FPGA AI Suite: IP Reference Manual

ID 768974
Date 3/29/2024
Public
Document Table of Contents

4. FPGA AI Suite Ahead-of-Time Splitter Utility

Typically, the FPGA AI Suite runtime OpenVINO™ plugin provisions the control-and-status registers (CSR) and memory needed to run inference. If you want to run inference on the FPGA AI Suite IP without using the OpenVINO™ Inference Engine, you can use the ahead-of-time (AOT) splitter utility (dla_aot_splitter) to extract the raw data that is provisioned in memory.

An example system where you might not want to use the OpenVINO™ Inference Engine is when you have an embedded soft processor that manages a FPGA AI Suite IP instance in the FPGA fabric. The embedded system can provision DDR memory with the files generate by the AOT splitter and configure the IP to use the provisioned data.

Review the source code provided in dla_aot_splitter_example to see how to use the files that are generated by the splitter utility in an application that performs inference without using the OpenVINO™ Inference Engine.