Concatenate two floating-point registers, shift the result right by t_shift elements, and return the result.
t_shift