Rather new to Elixir just played around with Saša Jurić book ( Erlang/OTP 27 Elixir 1.17.3 ) + (Livebook 1.18.1)
But now I want to get start coding with a IOT project .
BBB and RPi4 on the edge and AMD(ASUS) Ryzen 3 for now as Server
And I m just wondering what options are present to code with Rust and Elixir to contribute to both ecosystems
End goal is to build a more powerful but energy efficient ASUS/AMD system what is able to handle IOT workload from BBB and RPi4 edge devices
Maybe ST Nucleo as well should that be considered for Nerves
Only entry point I found so far was rustler_precompiled and I m just trying to compare it with PyO3/maturin for Python
Nerves runs Linux, so it doesn’t work on such limited boards.
rustler afaik works just fine on nerves, you can always compile on your own, no need to have precompiled packages. I’ve also run rust through ports on nerves.
Thanks for your clarification and I will give rustler a closer look
But in reference to ST Nucleo I believe they also got more powerful ones
Found NUCLEO-N657X0-Q and STM32N6570-DK and if I did not misread spec sheet both run with ARM Cortex m55
( https://uk.farnell.com/stmicroelectronics/nucleo-n657x0-q/development-brd-32bit-arm-cortex/dp/4587313 )
At least Zephyr should be available for ARM Cortex m55
ST even claimed that N6 system can run ML ( AI edge ) models 5x quicker then RPi4
But need to figure out details in connection with Nerves RPi4 or BBB will do for now
Even if you could get Linux running on a M55 (which it is not made for) you’d still have to build a nerves-system for a completely unsupported architecture which will require a deep understanding of things.
Thanks for your clarification
But I think I was misunderstood a bit
I do not have any intent to port nerves to a different platform .
But I knew that buildroot should work on ARM Cortex m55
Myself I m primary interested on workloads and never worked with buildroot.
Workloads on IOT but also on Server ( ASUS/AMD inclusive XILINX to be correct )
But concerning ST Nucleo I m really curios why ST claimes that their latest edge devices can do 5x better performance
compared to RPi4 with ML ( AI edge ) models https://www.st.com/resource/en/product_presentation/microcontrollers-stm32n6-series-overview.pdf
So to understand how Erlang,Elixir or rustler/ PyO3 deal with Coremark data would be of interest
Buildroot and Linux support MMU-less devices. This is not a direction taken by the Nerves core team. The GRiSP and AtomVM projects are better suited to running Erlang and Elixir on these devices. Possibly of interest, Buildroot recently integrated ARM FDPIC support which would help in any effort to run Nerves on an ARM-based no-MMU device.
I did have a look at AtomVM ( Release Notes — AtomVM 0.7.0-dev+git.a00098a0 documentation )
And I will try to get it running on my ST Nucleo ( NUCLEO H7A3ZI-Q )
as I believe it is close enough to the tested STM32F429ZI
But still do not understand the bit with ARM FDPIC support
I want to run Nerves primary for BBB and RPi2(4) not the NUCLEO-N657X0-Q mentioned
( its just a my purchase list for now )
But I m curios to get Buildroot with Vagrant running on Ubuntu 22.04 so that I m able to test Nerves for
BBB with different Capes
and GitHub - amd/aocl-data-analytics: AMD Optimized Data Analytics Library ( AMD Ryzen 3 and Ryzen 9 ) to test Elixir 1.18.1 on Server side
