search

libhal-google / libhal-armcortex

A collection of drivers and libraries for the Cortex M series ARM processors using libhal
Apache License 2.0
5 stars 3 forks source link
arm cortex cortex-m embedded embedded-systems microcontroller

readme

libhal-armcortex

✅ Checks Coverage Codacy Badge GitHub stars GitHub forks GitHub issues

libhal-armcortex is a processor library in the libhal ecosystem. It provides a set of drivers and functions for ARM Cortex processors, allowing developers to write portable and efficient embedded software.

📚 Software APIs & Usage

To learn about the available drivers and APIs see the libhal cortex_m APIs documentation page or look at the include/libhal-armcortex directory.

🧰 Setup

Following the 🚀 Getting Started instructions.

📡 Installing Profiles

Profiles define which platform you mean to build your project against. These profiles are needed for code and demos in this repo and for applications that wish to execute on an lpc40 device.

conan config install -sf conan/profiles/ -tf profiles https://github.com/libhal/libhal-armcortex.git

Note that running these functions is safe. THey simply overwrite the old files with the latest files. So running this for libhal-armcortex between this and other platform libraries is fine.

🏗️ Building Demos

To build demos, start at the root of the repo and execute the following command:

conan build demos -pr cortex-m4f -s build_type=Debug

This will build the demos for the cortex-m4f microcontroller in Debug mode. Replace cortex-m4f with any of the other profiles. Available profiles are:

🏁 Startup & Initialization

Startup function to initialize the data section. This is REQUIRED for systems that do not load themselves from storage into RAM. If the executable is executed from flash then hal::cortex_m::initialize_data_section() is required.

#include <libhal-armcortex/startup.hpp>

hal::cortex_m::initialize_data_section();

The arm-gnu-toolchain package provides the crt0.s startup file which initializes the BSS (uninitialized) section of memory. If this startup file is not used, then a call to hal::cortex_m::initialize_bss_section() is required.

#include <libhal-armcortex/startup.hpp>

hal::cortex_m::initialize_data_section();
hal::cortex_m::initialize_bss_section();

If the device has an FPU (floating point unit) then a call to hal::cortex_m::initialize_floating_point_unit() is required before any floating point unit registers or floating point instructions.

#include <libhal-armcortex/system_control.hpp>

hal::cortex_m::initialize_floating_point_unit();

Using conan/profiles

Learn more about Conan profiles in the Conan documentation.

The conan/profiles directory contains a set of predefined profiles for different ARM Cortex-M cores. These profiles are used to specify settings for conan in order to either compiler the library for a specific target or provide information to an application about which binaries to download from the conan package repository. Each profile sets the appropriate compiler flags for the corresponding core.

The profiles labeled thumbvX where X is the thumb instruction set version number. For example, the thumbv7 profile defines the mininum compiler flags for the cores capable of using thumbv7 such as Cortex-M3 and Cortex-M4. While the thumbv8 profile sets the compiler flags for the ARM Cortex-M23 and Cortex-M33 cores. Here is an example profile for thumbv7:

[settings]
compiler=gcc
compiler.cppstd=20
compiler.libcxx=libstdc++
compiler.version=12.3
os=baremetal
arch=thumbv7
arch.float_abi={{ float_abi }}
arch.fpu={{ fpu }}
arch.processor={{ cpu }}

[tool_requires]
arm-gnu-toolchain/12.2.1

[conf]
tools.build:cflags=["-mfloat-abi={{ float_abi }}", "-mcpu={{ cpu }}", "-mfpu={{ fpu }}", "-fno-exceptions", "-fno-rtti"]
tools.build:cxxflags=["-mfloat-abi={{ float_abi }}", "-mcpu={{ cpu }}", "-mfpu={{ fpu }}", "-fno-exceptions", "-fno-rtti"]
tools.build:exelinkflags=["--specs=nano.specs", "--specs=nosys.specs", "-mfloat-abi={{ float_abi }}", "-mcpu={{ cpu }}", "-mfpu={{ fpu }}", "-fno-exceptions", "-fno-rtti"]

Conan uses jinja2 templates to fill in the areas marked with {{ <variable name> }}. The complete profiles such as cortex-m0, cortex-m3, and cortex-m4f set the appropriate flags and import their respective thumbvX profile. For example lets take a look at cortext-m0 and cortex-m4f.

{% set cpu = "cortex-m0" %}
{% include "thumbv6" %}
{% set cpu = "cortex-m4" %}
{% set float_abi = "hard" %}
{% set fpu = "auto" %}
{% include "thumbv7" %}

When cortex-m4f is selected as the conan profile, the contents of the profile are expanded and all templates are replaced, allowing a library or application to be compiled for that processor.

Also notice how cortex-m0 does not have a field for float_abi? Thats because thumbv6 do not support floating point units and their instructions, so the floating point ABI compiler flags can be set to soft by default.

Supporting new Processors

To add additional cortex-mX processors, simply make a new file in this repo's conan/profile directory with the name of the processor following the format cortex-mX where X is the processor number. Modifiers can be added to the end to specify additional details like f for single precisions floating point hardware support, d for double precision floating point hardware support, and neon for devices that support the neon floatign point unit. Follow the same template as the other profiles and make sure that every jinja2 template is accounted for in the variable set before using the {% include "thumbvY" %} line.

What if the thumbv profile you need doesn't exist? Simply take the others as reference and add whichever fields are necessary to support that instruction set.

