[llvm] [VPlan] Separate out logic to manage IR flags to VPIRFlags (NFC). (PR #140621)

[via llvm-commits]

================
@@ -855,24 +845,11 @@ bool VPInstruction::isSingleScalar() const {
          getOpcode() == Instruction::PHI;
 }
 
-#if !defined(NDEBUG)
-bool VPInstruction::isFPMathOp() const {
-  // Inspired by FPMathOperator::classof. Notable differences are that we don't
-  // support Call, PHI and Select opcodes here yet.
-  return Opcode == Instruction::FAdd || Opcode == Instruction::FMul ||
-         Opcode == Instruction::FNeg || Opcode == Instruction::FSub ||
-         Opcode == Instruction::FDiv || Opcode == Instruction::FRem ||
-         Opcode == Instruction::FCmp || Opcode == Instruction::Select ||
-         Opcode == VPInstruction::WideIVStep;
-}
-#endif
-
 void VPInstruction::execute(VPTransformState &State) {
   assert(!State.Lane && "VPInstruction executing an Lane");
   IRBuilderBase::FastMathFlagGuard FMFGuard(State.Builder);
-  assert((hasFastMathFlags() == isFPMathOp() ||
-          getOpcode() == Instruction::Select) &&
-         "Recipe not a FPMathOp but has fast-math flags?");
+  assert(flagsValidForOpcode(getOpcode()) &&
+         "Set flags not supported for the provided opcode");
----------------
ayalz wrote:

Consistency: if desired both at construction and at execution (considering flags as mutable), should `VPWidenCastRecipe::execute()` also assert?

https://github.com/llvm/llvm-project/pull/140621