Has anyone tried new CUDA 5.0’s command line profiler? In earlier version it worked well for both CUDA and OpenCL. But in 5.0, when tested with GTX680, I can only produce the output file (cuda_profile_0.log) for cuda programs. For OpenCL programs, no profile output is produced. Am I missing any configuration or something like that? Or is that just there is no support for OpenCL profiling in CUDA 5.0?
Has anyone tried new CUDA 5.0’s command line profiler? In earlier version it worked
Share
OpenCL is not supported by nvprof. Please note that nvprof is a new tool in CUDA Toolkit 5.0 so it never supported OpenCL.