Just to circle back here a bit. the main root issue is that we are using runtime::DataType, which is supposely being concrete through out the TIR node.
This places restrictions on what we can normally represent. A more comprehensive update would change the PrimExpr's field to also an object, as per StructInfo in the relax. That would requires bit more thinking, which likely can get around the issues mentioned in the thread(of passing around runtime::DataType which is not an object). I think in short term making the protocol of lanes = -1 and lanes = -8(for vscale(8)) may not be a bad idea. The main reason is I cannot think of another possible use of the lanes field other than for the SVE. -- Reply to this email directly or view it on GitHub: https://github.com/apache/tvm-rfcs/pull/104#issuecomment-1845530627 You are receiving this because you are subscribed to this thread. Message ID: <apache/tvm-rfcs/pull/104/c1845530...@github.com>
