On some machines I simply can’t get the most vanilla example (deviceQuery) to work.
It’s an older 4.x CentOS that’s had everything maintained, with updated Kernel 2.6.18… The newest NVidia kernel modules build and load fine:
“NVRM: loading NVIDIA UNIX x86_64 Kernel Module 260.19.12 Fri Oct 8 11:17:08 PDT 2010”
Installing Cuda 3.1, the SDK examples build fine with no complaint. I set LD_LIBRARY_PATH to the cuda directory so it looks like everything’s going well, no complaints about LD stuff, but then upon running deviceQuery I get the output:
rosaleen : ./deviceQuery
CUDA Device Query (Runtime API) version (CUDART static linking)
cudaGetDeviceCount FAILED CUDA Driver and Runtime version may be mismatched.
Press to Quit…
The RW permissions on /dev/nvidia0 and /dev/nvidiactl are fine - though failing at this point I think is before those even matter.
Any suggestions would be greatly appreciated, as this is keeping my company from looking further into Cuda…