08-28-2019 09:13 AM
I try DPU Integration on Vivado and Petalinux 2019.1
I can vivado commpile, Petalinux commpile and I move img on Ultra96
ardunizp xilinx_dnndk_v3.1_190809.tar.gz and copy Ultra96 folder to SD card.
Power on Ultra96 and cd Ultra96 and ./install.sh , but it happns install error.
on Ultra96 , I move to Ulta96 forlder and run ./install.sh
But I cannot install drivers.
I try modbole and display message.
root@dpu3v1linux:~/Ultra96# modprobe dpu.ko
[ 259.777464] dpu: version magic '4.19.0-xilinx-v2019.1 SMP mod_unload aarch64' should be '4.19.0+ SMP mod_unload aarch64'
[ 259.790089] dpu: version magic '4.19.0-xilinx-v2019.1 SMP mod_unload aarch64' should be '4.19.0+ SMP mod_unload aarch64'
modprobe: can't load module dpu.ko (extra/dpu.ko): invalid module format
Why diffelent version?
09-01-2019 03:33 AM
09-05-2019 04:44 AM
We are having the same issue and error. Could you please provide futher details on how did you resolve this?
1. Did you use the bsp provided in the TRD for ZCU102 to build the petalinux project?
2. Which DPU driver file (dpu.ko) did you use?
Thanks in advance
09-05-2019 09:14 AM
Hi email@example.com - Can you please elaborate on what you're trying to do specifically? Are you trying to boot the pre-built DNNDKv3.1 Linux images and run the included examples on the Ultra96, or are you trying to go through a custom Vivado->Petalinux flow for Ultra96?
09-11-2019 07:40 AM
I'm trying to go through a custom Vivado->Petalinux flow for Ultra96
Sorry, it took me a while but finally i got a working solution by using the files and instructions given in https://github.com/Xilinx/Edge-AI-Platform-Tutorials/tree/master/docs/DPU-Integration
Now i have a design with dpu version 2.0 and instruction set 1.4.0. For instruction set 1.4.0 you need Dnndk v3.1. Dnndk v3.1 is provided here https://www.xilinx.com/products/design-tools/ai-inference/ai-developer-hub.html#edge
First i tried with trd dpu dpu version v3.0 from this site https://www.xilinx.com/products/design-tools/ai-inference/ai-developer-hub.html#edge but this won't work with kernel module of https://github.com/Xilinx/Edge-AI-Platform-Tutorials/tree/master/docs/DPU-Integration . And i had to use the kernel module source of https://github.com/Xilinx/Edge-AI-Platform-Tutorials/tree/master/docs/DPU-Integration because i couldn't find a binary dpu.ko for petalinux 2018.3
By know there's no petalinux 2019.1 bsp for ultra96 so i also can't use the dpu.ko provided for 2019.1
09-11-2019 08:11 AM
To use 2019.1, you need to use the libs/IP/driver from the latest DPU TRD for DNNDKv3.1 which use DPUv3.0. I'm working on an update to the DPU Integration Tutorial but it will be a few weeks before it's ready.
The current tutorial uses 2018.2, DNNDKv3.0, DPUv2.0 with instruction set 1.4.0 - if you can stick with that for the time being, that would be easiest. Otherwise, you can take the libs/IP/driver from the latest TRD and modify the tutorial to use them.