Home

Awesome

Real-Time-C++

<p align="center"> <a href="https://github.com/ckormanyos/real-time-cpp/actions"> <img src="https://github.com/ckormanyos/real-time-cpp/actions/workflows/real-time-cpp.yml/badge.svg" alt="Build Status"></a> <a href="https://github.com/ckormanyos/real-time-cpp/actions"> <img src="https://github.com/ckormanyos/real-time-cpp/actions/workflows/real-time-cpp-examples.yml/badge.svg" alt="Build Examples"></a> <a href="https://github.com/ckormanyos/real-time-cpp/actions"> <img src="https://github.com/ckormanyos/real-time-cpp/actions/workflows/real-time-cpp-snippets.yml/badge.svg" alt="Build Snippets"></a> <a href="https://github.com/ckormanyos/real-time-cpp/actions"> <img src="https://github.com/ckormanyos/real-time-cpp/actions/workflows/real-time-cpp-benchmarks.yml/badge.svg" alt="Build Benchmarks"></a> <a href="https://github.com/ckormanyos/real-time-cpp/issues?q=is%3Aissue+is%3Aopen+sort%3Aupdated-desc"> <img src="https://custom-icon-badges.herokuapp.com/github/issues-raw/ckormanyos/real-time-cpp?logo=github" alt="Issues" /></a> <a href="https://github.com/ckormanyos/real-time-cpp/actions/workflows/codeql.yml"> <img src="https://github.com/ckormanyos/real-time-cpp/actions/workflows/codeql.yml/badge.svg" alt="CodeQL"></a> <a href="https://scan.coverity.com/projects/ckormanyos-real-time-cpp"> <img src="https://scan.coverity.com/projects/24862/badge.svg" alt="Coverity Scan"></a> <a href="https://sonarcloud.io/summary/new_code?id=ckormanyos_real-time-cpp"> <img src="https://sonarcloud.io/api/project_badges/measure?project=ckormanyos_real-time-cpp&metric=alert_status" alt="Quality Gate Status"></a> <a href="https://github.com/ckormanyos/real-time-cpp/blob/master/LICENSE_1_0.txt"> <img src="https://img.shields.io/badge/license-BSL%201.0-blue.svg" alt="Boost Software License 1.0"></a> </p>

This is the companion code for the book C.M. Kormanyos, Real-Time C++: Efficient Object-Oriented and Template Microcontroller Programming, Fourth Edition (Springer, Heidelberg, 2021) ISBN 9783662629956.

This repository has several main parts.

GNU/GCC cross compilers and various additional tools running on Win*, optionally needed for certain builds as described below, can be found in the related ckormanyos/real-time-cpp-toolchains repository.

Details on the Reference Application

The reference application boots with a small startup code and subsequently initializes a skinny microcontroller abstraction layer (MCAL). Control is then passed to a simple multitasking scheduler that manages the LED application, calls a cyclic benchmark task and services the watchdog.

The LED application toggles a user-LED with a frequency of $\frac{1}{2}~\text{Hz}$ The result is LED on for one second, LED off for one second. The LED application runs cyclically and perpetually without break or pause.

The reference application is compatible with C++14, 17, 20, 23 and beyond.

Portability

The application software is implemented once and used uniformly on each supported target in the ref_app. Differences among the individual targets arise only in the lower software layers pertaining to chip-specific and board-specific startup/MCAL details.

In this way the project exhibits a high level of portability.

Supported Targets in the Reference Application

The reference application supports the following targets:

