Intel® Advisor Help

gpu-kernels

Model performance only for code regions running on a GPU.

Syntax

--gpu-kernels

--no-gpu-kernels

Default

Off (no-gpu-kernels)

Actions Modified

collect=projection

Usage

Use this option to model performance of a GPU-enabled application that uses Data Parallel C++, OpenMP* target, or OpenCL™ programming model.

Important

Set up system environment to enable GPU kernel profiling.

Use the --profile-gpu option when running Survey and Trip Counts and FLOP analyses.

Example

Model performance of a GPU-enabled application on a target device and analyze only GPU-enabled kernels.

$ advisor --collect=projection --gpu-kernels --project-dir=./advi_results

See Also