Intel® Advisor User Guide

ID 766448
Date 3/22/2024
Public

A newer version of this document is available.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

Document Table of Contents

gpu

With Offload Modeling perspective, analyze OpenCL™ and oneAPI Level Zero programs running on Intel® Graphics. With GPU Roofline perspective. create a Roofline interactive HTML report for data collected on GPUs.

GUI Equivalent

For Offload Modeling perspective: Analysis Workflow > Baseline Device > GPU

For Offload Modeling perspective: Project Properties > Analysis Target > Performance Modeling > GPU

Syntax

--gpu

--no-gpu

Default

Off (no-gpu)

Actions Modified

collect=offload

report=roofline

Usage

For Offload Modeling perspective:

Use this option if you want to run GPU-to-GPU modeling using the collection presets. This option automatically applies the --profile-gpu option to all analyses it runs.

For GPU Roofline perspective:

Prerequisites: Collect Roofline data with --profile-gpu option enabled.

The default Roofline interactive HTML report is created for data collected on CPUs. To create a report for data collected on GPUs, use this option.

Example

Run the GPU-to-GPU Offload Modeling with collection preset.

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

Generate a Roofline interactive HTML report for data collected on GPUs.

advisor --report=roofline --report-output=./out/roofline.html --gpu --project-dir=./advi_results

See Also