1) Is it worth it. Maybe, I try to avoid the temptation to implement 2 processors because i have run out of "freeboard". Instead i look for some algorithmic advantage from 2 asynchronous computational elements. If not i bump up the processing power. Analog's DSP are easy to implement (especially if your a microcontroller type) and inexpensive.

2)Pipelining is where a "work" propagates through a sequential chain of operation specific stages, each cell performing an sub-component of a larger algorith and then passes it on to the next stage.

Typically these stages are "registered" so that upon each major "clock" cycle the data transfers stage to stage. (if they are asynch the stages must be arranged such that the longest executing stage preceeds a stage of shorter execution time).

The advantage of a registered pipeline is that once it is "full" (each stage provided data) work emits from the pipeline each major clock cycle. So even if 12 operations are required after the initial latency of 12 stages data emits in 1/12th the time required if the work was done procedurally.

Pipelining is best suited for repeating non-conditional algorithms.

regards,
p