Skip to content

Problem when run with P100 in cluster

OpenLB – Open Source Lattice Boltzmann Code Forums on OpenLB General Topics Problem when run with P100 in cluster

Viewing 2 posts - 1 through 2 (of 2 total)
  • Author
    Posts
  • #9438
    thanhphatvt
    Participant

    Dear all,
    I use the P100 in our cluster and I use nvhpc/24.7 now when I build the model I’m having this problem
    “nvcc -O3 -std=c++17 -I/opt/nvidia/hpc_sdk/Linux_x86_64/22.3/comm_libs/openmpi/openmpi-3.1.5/include/ –generate-code=arch=compute_60,code=[compute_60,sm_60] –extended-lambda –expt-relaxed-constexpr -rdc=true -Xcudafe “–diag_suppress=implicit_return_from_non_void_function –display_error_number –diag_suppress=20014 –diag_suppress=20011″ -Xlinker –version-script=build/olbcuda.version -shared build/olbcuda.o -o libolbcuda.so
    mpic++ cavity3d.o -o cavity3d -L../../../external/lib -lpthread -lz -ltinyxml -L../../../build/lib -L . -lolbcuda -lolbcore -L/opt/nvidia/hpc_sdk/Linux_x86_64/22.3/cuda/11.6/lib64 -L/opt/nvidia/hpc_sdk/Linux_x86_64/22.4/REDIST/cuda/11.6/targets/x86_64-linux/lib/stubs/ -lcuda -lcudadevrt -lcudart
    /usr/bin/ld: cannot find -lcuda: No such file or directory
    make: *** [../../../default.mixed.mk:111: cavity3d] Error 2

    How can I fix it? Thank you so much!

    #9439
    Adrian
    Keymaster

    You can fix it by adding the path where CUDA shared libraries are located in the linker search paths. This is all the error message tells you: It can not find the CUDA library to link to (likely it also doesn’t find cudadevrt and cudart but exits beforehand).

    If the CUDA module on your cluster doesn’t do this (assuming you loaded it and did not manually extract the paths) you can expand the linker paths by extending e.g. LD_LIBRARY_PATH.

Viewing 2 posts - 1 through 2 (of 2 total)
  • You must be logged in to reply to this topic.