clGetPlatformIDs is hanging/not returning

Hi

I’m just getting started and have had no problems getting a simple OpenCL test program (written in C) compiling and linking on our Linux box using the local NVIDIA cuda dev install.

However, when I try running the test application it hangs at the clGetPlatformIDs call. :confused:

I’ve checked in /etc/OpenCL/vendors and can see a single nvidia.icd file containing a reference to libcuda.so - and this DLL exists and is readable in /usr/lib.

I’ve also confirmed that simple native cuda test programs compile and run on this machine (containing a Tesla C2050 card) without any issue.

So as far as I can tell, everything is set up fine and should be working.

Is anyone please able to advise as to why the call to clGetPlatformIDs is not returning?

Thanks in advance! :slight_smile:

you will not need any cuda libs for opencl. better have a look for your opencl.so i think (havent worked on linux for years now)

And if i’m remembering correctly, cuda 5 will not have opencl on board.

[QUOTE=clint3112;29293]you will not need any cuda libs for opencl. better have a look for your opencl.so i think (havent worked on linux for years now)

And if i’m remembering correctly, cuda 5 will not have opencl on board.[/QUOTE]

I don’t think that’s the issue. We have cuda 4.1 installed on our Linux box (with NVIDIA Tesla video card) and I notice that it comes with a bunch of sample OpenCL applications.

All of these build just fine right out of the box but exhibit the same problem.

They all hang on the first OpenCL call.