Possibly someone here can help me, have followed this guide and successfully running
- Win7 VM with AMD card passed throug
- Fedora 18 VM with Nvidia Quadro 2000 card passed through
Trying to run a custom weather forecast which uses the Cuda capability of the card on the F18 VM.
- Latest Nvidia drivers 319.32
- Cuda toolkit 5.5
- Most Cuda toolkit samples will compile and run
However the output from the forecast programme is as follows:
NCL VERSION 6
USING NCL wrf_user_getvar()
Module ncl_jack_fortran.so compiled at Tue Apr 24 14:30:48 2012, expires 2017-12-06
Drjack Cuda routines Init: Module ncl_jack.so compiled for Cuda computelevel 2.1
with opts " --generate-code arch=compute_20,code=sm_21 "
Found 1 cuda devices
using Device 262144: "Quadro 2000"
Cuda ERROR: invalid device ordinalcudaSetDevice failed Drjacks wrf_user init...Complete
Output files in ./NCL
The key item is the ":invalid device ordinal" with the device number as Device "262144" which should be 0 typically I'm told.
Could xen be causing this?
Are there further settings on gpu passthrough?