Important: Please read the Qt Code of Conduct -

adding CUDA to Qt

  • Hi, im developing a qt application which would benefit form GPU acceleration, therefore i picked CUDA. However im struggling with connecting Qt and CUDA. I get this error while i build my project: 0_1533481894888_wykop.png

    and due to the only one answer i have found it might be PATH variable issue: , but i had already set them to proper paths 0_1533481994078_wykop1.png

    My laptop is running on Ubuntu 18.04 with NVDIA Drivier 396.45. Cuda toolkit-9.2 was installed from runfile downloaded direct from nvidia website.

    Part of .pro file:

    And ss of nvcc and gcc:

    If somebody has a clue how to solve it or how can i move on ( i even consider switching to other SDK ) please let me know.

  • Lifetime Qt Champion

    Hi @m00lecule,

    I have the following tip for you: remove the QMAKE_LIBDIR line in the project file and change the LIBS line to the following:

    LIBS += -L$$CUDA_DIR/lib64 -lcuda -lcudart

    That way QtCreator will adjust the LD_LIBRARY_PATH correctly.

    If that is not enough, go to a terminal, cd to $$CUDA_DIR/lib64 and post the output of ldd and ldd


    PS: if possible, post text files as text here, that makes it way easier for me to copy and modify the values without introducing new errors. Thanks.

  • Lifetime Qt Champion


    To add to @aha_1980 , did you check the Run part of the Project panel ? Just to see whether your LD_LIBRARY_PATH variable is set correctly.

  • It turned out that making cuda.conf file in /etc/ containing path to cuda libs directory and running ldconfig solved my problem.

Log in to reply