无法在 Windows 11、VIsual Studio 2019 中使用 OpenMP 并行运行超过 1 个线程

Question

I created an empty C++ project using Visual Studio 2019 in Windows 11.

Then I ran this minimal piece of code to test multi-threading, using OpenMP:

#include <iostream>
#include <omp.h>


int main() {
    
    omp_set_dynamic(0); // ensure that the number of threads doesn't change due to system demands
    omp_set_num_threads(16);
    std::cout << "Max threads: " << omp_get_max_threads() << "\n";

    int id;
#pragma omp parallel private(id)
    {
        std::cout << "Number of threads running in parallel: " << omp_get_num_threads() << "\n";
        std::cout << "Thread ID in parallel region: " << omp_get_thread_num() << "\n";
    }

    return 0;
}

This is the output I am getting:

Max threads: 16
Number of threads running in parallel: 1
Thread ID in parallel region: 0

I first tried multithreading in a CMake project (used CMake 3.22 and the modern way of including OpenMP in a CMake project ) and attempted using both the Visual Studio 2019 and 2022 compilers. The result was the same, I couldn't get more than 1 thread to run in parallel. So, then I tried this simple example in Visual Studio 2019 and ran into the same issue.

I'm wondering if there is some Windows rule/setting that restricts the number of threads I can use. Btw, using Matlab 2021b and the Parallel Computing toolbox, I was able to use multithreading successfully (8/16 threads).

If you need any other information about the issue/my system, don't hesitate to ask it from me.

Thank you.

Edit:

I get the following output from CMake

"C:\Program Files\JetBrains\CLion 2022.1.1\bin\cmake\win\bin\cmake.exe" -DCMAKE_BUILD_TYPE=Debug "-DCMAKE_MAKE_PROGRAM=C:/Program Files/JetBrains/CLion 2022.1.1/bin/ninja/win/ninja.exe" -G Ninja -S C:\Users\parvanitis\Documents\MBI_code\cpp\ImagingFocus -B C:\Users\parvanitis\Documents\MBI_code\cpp\ImagingFocus\cmake-build-debug

• The CXX compiler identification is MSVC 19.32.31332.0
• The CUDA compiler identification is NVIDIA 11.6.124
• Detecting CXX compiler ABI info
• Detecting CXX compiler ABI info - done
• Check for working CXX compiler: C:/Program Files/Microsoft Visual Studio/2022/Community/VC/Tools/MSVC/14.32.31326/bin/Hostx64/x64/cl.exe
• skipped
• Detecting CXX compile features
• Detecting CXX compile features - done
• Detecting CUDA compiler ABI info
• Detecting CUDA compiler ABI info - done
• Check for working CUDA compiler: C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v11.6/bin/nvcc.exe - skipped
• Detecting CUDA compile features
• Detecting CUDA compile features - done
• Found Matlab: C:/Program Files/MATLAB/R2022a/extern/include (found version "9.12") found components: MAT_LIBRARY
• Found Git: C:/Program Files/Git/cmd/git.exe (found version "2.36.1.windows.1")
• Found OpenMP_CXX: -openmp (found version "2.0")
• Found OpenMP: TRUE (found version "2.0")
• Configuring done
• Generating done
• Build files have been written to: C:/Users/parvanitis/Documents/MBI_code/cpp/ImagingFocus/cmake-build-debug

[Finished]

Answer 1

OK, real rookie mistake(s) here.

First of all, I didn't add the /openmp flag to the CMAKE_CXX_FLAGS which is necessary in Windows (I thought find_package(OpenMP) was enough, while adding -fopenmp, as in Linux, was not recognized, so I skipped it).

Even after I did, it didn't work. The reason was that the project I was building was also using CUDA and I was calling the OpenMP-multithreaded code from a.cu file, which is of course compiled using NVCC, which is ignorant of the CXX flags, and most likely does not have any reason to support them awyway. So, I needed to move the OpenMP-multithreaded code to a.cpp file in order for it to work. I didn't mention my project was also using CUDA in the first place because I didn't think that would cause a problem (silly me).

Thanks for your help anyway Jérôme.