Intel® Advisor User Guide

ID 766448
Date 3/31/2023

A newer version of this document is available. Customers should click here to go to the newest version.

Document Table of Contents

Run Offload Modeling Perspective from GUI


  • For a SYCL, OpenMP* target, or OpenCL™ application, do one of the following:
  • In the graphical-user interface (GUI): Create a project and specify an analysis target and target options.

To configure and run the Offload Modeling perspective from the GUI:

  1. Select a baseline device from the drop-down. This is the device that your application runs on for the Intel® Advisor to collect performance data.
    • To analyze an application running on a CPU (for example, C, C++, or Fortran), make sure CPU is selected.
    • To analyze an application running on a GPU (for example, SYCL, OpenMP target, OpenCL), select the GPU baseline device.
      If you select GPU, make sure the GPU Profiling checkbox is enabled under Survey, Characterization, and Performance Modeling analyses.
  2. Configure the perspective and set analysis properties, depending on desired results.
    • Select a collection accuracy level with analysis properties preset for a specific result:

      • Low: Model your application performance for a target device and get the basic low-confidence information about potential speed-up and performance.
      • Medium: Model your application performance and data transfers between host and target devices.
      • High: Model your application performance, data transfers, and memory objects attribution to improve offload modeling accuracy. For application running on CPU, analyze loop-carried dependencies.
    • Select the analyses and properties manually to adjust the perspective flow to your needs. The accuracy level is set to Custom.

    The higher accuracy value you choose, the higher runtime overhead is added to your application. The Overhead indicator shows the overhead for the selected configuration. For the Custom accuracy, the overhead is calculated automatically for the selected analyses and properties.

    The Dependencies analysis (included in the high accuracy for CPU baseline device) adds the highest overhead and is not required if your application is highly parallelized or vectorized on a CPU or if you know that key hotspots in your application do not have loop-carried dependencies. You may need to run it for a CPU application if it has scalar loops/functions or you are not sure about dependencies in your code. See Check How Assumed Dependencies Affect Modeling for a workflow to learn about potential dependencies in your code.

    By default, accuracy is set to Low. See Offload Modeling Accuracy Presets for more details.

  3. Select a target platform from the Target Platform Model drop-down. This is a platform that the Intel Advisor models your application performance on. The following target platforms are available:



    pvc_xt_448xve (default)

    Intel® Data Center GPU Max 448


    Intel® Data Center GPU Max 512

    XeHPG 512

    Intel® Arc™ Graphics with 512 vector engines

    XeHPG 256

    Intel® Arc™ Graphics with 256 vector engines

    Gen11 GT2

    Intel® Iris® Plus Graphics

    XeLP Max 96

    Intel® Iris® Xe MAX Graphics

    XeLP GT2

    Intel® Iris® Xe Graphics

    Gen9 GT2

    Intel® HD Graphics 530

    Gen9 GT3e

    Intel® Iris® Graphics 550

    Gen9 GT4e

    Intel® Iris® Pro Graphics 580

    Multi-tile and multi-GPU analysis for pvc_xt_448xve and pvc_xt_512xve platforms is not supported at the moment.
  4. Click Run to run the perspective.

    While the perspective is running, you can do the following in the Analysis Workflow tab:

    • Control the perspective execution:
      • Stop data collection and see the already collected data: Click the button.
      • Pause data collection: Click the button.
      • Cancel data collection and discard the collected data: Click the button.
    • Expand an analysis with to control the analysis execution:
      • Pause the analysis: Click the button.
      • Stop the currently running analysis and start the next analysis selected: Click the button.
      • Interrupt execution of all selected analyses and see the already collected data: Click the button.

    After you run the Offload Modeling perspective, the collected Survey data becomes available for all other perspectives. If you switch to another perspective, you can skip the Survey step and run only perspective-specific analyses.

To run the CPU-to-GPU Offload Modeling perspective with the Medium accuracy from the command line interface:

advisor --collect=offload --project-dir=./advi_results -- ./myApplication

To run the GPU-to-GPU Offload Modeling perspective with the Medium accuracy from the command line interface:

advisor --collect=offload --gpu --project-dir=./advi_results -- ./myApplication

See Run Offload Modeling Perspective from Command Line for details. See Run GPU-to-GPU Performance Modeling from Command Line for details about the GPU-to-GPU Offload Modeling.

To generate command lines for selected perspective configuration, click the Command Line button.

Once the Offload Modeling perspective collects data, the report opens showing a Summary tab with performance metrics estimated for the selected target platform, such as estimated speedup, potential performance bottlenecks, and top offloaded loops. Depending on the selected accuracy level and perspective properties, continue to investigate the results. See Explore Offload Modeling Results