The libopencm3 project aims to create an open-source firmware library for various ARM Cortex-M microcontrollers.
Currently (at least partly) supported microcontrollers:
- ST STM32 F0xx/F1xx/F2xx/F30x/F37x/F4xx/F7xx/H7xx series
- ST STM32 G0xx G4xx L0xx L1xx L4xx series
- Atmel SAM3A/3N/3S/3U/3X series, as well as SAMDxx and friends
- NXP LPC1311/13/17/42/43
- Stellaris LM3S series (discontinued, without replacement)
- TI (Tiva) LM4F series (continuing as TM4F, pin and peripheral compatible)
- EFM32 Gecko series (only core support)
- Freescale Vybrid VF6xx
- Qorvo (formerly ActiveSemi) PAC55XX
- Synwit SWM050
- Nordic NRF51x and NRF52x
The library is written completely from scratch based on the vendor datasheets, programming manuals, and application notes. The code is meant to be used with a GCC toolchain for ARM (arm-elf or arm-none-eabi), flashing of the code to a microcontroller can be done using the OpenOCD ARM JTAG software.
The libopencm3 project is (and presumably, always will be) a work in progress. Not all subsystems of all microcontrollers are supported, yet, though some parts have more complete support than others.
Prior to version 0.8.0, the api was largely in flux. Attempts were made to provide backwards compatibility, but this was not always considered critical.
From 0.8.0 to 1.0, we'll atempt to follow semver, but EXPECT CHANGES, as we attempt to clear up old APIs and remove deprecated functions. The 0.8.0 tag was placed to provide the "old stable" point before all the new code started landing.
preview code often lands in the "wildwest-N" branches that appear and disappear in the repository. Pull requests marked as "merged-dev" will be in this branch, and will be closed when they merge to master. This is useful for bigger interdependent patch sets, and also allows review of merge conflicts in public.
From 1.0, expect to follow semver, with functions (and defines!) being deprecated for a release before being removed.
TIP: Include this repository as a Git submodule in your project to make sure your users get the right version of the library to compile your project. For how that can be done refer to the libopencm3-template repository.
Building requires Python (some code is generated).
For Ubuntu/Fedora:
- An arm-none-eabi/arm-elf toolchain.
For Windows:
Download and install:
- msys - http://sourceforge.net/projects/mingw/files/MSYS/Base/msys-core/msys-1.0.11/MSYS-1.0.11.exe
- Python - https://www.python.org/downloads/windows/ (any release)
- arm-none-eabi/arm-elf toolchain (for example this one https://launchpad.net/gcc-arm-embedded)
Run msys shell and set the path without standard Windows paths (adjusting to your version of Python), so Windows programs such as 'find' won't interfere:
export PATH="/c//Program Files/Python 3.9:/c/ARMToolchain/bin:/usr/local/bin:/usr/bin:/bin"
After that you can navigate to the folder where you've extracted libopencm3 and build it.
The most heavily tested toolchain is "gcc-arm-embedded" This used to be available at https://launchpad.net/gcc-arm-embedded
Other toolchains should work, but they have not been nearly as well tested. Toolchains targeting Linux, such as "gcc-arm-linux-gnu" or the like are not appropriate.
NOTE: GCC version 6 or later is required, as we're using attributes on enumerators to help mark deprecations.
$ make
If you have an arm-elf toolchain (uncommon) you may want to override the toolchain prefix (arm-none-eabi is the default)
$ PREFIX=arm-elf make
For a more verbose build you can use
$ make V=1
You can reduce the build time by specifying a particular MCU series
$ make TARGETS='stm32/f1 stm32/f4'
Supported targets can be listed using:
$ make list-targets
The build may be fine-tuned with a limited number of parameters, by specifying them as environment variables, for example:
$ VARIABLE=value make
-
FP_FLAGS
- Control the floating-point ABIIf the Cortex-M core supports a hard float ABI, it will be compiled with the best floating-point support by default. In cases where this is not desired, the behavior can be specified by setting
FP_FLAGS
.Currently, M4F cores default to
-mfloat-abi=hard -mfpu=fpv4-sp-d16
, and M7 cores defaults to double precision-mfloat-abi=hard -mfpu=fpv5-d16
if available, and single precision-mfloat-abi=hard -mfpu=fpv5-sp-d16
otherwise. Other architectures use no FP flags, in otherwords, traditional softfp.You may find which FP_FLAGS you can use in a particular architecture in the readme.txt file shipped with the gcc-arm-embedded package.
Examples:
$ FP_FLAGS="-mfloat-abi=soft" make # No hardfloat $ FP_FLAGS="-mfloat-abi=hard -mfpu=magic" make # New FPU we don't know of
-
CFLAGS
- Add to or supersede compiler flagsIf the library needs to be compiled with additional flags, they can be passed to the build system via the environment variable
CFLAGS
. The contents ofCFLAGS
will be placed after all flags defined by the build system, giving the user a way to override any default if necessary.Examples:
$ CFLAGS="-fshort-wchar" make # Compile lib with 2 byte wide wchar_t
The libopencm3 community has written and is maintaining a huge collection of examples, displaying the capabilities and uses of the library. You can find all of them in the libopencm3-examples repository:
https://github.com/libopencm3/libopencm3-examples
If you just wish to test your toolchain and build environment, a collection of mini blink projects is available too. This covers many more boards, but, as the name suggests, only demonstrates blinking LEDs.
https://github.com/libopencm3/libopencm3-miniblink
Simply pass -I and -L flags to your own project. See the libopencm3-template repository for a template repository using this library as a Git submodule, the most popular method of use. The libopencm3-examples is another example of this.
It is strongly advised that you do not attempt to install this library to any
path inside your toolchain itself. While this means you don't have to include
any -I
or -L
flags in your projects, it is very easy to confuse a multi-library
linker from picking the right versions of libraries. Common symptoms are
hardfaults caused by branches into ARM code. You can use arm-none-eabi-objdump
to check for this in your final ELF file. You have been warned.
See HACKING.
The libopencm3 code is released under the terms of the GNU Lesser General Public License (LGPL), version 3 or later.
See COPYING.GPL3 and COPYING.LGPL3 for details.
-
Developer mailing list (for patches and discussions): https://lists.sourceforge.net/lists/listinfo/libopencm3-devel
-
Commits mailing list (receives one mail per
git push
): https://lists.sourceforge.net/lists/listinfo/libopencm3-commits
- http://libopencm3.org - contains daily autogenerated API documentation