Target name (as used in build command)Target Description*(breadboard)
avrMICROCHIP(R) [former ATMEL(R)] AVR(R) ATmega328PX
atmega2560MICROCHIP(R) [former ATMEL(R)] AVR(R) ATmega2560
atmega4809MICROCHIP(R) [former ATMEL(R)] AVR(R) ATmegax4809X
am335xBeagleBone with Texas Instruments(R) AM335x ARM(R) A8
bcm2835_raspi_bRaspberryPi(R) Zero with ARM1176-JZFS(TM)
Debug/ReleasePC on Win* via MSVC x64 compiler Debug/Release
hostPC/Workstation on Win*/mingw64/*nix via host compiler
lpc11c24NXP(R) OM13093 LPC11C24 board ARM(R) Cortex(R)-M0+
nxp_imxrt1062Teensy 4.0 Board / NXP(R) iMXRT1062 ARM(R) Cortex(R)-M7X
riscvfe310SiFive RISC-V FE310 SoC
rl78Renesas(R) RL78/G13
rpi_pico_rp2040RaspberryPi(R) Pico RP2040 with dual ARM(R) Cortex(R)-M0+X
rpi_pico2_rp2350RaspberryPi(R) Pico2 RP2350 with dual ARM(R) Cortex(R)-M33X
rx63nRenesas(R) RX630/RX631
stm32f100STMicroelectronics(R) STM32F100 ARM(R) Cortex(R)-M3X
stm32f407STMicroelectronics(R) STM32F407 ARM(R) Cortex(R)-M4F
stm32f429STMicroelectronics(R) STM32F429 ARM(R) Cortex(R)-M4F
stm32f446STMicroelectronics(R) STM32F446 ARM(R) Cortex(R)-M4F
stm32h7a3STMicroelectronics(R) STM32H7A3 ARM(R) Cortex(R)-M7
stm32l100cSTMicroelectronics(R) STM32L100 ARM(R) Cortex(R)-M3X
stm32l152STMicroelectronics(R) STM32L152 ARM(R) Cortex(R)-M3
stm32l432STMicroelectronics(R) STM32L432 ARM(R) Cortex(R)-M4FX
v850es_fx2Renesas(R) Electronics V850es/Fx2 upd703231
wch_ch32v307WCH CH32v307 RISC-V board
wch_ch32v307_llvmWCH CH32v307 RISC-V board (but using an LLVM toolchain)
x86_64-w64-mingw32PC on Win*/mingw64 via GNU/GCC x86_x64 compiler
xtensa32Espressif (XTENSA) NodeMCU ESP32X

In this table, *(breadboard) means the board (or certain versions of it) can be readily used with a common breadboard. This may possibly need some very straightforward manual soldering/mounting of header-pins.

Getting Started with the Reference Application

It is easiest to get started with the reference application using one of the supported boards, such as avr (ARDUINO) or bcm2835_raspi_b (RaspberryPi ZERO) or am335x (BeagleBoneBlack), etc. The reference application can be found in the directory ref_app and its subdirectories.

The reference application uses cross-development and build systems are supported on:

Upon successful completion of the build, the resulting artifacts including HEX-files (such as ref_app.hex), map files, size reports, etc., are available in the bin directory.

Build with Bash Shell Script and GNU make

To get started with the reference application on *nix

Example build on *nix for target avr

We will now exemplify how to build the reference application on a command shell in *nix for target avr. This target system includes essentially any ARDUINO(R)-compatible board. This is also the board compatibility actually used with the homemade boards in the book.

Install gcc-avr if needed.

sudo apt install gcc-avr avr-libc

Clone or get the ckormanyos/real-time-cpp repository. Then build with:

cd real-time-cpp
cd ref_app
./target/build/build.sh avr rebuild

Example build on *nix for target stm32f446

We will now exemplify how to build the reference application on a command shell in *nix for an ARM(R) target. Consider, for example, the build variant target stm32f446. The NUCLEO-F446RE board from STMicroelectronics(R) can conveniently be used for this.

Install gcc-arm-none-eabi if needed.

sudo apt install gcc-arm-none-eabi

Clone or get the ckormanyos/real-time-cpp repository. Then build with:

cd real-time-cpp
cd ref_app
./target/build/build.sh stm32f446 rebuild

Example build on MacOS for target stm32f446

We will now exemplify how to build the reference application in a command shell in MacOS for an ARM(R) target. Consider, for example, the build variant target stm32f446. The NUCLEO-F446RE board from STMicroelectronics(R) can conveniently be used for this.

Clone or get the ckormanyos/real-time-cpp repository.

The default version 3.81 of GNUmake on MacOS can (now) be used. The make files used in this repository have been made compatible with it. For background information, see also issue 273.

