I’ve recently installed the CUDA toolkit and SDK on an AMD box running Scientific Linux 5, x86 (This is just RHEL5, recompiled from source by the friendly people at Fermilab).
As a regular user, I’ve unpacked the SDK in my home directory and now am trying to compile the example program. With,
[nmoore@buff NVIDIA_CUDA_SDK]$ echo $PATH
/usr/kerberos/bin:/usr/local/bin:/bin:/usr/bin:/usr/local/cuda/bin:/home/nmoore/bin:/sbin
[nmoore@buff NVIDIA_CUDA_SDK]$ echo $LD_LIBRARY_PATH
/usr/local/cuda/lib:
I compile the example and see lots of errors:
[nmoore@buff NVIDIA_CUDA_SDK]$ make
make[1]: Entering directory /home/nmoore/NVIDIA_CUDA_SDK/common' ar: creating ./../lib/libcutil.a a - obj/release/bank_checker.cpp_o a - obj/release/cmd_arg_reader.cpp_o a - obj/release/cutil.cpp_o a - obj/release/error_checker.cpp_o a - obj/release/stopwatch.cpp_o a - obj/release/stopwatch_linux.cpp_o a - obj/release/cutil_interop.cpp_o make[1]: Leaving directory /home/nmoore/NVIDIA_CUDA_SDK/common’
make -C projects/scan/
make[1]: Entering directory `/home/nmoore/NVIDIA_CUDA_SDK/projects/scan’
/usr/bin/ld: warning: libtlshook.so, needed by /usr/local/cuda/lib/libcudart.so, not found (try using -rpath or -rpath-link)
Strangely though, I see the library on my filesystem,
[nmoore@buff NVIDIA_CUDA_SDK]$ ls -1 /usr/local/cuda/lib/
cudarts.ptx
libcublasemu.so
libcublasemu.so.1
libcublasemu.so.1.0
libcublas.so
libcublas.so.1
libcublas.so.1.0
libcudart.so
libcudart.so.1
libcudart.so.1.0
libcuda.so
libcuda.so.1
libcuda.so.1.0
libcufftemu.so
libcufftemu.so.1
libcufftemu.so.1.0
libcufft.so
libcufft.so.1
libcufft.so.1.0
libtlshook.so
libtlshook.so.1
libtlshook.so.1.0
So, what should I be doing differently?
regards,
Nathan Moore