SIMD optimization for IIR filtering

Having 2 16-bit input channels with an actual full depth of 12 bits, I need the most efficient implementation of an IIR filter. ARM has created a SIMD implementation for a direct form 1 bi-quad cascade filtering. My question is there a clear and simple method for implementing an arbitrary order filter for a single channel/ two channels? For example, I would like to implement a 3rd order filter using SIMD