On 4/26/22 05:26, Giovanni Mascellani wrote:
They would interfere with the following add_array_load() patch, because they would cause indices to be of type uint1 instead of uint.
I've signed off on this because I think it's valuable on its own merits, but I think the correct solution here is that we shouldn't care about SCALAR versus VECTOR types past hlsl.y.
I.e. in this instance, we should be using is_vec1() in hlsl_offset_from_deref().