> I was merely suggesting a better comment describing the function you are > introducing. Oh. I know. Good wording.
> Thinking about it a bit more, I think you simply do not want to ever > push the extra VIEW_CONVERT_EXPR to the vector and in > evaluate_conditions_for_known_args always do a fold_convert to the > desired type (similarly like we do today). Yes. It is, but at cost of creating an extra memory slot. Feng
