[Board] Summarize Buildroot question

Wed Nov 8 10:37:55 PST 2023

Thank you to Jeff Miller, who called me a few weeks ago to check in.

Greetings all!

We are still trying to use Transceiver Toolbox to do over-the-air
demonstrations/development of open source transceiver designs at ORI that
use HDL Coder/Simulink. This is a summary of where we're at with an
ADRV9002/zcu102.

For Transceiver Toolbox, we see that the required versions are:
HDL Branch 2021_R2
MATLAB R2022b
Vivado 2021.2
Transceiver Toolbox 22.2.2
HDL Coder Workflow Advisor in MATLAB R2022b has only been tested to Vivado
2020.2

(How concerned about that should I be?)

When last we spoke, ADI's Transceiver Evaluation Software version was
0.23.1.

It's now 0.24.1

Today we installed 0.24.1, and it wants a device driver API version of
68.8.1. From what I can tell, this means Petalinux 2023.1, because the
readme for the Analog Devices linux directory that the 68.8.1 update is in
says that "The current master is based on xilinx v2023.1."

This is ahead of the Vivado/Petalinux version that Transceiver Toolbox and
HDL Coder require (see above)
Last time around, trying versions that don't match results only in pain.
There's yocto layers mismatches, API mismatches, build errors I've never
seen before, or HDL Coder workflow advisor simply doesn't work.

TES 0.23.1 demanded 68.5.0, which was in Petalinux 2022.2, and there was
*no API mismatch*... but this system was not and is not supported by
Transceiver Toolbox or HDL Coder.

TES 0.22.0 would match with 2021.1 Petalinux, 2021.1 HDL reference design,
2021.1 Vivado, Transceiver Toolbox, and MATLAB, and HDL Coder, and
Simulink. This is Device Driver version 68.0.6.

TES 0.22.0 is the only version not available in the Internet Archive, and
Analog Devices flatly refused to provide a copy, even after we offered to
sign a waiver and a senior application engineer personally placed the
request as a trouble ticket.

As most or all of you know, both TES and Transceiver Toolbox are from ADI.
TES is required to generate the configuration for the radio chip and recent
versions also produce a binary firmware file for the ARM.

I have talked to everyone I can think of, including many of you, Xilinx and
ADI booth folks at conferences like IMS, IWRC, etc., and have made multiple
attempts on the "engineer zone", to no avail, since June.

Jeff mentioned there may be another entity with the same challenges? Please
give my contact information to them if there's any chance we can get our
development boards off the ground. It's been a drag to have this be the
hardest part of our SDR development and to essentially stop over the air
demos with MATLAB/Simulink.

We have solid and useful designs, have educated a lot of people at the
non-profit, and have inspired folks to get PhDs, go back to school, get
promotions, get involved in STEM, and more. We want to keep doing all of
this and have things work over the air in the lab again. That hasn't
happened with the 9002/9009 boards we have.

I've cleared out all other work and will be able to spend time working on
this. Advice or updates about things I've missed or misunderstood are
deeply appreciated.

Thank you,
-Michelle Thompson
ORI
+1 858 229 3399

On Wed, Jul 12, 2023 at 5:25 PM Jeff Miller <jmiller at mathworks.com> wrote:

> Michelle,
>
>
>
> Thanks for the details, opening up to see if anyone else on this thread
> can comment.
>
>
>
> All the best,
>
>
>
>
>
> *Jeff Miller*
>
> *Principal Applications Engineer*
>
> *MathWorks*
>
> *jmiller at mathworks.com <jmiller at mathworks.com>*
>
> *508-647-4207*
>
> <https://www.linkedin.com/in/jeff-miller-b818643/>
>
>
>
> https://www.linkedin.com/in/jeff-miller-b818643/
>
>
>
> *From:* Michelle Thompson <mountain.michelle at gmail.com>
> *Sent:* Wednesday, July 12, 2023 7:54 PM
> *To:* Jeff Miller <jmiller at mathworks.com>
> *Cc:* Scott Erickson <serickso at mathworks.com>; TJ Moor <
> tjmoor at mathworks.com>; Robin Getz <rgetz at mathworks.com>; Radu David <
> rdavid at mathworks.com>; Prerak Pradhan <ppradhan at mathworks.com>
> *Subject:* Re: Summarize Buildroot question
>
>
>
> Greetings Jeff,
>
> Had to travel for work, but am back in San Diego now. The team worked on
> the specification and did not make any headway on the build.
>
> I updated the post here:
> https://www.mathworks.com/matlabcentral/answers/1965389-errors-with-mathworks-buildroot-image-compilation
>
> There were some additional machinations required to include the .dts
> files. I wrote these down but can't find them, which is unfortunate. The
> current problem is "how to get the device tree for the radio card into the
> build".
>
> There was a "missing destination file" error in genl that we never solved
> for the zc706+ADRV9009 attempt. From my notes, it looks like the missing
> destination file error didn't happen on the zcu102+ADRV9002. Therefore, I
> didn't include it. It seems like there's a missing argument somewhere in a
> makefile. We think this because of the lines:
>
> install -m 0755 genl /home/abraxas3d...
> install -m 0755   /home/abraxas3d...
>
> The missing destination file operand after '/home/abraxas3d...' does make
> it look like there is a syntax error after the genl.
>
> We searched for packages that contained genl and tried to grep for
> artifacts with grep -r -w genl * | grep -v output and scoured through
> iproute2/nl_protos but turned nothing up there either.
>
> That's the Mathworks buildroot experience summary up to now.
>
> The problem with Analog Devices TES software: it is only being available
> in the very latest and newest version. If we have to use Vivado 2019.1 or
> maybe Vivado 2021.1, this means that the API version in the TES radio card
> profiles (which are required to configure the radio chip and can only be
> obtained from TES) will never match the linux versions required for HDL
> Coder and MATLAB and Mathworks Buildroot.
>
> We have done all we can think of to do to get a slightly older version of
> TES. The version we need is (of course) the only one not available from the
> Internet Archive, not available from Analog Devices even when appealing in
> person at IMS2023, and no one we've asked has a copy lying around. We have
> an API version of 68.0.6 from dmesg, using Petalinux to produce a linux
> build for the zcu102+ADRV9002. We need 68.5.0 in order to use the current
> version of TES, which is 0.23.1. Yes, version zero! It's not even funny.
>
> In order to produce profiles that work with the API version that we have
> to use in order to support Transceiver Toolbox, which is from the same
> company that produces TES, we would need TES version 0.22.0.
>
> Making it a hard requirement to exactly match internal API versions (which
> we assume have something to do with libiio implementation) and not
> providing any utility software except the very latest version honestly
> baffled me. I cannot figure out how to get the new API in the slightly
> older version of Petalinux. Or, into Mathworks Buildroot. Maybe they mean
> for people to only use Kuiper. However, Kuiper is pre-built. They do not
> intend for you to build your own linux and do not support any efforts to do
> so. We tried that path too.
>
> So, we want to solve this and document it, like we spoke about on the
> call. These products really do need to work together better because they
> are all extremely powerful. What can we do to best help, from here?
>
>
> -Michelle Thompson
>
>
>
>
>
>
>
> On Tue, Jun 20, 2023 at 1:16 PM Jeff Miller <jmiller at mathworks.com> wrote:
>
> Michelle,
>
>
>
> Nice talking to you today, if you could summarize your questions and where
> you currently are working through.
>
> -Jeff
>
>
>
> MathWorks buildroot, need to make HDL coder work well with
> FPGA-in-the-loop, experience with Petalinux
>
> Starting building from repo and struggling to build from ZCU
>
> Have been getting help from Prerak, and Radu
>
>
>
> *Jeff Miller*
>
> *Principal Applications Engineer*
>
> *MathWorks*
>
> *jmiller at mathworks.com <jmiller at mathworks.com>*
>
> *508-647-4207 *
>
> <https://www.linkedin.com/in/jeff-miller-b818643/>
>
>
>
> https://www.linkedin.com/in/jeff-miller-b818643/
>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openresearch.institute/pipermail/board-openresearch.institute/attachments/20231108/1cce242e/attachment-0001.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image001.png
Type: image/png
Size: 6313 bytes
Desc: not available
URL: <http://lists.openresearch.institute/pipermail/board-openresearch.institute/attachments/20231108/1cce242e/attachment-0001.png>