posted 15 July 2005 17:23            

---

is it correct for a feedback network only to live on one processor?

for example: a signal path that starts with an input, runs through processing and then is written into memorywriter and which is added back into the input.

given that you can't break the entire process into branches that very well may be how it is, but i thought i'd check in and make sure that is correct.

my work around has to be been use IO routing, but that introduces latency which may or may not matter. on the plus side is routing via IO encourages me to incorporate more external processsing which is often a good thing.

IP: Logged

posted 15 July 2005 19:28            

---

You could try using this scheme to do cross-processor feedback loops:

<FeedbackLoopOutput + source> --> [some processining] --->FeedbackLoopInput

IP: Logged

All times are CT (US)