• MorphoOptimizationProject_usingVTune

THIS PAGE HAS BEEN MOVED TO SHAREPOINT!

Please refer to this site/make edits here for the most updated information: https://partnershealthcare.sharepoint.com/sites/LCN/SitePages/Morpho-Optimization-Project.aspx

Parent: See MorphoOptimizationProject_improvingParallelism for an example of using it. This page contains more specific details and more sophisticated uses.

Setting PATH etc.

source /opt/intel/parallel_studio_xe_2019/bin/psxevars.sh

Gathering high resolution data

Obviously the more samples per second are gathered, the larger the data files are going to be. So the solution is to control the start time, the duration, and the resolution to see the interval needed at the resolution needed.

The VTune collection command has three command switches to do this

1. --resume-after 14.9
2. --duration 15
3. -knob sampling-interval=0.1

The resume-after is measured in seconds, commencing when the start or attach is done. It can be a decimal number such as 4.5.

The duration is the end time, also measured in seconds, and commencing at the same place as the resume-after. Currently it must be an integer.

The sampling-interval is measured in milliseconds and can be a decimal number. Collections at 0.1 ms are possible.

Gathering high resolution data with a high resume-after

I have reported a bug to Intel because some data is gathered starting with the start or attach. This bug means the following method needs to be employed to gather high resolution data more than a few seconds into the execution.

On one terminal start the application and then pause it when it reaches the interesting point. Sadly this will require either watching the output or adding some kind of stalling mechanism to the code.

../mris_fix_topology -mgz -sphere qsphere.nofix -ga -seed 1234 bert lh
^Z

On another terminal, enter the collection command with the -target switch, but do not press ENTER yet

amplxe-cl -collect advanced-hotspots \
          -knob collection-detail=stack-sampling -knob analyze-openmp=true \
          -knob enable-characterization-insights=false \
          -data-limit=5000 -finalization-mode=full \
          --duration 1 -knob sampling-interval=0.1 \
          -target-process=mris_fix_topology

Once the command is ready, resume the application in the first terminal

fg

and immediately press ENTER in the second terminal to attach the collector to it and gather the data