AM625: About IPC (RemoteProc and RPMsg)

Hello Daisuke-san,

Links to documentation:

Please note that the RemoteProc driver used to initialize PRU cores and the RPMsg driver used to communicate with PRU cores is slightly different than the RemoteProc and RPMsg drivers used to initialize and communicate with M4F cores and R5F cores (for AM64x). So some details in the PRU documentation you linked will not apply to M4F and R5F.

In addition to the links you posted, I am also adding information on IPC to the AM62x & AM64x Linux Academies. So far IPC has been added to the Tour of TI Linux: https://dev.ti.com/tirex/explore/node?node=A__ACKtJpWz8b53RwT-PJ4qmA__linux_academy_am62x__XaWts8R__LATEST . However, I am also writing a "Multicore Development" section that will talk about IPC in more detail.

What is IPC? 

IPC just means "inter-processor communication". There are many different ways to implement IPC. The version of IPC that TI currently provides is called RPMsg. RPMsg is pretty easy to get working, but it may or may not be the best form of IPC for your usecase. It is up to the customer to determine if RPMsg is the right kind of IPC for their design, or if they want to implement a different kind of IPC.

Is TI's RemoteProc & RPMsg compliant with different frameworks? 

Yes, TI's implementation of RemoteProc & RPMsg are compliant with the Linux framework for PRU, M4F, and R5F.

OpenAMP is a different implementation of IPC. At this point, TI has released any OpenAMP code.

Is there anything else to keep in mind? 

1) A multicore system with IPC needs to make sure that if there are throughput or latency requirements for the IPC, or overall cycle time requirements, that all those requirements are met. See more discussion here: https://e2e.ti.com/support/processors-group/processors/f/processors-forum/1085663/faq-sitara-multicore-system-design-how-to-ensure-computations-occur-within-a-set-cycle-time

2) RPMsg only allows for 496 bytes of data to be passed per message (plus a 16 byte header, for 512 bytes total). What if we want to use a shared memory location to pass large amounts of data?

We are releasing a shared memory example where RPMsg is used to notify between cores within the next couple of weeks! Let me know if you want to be notified when the git repo becomes publicly available.

Regards,

Nick

Hello Daisuke-san,

The shared memory example is now live:
https://git.ti.com/cgit/rpmsg/rpmsg_char_zerocopy/

Le me know if you have any feedback or questions.

Revards,

Nick