I am trying to utilize the C++ API for OpenCL. I have installed my NVIDIA drivers and I have tested that I can run the simple vector addition program provided on 'thebigblob’s ‘getting started with opencl’ post. I can compile this program with following gcc call and the program runs without problem.
gcc main.c -o vectorAddition -l OpenCL -I/usr/local/cuda-7.0/include
However, I would very much prefer to use the C++ API as opposed the very verbose host files needed for C.
I downloaded the C++ bindings from Khronos (version 1.1) and placed the
cl.hpp file in the same location as my other
cl.h file. I then tried the code provided in the ‘Intro to OpenCL tutorial’. I set LD_LIBRARY_PATH = /usr/local/cuda-7.0/lib64 and copiled with the following call.
g++ lesson1.cpp -o hello_world -l OpenCL -I/usr/local/cuda-7.0/include
It compile fine but when I try and run ./hello_world I receive the following error:
ERROR: cl::Platform::get (-1)
Have I missed something?
I can’t figure how to edit my original post so here is an update regarding this problem.
I learned that my graphics card cannot use the most recent NVIDIA driver so I purged and installed the nvidia-340 drivers. I then installed cuda-6.5 which is compatible with the nvidia-340 driver. Modifying the initial compilation above it is now:
gcc main.c -o vectorAddition -l OpenCL -I/usr/local/cuda-6.5/include
Again this compiles but I also modified it to return the number of platforms found which strangely returns 0. Have I missed something here that my code cannot find my graphics card?
My workflow from clean install:
- sudo apt-get install nvidia-340
- Download 6.5 cuda .run file and install (skipping driver install)
- Update PATH and LD_LIBRARY_PATH environmental variables
- compile main.c and execute
I did manage to solve my problem. It appears necessary to avoid doing any installation with apt-get calls. After purging all cuda and nvidia implementations I reinstalled everything with the cuda run file. This got everything to install correctly and the files compiled and executed correctly.