HI Ti team,
I am working on the AM64x platform with communication between A53 Core0 (Linux) and R5F Core0 (MCU+ SDK).
Currently, we are using the standard RPMessage/RPMsg framework for data transfer. However, our application requires higher throughput and lower latency for exchanging large amounts of data between the A53 and R5F.
Our target is to achieve:
- Transfer size: 1 KB
- Transfer latency: 50 µs (A53 -> R5F, R5F-> A53)
Could you please advise:
- What is the recommended IPC framework for achieving the best performance between A53 Core0 and R5F Core0?
- Is RPMessage/RPMsg still the recommended approach, or should we use a shared-memory (zero-copy) mechanism with mailbox notifications?
- Does TI provide any reference implementation or example project for a zero-copy framework on AM64x?
- Are there any recommended optimizations (e.g., shared memory configuration, VirtIO/RPMsg tuning, mailbox usage, DMA) to improve throughput beyond the standard RPMsg implementation?
- Are there any benchmark results comparing standard RPMsg with a shared-memory/zero-copy implementation?
I am using SDK 09.02.01.10 RT-LINUX and for R5F MCU+SDK 10_01_00_32 Version and I want to optimize the IPC by adapting the new framework to be impacted in the same given SDK versions...Kindly advise on this..
Any recommendations or reference examples would be greatly appreciated.
Thank you.
Ganesh D
