Profile Applications#

To get the most out of your AI system using OpenVINO, OVMS, or DL Streamer, you may consider profiling it. You can do it with Intel® VTune™ Profiler that will enable you to analyze CPU/GPU usage, kernel execution, threading, memory bottlenecks, and overall performance of your code. Supporting multiple programming languages, it will help you identify inefficient algorithm and hardware usage, as well as give tuning advice.

Check out its User Guide to learn more about its features and workflow, or Get Started right away. The process is easy, for example, based on the DL Streamer Profiling Guide:

  • Download Intel® VTune™ Profiler and install it.

  • Make sure your system has the appropriate Intel CPU/GPU drivers for your hardware.

  • Since ITT instrumentation is enabled in the default builds of DL Streamer, just use the following command:

    GST_DEBUG=3 ./your_pipeline.sh
vtune -collect hotspots -result-dir dlstreamer_results ./your_pipeline.sh

    or, for pipelines that require heavy GPU/accelerator, such as OpenVINO on GPU:

    vtune -collect gpu-hotspots -result-dir dlstreamer_gpu ./your_pipeline.sh

This Page