Stream processing has two outstanding aspects.
One is arithmetic, another is bandwidth. Arithmetic intensity, which is the ratio of arithmetic to bandwidth,
should be improved as soon as possible.
One way to raise arithmetic intensity is multi-stage memory structure and specific computing path [10].
This way has two problems, its need huge amount of registers to build SRF and flexibility will be poorer while pursuing
locality.
This section explores not only locality, but also flexibility of processor by considering processing efficiency and general aspects of processing.
We observe and summarize multi-stream processing of communication standards at first.
Then a reasonable stream storage subsystem and specific data paths which can accelerate the processing are put up. Lastly we construct a processor by putting both of these subsystems into a host system.