I want to create a profile for use with the Cuda Visual Profiler, but without doing it interactively. I would like to be able to submit a job to the batch system which collects all the data needed for the profiler to view the output. Is this possible?
CUDA does support simple profiling controlled by environment variables which dumps to a log file (see the file “CUDA_profiler.txt” in the SDK). Perhaps you could use this for batch-mode profiling, although I don’t think you can visualize the output with the Visual Profiler.