Profile applications#

Section: vTune

  • Identify other analysis and telemetry tools in use with OpenVINO and DLS

While we have an article in DLStreamer documentation about profiling, we do not have such resources for OpenVINO and OVMS

–>

You can profile OpenVINO, OVMS, and DL Streamer applications using Intel VTune Profiler to analyze CPU/GPU usage, kernel execution, threading, memory bottlenecks, and overall performance.

Preparing for Profiling#

## Profiling OpenVINO apps
Aliqua sit id proident officia ad ex aute deserunt pariatur laboris sunt aliquip esse. Duis officia amet labore adipisicing eu excepteur aliqua. Exercitation laboris laborum nulla sunt dolor ipsum dolore eu. Ea excepteur ea ex et ea ex do pariatur officia culpa commodo mollit sunt.

To learn more, please refer to [XXX Profiling Guide](https://docs.openedgeplatform.intel.com/2025.1/edge-ai-libraries/dl-streamer/dev_guide/profiling.html)

## Profiling OVMS apps
Et labore dolore laboris est. Nostrud minim veniam excepteur adipisicing ipsum. Voluptate ipsum eu culpa nostrud amet aliquip occaecat. Esse deserunt proident nostrud mollit eu minim Lorem reprehenderit nostrud reprehenderit pariatur. Anim quis esse aliquip aliquip tempor excepteur sit et.

To learn more, please refer to [XXX Profiling Guide](https://docs.openedgeplatform.intel.com/2025.1/edge-ai-libraries/dl-streamer/dev_guide/profiling.html)

Profiling DLStreamer apps#

Intel VTune™ uses ITT interface to capture custom tasks and visualize them on the Profile tab. ITT instrumentation is enabled in the default builds of Deep Learning Streamer.

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

For GPU/accelerator-heavy pipelines (e.g., OpenVINO on GPU), use:

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

To learn more, please refer to DLStreamer Profiling Guide