Hello list,

I'd like some pointers on generating the RTL files for a processor I am interested in, the STM32L432KC (which is available for ~$15 with JTAG on a "Nucleo" board from STMicroelectronics).

The CMSIS (Cortex Microcontroller Software Interface Standard) files, as they come from STM, use structures to represent the registers. The example RTL files for STM devices seem to follow this pattern fairly well, but I would like to know about any discrepancies; I opened one file and think it was structured more closely to the way libopencm3 does things, but I can't find it again. This may have been the file for the NXP part listed on the Wiki.

How much was converted by hand, and how much can be automated? M4 devices are noticeably more complicated, and even though this is a hobby project I am worried about the time investment required to get my device working with FPC.

What complicates things is the way libopencm3 has their headers structures is more standard. They avoid using structures that represent the registers, instead using faux namespacing with lots of underscores in macro names.

Cheers,
R0b0t1

The process is completely automated and is based on converting the header files that come in the CMSIS packages of the processors.

I will send you the file for that chip via pm, you will also have to tweak compiler/systems/t_embed.pas compiler/arm/cpuinfo.pas but this is pretty straightforward, only extend both structs for the processors.

There is a second class of Headerfiles that were done half automated (afaik) by Jeppe Johansen that covers the STM32F7 series. Those Headers more closely match the STM32 code C-code examples but are a lot less portable to other chips (Microchip etc...)

Michael

What are you planing to implement? Do you need the RAM and FLASH of the stm32l432 or the low power features?

If not then I'd suggest to start wit a more simple CPU like the STM32F303K8 or, if you are okay with standard size nucleo boards the STM32F401RE or STM32F411RE are a good choice.

On the low energy chips the configuration is more demanding and for the other chips mentioned above there's already plenty of code available to re-use/re-purpose

Michael

Michael Ring wrote

The process is completely automated and is based on converting the header files that come in the CMSIS packages of the processors.

Excellent! What about the startup assembly files? I see an equivalent, it is autogenerated too?

I will send you the file for that chip via pm, you will also have to tweak compiler/systems/t_embed.pas compiler/arm/cpuinfo.pas but this is pretty straightforward, only extend both structs for the processors.

Yes, I forgot about this, but it does sound easy. Which chip?

Where are these?

I have some interest in exploring a wide variety of platforms with FPC. When doing so using C, I have unfortunately found that portability only half-exists between chips of the same family from the same manufacturer.

My interest in FPC is partly an interest in increased portability, but that may need to be achieved in some other way than a HAL. This may be due to how peripheral mappings are supplied. Perhaps there is a better way I do not know about. Large projects such as
https://github.com/qmk/qmk_firmware manage.

I will be following up with you off list, since you do not seem to mind.

Cheers,
R0b0t1

On 28 Feb 2018 4:22 am, "R0b0t1" wrote:

I will be following up with you off list, since you do not seem to mind.

I'm also interested in the general topic of incorporating embedded controllers (avr at the moment) into fpc. It would be useful to me if I could follow at least the general highlights for the porting process, not necessarily all the platform specific details.

Nope, those were created manually (from Jeppe), the one relevant for the STM32L432KC is cortexm4f_start.inc
It will automagically be included when you add definitions for the chip.

STM32L432KC

if you plan to some some serious development I can provide you a patch file for including the whole stm32l4 family

Take a look at github:

https://github.com/Laksen/fp-stm32f7xx_hal

great for stm32f7, but changing this to stm32l4 is a heroic effort ;-)

I have done my own platform independant lib for stm32(f0)/f1/f3/f4 and pic32 with a focus on low memory consumption, if you are interested I can put it on github, still some loose ends as I only have implemented what I needed.

I will be following up with you off list, since you do not seem to mind.

let's stay on list, there are others like Christo who may be interested.

Michael

I am planning serious development. If you can provide the patch that would be great, but if it is any amount of work I may be able to do it myself with some pointers.

It is my intent to stick with the L4 part I am using, but there have been many issues with it so far. Moving to an F3 or F4 may happen, but only after I attempt some actual use of the L4. Maybe eventually working libraries can be built up.

Right, I find myself needing to refer to STM's official HAL. Their normal documentation is lacking.

Please do, thanks.

I have uploaded my Code to github:

https://github.com/michael-ring/mbf.git

In the Patches directory you will find a patch to add all STM32L4 chips to a vanilla fpc-trunk installation.

Please pm me for questions or use github, documentation is still very sparse ;-)

The Blinky Example is prepared for the Nucleo-L432 Board, it does of course not compile successfully because initial code for switching CPU Frequency and accessing GPIO are missing for the L4 family, but this is easy to add, L4 is similar to F0 when I remember correctly.

Michael

Me too, especially STM32F407 & 446.

Well - as mentioned, it may be best to pick "standard" parts (did you have a preference?). This is an issue even with C or C++; many parts, especially from STM32, are very poorly documented. In C I am still having issues with my controller than I can only fix by copying, verbatim, the autogenerated code. At this point I suspect order of hardware initialization matters where no order dependency is documented.

This is kind of sad, because most of the popular STM32 parts are rather old. For other manufacturers the situation can be even worse. It is also sad because I want to try various products to see which is best (e.g. PIC32 parts are very cheap but still performant), but my experience as far as C goes is that all development environments and vendor provided libraries are terrible.

Also, thank you Michael, I may be able to give an update with my progress soon. It seems wise to iron out the initialization and hardware setup in C first.

Without much cheer,
R0b0t1

Whoops.

Both STM32F407 & 446 Chips are already in fpc-trunk, so no patches are needed to use those chips, mbf should work ok for you, I have a development kit for STM32F407VC (Mikroe EasyMXPro V7) and I should also have a NUCLEOF446RE Board somewhere here, so in case something does not work I may be able to help.

Michael

@R0b0t1:

FYI, I have now implemented Clock Configuration and gpio for stm32l4 chips, the blinky program works now, please pull latest mdf libs from github.

Why do you think STM32 chips are poorly documented?

The Reference Manuals are pretty good an the code generated by STM32CubeMX is very helpful to better understand the inner working of the Chips.

Michael

Thank you Michael, it may be some time before I can work with this, much to my dismay.

The generated code has been about the only thing that has helped me. Trying to use anything but the generated code tends to lead to lots of pain. Specifically the L4 (or just generally newer) parts seem to have lots of changes that break compatibility with existing projects or hardware documentation efforts.

I am lacking time, but what time I can devote to this isn't getting me very far. Hopefully some day :).

Cheers,
R0b0t1