Intel® Advisor Help
Any type of analysis and data collection that can be performed through the Intel® Advisor GUI can also be performed through the command line, using the same workflow. Commands are most commonly used in scripts to run analysis remotely. The advixe-cl command line tool is the only way to examine your MPI application for parallelism opportunities.
To perform an analysis through the command line, use the advixe-cl command tool with the collect action and the analysis type that you want to run.
The collect action runs the analysis, collects and finalizes data, then creates a result in your project directory. The project directory contains three experiment directories, one for each of the three analysis tools: Survey, Suitability and Dependencies. Each experiment directory contains a single result.
Specify the search directory when performing a collect action, as described in Specifying Directories.
When running analysis remotely, you may want to create an option file, then specify your option file on the command line using the option-file action-option.
This command executes the actions that are specified in an option file named my_suitability_analysis.
$ advixe-cl -report suitability -option-file ../my_suitability_analysis
To view a result, you can open your project in the GUI, or use the report action and write the output to a file.
Double-click on a project to open it in the GUI, or use this command to launch the GUI:
$ advixe-gui ./myProject
By default, a report is written to standard output in text format, but the advixe-cl tool provides a number of options you can use when generating a report.
To write a report to a file, use the report-output option.
This example generates a Dependencies report with tab delimiters and saves it as advisor-Dependencies.csv.
$ advixe-cl -report Dependencies -project-dir ./advi -format csv -csv-delimiter tab -report-output=./out/advisor-Dependencies.csv
To begin the Intel Advisor workflow, perform a survey. Run the collect action with the survey analysis type. After running the survey, you can generate a Survey report, using the report action with the survey report type.
For more information, see Viewing Data from the Survey Tool.
The Suitability tool examines your running program to provide approximate performance characteristics for annotated parallel sites. Suitability data helps to predict the potential for performance gains when running the parallel program on multiple CPUs, and the potential impact of parallel overhead.
To perform a Suitability analysis, run the collect action with the suitability analysis type. There are a number of report types supported for suitability results. Help for the report action lists all compatible report types.
For more information, see Viewing Data from the Suitability Tool.
The Dependencies tool analyzes your running program to predict data sharing problems.
Run a collect action with the Dependencies analysis type. To generate a report, use the report action with the Dependencies report type.
For more information, see Viewing Data from the Dependencies Tool.
After annotating your code, run the Annotation report to get a list of annotations and their types. Use the collect action with either a suitability or Dependencies analysis type for collection. Use the report action with the annotation report type to generate a report.
For more information, see Window Annotation Report.
A Summary report provides an overview of the analysis result. Use any of the three analysis tools during collection: survey, suitability or Dependencies. Use the summary report type when generating your report.
For more information, see Window Summary.
MPI analysis uses the same general workflow, but somewhat different command syntax, since the MPI process is called first and then spawns the advixe-cl tool. The result can be viewed in the GUI or on the command line. For more information, see Analyzing MPI Workloads.