Intel® Advisor Help
Model performance only for code regions running on a GPU.
--gpu-kernels --no-gpu-kernels |
Off (no-gpu-kernels)
collect=projection
Use this option to model performance of a GPU-enabled application that uses Data Parallel C++, OpenMP* target, or OpenCL™ programming model.
Set up system environment to enable GPU kernel profiling.
Use the --profile-gpu option when running Survey and Trip Counts and FLOP analyses.
Model performance of a GPU-enabled application on a target device and analyze only GPU-enabled kernels.
$ advisor --collect=projection --project-dir=./advi --gpu-kernels