- The oneVPL dispatcher automatically finds all available accelerators during runtime.
- Dispatcher uses the selected accelerator context to initialize a session.
- oneVPL configures the video component at the start of the session.
- oneVPL processing loop is launched. The processing loop handles work asynchronously.
- If the application chooses to let oneVPL manage working memory, then memory allocation will be implicitly managed by the video calls in the processing loop.
- After work is done, oneVPL uses a clear call to clean up all resources.