TMS320F28377D-EP: Shared memory arbitration - does fixed or round-robin incrment on subsequenct cycles?

Hi Karl,

Let's say that a global RAM block has been configured to allow full access from both sub-sytems (as above), and that the current round-robin state is for CPU1 accesses.

We can not have full access from both subsystem (CPU1 & CPU2) for global RAMs. Write and fetch access is allowed from one CPU only (based on master ownership). Only READ is allowed from both the CPU.

Since the arbitration logic is round robin for access from different masters, the order will depend on which master made the last access to that specific RAM block. E.g, if CPU1 READ was the previous access and CPU1/CPU2 read comes next in same cycle, CPU2 read will get priority.

Hope this is clear.

Regards,

Vivek Singh 

Hi Vivek,

Yes, sorry, 'full access' was the wrong term to use. I understand that only one sub-system can have read/CPU fetch/write access, whilst the other sub-system has only read access. Perhaps I should have said 'with no additional protections applied'.

You say "arbitration is round robin for access from different masters", but equally arbration is fixed-priority for the same master. So that statement alone doesn't answer my question.

But your example states if a CPU1 access is currently being processed, then the next access granted will be from CPU2, even if other CPU1 requests are queued.

So IMO, round robin appears to take priority. Forgive me if my terminology is poor.

Thanks for such a quick answer.

Karl