MARCO is a prototype compiler for the Modelica language. It is written in C++ and leverages both the LLVM and MLIR infrastructures.
MARCO runs natively on Linux and macOS. At the moment, Windows is not supported. Directions to obtain the required dependencies and compile the project can be found here.
After installation, two executables should appear in the bin
directory inside the chosen installation path.
marco
The marco
executable represents the main entry point for the compiler.
The list of options can be obtained by passing the --help
parameter.
As an example, consider the following model involving a scalar variable and its derivative:
model SimpleFirstOrder
Real x(start = 0, fixed = true);
equation
der(x) = 1 - x;
end SimpleFirstOrder;
The following command takes the source file, named SimpleFirstOrder.mo
, and generate a binary file named simulation
.
The numerical integration method is chosen with the --solver
flag, and in this case consists in the explicit Euler method.
The <marco_runtime_lib_folder>
argument has to be replaced with the path containing the installed MARCO runtime libraries (tipically, the lib
folder inside the specified installation path).
marco SimpleFirstOrder.mo -o simulation --model=SimpleFirstOrder --solver=euler-forward -L <marco_runtime_lib_folder> -Wl,-rpath,<marco_runtime_lib_folder>
As the name suggests, the simulation
binary file implements the simulation that computes the evolution of the system.
Additional parameters can be also be provided when running the simulation. Their list can be obtained using the --help
argument.
./simulation --time-step=0.5 --end-time=5 ✔
"time","x"
0.000000000,0.000000000
0.500000000,0.500000000
1.000000000,0.750000000
1.500000000,0.875000000
2.000000000,0.937500000
2.500000000,0.968750000
3.000000000,0.984375000
3.500000000,0.992187500
4.000000000,0.996093750
4.500000000,0.998046875
5.000000000,0.999023438
modelica-opt
The modelica-opt
executable is intended to be used mainly for debugging purposes.
It accepts only MLIR code and applies the user-requested pipeline of transformations, thus allowing the user to check their correct behaviour.
Contributions are welcome and encouraged, both in the form of pull requests and issues reporting bugs, errors, or possible improvements.