[LoopVectorizer] Add support for chaining partial reductions

llvm/lib/Target/AArch64/AArch64TargetTransformInfo.h

@@ -368,7 +368,7 @@ class AArch64TTIImpl : public BasicTTIImplBase<AArch64TTIImpl> {
     InstructionCost Invalid = InstructionCost::getInvalid();
     InstructionCost Cost(TTI::TCC_Basic);
 
-    if (Opcode != Instruction::Add)
+    if (Opcode != Instruction::Add && Opcode != Instruction::Sub)
       return Invalid;
 
     if (InputTypeA != InputTypeB)

llvm/lib/Transforms/Vectorize/LoopVectorize.cpp

@@ -8790,12 +8790,12 @@ VPReplicateRecipe *VPRecipeBuilder::handleReplication(Instruction *I,
 /// are valid so recipes can be formed later.
 void VPRecipeBuilder::collectScaledReductions(VFRange &Range) {
   // Find all possible partial reductions.
-  SmallVector<std::pair<PartialReductionChain, unsigned>, 1>
+  SmallVector<std::pair<PartialReductionChain, unsigned>>
       PartialReductionChains;
-  for (const auto &[Phi, RdxDesc] : Legal->getReductionVars())
-    if (std::optional<std::pair<PartialReductionChain, unsigned>> Pair =
-            getScaledReduction(Phi, RdxDesc, Range))
-      PartialReductionChains.push_back(*Pair);
+  for (const auto &[Phi, RdxDesc] : Legal->getReductionVars()) {
+    if (auto SR = getScaledReduction(Phi, RdxDesc.getLoopExitInstr(), Range))
+      PartialReductionChains.append(*SR);
+  }
 
   // A partial reduction is invalid if any of its extends are used by
   // something that isn't another partial reduction. This is because the
@@ -8823,26 +8823,41 @@ void VPRecipeBuilder::collectScaledReductions(VFRange &Range) {
   }
 }
 
-std::optional<std::pair<PartialReductionChain, unsigned>>
-VPRecipeBuilder::getScaledReduction(PHINode *PHI,
-                                    const RecurrenceDescriptor &Rdx,
+std::optional<SmallVector<std::pair<PartialReductionChain, unsigned>>>
+VPRecipeBuilder::getScaledReduction(Instruction *PHI, Instruction *RdxExitInstr,
                                     VFRange &Range) {
+
+  if (!CM.TheLoop->contains(RdxExitInstr))
+    return std::nullopt;
+
   // TODO: Allow scaling reductions when predicating. The select at
   // the end of the loop chooses between the phi value and most recent
   // reduction result, both of which have different VFs to the active lane
   // mask when scaling.
-  if (CM.blockNeedsPredicationForAnyReason(Rdx.getLoopExitInstr()->getParent()))
+  if (CM.blockNeedsPredicationForAnyReason(RdxExitInstr->getParent()))
     return std::nullopt;
 
-  auto *Update = dyn_cast<BinaryOperator>(Rdx.getLoopExitInstr());
+  auto *Update = dyn_cast<BinaryOperator>(RdxExitInstr);
   if (!Update)
     return std::nullopt;
 
   Value *Op = Update->getOperand(0);
   Value *PhiOp = Update->getOperand(1);
-  if (Op == PHI) {
-    Op = Update->getOperand(1);
-    PhiOp = Update->getOperand(0);
+  if (Op == PHI)
+    std::swap(Op, PhiOp);
+
+  SmallVector<std::pair<PartialReductionChain, unsigned>> Chains;
+
+  if (auto *OpInst = dyn_cast<Instruction>(Op)) {
+    if (auto SR0 = getScaledReduction(PHI, OpInst, Range)) {
+      Chains.append(*SR0);
+      PHI = SR0->rbegin()->first.Reduction;
+
+      Op = Update->getOperand(0);
+      PhiOp = Update->getOperand(1);
+      if (Op == PHI)
+        std::swap(Op, PhiOp);
+    }
   }
   if (PhiOp != PHI)
     return std::nullopt;
@@ -8860,12 +8875,16 @@ VPRecipeBuilder::getScaledReduction(PHINode *PHI,
   Instruction *ExtA = cast<Instruction>(BinOp->getOperand(0));
   Instruction *ExtB = cast<Instruction>(BinOp->getOperand(1));
 
+  // Check that the extends extend from the same type.
+  if (A->getType() != B->getType())
+    return std::nullopt;
+
   TTI::PartialReductionExtendKind OpAExtend =
       TargetTransformInfo::getPartialReductionExtendKind(ExtA);
   TTI::PartialReductionExtendKind OpBExtend =
       TargetTransformInfo::getPartialReductionExtendKind(ExtB);
 
