Partitioning

Partitioning#

This section comprises two pages: Summary and Graph.

Summary#

The Summary page gives an overview of how the model’s operators have been assigned to AMD Vitis™ AI CPU and NPU along with charts capturing GigaOp (GOP) offloading by operator type.

There is also a table titled “CPU Because” that shows the reasons why certain operators were not offloaded to the NPU.

../../_images/AIAnalyzer_PartitioningSummary.png

There are 6 charts displayed on the Summary page (charts or table):

  • Operator Count: Displays the number of operators assigned to the CPU and NPU, respectively.

  • NPU Operators: Displays the operators assigned to the NPU, broken down by operator type.

  • CPU Operators: Displays the operators assigned to the CPU, broken down by operator type.

  • GOPs: Displays the total GOPs assigned to the CPU and NPU, respectively.

  • CPU GOPs by Operator Type: Displays the GOPs assigned to the CPU, broken down by operator type.

  • NPU GOPs by Operator Type: Displays the GOPs assigned to the NPU, broken down by operator type. (Not currently visible)

Graph#

The Graph page shows an interactive diagram of the partitioned ONNX model, showing graphically how the layers are assigned to the NPU and the CPU.

Toolbar#

  • You can choose to show or hide individual NPU partitions, if any, with the “Filter by Partition” button.

  • You can show or hide a panel that displays properties for selected objects through the “Show Properties” toggle button.

  • You can show or hide the model table through the “Show Table” toggle button.

  • You can show or hide the Processor, Partition, and Instance Name through the “Filter Graph” toggle button.

Filter Graph#

  • Show Processor: Separates operators that run on CPU and NPU respectively.

  • Show Partition: Separates operators running on the NPU by their respective NPU partition, if any.

  • Show Instance Name: Displays the full hierarchical name for the operators in the ONNX model.

All objects in the graph have properties that can be viewed to the right of the graph.

Model Table#

The table following the graph lists all objects in the partitioned ONNX model:

  • Processor (NPU or CPU)

  • Function (Layer)

  • Operator

Contents