Revert "[LoopVectorizer] Add support for chaining partial reductions (#120272)" (#124198)

Introduced stack buffer overflow, see #120272.

`getScaledReduction` can return empty vector, and there is not check for
that.

This reverts commit c9b7303.
This reverts commit caf0540.

Author: vitalybuka

llvm/lib/Transforms/Vectorize/LoopVectorize.cpp

@@ -8682,12 +8682,12 @@ 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>>
+  SmallVector<std::pair<PartialReductionChain, unsigned>, 1>
       PartialReductionChains;
-  for (const auto &[Phi, RdxDesc] : Legal->getReductionVars()) {
-    if (auto SR = getScaledReduction(Phi, RdxDesc.getLoopExitInstr(), Range))
-      PartialReductionChains.append(*SR);
-  }
+  for (const auto &[Phi, RdxDesc] : Legal->getReductionVars())
+    if (std::optional<std::pair<PartialReductionChain, unsigned>> Pair =
+            getScaledReduction(Phi, RdxDesc, Range))
+      PartialReductionChains.push_back(*Pair);
 
   // A partial reduction is invalid if any of its extends are used by
   // something that isn't another partial reduction. This is because the
@@ -8715,44 +8715,26 @@ void VPRecipeBuilder::collectScaledReductions(VFRange &Range) {
   }
 }
 
-std::optional<SmallVector<std::pair<PartialReductionChain, unsigned>>>
-VPRecipeBuilder::getScaledReduction(Instruction *PHI, Instruction *RdxExitInstr,
+std::optional<std::pair<PartialReductionChain, unsigned>>
+VPRecipeBuilder::getScaledReduction(PHINode *PHI,
+                                    const RecurrenceDescriptor &Rdx,
                                     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(RdxExitInstr->getParent()))
+  if (CM.blockNeedsPredicationForAnyReason(Rdx.getLoopExitInstr()->getParent()))
     return std::nullopt;
 
-  auto *Update = dyn_cast<BinaryOperator>(RdxExitInstr);
+  auto *Update = dyn_cast<BinaryOperator>(Rdx.getLoopExitInstr());
   if (!Update)
     return std::nullopt;
 
   Value *Op = Update->getOperand(0);
   Value *PhiOp = Update->getOperand(1);
-  if (Op == PHI)
-    std::swap(Op, PhiOp);
-
-  SmallVector<std::pair<PartialReductionChain, unsigned>> Chains;
-
-  // Try and get a scaled reduction from the first non-phi operand.
-  // If one is found, we use the discovered reduction instruction in
-  // place of the accumulator for costing.
-  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 (Op == PHI) {
+    Op = Update->getOperand(1);
+    PhiOp = Update->getOperand(0);
   }
   if (PhiOp != PHI)
     return std::nullopt;
@@ -8775,7 +8757,7 @@ VPRecipeBuilder::getScaledReduction(Instruction *PHI, Instruction *RdxExitInstr,
   TTI::PartialReductionExtendKind OpBExtend =
       TargetTransformInfo::getPartialReductionExtendKind(ExtB);
 
-  PartialReductionChain Chain(RdxExitInstr, ExtA, ExtB, BinOp);
+  PartialReductionChain Chain(Rdx.getLoopExitInstr(), ExtA, ExtB, BinOp);
 
   unsigned TargetScaleFactor =
       PHI->getType()->getPrimitiveSizeInBits().getKnownScalarFactor(
@@ -8790,9 +8772,9 @@ VPRecipeBuilder::getScaledReduction(Instruction *PHI, Instruction *RdxExitInstr,
             return Cost.isValid();
           },
           Range))
-    Chains.push_back(std::make_pair(Chain, TargetScaleFactor));
+    return std::make_pair(Chain, TargetScaleFactor);
 
-  return Chains;
+  return std::nullopt;
 }
 
 VPRecipeBase *
@@ -8887,14 +8869,12 @@ VPRecipeBuilder::tryToCreatePartialReduction(Instruction *Reduction,
          "Unexpected number of operands for partial reduction");
 
   VPValue *BinOp = Operands[0];
-  VPValue *Accumulator = Operands[1];
-  VPRecipeBase *BinOpRecipe = BinOp->getDefiningRecipe();
-  if (isa<VPReductionPHIRecipe>(BinOpRecipe) ||
-      isa<VPPartialReductionRecipe>(BinOpRecipe))
-    std::swap(BinOp, Accumulator);
-
-  return new VPPartialReductionRecipe(Reduction->getOpcode(), BinOp,
-                                      Accumulator, Reduction);
+  VPValue *Phi = Operands[1];
+  if (isa<VPReductionPHIRecipe>(BinOp->getDefiningRecipe()))
+    std::swap(BinOp, Phi);
+
+  return new VPPartialReductionRecipe(Reduction->getOpcode(), BinOp, Phi,
+                                      Reduction);
 }
 
 void LoopVectorizationPlanner::buildVPlansWithVPRecipes(ElementCount MinVF,

llvm/lib/Transforms/Vectorize/VPRecipeBuilder.h

@@ -142,8 +142,8 @@ 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<SmallVector<std::pair<PartialReductionChain, unsigned>>>
-  getScaledReduction(Instruction *PHI, Instruction *RdxExitInstr,
+  std::optional<std::pair<PartialReductionChain, unsigned>>
+  getScaledReduction(PHINode *PHI, const RecurrenceDescriptor &Rdx,
                      VFRange &Range);
 
 public:

llvm/lib/Transforms/Vectorize/VPlan.h

@@ -2455,10 +2455,7 @@ class VPPartialReductionRecipe : public VPSingleDefRecipe {
       : VPSingleDefRecipe(VPDef::VPPartialReductionSC,
                           ArrayRef<VPValue *>({Op0, Op1}), ReductionInst),
         Opcode(Opcode) {
-    [[maybe_unused]] auto *AccumulatorRecipe =
-        getOperand(1)->getDefiningRecipe();
-    assert((isa<VPReductionPHIRecipe>(AccumulatorRecipe) ||
-            isa<VPPartialReductionRecipe>(AccumulatorRecipe)) &&
+    assert(isa<VPReductionPHIRecipe>(getOperand(1)->getDefiningRecipe()) &&
            "Unexpected operand order for partial reduction recipe");
   }
   ~VPPartialReductionRecipe() override = default;