Creating platform profiles

Platform profiles take a processor profile and combine it with platform specific details. These details must include setting the global platform option and any other profile related details that can be used for a build.

Lets consider two examples, the lpc4078 profile.

lpc4078 has a profile fragment called lpc40 which specifies generic info.

[options]
*:platform={{ platform }}

[buildenv]
LIBHAL_PLATFORM={{ platform }}
LIBHAL_PLATFORM_LIBRARY=lpc40

lpc4078's profile looks like this:

{% include "cortex-m4f" %}
{% set platform = "lpc4078" %}
{% include "lpc40" %}

The lpc4078 microcontroller uses a cortex-m4f processor, so that is included in this profile. Next we set the platform variable. All libhal platform profiles MUST have a platform name specified in their [options] section of their profile. And finally the profile includes the lpc40 which provides the final portion of the profile.

Note that the conan profiles always override duplication information with the last occurrences of the duplicate.

A profile should be made for every supported microcontroller in a platform library's chip family.

By creating and using these profiles, you can easily compile the library for different platforms and MCUs while ensuring the correct compiler flags are used.

How Linker Scripts Work

Linker scripts are used to control the memory layout of the final binary file. They define the memory regions and sections of the binary, and specify where and how the linker should place the code and data.

In libhal-armcortex, there is a standard linker script named standard.ld located in the linker_scripts/libhal-armcortex directory. This script is included by other linker scripts in the library to provide a common base configuration. When lihal-armcortex is added as a dependancy of a library or application the path to the linker scripts in linker_scripts directory are added to the linker flags of the build, making them accessible within other linker scripts via the INCLUDE "libhal-armcortex/standard.ld command.

Here is the content of standard.ld:

INCLUDE "libhal-armcortex/third_party/standard.ld"

This script includes another script third_party/standard.ld which contains the actual linker commands. This script has 4 variables that must be defined for the linker script to work as intended. These variables are

Currently, libhal only provides standard.ld which supports a single memory mapped flash and ram block.

Additional linker scripts for multi-ram, multi-flash, and execute from RAM only systems are planned to be provided at a later date when systems with those requirements appear in the ecosystem.

Using Linker Scripts in a Platform Library

Linker scripts are used in platform libraries to define the memory layout specific to the platform. For example, in the libhal-lpc40 library, there are several linker scripts in the linker_scripts/libhal-lpc40 directory. Each of these scripts corresponds to a specific model of the lpc40 series of MCUs.

Here is an example of the lpc4076.ld linker script from libhal-lpc40:

__flash = 0x00000000;
__flash_size = 256K;
__ram = 0x10000000;
__ram_size = 64K;
__stack_size = 1K;

INCLUDE "libhal-armcortex/standard.ld"

This script defines the memory layout for the lpc4076 MCU. It specifies the start addresses and sizes of the flash and RAM memory regions, as well as the stack size. It then includes the standard linker script from libhal-armcortex to provide the common base configuration.

To create a new platform profile, you would create a new linker script similar to this one, but with the memory layout specific to your platform. You would then use this linker script when building your platform library.

Every supported microcontroller in the platform library should have an associated linker script with its name like so platform_name.ld. The platform library should have a package_info() section like this in their conanfile.py:

def package_info(self):
    self.cpp_info.set_property("cmake_target_name", "libhal::lpc40")
    self.cpp_info.libs = ["libhal-lpc40"]

    if self._bare_metal and self._is_me:
        linker_path = os.path.join(self.package_folder, "linker_scripts")
        link_script = "-Tlibhal-lpc40/" + str(self.options.platform) + ".ld"
        self.cpp_info.exelinkflags = ["-L" + linker_path, link_script]

This will add the platform linker scripts to the linker script flags. The self._bare_metal property determines if the OS is equal to baremetal. The self._is_me property determines if the platform option is one of the microcontrollers supported by this library. This together check if the system is being cross built and targeting a platform we have a linker script for. In other cases, if this isn't a bare metal build, for example a build for unit testing, then we want to ensure that the linker scripts are not added to the linker flags. Will cause an error because the linkers for applications on a OS like linux or mac will not match the one in the bare metal case.

Customizing Profiles & Linker Scripts

In libhal, customization of profiles and linker scripts is a straightforward process. This allows developers to tailor the build process to their specific needs and hardware configurations.

To customize a profile, you need to create a new profile. You have the option to simple include one of the platform's pre-made profile files. Once you finished making your profile, you then, set the platform option name to anything other than the available platforms in the library. It's important to note that Conan profiles always prioritize and overwrite the fields based on the last one defined. This means that any settings in your custom profile will overwrite the corresponding settings in the included platform profile.

For example, the libhal-lpc40 library has several predefined platforms: lpc4072, lpc4074, lpc4076, lpc4078, and lpc4088. If you want to customize the build process for this library, you could create a new profile named lpc40xx. This profile would include the lpc40 platform profile and set the platform option to lpc40xx.

include(lpc4078)

[options]
platform=lpc40xx

Setting the platform option to lpc40xx causes the recipe to skip adding a linker to the linker flags. This allows you, as the developer, to choose your own linker script for your build. This is required when the platform option is set to a value not recognized by the library.

By customizing profiles and linker scripts in this way, you can easily adapt the build process to different hardware configurations and project requirements.

Contributing

See CONTRIBUTING.md for details.

License

Apache 2.0; see LICENSE for details.

Disclaimer

This project is not an official Google project. It is not supported by Google and Google specifically disclaims all warranties as to its quality, merchantability, or fitness for a particular purpose.