[llvm] [RISCV] Don't cost vector arithmetic fp ops as cheaper than scalar (PR #99594)

[Luke Lau via llvm-commits]

================
@@ -1739,8 +1738,14 @@ InstructionCost RISCVTTIImpl::getArithmeticInstrCost(
                                                            Op1Info, Op2Info,
                                                            Args, CxtI);
   }
-  return ConstantMatCost +
-         LT.first * getRISCVInstructionCost(Op, LT.second, CostKind);
+
+  InstructionCost InstrCost = getRISCVInstructionCost(Op, LT.second, CostKind);
+  // We use BasicTTIImpl to calculate scalar costs, which assumes floating point
+  // ops are twice as expensive as integer ops. Do the same for vectors so
+  // scalar floating point ops aren't cheaper than their vector equivalents.
+  if (Ty->isFPOrFPVectorTy())
----------------
lukel97 wrote:

But putting the doubling logic in getRISCVInstructionCost means we would end up doubling the cost everywhere, I think we only need to double it in RISCVTTIImpl::getArithmeticInstrCost where we invoke BasicTTIImpl:: getArithmeticInstrCost?

https://github.com/llvm/llvm-project/pull/99594