Do you have a specific DAG usage in mind for VPSHUFBITQMB? If not maybe wait off on the opcode and just use an intrinsic?
If it supports masking we can't use the intrinsic in the tablegen as it would go against our normal lowering of intrinsics.
intrinsics support (3/3)
Should this have SDTCisSameNumEltsAs<0,1> as well?
adopted a stricter type constraints on bits shuffle node, as suggested by Craig