Build the target with a direct manual call to make.

cd real-time-cpp
cd ref_app
make -f target/app/make/app_make.gmk rebuild TGT=stm32f446

If the toolchain is needed then it must be installed or retrieved prior to building the target of the reference application.

You can obtain via wget (or optionally install) the gcc-arm-none-eabi toolchain if needed. In this case, I have found it convenient to use a modern gcc-arm-none-eabi for MacOS which can be found at Arm GNU Toolchain Downloads.

The arm-non-eabi toolchain can be fetched via wget and successfully used locally in the shell. If this is desired, follow the step-by-step procedure below.

Step 1: Make a local directory (such as macos-gnu-arm-toolchain) and cd into it.

cd real-time-cpp
mkdir -p macos-gnu-arm-toolchain
cd macos-gnu-arm-toolchain

Step 2: Fetch the toolchain's tarball with wget, unpack it and add the compiler's bin-directory to the shell's executable path.

wget --no-check-certificate https://developer.arm.com/-/media/Files/downloads/gnu/12.2.rel1/binrel/arm-gnu-toolchain-12.2.rel1-darwin-x86_64-arm-none-eabi.tar.xz
tar -xvf arm-gnu-toolchain-12.2.rel1-darwin-x86_64-arm-none-eabi.tar.xz
PATH=$(pwd)/arm-gnu-toolchain-12.2.rel1-darwin-x86_64-arm-none-eabi/bin:$PATH

Step 3: Optionally echo the PATH for a quick path-check. It can also be helpful to query arm-non-eabi-g++'s version. This is expected to verify that the toolchain is correctly added to this shell's local PATH.

echo $PATH
arm-none-eabi-g++ -v

Now simply use the commands to build the target with a direct call to make (which is the same as shown above for the *nix case).

cd real-time-cpp
cd ref_app
make -f target/app/make/app_make.gmk rebuild TGT=stm32f446

Build with VisualStudio(R) Project and CMD Batch

To get started with the reference application on Win*

The ref_app build in Microsoft(R) VisualStudio(R) makes heavy use of cross development using a project workspace of type External Makefile. GNUmake is invoked via batch file in the build process. It subsequently runs in combination with several Makefiles.

To build any ref_app target other than Debug or Release for Win32, a cross-compiler (GNU/GCC cross compiler) is required. See the text below for additional details.

GNU/GCC cross compilers running on Win* intended for the reference application on VisualStudio(R) can be found in the toolchains repository, ckormanyos/real-time-cpp-toolchains. The toolchains repository contains detailed instructions on installing, moving and using these ported GNU/GCC compilers.

Note on GNUmake for Win*: A GNUmake capable of being used on Win* can be found in the ckormanyos/make-4.2.1-msvc-build repository. If desired, clone or get the code of this repository. Build make-4.2.1 in its x64 Release configuration with MSVC (i.e., VC 14.2 or later, Community Edition is OK).

Build with Cross-Environment CMake

Cross-Environment CMake can build the reference application. For this purpose, CMake files have also been created for each supported target.

Consider, for instance, building the reference application for the avr target with CMake. The pattern is shown below.

cd real-time-cpp
mkdir build
cd build
cmake ../ref_app -DTRIPLE=avr -DTARGET=avr -DCMAKE_TOOLCHAIN_FILE=../ref_app/cmake/gcc-toolchain.cmake
make -j ref_app

We will now consider, for instance, building the reference application for one of the supported ARM(R) targets with CMake. The pattern is shown below. In this case, we need to identify the following make options:

-DTRIPLE=avr -DTARGET=avr

Switch these options to the ones intended for the stm32f446 ARM(R)-based target being built.

-DTRIPLE=arm-none-eabi -DTARGET=stm32f446

Let's clarify the commands in their entirety in order to run a CMake build for stm32f446 (i.e., ST Microelectronics(R) STM32F446 ARM(R) featuring Cortex(R)-M4F).

cd real-time-cpp
mkdir build
cd build
cmake ../ref_app -DTRIPLE=arm-none-eabi -DTARGET=stm32f446 -DCMAKE_TOOLCHAIN_FILE=../ref_app/cmake/gcc-toolchain.cmake
make -j ref_app

When building with CMake for other targets, follow the standard *nix pattern to build. Also building with CMake for x86_64-w64-mingw32 or host from MSYS, Cygwin or any similar *nix-like shell or console should work too.

