cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 
qwqw0001
Contributor
Contributor
717 Views
Registered: ‎07-15-2019

VITIS-AI and HLS

Jump to solution

hi,

   I have used VITIS-AI to deploy my model on ZCU102 , but I found my post-processing is too low. My model on DPU only need 0.02s,but my post-processing  need 2s! So I want to use HLS to accelerate my post-processing, do you have any advice or tutorials?

    In addition, I have used HLS on PYNQ, so I only know to allocate a buffer through xlnk.cma_array, and then assign the physical_address of the buffer to the address of the IP core generated by HLS through mmio.write.    If I want to use ZCU102,how do I connect my HLS and DPU, or how to assign the DPU result to the HLS IP core?

    Is it feasible for HLS ip core to process data through handshake signals such as start, idle, ready, and done on the PL side without ARM?

   ylh

Tags (1)
0 Kudos
1 Solution

Accepted Solutions
idiotic_genius
Adventurer
Adventurer
667 Views
Registered: ‎07-20-2017
Yes deploying you post-processing to PL will definitely improve the overall model speed. For the first portion of the question check this:
https://xilinx.github.io/Vitis_Libraries/vision/2020.1/overview.html#x-mlpipeline
This shows the preprocessing done in PL using Vitis Vision Libraries.
You should be able to access the HLS results in Linux environment and pass the output to the DPU through APIs. I think you will have to write applications using DPU APIs. This tutorial might help: https://xilinx-wiki.atlassian.net/wiki/spaces/A/pages/235602018
It uses ZCU106 VCU to encode the Region of Interest extracted by DPU.
Regards, abhidan@logictronix.com
Please mark the Answer as "Accept as solution" if information provided solves your query. Give Kudos if you think it was helpful and reply oriented.

View solution in original post

0 Kudos
1 Reply
idiotic_genius
Adventurer
Adventurer
668 Views
Registered: ‎07-20-2017
Yes deploying you post-processing to PL will definitely improve the overall model speed. For the first portion of the question check this:
https://xilinx.github.io/Vitis_Libraries/vision/2020.1/overview.html#x-mlpipeline
This shows the preprocessing done in PL using Vitis Vision Libraries.
You should be able to access the HLS results in Linux environment and pass the output to the DPU through APIs. I think you will have to write applications using DPU APIs. This tutorial might help: https://xilinx-wiki.atlassian.net/wiki/spaces/A/pages/235602018
It uses ZCU106 VCU to encode the Region of Interest extracted by DPU.
Regards, abhidan@logictronix.com
Please mark the Answer as "Accept as solution" if information provided solves your query. Give Kudos if you think it was helpful and reply oriented.

View solution in original post

0 Kudos