FPGA AI Suite: Design Examples User Guide

ID 848957
Date 4/22/2025
Public

A newer version of this document is available. Customers should click here to go to the newest version.

Document Table of Contents

8.1.2. [OFS-PCIE] Build Options

To build the runtime for the OFS for PCIe* Attach design example:
  1. Ensure that you have created a working directory as described in "Creating a Working Directory" in the FPGA AI Suite Getting Started Guide .
  2. Ensure that the OPAE_SDK_ROOT environment variable is set in your build environment. For example, export OPAE_SDK_ROOT=/usr/.
  3. Run one of the following sets of commands to build the runtime, depending on your Agilex™ 7 PCIe* Attach OFS board:
    • Agilex™ 7 FPGA I-Series Development Kit
      cd $COREDLA_WORK/runtime
      
      ./build_runtime.sh -target_agx7_i_dk
    • Intel® FPGA SmartNIC N6001-PL Platform
      cd $COREDLA_WORK/runtime
      
      ./build_runtime.sh -target_agx7_n6001

    For other build_runtime.sh options, refer to [PCIE] Build Options.

FPGA AI Suite hardware is compiled to include one or more IP instances, with the same architecture for all instances. Each instance accesses data from a unique bank of DDR.

The Agilex™ 7 FPGA I-Series Development Kit and Intel® FPGA SmartNIC N6001-PL Platform both have four DDR banks (two onboard and two DIMM slots) and support up to four FPGA AI Suite IP instances. Install the required DIMMs into each slot that is used. For example, to support four FPGA AI Suite IP instances, you must have both DIMM slots fitted with 8 GB DIMMs. To support two FPGA AI Suite IP instances, you must have the first DIMM slot fitted with an 8GB DIMM.

The four DDR banks are ordered as follows:
  1. Onboard DDR 0
  2. DIMM slot 0
  3. Onboard DDR 1
  4. DIMM slot 1
Each DIMM slot can support two FPGA AI Suite IP instances.

The runtime automatically adapts to the correct number of instances.

If the FPGA AI Suite runtime uses two or more instances, then the image batches are divided between the instances to execute two or more batches in parallel on the FPGA device.