ResultTy, DebugLoc DL, const Twine &Name = "")
       : VPInstruction(Opcode, Operands, DL, Name), ResultTy(ResultTy) {}
+  VPInstructionWithType(unsigned Opcode,
+                        std::initializer_list<VPValue *> Operands,
+                        Type *ResultTy, FastMathFlags FMFs, DebugLoc DL = {},
+                        const Twine &Name = "")
+      : VPInstruction(Opcode, Operands, FMFs, DL, Name), ResultTy(ResultTy) {}
 
   static inline bool classof(const VPRecipeBase *R) {
     // VPInstructionWithType are VPInstructions with specific opcodes requiring
     // type information.
-    return R->isScalarCast();
+    auto *VPI = dyn_cast<VPInstruction>(R);
+    return R->isScalarCast() ||
+           (VPI && VPI->getOpcode() == VPInstruction::WideIVStep);
   }
 
   static inline bool classof(const VPUser *R) {

llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp

@@ -873,7 +873,8 @@ bool VPInstruction::isFPMathOp() const {
   return Opcode == Instruction::FAdd || Opcode == Instruction::FMul ||
          Opcode == Instruction::FNeg || Opcode == Instruction::FSub ||
          Opcode == Instruction::FDiv || Opcode == Instruction::FRem ||
-         Opcode == Instruction::FCmp || Opcode == Instruction::Select;
+         Opcode == Instruction::FCmp || Opcode == Instruction::Select ||
+         Opcode == VPInstruction::WideIVStep;
 }
 #endif
 
@@ -928,6 +929,7 @@ bool VPInstruction::opcodeMayReadOrWriteFromMemory() const {
   case VPInstruction::LogicalAnd:
   case VPInstruction::Not:
   case VPInstruction::PtrAdd:
+  case VPInstruction::WideIVStep:
     return false;
   default:
     return true;
@@ -1097,9 +1099,19 @@ void VPInstructionWithType::print(raw_ostream &O, const Twine &Indent,
                                   VPSlotTracker &SlotTracker) const {
   O << Indent << "EMIT ";
   printAsOperand(O, SlotTracker);
-  O << " = " << Instruction::getOpcodeName(getOpcode()) << " ";
-  printOperands(O, SlotTracker);
-  O << " to " << *ResultTy;
+  O << " = ";
+
+  switch (getOpcode()) {
+  case VPInstruction::WideIVStep:
+    O << "wide-iv-step";
+    printOperands(O, SlotTracker);
+    break;
+  default:
+    assert(Instruction::isCast(getOpcode()) && "unhandled opcode");
+    O << Instruction::getOpcodeName(getOpcode()) << " ";
+    printOperands(O, SlotTracker);
+    O << " to " << *ResultTy;
+  }
 }
 #endif
 

llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp

@@ -1019,6 +1019,16 @@ static void simplifyRecipe(VPRecipeBase &R, VPTypeAnalysis &TypeInfo) {
       TypeInfo.inferScalarType(R.getOperand(1)) ==
           TypeInfo.inferScalarType(R.getVPSingleValue()))
     return R.getVPSingleValue()->replaceAllUsesWith(R.getOperand(1));
+
+  if (match(&R, m_VPInstruction<VPInstruction::WideIVStep>(m_VPValue(X),
+                                                           m_SpecificInt(1)))) {
+    Type *WideStepTy = TypeInfo.inferScalarType(R.getVPSingleValue());
+    if (TypeInfo.inferScalarType(X) != WideStepTy) {
+      X = new VPWidenCastRecipe(Instruction::Trunc, X, WideStepTy);
+      X->getDefiningRecipe()->insertBefore(&R);
+    }
+    R.getVPSingleValue()->replaceAllUsesWith(X);
+  }
 }
 
 void VPlanTransforms::simplifyRecipes(VPlan &Plan, Type &CanonicalIVTy) {
@@ -2367,21 +2377,62 @@ void VPlanTransforms::createInterleaveGroups(
   }
 }
 
