On Wed May 24 11:09:03 2023 +0000, Giovanni Mascellani wrote:
Maybe you should skip `HALF` too, so that the dot product operation is emitted?
You are right, I forgot about `HALF`. Also, we may want to keep the lowering for `DOUBLE`s, but that made me realize it is perhaps better to change the name of the pass to `lower_non_float_dot()`.