Hello,everyone:
If I want to port lwip to octeon processors running as simple executive progress , which software architecture is preferred , run-to-complete or pipeline ?
I have learned about that run-to-completion is more intuitive and
less complicated. It's also more flexible because there are
no dependencies between processing of different pipestages. While pipeline is supposed to have higher performance and more modular.
If pipeline is used , Is it proper that divide lwip process procedure into several stages according to the tcp/ip layer model?
If run-to-completion is used , is it a must that after each stage(one of tcp/ip layer) finish processing , it switch tags ,re-queued and de-sheduled?
Regards,
Tom