C/C++ Deployment

This section describes how to deploy generated code in a C/C++ environment.

warning

Include the generated source files and link the dynamic or static library to call the generated code. Note the following when calling the generated code:

• The calling platform and compiler must be consistent with the option.platform setting during code generation
• When the calling platform is Linux/Mac, you need to link the math library (-lm) to use functions such as sin, cos, tan, sqrt
• When the calling platform is Windows, you need to link iphlpapi (-liphlpapi, or add iphlpapi.lib in Visual Studio) to obtain MAC address and other information

Using vehicle as an example, assuming platform is 'linux-x86_64-gcc' and lib_type is 'static', the code directory structure is as follows:

working_dir/
├── vehicle/
│   ├── vehicle_prob.c
│   ├── vehicle_prob.h
│   ├── vehicle_solver.h
│   ├── libvehicle_solver_static.a
├── vehicle_demo.cpp
.

Compilation and Execution

Using CMake

Below is a CMake snippet for compiling a demo (vehicle_demo.cpp) using the static library:

CMakeLists.txt

# ${VEHICLE_CODEGEN_DIR} is the dir of the generated code
set(VEHICLE_CODEGEN_DIR ${CMAKE_CURRENT_SOURCE_DIR}/vehicle) 
# ${LIBS} is the required libs (math or iphlpapi)
set(LIBS m) 
add_executable(vehicle_demo vehicle_demo.cpp ${VEHICLE_CODEGEN_DIR}/vehicle_prob.c)
target_include_directories(vehicle_demo PRIVATE ${VEHICLE_CODEGEN_DIR})
target_link_libraries(vehicle_demo ${VEHICLE_CODEGEN_DIR}/libvehicle_solver_static.a ${LIBS})

Command Line

Below is a command line example for compiling with g++:

g++ -o vehicle_demo vehicle_demo.cpp vehicle/vehicle_prob.c -I./vehicle -L./vehicle -lvehicle_solver_static -lm

After compilation is complete, run the generated executable.

./vehicle_demo

Visual Studio

When using Visual Studio, you need to set the platform to windows-x86_64-msvc for code generation. Below are the steps to deploy the generated code in Visual Studio.

Create a new C++ project in Visual Studio, add the generated source file (vehicle_prob.c) and demo file (vehicle_demo.cpp) to the project, then configure the following settings in the project properties:

• In "C/C++ > General > Additional Include Directories", add the generated code directory (vehicle/)
• In "Linker > General > Additional Library Directories", add the generated code directory (vehicle/)
• In "Linker > Input > Additional Dependencies", add the generated static library filename (libvehicle_solver_static.lib)
• In "Linker > Input > Additional Dependencies", add iphlpapi.lib

Then you can compile and run the project.

Multi-threaded Deployment

When multiple problems need to be solved simultaneously, multi-threading can be used to accelerate solving.

warning

An independent workspace and output must be allocated for each thread; otherwise, data races will occur.

Below is a code snippet example using OpenMP for parallel solving (NUM_PROBS is the number of problems):

vehicle_demo.cpp

...
#pragma omp parallel for
for (int worker = 0; worker < NUM_PROBS; worker++) {
    ws[worker].thread_id = worker;
    vehicle_solve(&prob[worker], &option, &ws[worker], &output[worker]);
}
...

Here we use thread_id to identify each thread. This value is passed to the function parameters of prob and can be accessed via _solverinfo->thread_id:

vehicle_prob.c

void vehicle_l(const Vehicle_SolverInfo *_solverinfo, ...)
{
    ...
    int thread_id = _solverinfo->thread_id;
    ...
}

tip

This value is not required, but when using multi-threading, it can be used to distinguish different threads.

For example, when the optimization problem functions are user-defined external functions, this value can be used to distinguish different threads to avoid data races.