STM32MP1

[patacongo]

Thoughts?  Should NuttX be playing in STM32MP1 platforms? https://www.st.com/en/microcontrollers-microprocessors/stm32mp1-series.html?ecmp=tt10609_gl_link_feb2019&querycriteria=productId=SS2003  There is a lot of press and, apparently, enthusiasm for the STM32MP1.  I just saw this one today:  https://hackaday.com/2019/02/27/new-part-day-the-stm32-that-runs-linux/

NuttX does support multi-core Cortex-A7 and it does support Cortex-M4 so all of the technology is in place to support a NuttX port to the STM32MP1.

Historically, however, NuttX has not been popular well on platforms designed for Linux support.  After all, if you have the resources to run Linux, then why would you not use Linux?  Sure, you could get by with fewer resources with NuttX, but I think most people feel that features offered by Linux outweigh the cost of the additional resources.

A NuttX Cortex-M4-only STM32MP1 would probably be a pretty easy thing.  But it is not a very satisfying project working with one of the many CPUs available.  And the CPUs are not independent of each other:  They share some start-up philosophy and must cooperate to be usable.  Most of the complexity would be in the AMP design to interact with the Cortex-A7's.  You would have to either adopt a standard like OpenAMP, or else have control of both sides of the interface.  Coordinating that with a separate Linux project would be a pain in the ass.

This chip is not so different then others in the i.MX family and no one has ever shown interest in NuttX with those.

So, it is hard to ignore a new STM32 family, but I don't see anyway to go forward with this one.

[patacongo]

Thoughts?  Should NuttX be playing in STM32MP1 platforms? https://www.st.com/en/microcontrollers-microprocessors/stm32mp1-series.html?ecmp=tt10609_gl_link_feb2019&querycriteria=productId=SS2003  There is a lot of press and, apparently, enthusiasm for the STM32MP1.  I just saw this one today:  https://hackaday.com/2019/02/27/new-part-day-the-stm32-that-runs-linux/

Looks like Octavo will be providing STM32MP1 parts in a module:  https://octavosystems.com/octavo_products/osd32mp15x/ , https://octavosystems.com/2019/02/25/stm32mp1-system-in-package/

[matous....@datavision.cz]

Hello,

I think in some specific cases could be nice to have a powerful real-time machine, i.e., ECU, airspace application, etc. I do not know the news in the real-time Linux kernel but a few years ago did not exist a common good solution. I think NuttX is more deterministic then Linux kernel with and without a real-time patch. I saw a few these real-time system based on RTEMS. NuttX could be alternative:)

Sony CXD5602 on Spresense dev board has 6 Cortex-M4 cores. Do you know how the communicate to each other? Is it the same technology OpenAMP as in STM32MP1?

Mat.

Dne středa 27. února 2019 23:38:06 UTC+1 patacongo napsal(a):

[Gregory Nutt]

> Sony CXD5602 on Spresense dev board has 6 Cortex-M4 cores. Do you know
> how the communicate to each other? Is it the same technology OpenAMP
> as in STM32MP1?

I personally don't know anything about the Spresense hardware. That is
not supported by upstream NuttX, but by a Sony-maintained fork of NuttX
(which I hope comes upstream in the near future). Perhaps it is this:
https://github.com/sonydevworld/spresense-nuttx

I had assumed that it used SMP vs AMP but I think you would have to ask
someone in one of the Spresense lists.

[pavel.ionut...@gmail.com]

Real-time maybe, Automotive/Airspace/Medical, probably not. RTEMS and the bunch are qualified, they are not used because they are "better" than x. You will be amazed on the required paperwork that must be done before putting some operating system inside a product that has a chance to inflict potential injury/endanger well-being. First of all, i like nuttx and linux, but they are nowhere near compliant with a safety process, and no management will put these inside a safety critical application unless they cover their ass with miles of paperwork, which is unlikely.