What do you guys think about having `Layout` as a member of `TensorType`? Currently `Type` basically means dtype and shape. I think it is very useful to have `Layout` there as well. If thats the case, the `Legalize` API will get arg_dtypes, and thus layout, enabling transformation based on input layouts.
This also means `infer_type` gets more complicated, as it will have to infer layout as well. However, I think it should be fine to have some Layouts as undefined/non-inferable (something like that). @tqchen @merrymercy @zhiics -- You are receiving this because you are subscribed to this thread. Reply to this email directly or view it on GitHub: https://github.com/dmlc/tvm/issues/3670#issuecomment-518884668
