Intel® Inspector User Guide for Linux* OS

ID 767796
Date 7/13/2023
Public

A newer version of this document is available.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

Document Table of Contents

Choose Data Sets

When you run an analysis, Intel® Inspector executes an application. Data set size and workload have a direct impact on application execution time and analysis speed.

There are three criteria, depending on which you can choose the smallest data sets available:

  • Intel Inspector maintains shadow memory structure to keep extra information on allocated memory blocks and/or synchronization primitives used by the application. Depending on the selected analysis type, the size of those extra details can consume significant amount of process memory in addition to application memory usage. The system should have enough physical memory to run application with extra memory overhead.
  • During threading analysis, Intel Inspector verifies correctness of every potentially shared memory access. If address which is being verified is simultaneously accessed by another thread, the checks will be serialized. The more such intersections at time, the more overhead analysis will add to the application. Reducing number of threads (e.g. by defining OMP_NUM_THREAD variable for OpenMP), might significantly reduce analysis time without significant loss in analysis quality.
  • Intel Inspector intercepts those call for analysis and serialize it in many cases across threads. Reducing frequency of those events in the application will decrease analysis overhead in general and can improve performance of the application when it is run without analysis.

To choose a data set:

Do one of the following:

To Choose a Data Set

Do This

Before running an analysis

  1. Edit source code to select the appropriate data set.

  2. Rebuild the application.

During analysis

Interact with the application during execution to select the appropriate data set.