**** BEGIN LOGGING AT Wed Apr 07 02:59:57 2021
Apr 07 04:57:52 * satacker[m]  < https://matrix.org/_matrix/media/r0/download/matrix.org/QAaAZZeoSLrpaEBhVCJQQMYI/message.txt >
Apr 07 04:58:17 <satacker[m]> * Abhishek  could you let me know if the beagle config proposal is good to go or there are any more things i need to look for?
Apr 07 05:29:50 <Abhishek[m]_> satacker: I’ve not been well in the past few days. Will have limited availability
Apr 07 05:30:02 <Abhishek[m]_> Shall I look at your Google doc?
Apr 07 05:30:59 <Abhishek[m]_> <Abhishek[m]_ "satacker: I’ve not been well in "> Also FYI for other students who would like their proposals reviewed
Apr 07 05:31:23 <DhruvaGole[m]> Take care Abhishek !
Apr 07 05:31:45 <Abhishek[m]_> Thanks Dhruva
Apr 07 05:38:29 <satacker[m]> <Abhishek[m]_ "satacker: I’ve not been well in "> Yes docs.
Apr 07 07:09:24 <ArchismanDey[m]> vedant16: I have added the ideas about adding chars, arrays, and sending things other than ints through rpmsg. Please review my proposal here: https://elinux.org/BeagleBoard/GSoC/2021_Proposal/simpPRU_Improvements Abhishek: Sorry to hear that, hope you get well soon and please look at my proposal whenever you feel better.
Apr 07 07:10:07 <vedant16[m]> cool Archisman Dey I'll look later
Apr 07 07:10:23 <vedant16[m]> Abhishek: Take care and hope you recover soon :)
Apr 07 07:11:42 <Abhishek[m]_> Archisman Dey: copy your proposal in a Google doc and share it with me via email please - is useful to keep content and comments in one place.
Apr 07 07:12:14 <Abhishek[m]_> Keep the eLinux version in sync with the doc if you make changes
Apr 07 07:12:19 <ArchismanDey[m]> <Abhishek[m]_ "Archisman Dey: copy your proposa"> Sure, I will
Apr 07 07:12:53 <Abhishek[m]_> Note that elinux is the official version that will be considered while evaluating
Apr 07 07:13:14 <Abhishek[m]_> GDocs is something I am doing for projects I am mentoring
Apr 07 07:14:51 <Abhishek[m]_> You can get my email from a fellow mentor or satacker
Apr 07 07:15:12 <ArchismanDey[m]> <Abhishek[m]_ "GDocs is something I am doing fo"> Okay, understood
Apr 07 08:46:42 <lorforlinux[m]> Get well soon Abhishek
Apr 07 08:54:22 <ArchismanDey[m]> Abhishek: shared to you via email
Apr 07 08:59:21 <ArchismanDey[m]> Here's the link with comment access if anyone else wants to see: https://docs.google.com/document/d/1SOOft2ARtzk0hy2NcCcSO52mhFWA7P6t85m6N3ZA-Co/edit?usp=sharing
Apr 07 10:10:25 * Abhishek[m]_ looks at Etherpad
Apr 07 10:11:07 <Abhishek[m]_> It’s an OSS Google docs alternative
Apr 07 10:11:26 <Abhishek[m]_> Maybe other mentors will find it better
Apr 07 10:18:34 <satacker[m]> Abhishek thank you so much for all the suggestions.  I had overlooked many things. I Feel 😇.  Will make the changes (dig more, add recipes,  documentation of the details you mentioned, etc.) and reply to your comments there itself.
Apr 07 14:10:36 <Abhishek[m]_> Archisman Dey: have seen your proposal. Will get back with comments tomorrow
Apr 07 14:11:06 <ArchismanDey[m]> <Abhishek[m]_ "Archisman Dey: have seen your pr"> Thanks!
Apr 07 15:51:42 <jkridner> DhruvaGole[m]: how are your proposals going?
Apr 07 15:53:00 <jkridner> Abhishek[m]_: the closed-source aspect of GDocs is a bit of the problem for me, but the bigger thing for me gets to be if I have 4, 5 or more of them open along with other stuff it starts to bog my computer down.
Apr 07 15:53:04 <DhruvaGole[m]> Hey jkridner  Here it is... https://elinux.org/BeagleBoard/GSoC/2021_Proposal/DhruvaGole
Apr 07 15:56:38 <DhruvaGole[m]> jkridner I have made many additions since you last reviewed it, please let me know if it makes somewhat sense now?
Apr 07 16:06:37 <Abhishek[m]_> jkridner: It’s not an issue on your side either, there’s a bit of spec bloat happening afaik
Apr 07 16:07:26 <vedant16[m]> Abhishek: even CAD software has been made online (fusion360), we are long when OS will run on browser itself :la
Apr 07 16:07:32 <vedant16[m]> 😂
Apr 07 16:07:35 <Abhishek[m]_> If you have a mac, trying Safari might give better results
Apr 07 16:07:43 <Abhishek[m]_> vedant16: heard of onshape?
