Analyzing AI Engine Status in Hardware Emulation - 2022.2 English

AI Engine Tools and Flows User Guide (UG1076)

Document ID
UG1076
Release Date
2022-10-19
Version
2022.2 English

Generating and Analyzing AI Engine Status

During hardware emulation, status output can be generated periodically. The status and warnings can be analyzed in the Vitis™ Analyzer for debug purposes.

  1. Add the following options in your xrt.ini to output the AI Engine status. This is a one-time set up that results in periodic output of status data, including deadlock detection.
    • aie_status=true: Turns on the output generation feature.
    • aie_status_interval_us: Specifies the interval of the AI Engine status output.
      Example:
      [Debug]
      aie_status=true
      aie_status_interval_us=1000

    The following files are generated by hardware emulation.

    • xrt.run_summary: A summary file generated by hardware emulation that can be opened in Vitis Analyzer.
    • aie_status_edge.json: A status of AI Engine and AI Engine memory.
    • aieshim_status_edge.json: A status of AI Engine interface tiles.
    • summary.csv: A summary CSV file.
  2. The summary and JSON files must be copied from the QEMU Linux system to your local system using the scp command. The status files must be in a local directory prior to the analysis in Vitis Analyzer. For more details on how to copy these files to your local system, see Running Emulation on an Embedded Processor Platform in Vitis Unified Software Platform Documentation: Application Acceleration Development (UG1393)
  3. Analyze the summary and JSON files in Vitis Analyzer. For details, see Analyzing AI Engine Status in Vitis Analyzer.