Robin- > Now we have ported PJSIP to the MIPS embedded platform under Linux > environment, and now we want to integrate the PJSIP stack with our DSP > device since there's no sound device on the board. Our DSP has the ability > to pack data to RTP and decode from RTP flow, in that case, what's the best > practice should we take? The codec takes the real work, so simply call encode() and decode() functions using your DSP. Echo-can maybe is another function to move there at some point, if needed... In the tightly coupled architecture that you describe (similar to an SoC), where the MIPS CPU has the IP/UDP interface and the DSP is "attached" to the MIPS in some way (shared mem, FIFO, etc), it makes sense to only run "compute intensive" code on the DSP, and keep everything else on the MIPS. -Jeff