The following command sequence will build for the native host on a *nix-like shell or console.

cd real-time-cpp
mkdir build
cd build
cmake ../ref_app -DTARGET=host -DCMAKE_TOOLCHAIN_FILE=../ref_app/cmake/gcc-toolchain.cmake
make -j ref_app

Build with MICROCHIP's ATMEL Studio

There is also a workspace solution for ATMEL(R) AtmelStudio(R) 7. It is called ref_app.atsln and is also located in the ref_app directory. There are ATMEL Studio projects for both the reference application as well as for each of the examples. ATMEL Studio projects in this repository support the AVR target only.

If you decide to use ATMEL Studio, you do not need to use or include any additional libraries for these projects (other than those that are ordinarily installed during the standard installation of ATMEL Studio).

Target Details

Target details including startup code and linker definition files can be found in the ref_app/target directory and its subdirectories. There are individual subdirectories for each supported target microcontroller system.

The MICROCHIP(R) [former ATMEL(R)] AVR(R) configuration called target avr runs on a classic ARDUINO(R) compatible board. The program toggles the yellow LED on portb.5.

The MICROCHIP(R) [former ATMEL(R)] AVR(R) configuration called target atmega2560 runs on the ARDUINO(R) MEGA compatible board. The program toggles the orange LED on portb.7. At the moment, the environment and build for this target are set up for $64~\text{kByte}$ program code. If the fully available $128~\text{kByte}$ code space needs to be used, then adaptions to the compiler switches, linker file, startup-code and interrupt-vector table will likely be necessary. For this potential adaption, see also issue 593.

The MICROCHIP(R) [former ATMEL(R)] ATmega4809 configuration called target atmega4809 runs on an ARDUINO(R) EVERY compatible board clocked with the internal resonator at $20~\text{MHz}$. The program toggles the yellow LED on porte.2 (i.e., D5).

The Espressif (XTENSA) NodeMCU ESP32 implementation uses a subset of the Espressif SDK to run the reference application with a single OS task exclusively on 1 of its cores.

The NXP(R) OM13093 LPC11C24 board ARM(R) Cortex(R)-M0+ configuration called "target lpc11c24" toggles the LED on port0.8.

The ARM(R) Cortex(R)-M3 configuration (called target stm32f100) runs on the STM32VL-DISCOVERY board commercially available from ST Microelectronics(R). The program toggles the blue LED on portc.8.

The second ARM(R) Cortex(R)-M3 configuration (called target stm32l100c) runs on the STM32L100-DISCOVERY board commercially available from ST Microelectronics(R). The program toggles the blue LED on portc.8.

The third ARM(R) Cortex(R)-M3 configuration (called target stm32l152) runs on the STM32L152C-DISCOVERY board commercially available from ST Microelectronics(R). The program toggles the blue LED on portb.6.

The first ARM(R) Cortex(R)-M4F configuration (called target stm32f407) runs on the STM32F4-DISCOVERY board commercially available from ST Microelectronics(R). The program toggles the blue LED on portd.15.

Another ARM(R) Cortex(R)-M4F configuration (called target stm32f446) runs on the STM32F446-NUCLEO-64 board commercially available from ST Microelectronics(R). The program toggles the green LED on porta.5. An Ozone debug file is supplied for this system for those interested.

The first ARM(R) Cortex(R)-M7 configuration (called target stm32h7a3) runs on the STM32H7A3-NUCLEO-144 board commercially available from ST Microelectronics(R). The program toggles the green LED on portb.0.

The ARM(R) A8 configuration (called target am335x) runs on the BeagleBone board (black edition). For the white edition, the CPU clock needs to be reduced from $900~\text{MHz}$ to something like $600~\text{MHz}$. This project creates a bare-metal program for the BeagleBone that runs independently from any kind of *nix distro on the board. Our program is designed to boot the BeagleBone from a raw binary file called MLO stored on a FAT32 SDHC microcard. The binary file includes a special boot header comprised of two 32-bit integers. The program is loaded from SD-card into RAM memory and subsequently executed. When switching on the BeagleBone black, the boot button (S2) must be pressed while powering up the board. The program toggles the first user LED (LED1 on port1.21).

