Since CUDA 3.1, computeprof is included in the Mac distro, too. However, I was never able to profile an OpenCL application (neither mine nor the SDK examples). On the other hand, Nvidia seems not to ship a Mac version of OpenCL, so the only OpenCL platform available is Apple.
Is OpenCL supposed to work with computeprof on MacOS? If not, this is perhaps important enough to put it into release notes (the only pecularity for MacOS mentioned here is that help does not work…)
Tried: CUDA 3.1 (driver 3.1.10, 3.1.14, 3.1.17) on MacBook Pro (8600M GT) and Mac Pro (8880 GT). Also 3.2RC on the latter.
Any solution found for this??? I am having the same issue trying to get computeprof to work on the OpenCL SDK examples. I was able to get it running fine on a Ubuntu 9.04 box.
Computeprof runs fine with the CUDA example but I am getting an error: “Error in reading profiler output” running for an OpenCL example.
Another note, Nvidia does ship an OpenCL implementation with its SDK, you should be able to fine it in /Developer/GPU Computing/OpenCL
Any solution found for this??? I am having the same issue trying to get computeprof to work on the OpenCL SDK examples. I was able to get it running fine on a Ubuntu 9.04 box.
Computeprof runs fine with the CUDA example but I am getting an error: “Error in reading profiler output” running for an OpenCL example.
Another note, Nvidia does ship an OpenCL implementation with its SDK, you should be able to fine it in /Developer/GPU Computing/OpenCL
Any solution found for this??? I am having the same issue trying to get computeprof to work on the OpenCL SDK examples. I was able to get it running fine on a Ubuntu 9.04 box.
Computeprof runs fine with the CUDA example but I am getting an error: “Error in reading profiler output” running for an OpenCL example.
Another note, Nvidia does ship an OpenCL implementation with its SDK, you should be able to fine it in /Developer/GPU Computing/OpenCL