How xilinx/plnx-aie-examples works
Project type and purpose
• This is an embedded hardware–software co-design example repository for Xilinx/AMD devices. It demonstrates how to build and run an AIE-accelerated GeMM (General Matrix Multiply) application on a programmable SoC platform using Vitis and PetaLinux.
• Key purpose: illustrate end-to-end flow from AIE graph compilation and PL bitstream generation to PS application deployment (baremetal or Linux) and runtime data movement via GMIO.
File structure
• Top level
• designs/xgemm-gmio – the full AIE-GeMM design
• platforms – platform creation Tcl and Makefile for board support
• scripts/build_*.sh, settings.sh – build-flow bootstrap
• designs/xgemm-gmio subfolders
• aie/ – AI Engine dataflow graph code and kernels
• hw/ – PL system configuration (system.cfg) and Makefile to assemble xclbin
• ps/ – PS application for baremetal and Linux (linker scripts, Makefiles, xrt.ini)
• sw/ – helper scripts or host-side software (if any)
Extract architectural insights from README
• AIE graph defined in xgemm.h/xgemm.cpp, built with kernels in aie/kernels.
• system.cfg bridges PL (AXI, GMIO) and AIE routers.
• Two PS OS targets: Linux (using XRT, PetaLinux BSP) and baremetal.
• Build steps generate base platform, compile AIE, create PL bitstream, build PetaLinux, cross-compile PS, and package boot images.
Essential elements
• Clear labels: “Host Tools”, “PS (ARM+A53)”, “PL (FPGA)”, “AIE Array”, “DDR Memory”
• Rectangles in blue for hardware blocks (PL, AIE, DDR)