[LoopVectorizer][AArch64] Add support for partial reduce subtraction

Author: NickGuy-Arm

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/VPlanRecipes.cpp

@@ -317,13 +317,20 @@ 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");
 
+  unsigned Opcode = getOpcode();
+
+  if (Opcode == Instruction::Sub) {
+    bool HasNSW = cast<Instruction>(BinOpVal)->hasNoSignedWrap();
+    BinOpVal = Builder.CreateNeg(BinOpVal, "", HasNSW);
+    Opcode = Instruction::Add;
+  }
+
+  assert(Opcode == Instruction::Add && "Unhandled partial reduction opcode");
+
   Type *RetTy = PhiVal->getType();
 
   CallInst *V = Builder.CreateIntrinsic(