Intel® Advisor Help
In some cases, pre-configured accuracy levels of the Intel Advisor are not enough to accurately model performance of your application. You might want to use some advanced custom modeling strategies or fine-tune offload modeling parameters to better adjust the model to your application and for your goal.
Manage invocation taxes to control how to model kernel launch for your application,
Enforce offloading for specific loops to get modeled performance results on a target for specific loops even if the estimated speedup for them is low.
Check how assumed dependencies affect modeling to decide if you need to run the Dependencies analysis to get more accurate modeling results.