-  PartialReductionChain Chain(Rdx.getLoopExitInstr(), ExtA, ExtB, BinOp);
+  PartialReductionChain Chain(RdxExitInstr, ExtA, ExtB, BinOp);
 
   unsigned TargetScaleFactor =
       PHI->getType()->getPrimitiveSizeInBits().getKnownScalarFactor(
@@ -8880,9 +8899,9 @@ VPRecipeBuilder::getScaledReduction(PHINode *PHI,
             return Cost.isValid();
           },
           Range))
-    return std::make_pair(Chain, TargetScaleFactor);
+    Chains.push_back(std::make_pair(Chain, TargetScaleFactor));
 
-  return std::nullopt;
+  return Chains;
 }
 
 VPRecipeBase *
@@ -8979,7 +8998,9 @@ VPRecipeBuilder::tryToCreatePartialReduction(Instruction *Reduction,
 
   VPValue *BinOp = Operands[0];
   VPValue *Phi = Operands[1];
-  if (isa<VPReductionPHIRecipe>(BinOp->getDefiningRecipe()))
+  VPRecipeBase *BinOpRecipe = BinOp->getDefiningRecipe();
+  if (isa<VPReductionPHIRecipe>(BinOpRecipe) ||
+      isa<VPPartialReductionRecipe>(BinOpRecipe))
     std::swap(BinOp, Phi);
 
   return new VPPartialReductionRecipe(Reduction->getOpcode(), BinOp, Phi,

llvm/lib/Transforms/Vectorize/VPRecipeBuilder.h

@@ -144,8 +144,8 @@ class VPRecipeBuilder {
   /// Returns null if no scaled reduction was found, otherwise a pair with a
   /// struct containing reduction information and the scaling factor between the
   /// number of elements in the input and output.
-  std::optional<std::pair<PartialReductionChain, unsigned>>
-  getScaledReduction(PHINode *PHI, const RecurrenceDescriptor &Rdx,
+  std::optional<SmallVector<std::pair<PartialReductionChain, unsigned>>>
+  getScaledReduction(Instruction *PHI, Instruction *RdxExitInstr,
                      VFRange &Range);
 
 public:

llvm/lib/Transforms/Vectorize/VPlan.h

@@ -2453,13 +2453,16 @@ class VPPartialReductionRecipe : public VPSingleDefRecipe {
       : VPSingleDefRecipe(VPDef::VPPartialReductionSC,
                           ArrayRef<VPValue *>({Op0, Op1}), ReductionInst),
         Opcode(Opcode) {
-    assert(isa<VPReductionPHIRecipe>(getOperand(1)->getDefiningRecipe()) &&
+    auto *DefiningRecipe = getOperand(1)->getDefiningRecipe();
+    assert((isa<VPReductionPHIRecipe>(DefiningRecipe) ||
+            isa<VPPartialReductionRecipe>(DefiningRecipe)) &&
            "Unexpected operand order for partial reduction recipe");
   }
   ~VPPartialReductionRecipe() override = default;
 
   VPPartialReductionRecipe *clone() override {
-    return new VPPartialReductionRecipe(Opcode, getOperand(0), getOperand(1));
+    return new VPPartialReductionRecipe(Opcode, getOperand(0), getOperand(1),
+                                        getUnderlyingInstr());
   }
 
   VP_CLASSOF_IMPL(VPDef::VPPartialReductionSC)

llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp

@@ -317,13 +317,22 @@ void VPPartialReductionRecipe::execute(VPTransformState &State) {
   State.setDebugLocFrom(getDebugLoc());
   auto &Builder = State.Builder;
 
-  assert(getOpcode() == Instruction::Add &&
-         "Unhandled partial reduction opcode");
-
   Value *BinOpVal = State.get(getOperand(0));
   Value *PhiVal = State.get(getOperand(1));
   assert(PhiVal && BinOpVal && "Phi and Mul must be set");
 
+  auto Opcode = getOpcode();
+
+  // Currently we don't have a partial_reduce_sub intrinsic,
+  // so mimic the behaviour by negating the second operand
+  if (Opcode == Instruction::Sub) {
+    BinOpVal = Builder.CreateSub(Constant::getNullValue(BinOpVal->getType()),
+                                 BinOpVal);
+    Opcode = Instruction::Add;
+  }
+
+  assert(Opcode == Instruction::Add && "Unhandled partial reduction opcode");
+
   Type *RetTy = PhiVal->getType();
 
   CallInst *V = Builder.CreateIntrinsic(