nhma20/U96_DPU

Implementing Deep Learning Processing Unit (DPU) on the Ultra96V2 board

U96_DPU

Implementing Deep Learning Processing Unit (DPU) on the Ultra96V2 board

Requires:

• Vitis/Vivado 2021.1 Tools
• Petalinux 2021.1

Build process from: https://www.hackster.io/AlbertaBeef/vitis-ai-1-4-flow-for-avnet-vitis-platforms-5ebc3f

0. Clone all needed repos:

   git clone https://github.com/Avnet/bdf

   git clone -b 2021.1 https://github.com/Avnet/hdl

   git clone -b 2021.1 https://github.com/Avnet/petalinux

   git clone -b 2021.1 https://github.com/Avnet/vitis

   In /vitis/projectMakefile.mk add "-j 10" after each $(MAKE) to avoid crashing. Also, in line 144 source ./add_petalinux_packages.sh replace source with . or bash. This file determines what happens in each "step" (xsa, plnx, dpu etc.)

1. In vitis/ directory do:

   make u96v2_sbc step=xsa (may have to add new line 95 in hdl/scripts/make.tcl return 10 to avoid using too many ressources and crashing machine)

   This creates new directory hdl/projects/u96v2_sbc_base_2021_1 with Vivado project which can be opened.

2. Next, with petalinux installed as ~/petalinux/, do:

   source ../../petalinux/settings.sh

   make u96v2_sbc step=plnx

3. Next do:

   make u96v2_sbc step=sysroot

4. Next do:

   make u96v2_sbc step=pfm

5. Next step can take a long time. Do:

   make u96v2_sbc step=dpu

• 1x B2304, low RAM, low DSP, 2x gated clk
• clk_dsp = 2x clk_dpu