User Guide

Contents

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.
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 --project-dir=./advi --gpu-kernels

Product and Performance Information

1

Performance varies by use, configuration and other factors. Learn more at www.Intel.com/PerformanceIndex.