I met some issues on cuda code running on linux.
I am not sure they are related to the super or not.
I checked the LD_LIBRARY_PATH, and both RC3.1 and RC4.0 have the same issues.
anyone can help me?
thank you very much.
issue 1: when run SDK exmaple, it shows 'Runtime API error : all CUDA-capable devices are busy or unavailableâ€™
[jwang@super-496 ~]$ cd NVIDIA_GPU_Computing_SDK_4.0/C/bin/linux/release
[jwang@super-496 release]$ echo $LD_LIBRARY_PATH
[jwang@super-496 release]$ ./histogram
CUDA device [Tesla C1060] has 30 Multi-Processors, Compute 1.3
…allocating CPU memory.
…generating input data
…allocating GPU memory and copying input data
main.cpp(102) : cudaSafeCall() Runtime API error : all CUDA-capable devices are busy or unavailable.
issue 2: my cuda code worked correctly two weeks ago, but now shows error of â€œinvalid device functionâ€ after launch the kernel, even launch a very simple or empty kernel.
[jwang@super-496 HHT]$ ./Test_HHT_CUDA
Cuda error: INDEX_CUDA: invalid device function.