massive jit.net flow vs soundflower routing audio ?

Hi there,
I have couple of ways to do what I want to do.
I only tried one but I'd like to know, globally, which one could be the less worst approach if not the best ;)

My system is based on jit.catch~ popping out matrix to the visual generator of my system.

Options could be (non exhaustive of course):
- jit.catch~ in M4L popping out matrices via jit.net. to visual generator
- soundflower (64 channels version) routing audio into visual generator that also contains jit.catch~
- all in M4L (never my preferred approaches)

I have no more but no less than 16 channels

Both first seems efficient, but I'd like to have your opinions here.

Thanks.