Apr 07 16:07:51 <vedant16[m]>  * Abhishek: even CAD software has been made online (fusion360), we are not long when OS will run on browser itself :la
Apr 07 16:08:02 <vedant16[m]> Yupp, tried it. it lagged like hell
Apr 07 16:08:11 <vedant16[m]>  * Abhishek: even CAD software has been made online (fusion360), we are not long when OS will run on browser itself
Apr 07 16:08:38 <Abhishek[m]_> jkridner: Etherpad maybe?
Apr 07 16:12:36 <jkridner> +1
Apr 07 16:13:23 <Abhishek[m]_> Want to install on your server and provide for students?
Apr 07 16:13:26 <vedant16[m]> why does it have a download option ?
Apr 07 16:13:41 <Abhishek[m]_> Run locally?
Apr 07 16:13:42 <vedant16[m]> ohh, so it is host your own thing
Apr 07 16:13:51 <vedant16[m]> <Abhishek[m]_ "Run locally?"> okay
Apr 07 17:05:40 <jkridner> reverted whoever edited the proposal.
Apr 07 17:06:32 <jkridner> https://elinux.org/BeagleBoard/GSoC_2021/OpenGLES_acceleration_for_DL
Apr 07 17:07:16 <jkridner> YadnikBendale[m]: ^^^^
Apr 07 17:28:44 <YadnikBendale[m]> jkridner: Yes have been making the proposal since a while,I am yet to add the description and timeline after discussing a few things with ds2
Apr 07 17:29:02 <cwicks[m]> Hello Mentors and Students!!!
Apr 07 17:29:07 <jkridner[m]1> I moved it to a reasonable place. Please don't edit the template. :-D
Apr 07 17:29:47 <YadnikBendale[m]> <jkridner[m]1 "I moved it to a reasonable place"> I am very sorry I didn't notice
Apr 07 17:29:47 <DhruvaGole[m]> Hello cwicks!
Apr 07 17:29:47 <cwicks[m]> Yadnik Bendale: I grabbed your edits into an email to send to you in case you want them to start your template again
Apr 07 17:30:04 <cwicks[m]> Oh - I se jkridner did something better
Apr 07 17:30:32 <YadnikBendale[m]> <cwicks[m] "Yadnik Bendale: I grabbed your e"> Thank you
Apr 07 17:31:27 <YadnikBendale[m]> Thanks a lot jkridner.I am very sorry for that I didn't notice where i was editing
Apr 07 17:33:19 <cwicks[m]> Hello Everyone - time for our meeting - my name is Cathy Wicks, I am an admin for BeagleBoard.org for GSoC.  While I am not a coder myself, I do projects with BeagleBoards and have been a Beagler since the beginning.  I am a huge fan  Beagleboard.org in education and with students.  My goal in this group is to assist students in connecting with mentors to flush out the best possible proposals to be accepted by GSoC for
Apr 07 17:33:19 <cwicks[m]> this summer.
Apr 07 17:34:34 <cwicks[m]> I'd like to go around the "chat room" and have the students introduce or re-introduce themselves.  Tell us where you live, what you are interested in and if you have connected with a mentor yet.  Please include a link to your ongoing proposal.
Apr 07 17:38:07 * DhruvaGole[m]  < https://matrix.org/_matrix/media/r0/download/matrix.org/bBZcwZXnbqBLuylxEyaQOdkn/message.txt >
Apr 07 17:38:08 <jkridner> hi giuliomoro !
Apr 07 17:38:40 <cwicks[m]> Same with Mentors - please tell us where you are, what your interests are and if you have been working with GSoC in the past as a student and/or mentor
Apr 07 17:38:43 <jkridner> DhruvaGole[m]: your message was so long, it only showed up in Matrix and not on IRC, FYI.
Apr 07 17:38:44 <DhruvaGole[m]> jkridner: did you get time to go through my updated proposal?
Apr 07 17:38:56 <jkridner> not yet. trying to do so in the background.
Apr 07 17:39:07 <DhruvaGole[m]> okayy
Apr 07 17:39:17 <DhruvaGole[m]> <jkridner "Dhruva Gole: your message was so"> oh I see I will split it up
Apr 07 17:39:23 <DhruvaGole[m]> Hello! My name is Dhruva Gole, currently enrolled inn Third Year B. Tech EE in VJTI, Mumbai, India.
Apr 07 17:39:23 <DhruvaGole[m]> I have interest in IOT and robotics mainly in the embedded segment.
Apr 07 17:39:30 <DhruvaGole[m]> I have been in touch with
Apr 07 17:39:30 <DhruvaGole[m]> vedant16 , jkridner and lorforlinux so far, and they have been guiding me through my proposal for GSoC21.
Apr 07 17:39:39 <DhruvaGole[m]> Here is the link to my proposal: https://elinux.org/BeagleBoard/GSoC/2021_Proposal/DhruvaGole
Apr 07 17:40:47 <ArchismanDey[m]> <cwicks[m] "I'd like to go around the "chat "> Hi, I am Archisman from India, interested in the adding features to simpPRU idea. I have a draft proposal at https://elinux.org/BeagleBoard/GSoC/2021_Proposal/simpPRU_Improvements, my mentor Abhishek  has looked at it and has said he will get back with comments tomorrow.
