Intel® Advisor Help
Threading perspective includes several steps that you are recommended to run one by one:
Collect performance metrics and find candidates for parallelizing using a Survey analysis.
Annotate manually loops/functions to model parallelization for.
Model parallel design options and estimate speedup for the annotated loops using a Suitability analysis.
Check for loop-carried dependencies to make sure the loops/functions are safe to parallelize.
Set Intel Advisor environment variables with an automated script to enable the advisor command line interface (CLI).
Run the Survey analysis:
advisor --collect=survey --project-dir=<project-dir> -- <target-application> [<target-options>]
Run the Characterization analysis to collect trip counts and/or FLOP data:
advisor --collect=tripcounts --project-dir=<project-dir> [--flop] [--stacks] -- <target-application> [<target-options>]
where:
--flop is an option to collect data about floating-point and integer operations, memory traffic, and mask utilization metrics for AVX-512 platforms.
--stacks is an option to enable advanced collection of callstack data.
View the Survey report to identify candidates for parallelization. For example, run the following command to print the report in command line:
advisor --report=survey --project-dir=<project-dir>
Consider analyzing loops/functions with high total time.
In the application source code, annotate loops/functions of interest to model parallelization for. Rebuild the application.
Run the Suitability analysis for the annotated loops/functions:
advisor --collect=suitability --project-dir=<project-dir> -- <target-application> [<target-options>]
Run the Dependencies analysis to check for loop-carried dependencies in the annotated loops:
advisor --collect=dependencies --project-dir=<project-dir> [--filter-reductions] -- <target-application> [<target-options>]
where:
--filter-reductions is an option to mark all potential reductions with a specific diagnostic.
See advisor Command Line Interface Reference for more options.
Example
Run the Survey, Trip Counts, Suitability, and Dependencies analyses assuming you have annotated the loops.
advisor --collect=survey --project-dir=./advi -- myApplication
advisor --collect=tripcounts --project-dir=./advi --flop --stacks -- myApplication
advisor --collect=suitability --project-dir=./advi -– myApplication
advisor --collect=dependencies --project-dir=./advi --filter-reductions -– myApplication
Intel Advisor provides several ways to work with the Threading results.
View Results in CLI
You can print the results collected in the CLI and save them to a .txt, .csv, or .xml file.
Run the following command:
advisor --report=<analysis-type> --project-dir=<project-dir> --format=<format>
where:
<analysis-type> is the analysis you want to generate the results for. For example, survey for the Survey report, suitability for the Suitability report, or dependencies for the Dependencies report.
--format=<format> is a file format to save the results to. <format> is text (default), csv, xml.
If you generate the Suitability report, you can use additional options to control the result view:
--target-system=[cpu | xeon-phi | offload-to-xeon-phi] is a platform to model parallelization on.
--threading-model=[tbb | cilk | openmp | tpl | other] is a threading model to use.
--reduce-site-overhead=<string> is a list of annotated loops/functions to check if you can reduce overhead.
See advisor Command Line Interface Reference for more options.
For example, to generate the Suitability report for the OpenMP* threading model:
advisor --report=suitability --project-dir=./advi -–threading-model=openmp
You should see a similar result:
Target CPU Count: 8 Threading Model: OpenMP* Maximum gain for all sites: 6.10998 All Sites Site Label Source Location Impact to Total Serial Time Total Parallel Time Site Gain Average Serial Time Average Parallel Time Number of Program Gain Instances __________________________________________________________________________________________________________________________________________________________________ solve nqueens_serial.cpp:154 6.11x 4.080s 0.631s 6.47x 4.080s 0.631s 1 Site Details Annotation Annotation Label Source Location Number of Instances Maximum Instance Average Serial Minimum Instance Total Serial Time Time Time Time __________________________________________________________________________________________________________________________________________________________________ Selected Site solve nqueens_serial.cpp:154 1 4.080s 4.080s 4.080s 4.080s Task setQueen nqueens_serial.cpp:156 14 0.477s 0.267s 0.020s 3.734s Lock ? 365596 < 0.001s < 0.001s < 0.001s 0.100s Site Options Benefit Loss If Site Option Done? If Done Not Done Recommended _________________________________________________________________________ solve Reduce Site Overhead No solve Reduce Task Overhead No solve Reduce Lock Overhead No solve Reduce Lock Contention 0.16x No solve Enable Task Chunking No
The result is also saved into a text file advisor-survey.txt located at <project-dir>/eNNN/hsNNN.
advisor --report=joined --report-output=<path-to-csv>
where --report-output=<path-to-csv> is a path and a name for a .csv file to save the report to. For example, /home/report.csv. This option is required to generate a joined report.
See CPU and Memory Metrics for more information about the metrics reported.
View Results in GUI
When you run Intel Advisor CLI, a project is created automatically in the directory specified with --project-dir. All the collected results and analysis configurations are stored in the .advixeproj project, that you can view in the Intel Advisor.
To open the project in GUI, you can run the following command:
advisor-gui <project-dir>
You first see a Threading Summary report that includes the overall information about loops/functions performance in your code and the annotated parallel sites:
Performance metrics of your program and top five time-consuming loops/functions
Optimization recommendations for the whole application
Estimated performance gain for annotated loops/functions when parallelized
Save a Read-only Snapshot
A snapshot is a read-only copy of a project result, which you can view at any time using the Intel Advisor GUI. To save an active project result as a read-only snapshot:
advisor --snapshot --project-dir=<project-dir> [--cache-sources] [--cache-binaries] -- <snapshot-path>
where:
--cache-sources is an option to add application source code to the snapshot.
--cache-binaries is an option to add application binaries to the snapshot.
<snapshot-path is a path and a name for the snapshot. For example, if you specify /tmp/new_snapshot, a snapshot is saved in a tmp directory as new_snapshot.advixeexpz. You can skip this and save the snapshot to a current directory as snapshotXXX.advixeexpz.
To open the result snapshot in the Intel Advisor GUI, you can run the following command:
advisor-gui <snapshot-path>
You can visually compare the saved snapshot against the current active result or other snapshot results.
Continue to model threading results. For details about the metrics reported, see CPU and Memory Metrics.