[llvm] [RISCV][TTI] Implement instruction cost for vp.reduce.* (PR #114184)

[Elvis Wang via llvm-commits]

================
@@ -1191,6 +1191,40 @@ RISCVTTIImpl::getIntrinsicInstrCost(const IntrinsicCostAttributes &ICA,
     return getCmpSelInstrCost(Instruction::Select, ICA.getReturnType(),
                               ICA.getArgTypes()[0], CmpInst::BAD_ICMP_PREDICATE,
                               CostKind);
+  case Intrinsic::vp_reduce_add:
+  case Intrinsic::vp_reduce_fadd:
+  case Intrinsic::vp_reduce_mul:
+  case Intrinsic::vp_reduce_fmul:
+  case Intrinsic::vp_reduce_and:
+  case Intrinsic::vp_reduce_or:
+  case Intrinsic::vp_reduce_xor: {
+    std::optional<Intrinsic::ID> RedID =
+        VPIntrinsic::getFunctionalIntrinsicIDForVP(ICA.getID());
+    assert(RedID.has_value());
+    unsigned RedOp = getArithmeticReductionInstruction(*RedID);
+    if (RedOp == Instruction::FAdd || RedOp == Instruction::FMul)
+      return getArithmeticReductionCost(RedOp,
+                                        cast<VectorType>(ICA.getArgTypes()[1]),
+                                        ICA.getFlags(), CostKind);
+    return getArithmeticReductionCost(
+        RedOp, cast<VectorType>(ICA.getArgTypes()[1]), std::nullopt, CostKind);
----------------
ElvisWang123 wrote:

Fixed.
Before #114180 landed , passing FMF flags into `getArightmeicReductionCost()` with int reductions will always return the cost of `oredered fadd reduction`.

https://github.com/llvm/llvm-project/pull/114184