Can't register a custom board for the HDL Deep Learning Toolbox

I was trying to deploy a neural network in my FPGA, since it was not one of the natively supported ones I tried following the steps from this example in the documentation (https://www.mathworks.com/help/deep-learning-hdl/ug/define-custom-board-and-reference-design-for-dl-ip-core-workflow.html). It failed with my board with this error:

Could not apply model's TargetPlatform settings for the following reason:

Invalid value "Trenz TE0729 Starter Kit" for option "Board"

Valid values for "Board" are: Choose a platform; Generic Deep Learning Processor; Generic Deep Learning Processor Intel; Generic Deep Learning Processor Xilinx; Xilinx Kintex-Ultrascale KCU105 evaluation board; Xilinx Zynq UltraScale+ MPSoC ZCU102 Evaluation Kit; Xilinx Zynq ZC706 evaluation kit; Get more...;

Default setting is applied instead.

Please consider making these changes from the model or the HDL Workflow Advisor, and using 'Export to Script...' to

automatically generate the appropriate 'hdlset_param' commands.

I also tried running the example as is, and I believe it works because it defines an already compatible board (the Xilinx Kintex-Ultrascale KCU105 evaluation board). When changing the board name to "Xilinx Kintex-Ultrascale KCU1056 evaluation board" the same error occurs.

Does matlab actually have the option to register a custom board? Am I missing a step?

[Vikram Venkatesh]

Trying out the example with KCU105 should work without having to create any files. Could you try out these two commands and see if you are able to change the board:

hPC = dlhdl.ProcessorConfig;

hPC.TargetPlatform = 'Xilinx Kintex-Ultrascale KCU105 evaluation board';

------------------

To use the "Trenz TE0729 Starter Kit", you will have to create the 4 files shown in the example for this target. You will also have to create the depicted block design in Vivado and export it as a .tcl file. Once done, you will need to save these files in the format below. The folder names can be modified per your convenience, but do make sure to update the modified folder name in the corresponding hdlcoder_ref_design_customization.m and hdlcoder_board_customization.m files.

Answer 1

[Kiran Kintali]

>> Does MATLAB have the option to register a custom board?

Yes, You can see the doc here https://www.mathworks.com/help/hdlcoder/xilinx-fpga-soc-devices.html

To extend support to new hardware, see Deploy IP Core on Custom Hardware.