On Wed, Nov 09, 2022 at 04:43:56PM +0100, Aldy Hernandez wrote:
> Heh. I just ran into this while debugging the problem reported by Xi.
>
> We are solving NAN = op1 - VARYING, and trying to do it with op1 = NAN
> + VARYING, which returns op1 = NAN (incorrectly).
>
> I suppose in the above case op1 should ideally be
> [-INF,-INF][+INF,+INF]+-NAN, but since we can't represent that then
> [-INF,+INF] +-NAN, which is actually VARYING. Do you agree?
Yes.
For implementation, perhaps just on the result of the reverse ops
of whatever you do there currently (right before returning)
call some function common to all binary ops that will handle
the reverse op handling of NANs I've described.
If lhs can't be NAN, then clear_nan on the result, if
lhs is known to be NAN and other op can't be NAN,
have result be known NAN (with VARYING sign),
otherwise set maybe +-NAN on the result.
Jakub
