This allows the peephole optimizer to know that a MVE_VMOV_to_lane_32 is the same as an insert subreg, allowing it to optimize some redundant lane moves.
Thanks for taking a look.
VSETLNi32 handles D regs, so ssub_0 or ssub_1. Because MVE_VMOV_to_lane_32 is working on Q regs it can be ssub_2 or ssub_3 as well. There are other places in llvm that assume that ssub_0 + x == ssub_x, so that seems safe to use.