The ARM(R) 1176-JZF-S configuration (called target bcm2835_raspi_b) runs on the RaspberryPi(R) Zero (PiZero) single core controller. This project creates a bare-metal program for the PiZero. This program runs independently from any kind of *nix distro on the board. Our program is designed to boot the PiZero from a raw binary file. The raw binary file is called kernel.img and it is stored on a FAT32 SDHC microcard. The program objcopy can be used to extract raw binary from a ELF-file using the output flags -O binary. The kernel.img file is stored on the SD card together with three other files: bootcode.bin, start.elf and (an optional) config.txt, all described on internet. A complete set of PiZero boot contents for an SD card running the bare-metal reference application are included in this repo. The program toggles the GPIO status LED at GPIO index 0x47.

The rpi_pico_rp2040 target configuration employs the RaspberryPi(R) Pico RP2040 with dual-core ARM(R) Cortex(R)-M0+ clocked at $133~\text{MHz}$. The low-level startup boots through core 0. Core 0 then starts up core 1 (via a specific protocol). Core 1 subsequently carries out the blinky application, while core 0 enters an endless, idle loop. Ozone debug files are supplied for this system for those interested. Reverse engineering of the complicated (and scantly documented) dual-core startup originated in and have been taken from (with many thanks) from the Blinky_Pico_dual_core_nosdk repo.

The rpi_pico2_rp2350 target configuration employs the RaspberryPi(R) Pico2 RP2350 with dual-core ARM(R) Cortex(R)-M33 clocked at $150~\text{MHz}$. It has essentially the same boot structure as the 2040. Similarly the dual-core startup was pioneered by the efforts revealed in the modernized Blinky_Pico2_dual_core_nosdk repo.

Target v850es_fx2 uses a classic Renesas(R) V850es/Fx2 core. The upd703231 microcontroller derivative on an F-Line Drive It starter kit is used.

The riscvfe310 target utilizes the SiFive RISC-V FE310 SoC on Spark Fun's commercially available Red Thing Plus Board. The blue LED on port GPIO0.5 is toggled.

The adaption for wch_ch32v307 runs on the WCH CH32v307 board. It uses the RISC-V CH32v307 microcontroller from Nanjing Qinheng Microelectronics Co., Ltd. The blue LED1 manually connected to port GPIOC.0 via wire-connection provides the blinky toggle. The similar adaption wch_ch32v307_llvm is essentially the same except it uses an LLVM RISC-V toolchain instead of GCC RISC-V.

Target nxp_imxrt1062 runs on the Teensy 4.0 board from Spark Fun. The orange user-LED is toggled.

For other compatible boards, feel free contact me directly or submit an issue requesting support for your desired target system.

Benchmarks

Benchmarks provide scalable, portable means for identifying the performance and the performance class of the microcontroller. For more information, see the detailed information on the benchmarks pages.

All Bare-Metal

Projects in this repo are programmed OS-less in naked, bare-metal mode making use of self-written startup code. No external libraries other than native C++ and its own standard libraries are used.

Consider, for instance, the BeagleBone Black Edition (BBB, also known as target am335x) which is one of several popular target systems supported in this repository. The projects on this board boot from the binary image file MLO on the SD card. Like all other projects in this repository, the BBB projects perform their own static initialization and chip initialization (i.e., in this particular case chip initialization of the ARM(R) 8 AM335x processor). The BBB projects, following initialization, subsequently jump to main() which initializes the am335x MCAL and starts our self-written multitasking scheduler.

The image below depicts the bare-metal BeagleBone Black Edition in action. In this bare-metal operation mode, there is no running *nix OS on the BBB, no keyboard, no mouse, no monitor, no debug interface and no emulator.

The microcontroller on the board is cyclically performing one of the benchmarks mentioned above. The first user LED is toggled on port1.21 in multitasking operation and the oscilloscope captures a real-time measurement of the benchmark's time signal on digital I/O port1.15, header pin P8.15 of the BBB.

