Author Topic: Comparing the Heterogeneous/Asymmetric MPU options  (Read 2637 times)

0 Members and 1 Guest are viewing this topic.

Offline tmadness

  • Regular Contributor
  • *
  • Posts: 80
  • Country: us
Comparing the Heterogeneous/Asymmetric MPU options
« on: December 17, 2021, 07:09:04 pm »
I have been scoping the market for Heterogeneous/Asymmetric MPUs. My workload is real-time enough to justify a pretty fast separate MCU, however due to large amounts of data, and interfacing reasons I need a high level OS. The "traditional" option is to put a FPGA and a standard MPU, however, my expertise lies more in the MCU side of stuff and the real time aspects of my project don't need anything specially from the FPGA.
My solution, therefore, is to go for a integrated, "low development cost" alternative in the form of a MPU with a Heterogeneous/Asymmetric architecture (cortex A + cortex M typically). There seem to be various options available from all the big players, but two seem to really peek my interest: the TI Sitara AM572X (AKA beagle bone AI) and NXP IMX 8 Plus. There are advantages on both sides, eg TI has more real-time cores, NXP has more recent powerful cortex A cores etc.
But I'd like to know how quality the development ecosystem for these processors are.
How is the documentation? I have worked on the NXP RT MCUs and the documentation is decent, although some of their peripherals get really complicated.
How similar or dissimilar are the peripherals on the NXP IMX 8's cortex M7 compared to the IMX RT series?
How well have the data sharing / communications interfaces been implemented between the cores? I plan on using RPMsg and shared memory how has your experience been with these use cases?
Anything else I should know or care about?

Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo