sdk samples not installing

I’m trying to install the Cuda sdk samples on Scientific Linux 5.5 (32-bit).
I’ve installed the package into my home directory, but
when I run make I see the following:

[fbiett@fermi0 C]$ make
make[1]: Entering directory /home/fbiett/NVIDIA_GPU_Computing_SDK/C/common' make[1]: Leaving directory /home/fbiett/NVIDIA_GPU_Computing_SDK/C/common’
make[1]: Entering directory /home/fbiett/NVIDIA_GPU_Computing_SDK/C/common' make[1]: Leaving directory /home/fbiett/NVIDIA_GPU_Computing_SDK/C/common’
make[1]: Entering directory /home/fbiett/NVIDIA_GPU_Computing_SDK/C/common' make[1]: Leaving directory /home/fbiett/NVIDIA_GPU_Computing_SDK/C/common’
make[1]: Entering directory /home/fbiett/NVIDIA_GPU_Computing_SDK/shared' make[1]: Leaving directory /home/fbiett/NVIDIA_GPU_Computing_SDK/shared’
make -C src/vectorAddDrv/
make[1]: Entering directory /home/fbiett/NVIDIA_GPU_Computing_SDK/C/src/vectorAddDrv' obj/i386/release/vectorAdd.cpp.o: In function Cleanup(bool)’:
vectorAdd.cpp:(.text+0x18e): undefined reference to cuCtxDetach' vectorAdd.cpp:(.text+0x204): undefined reference to cuMemFree’
vectorAdd.cpp:(.text+0x219): undefined reference to cuMemFree' vectorAdd.cpp:(.text+0x234): undefined reference to cuMemFree’
obj/i386/release/vectorAdd.cpp.o: In function main': vectorAdd.cpp:(.text+0x279): undefined reference to cuInit’
vectorAdd.cpp:(.text+0x293): undefined reference to cuDeviceGetCount' vectorAdd.cpp:(.text+0x2ba): undefined reference to cuDeviceGet’
vectorAdd.cpp:(.text+0x2df): undefined reference to cuCtxCreate' vectorAdd.cpp:(.text+0x604): undefined reference to cuModuleLoadDataEx’
vectorAdd.cpp:(.text+0x643): undefined reference to cuModuleGetFunction' vectorAdd.cpp:(.text+0x6dc): undefined reference to cuMemAlloc’
vectorAdd.cpp:(.text+0x700): undefined reference to cuMemAlloc' vectorAdd.cpp:(.text+0x724): undefined reference to cuMemAlloc’
vectorAdd.cpp:(.text+0x752): undefined reference to cuMemcpyHtoD' vectorAdd.cpp:(.text+0x780): undefined reference to cuMemcpyHtoD’
vectorAdd.cpp:(.text+0x7bc): undefined reference to cuParamSetv' vectorAdd.cpp:(.text+0x7f5): undefined reference to cuParamSetv’
vectorAdd.cpp:(.text+0x82e): undefined reference to cuParamSetv' vectorAdd.cpp:(.text+0x85b): undefined reference to cuParamSeti’
vectorAdd.cpp:(.text+0x880): undefined reference to cuParamSetSize' vectorAdd.cpp:(.text+0x8b5): undefined reference to cuFuncSetBlockShape’
vectorAdd.cpp:(.text+0x8e2): undefined reference to cuLaunchGrid' vectorAdd.cpp:(.text+0x910): undefined reference to cuMemcpyDtoH’
vectorAdd.cpp:(.text+0xa68): undefined reference to cuModuleLoad' collect2: ld returned 1 exit status make[1]: *** [../../bin/linux/release/vectorAddDrv] Error 1 make[1]: Leaving directory /home/fbiett/NVIDIA_GPU_Computing_SDK/C/src/vectorAddDrv’
make: *** [src/vectorAddDrv/Makefile.ph_build] Error 2

Can anyone tell me what’s wrong?

I have my $PATH and $LD_LIBRARY_PATH set up correctly.

Thanks

Same prob here http://forums.nvidia.com/index.php?showtop…p;#entry1120607 but unfortunately no reply so far…

Same prob here http://forums.nvidia.com/index.php?showtop…p;#entry1120607 but unfortunately no reply so far…

Copy of reply from another thread.

I do not know about Ubuntu (using Debian) but will try to help.

I can think about two possible causes of those problems

  1. You miss some packages.
    Install all available NVIDIA packages. Here is list from my system:
    libcuda1
    libcuda1-dev
    libgl1-nvidia-alternatives
    libgl1-nvidia-dev
    libgl1-nvidia-glx
    libglx-nvidia-alternatives
    libnvidia-compiler
    nvidia-glx
    nvidia-glx-dev
    nvidia-kernel-dkms
    nvidia-kernel-source
    nvidia-libopencl1
    nvidia-libopencl1-dev
    nvidia-opencl-common
    nvidia-opencl-icd
    nvidia-smi
    nvidia-vdpau-driver
    nvidia-xconfig

The mos important is libcuda1 and libcuda1-dev, libnvidia-compiler (used to compile PTX to CUBIN).

  1. Conflict between driver version and CUDA toolkit version.
    Drivers 195.x (available in Debian testing and unstable and Ubuntu) do not cooperate well with CUDA toolkit 3.1. To have working CUDA 3.1 you need drivers 256.53 (available in Debian experimental).

Please reply whether any of those help to solve your problems/

Copy of reply from another thread.

I do not know about Ubuntu (using Debian) but will try to help.

I can think about two possible causes of those problems

  1. You miss some packages.
    Install all available NVIDIA packages. Here is list from my system:
    libcuda1
    libcuda1-dev
    libgl1-nvidia-alternatives
    libgl1-nvidia-dev
    libgl1-nvidia-glx
    libglx-nvidia-alternatives
    libnvidia-compiler
    nvidia-glx
    nvidia-glx-dev
    nvidia-kernel-dkms
    nvidia-kernel-source
    nvidia-libopencl1
    nvidia-libopencl1-dev
    nvidia-opencl-common
    nvidia-opencl-icd
    nvidia-smi
    nvidia-vdpau-driver
    nvidia-xconfig

The mos important is libcuda1 and libcuda1-dev, libnvidia-compiler (used to compile PTX to CUBIN).

  1. Conflict between driver version and CUDA toolkit version.
    Drivers 195.x (available in Debian testing and unstable and Ubuntu) do not cooperate well with CUDA toolkit 3.1. To have working CUDA 3.1 you need drivers 256.53 (available in Debian experimental).

Please reply whether any of those help to solve your problems/