Continuous Integration (CI)

Continuous integration uses GitHub Actions programmed in YAML. The CI script exercises various target builds, example builds and benchmark builds/runs on GitHub Actions' instances of ubuntu-latest, windows-latest and macos-latest using GNUmake, CMake or MSBuild depending on the particular OS/build/target-configuration.

Build Status

At the moment, there are distinct and separate, major individual builds. Each build emphasizes different capabilities of the companion code.

Here are the build status badges.

<p align="center"> <a href="https://github.com/ckormanyos/real-time-cpp/actions"> <img src="https://github.com/ckormanyos/real-time-cpp/actions/workflows/real-time-cpp.yml/badge.svg" alt="Build Status"></a> <a href="https://github.com/ckormanyos/real-time-cpp/actions"> <img src="https://github.com/ckormanyos/real-time-cpp/actions/workflows/real-time-cpp-examples.yml/badge.svg" alt="Build Examples"></a> <a href="https://github.com/ckormanyos/real-time-cpp/actions"> <img src="https://github.com/ckormanyos/real-time-cpp/actions/workflows/real-time-cpp-snippets.yml/badge.svg" alt="Build Snippets"></a> <a href="https://github.com/ckormanyos/real-time-cpp/actions"> <img src="https://github.com/ckormanyos/real-time-cpp/actions/workflows/real-time-cpp-benchmarks.yml/badge.svg" alt="Build Benchmarks"></a> </p>

The build status badges represent the state of the nightly CI builds and tests.

Modern avr-gcc Toolchain

The repo ckormanyos/avr-gcc-build builds up-to-date avr-gcc toolchains for x86_64-linux-gnu and x86_64-w64-mingw32. Shell and YAML scripts build avr-gcc directly from source on GHA runner(s). In addition, occasional GitHub-releases provide pre-built avr-gcc toolchains for x86_64-linux-gnu and x86_64-w64-mingw32.

This repo is a great place to learn how to build your own avr-gcc toolchain from source. The straightforward, well-described shell and YAML scripts are easy to understand, use or adapt.

As mentioned above, a much more detailed and wider scope of embedded toolchains are described in ckormanyos/real-time-cpp-toolchains. These include the afore-mentioned avr-gcc toolchain as well as others (some hard-to-find elsewhere).

GNU/GCC Compilers

The reference application and the examples (also the code snippets) can be built with GNU/GCC compilers and GNUmake on *nix. GNU/GCC cross compilers and GNUmake on *nix are assumed to be available in the standard executable path, such as after standard get-install practices.

Some ported GNU/GCC cross compilers for Win* are available in the toolchains repository, real-time-cpp-toolchains. These can be used with the microcontroller solution configurations in the reference application when developing/building within Microsoft(R) VisualStudio(R). Various other GNU tools such as GNUmake, SED, etc. have been ported and can be found there. These are used in the Makefiles When building cross embedded projects such as ref_app on Win*.

In the reference application on Win*, the Makefiles use a self-defined, default location for the respective tools and GNU/GCC toolchains. The toolchain default location on Win* is ./ref_app/tools/Util/msys64/usr/local. This particular toolchain location is inspired by the msys2/mingw64 system.

Toolchains intended for cross MSVC/GCC builds on Win* should be located there. These toolchains are not part of this repository and it is necessary to get these toolchains separately when using the supported Win* builds when optionally using VisualStudio(R) Projects with CMD Batch.

Detailed instructions on getting and using the toolchains for cross MSVC/GCC builds on Win* are available in the real-time-cpp-toolchains repository. These instructions provide guidance on using these toolchains when selecting the Microsoft(R) VisualStudio(R) project (via the usual, above-described MSVC/Win*-way) to build the reference application.

C++ Language Adherence

A GNU/GCC port (or other compiler) with a high level of C++14 (or higher) awareness and adherence such as GCC 5 through 13 (higher generally being more advantageous) or MSVC 14.2 or higher is required for building the reference application (and the examples and code snippets).

Some of the code snippets demonstrate language elements not only from C++14, but also from C++17, 20, 23 and beyond. A compiler with C++17 support or even C++20, 23 support (such as GCC 13, clang 15, MSVC 14.3, or higher) can, therefore, be beneficial for success with all of the code snippets.

Licensing