User Guide

Contents

CPU Roofline Accuracy Levels in Command Line

For each perspective,
Intel® Advisor
has several levels of collection accuracy. Each accuracy level is a set of analyses and properties that control what data is collected and the level of collection details. The higher accuracy value you choose, the higher runtime overhead is added.
In CLI, each accuracy level corresponds to a set of commands with specific options that you should run one by one to get a desired result.
The following accuracy levels are available:
Comparison / Accuracy Level
Low
Medium
Overhead
5 - 10x
15 - 50x
Goal
Analyze how well your application uses memory and compute resources of a CPU and determine the main limiting factor (memory bandwidth or compute capacity)
Analyze how well your application uses CPU memory at different cache levels in more details
Analyses
Survey + Characterization (FLOP)
Survey + Characterization (Trip Counts and FLOP with call stacks for all memory levels) + Memory Access Patterns
Result
Cache-aware CPU Roofline for L1 cache
Memory-level CPU Roofline with call stacks (for L1, L2, L3, DRAM)
Memory Access Patterns
You can generate commands for a desired accuracy level from the
Intel Advisor
GUI. See Generate Command Lines from GUI for details.
There is a variety of techniques available to minimize data collection, result size, and execution overhead. Check Minimize Analysis Overhead.

Low Accuracy

To run the
CPU / Memory Roofline Insights
perspective with the low accuracy:
advisor --collect=roofline --project-dir=./advi –- myApplication

Medium Accuracy

To run the
CPU / Memory Roofline Insights
perspective with the medium accuracy:
  1. Generate the Memory-level Roofline report with call stacks:
    advisor --collect=roofline --project-dir=./advi -–stacks --enable-data-transfer-analysis –- myApplication
  2. Mark up loops/functions to run the Memory Access Patterns analysis for. For example, mark up loops that have the
    Possible Inefficient Memory Access Pattern
    issue:
    advisor --mark-up-loops --project-dir=./advi --select=has-issue -- myApplication
  3. Run the Memory Access Pattern analysis for the marked loops:
advisor --collect=map --project-dir=./advi -– myApplication
You can view the results in the
Intel Advisor
GUI or generate an interactive HTML report.

Product and Performance Information

1

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