[llvm] [LV] Cache call vectorization decisions (PR #66521)

[David Sherwood via llvm-commits]

================
@@ -6970,6 +6968,122 @@ void LoopVectorizationCostModel::setCostBasedWideningDecision(ElementCount VF) {
   }
 }
 
+void LoopVectorizationCostModel::setVectorizedCallDecision(ElementCount VF) {
+  if (VF.isScalar())
+    return;
+
+  for (BasicBlock *BB : TheLoop->blocks()) {
+    // For each instruction in the old loop.
+    for (Instruction &I : *BB) {
+      CallInst *CI = dyn_cast<CallInst>(&I);
+
+      if (!CI || CallWideningDecisions.contains(std::make_pair(CI, VF)))
+        continue;
+
+      InstructionCost ScalarCost = InstructionCost::getInvalid();
+      InstructionCost VectorCost = InstructionCost::getInvalid();
+      InstructionCost IntrinsicCost = InstructionCost::getInvalid();
+      TTI::TargetCostKind CostKind = TTI::TCK_RecipThroughput;
+
+      Function *ScalarFunc = CI->getCalledFunction();
+      Type *ScalarRetTy = CI->getType();
+      SmallVector<Type *, 4> Tys, ScalarTys;
+      bool MaskRequired = Legal->isMaskRequired(CI);
+      for (auto &ArgOp : CI->args())
+        ScalarTys.push_back(ArgOp->getType());
+
+      // Compute corresponding vector type for return value and arguments.
+      Type *RetTy = ToVectorTy(ScalarRetTy, VF);
+      for (Type *ScalarTy : ScalarTys)
+        Tys.push_back(ToVectorTy(ScalarTy, VF));
+
+      // An in-loop reduction using an fmuladd intrinsic is a special case;
+      // we don't want the normal cost for that intrinsic.
+      if (RecurrenceDescriptor::isFMulAddIntrinsic(CI))
+        if (auto RedCost = getReductionPatternCost(CI, VF, RetTy, CostKind)) {
+          setCallWideningDecision(CI, VF, CM_IntrinsicCall, nullptr,
+                                  getVectorIntrinsicIDForCall(CI, TLI),
+                                  std::nullopt, *RedCost);
+          return;
----------------
david-arm wrote:

Do you really want to return here? Shouldn't this be `continue` instead? We might have both a fmuladd intrinsic and a sin call in the loop, for example. Is it worth adding a test case for this somehow, since clearly no tests are failing because of it?

https://github.com/llvm/llvm-project/pull/66521