-void VPlanTransforms::convertToConcreteRecipes(VPlan &Plan) {
+void VPlanTransforms::convertToConcreteRecipes(VPlan &Plan,
+                                               VPTypeAnalysis &TypeInfo) {
+  using namespace llvm::VPlanPatternMatch;
+
   for (VPBasicBlock *VPBB : VPBlockUtils::blocksOnly<VPBasicBlock>(
            vp_depth_first_deep(Plan.getEntry()))) {
-    for (VPRecipeBase &R : make_early_inc_range(VPBB->phis())) {
-      if (!isa<VPCanonicalIVPHIRecipe, VPEVLBasedIVPHIRecipe>(&R))
+    for (VPRecipeBase &R : make_early_inc_range(*VPBB)) {
+      if (isa<VPCanonicalIVPHIRecipe, VPEVLBasedIVPHIRecipe>(&R)) {
+        auto *PhiR = cast<VPHeaderPHIRecipe>(&R);
+        StringRef Name =
+            isa<VPCanonicalIVPHIRecipe>(PhiR) ? "index" : "evl.based.iv";
+        auto *ScalarR = new VPInstruction(
+            Instruction::PHI, {PhiR->getStartValue(), PhiR->getBackedgeValue()},
+            PhiR->getDebugLoc(), Name);
+        ScalarR->insertBefore(PhiR);
+        PhiR->replaceAllUsesWith(ScalarR);
+        PhiR->eraseFromParent();
+        continue;
+      }
+
+      VPValue *VectorStep;
+      VPValue *ScalarStep;
+      if (!match(&R, m_VPInstruction<VPInstruction::WideIVStep>(
+                         m_VPValue(VectorStep), m_VPValue(ScalarStep))))
         continue;
-      auto *PhiR = cast<VPHeaderPHIRecipe>(&R);
-      StringRef Name =
-          isa<VPCanonicalIVPHIRecipe>(PhiR) ? "index" : "evl.based.iv";
-      auto *ScalarR = new VPInstruction(
-          Instruction::PHI, {PhiR->getStartValue(), PhiR->getBackedgeValue()},
-          PhiR->getDebugLoc(), Name);
-      ScalarR->insertBefore(PhiR);
-      PhiR->replaceAllUsesWith(ScalarR);
-      PhiR->eraseFromParent();
+      auto *VPI = cast<VPInstruction>(&R);
+      VPBuilder Builder(VPI->getParent(), VPI->getIterator());
+      Type *IVTy = TypeInfo.inferScalarType(VPI);
+      if (TypeInfo.inferScalarType(VectorStep) != IVTy) {
+        Instruction::CastOps CastOp = IVTy->isFloatingPointTy()
+                                          ? Instruction::UIToFP
+                                          : Instruction::Trunc;
+        VectorStep = Builder.createWidenCast(CastOp, VectorStep, IVTy);
+      }
+
+      auto *ConstStep =
+          ScalarStep->isLiveIn()
+              ? dyn_cast<ConstantInt>(ScalarStep->getLiveInIRValue())
+              : nullptr;
+      assert(!ConstStep || ConstStep->getValue() != 1);
+      if (TypeInfo.inferScalarType(ScalarStep) != IVTy) {
+        ScalarStep =
+            Builder.createWidenCast(Instruction::Trunc, ScalarStep, IVTy);
+      }
+
+      std::optional<FastMathFlags> FMFs;
+      if (IVTy->isFloatingPointTy())
+        FMFs = VPI->getFastMathFlags();
+
+      unsigned MulOpc =
+          IVTy->isFloatingPointTy() ? Instruction::FMul : Instruction::Mul;
+      VPInstruction *Mul = Builder.createNaryOp(
+          MulOpc, {VectorStep, ScalarStep}, FMFs, R.getDebugLoc());
+      VectorStep = Mul;
+      VPI->replaceAllUsesWith(VectorStep);
+      VPI->eraseFromParent();
     }
   }
 }

llvm/lib/Transforms/Vectorize/VPlanTransforms.h

@@ -177,7 +177,7 @@ struct VPlanTransforms {
                                          VPRecipeBuilder &RecipeBuilder);
 
   /// Lower abstract recipes to concrete ones, that can be codegen'd.
-  static void convertToConcreteRecipes(VPlan &Plan);
+  static void convertToConcreteRecipes(VPlan &Plan, VPTypeAnalysis &TypeInfo);
 
   /// Perform instcombine-like simplifications on recipes in \p Plan. Use \p
   /// CanonicalIVTy as type for all un-typed live-ins in VPTypeAnalysis.

llvm/lib/Transforms/Vectorize/VPlanUnroll.cpp

@@ -155,33 +155,13 @@ void UnrollState::unrollWidenInductionByUF(
   if (isa_and_present<FPMathOperator>(ID.getInductionBinOp()))
     FMFs = ID.getInductionBinOp()->getFastMathFlags();
 
-  VPValue *VectorStep = &Plan.getVF();
-  VPBuilder Builder(PH);
-  if (TypeInfo.inferScalarType(VectorStep) != IVTy) {
-    Instruction::CastOps CastOp =
-        IVTy->isFloatingPointTy() ? Instruction::UIToFP : Instruction::Trunc;
-    VectorStep = Builder.createWidenCast(CastOp, VectorStep, IVTy);
-    ToSkip.insert(VectorStep->getDefiningRecipe());
-  }
-
   VPValue *ScalarStep = IV->getStepValue();
-  auto *ConstStep = ScalarStep->isLiveIn()
-                        ? dyn_cast<ConstantInt>(ScalarStep->getLiveInIRValue())
-                        : nullptr;
-  if (!ConstStep || ConstStep->getValue() != 1) {
-    if (TypeInfo.inferScalarType(ScalarStep) != IVTy) {
-      ScalarStep =
-          Builder.createWidenCast(Instruction::Trunc, ScalarStep, IVTy);
-      ToSkip.insert(ScalarStep->getDefiningRecipe());
-    }
+  VPBuilder Builder(PH);
+  VPInstruction *VectorStep = Builder.createNaryOp(
+      VPInstruction::WideIVStep, {&Plan.getVF(), ScalarStep}, IVTy, FMFs,
+      IV->getDebugLoc());
 
-    unsigned MulOpc =
-        IVTy->isFloatingPointTy() ? Instruction::FMul : Instruction::Mul;
-    VPInstruction *Mul = Builder.createNaryOp(MulOpc, {VectorStep, ScalarStep},
-                                              FMFs, IV->getDebugLoc());
-    VectorStep = Mul;
-    ToSkip.insert(Mul);
-  }
+  ToSkip.insert(VectorStep);
 
   // Now create recipes to compute the induction steps for part 1 .. UF. Part 0
   // remains the header phi. Parts > 0 are computed by adding Step to the