Apr 07 17:41:52 <jkridner> I'm Jason Kridner. I helped found beagleboard.org. I'm most interested in stuff like improved Zephyr support for BeagleConnect Freedom.... like stuff that involved getting that support upstream.
Apr 07 17:44:38 <DhruvaGole[m]> For any new people who have joined and need some background into Beagle
Apr 07 17:44:56 <DhruvaGole[m]> https://www.youtube.com/watch?v=E5Ku8s75l14 this is a nice video, I believe cwicks is a part of it?
Apr 07 17:46:30 <ArchismanDey[m]> hendersa, pratimu : I remember you mentioned you were willing to look at proposals related to the PRU, feel free to look at my proposal for adding features to simpPRU: https://elinux.org/BeagleBoard/GSoC/2021_Proposal/simpPRU_Improvements
Apr 07 17:46:33 <DhruvaGole[m]>  * https://www.youtube.com/watch?v=E5Ku8s75l14 this is a nice introductory video to BB products and BBAI specifically, I believe cwicks is a part of the vdo?
Apr 07 17:46:51 <ArchismanDey[m]>  * hendersa, pratimu : I remember you mentioned last week that you were willing to look at proposals related to the PRU, feel free to look at my proposal for adding features to simpPRU: https://elinux.org/BeagleBoard/GSoC/2021_Proposal/simpPRU_Improvements
Apr 07 17:48:26 <YadnikBendale[m]> Hello to one and all,I am Yadnik Bendale from Dwarkadas J Sanghvi College of Engineering,Mumbai.I have interests in embedded systems,IOT and FPGA's and have worked on various projects in these domains.I am working on the OpenGLES acceleration for DL project and I have been brainstorming various approaches with my mentor,ds2 I have my draft ready but have to add the description and timeline after discussing a few
Apr 07 17:48:26 <YadnikBendale[m]> things with him.https://elinux.org/BeagleBoard/GSoC_2021/OpenGLES_acceleration_for_DL
Apr 07 17:48:55 <pratimu[m]> Archisman Dey: will do
Apr 07 17:50:41 <cwicks[m]> <DhruvaGole[m] "For any new people who have join"> Yes you found me!  That's pdp7 doing the main portion of the interview.
Apr 07 17:51:08 <vedant16[m]> Hi all
Apr 07 17:51:49 <ArchismanDey[m]> vedant16: Hi, look at my proposal if you are free!
Apr 07 17:52:11 <vedant16[m]> Yes doing that rn
Apr 07 17:53:24 <satacker[m]> Hi 😀
Apr 07 17:54:55 <satacker[m]> Abhishek has reviewed my proposal on GDocs which is in sync with elinux wiki, I am making the suggested chages
Apr 07 17:55:00 <satacker[m]> changes*
Apr 07 17:58:20 <jduchniewicz1> Hello my name is Jakub Duchniewicz, I am MSc student in University of Turku/KTH Royal Academy currently residing in Turku Finland.  l am interested in embedded systems, game engines, FPGAs, audio processing, Deep Learning and accelerating calculations using heterogenuous processors. I am working on my project with @ds2
Apr 07 17:59:13 <jduchniewicz1> I am having some troubles with the proposal (being on a track with No success prospects) so I need discussion with @ds2
Apr 07 17:59:39 <lorforlinux[m]> <DhruvaGole[m] "So lorforlinux I feel like [my p"> I'll go through it today and let you know :)
Apr 07 18:00:17 <lorforlinux[m]> Hey everybody :)
Apr 07 18:02:47 <jkridner> DhruvaGole[m]: re "But this symlink is at gpio6_31 of BBAI", do you really mean "symlink"?
Apr 07 18:03:18 <DhruvaGole[m]> I mean it points to
Apr 07 18:03:19 <jkridner> giuliomoro: can you introduce yourself too?
Apr 07 18:04:27 <jkridner> I know there are other mentors lurking cwicks[m]  :-)
Apr 07 18:04:46 * jkridner is in panic mode regarding how few proposals are pending. :-(
Apr 07 18:04:56 <jkridner> final proposals are due, what, the 17th?
Apr 07 18:05:08 <DhruvaGole[m]> no
Apr 07 18:05:11 <DhruvaGole[m]> earlier
Apr 07 18:05:17 <YadnikBendale[m]> 13th April is the last date
Apr 07 18:05:19 <jkridner> 13th!!!!
Apr 07 18:05:25 <jkridner> and, they CAN NOT be late....
Apr 07 18:05:29 <jkridner> not even 1 minute late.
Apr 07 18:05:38 <DhruvaGole[m]> yep
Apr 07 18:05:42 <ArchismanDey[m]> less than 1 week left
Apr 07 18:05:49 * lorforlinux[m]  < https://matrix.org/_matrix/media/r0/download/matrix.org/jgwzuCXdEPuulUuGVNFnfEtx/message.txt >
Apr 07 18:06:14 <jkridner> DhruvaGole[m]: are you focusing in on the librobotcontrol project more so than Blynk?
Apr 07 18:06:25 * satacker[m]  < https://matrix.org/_matrix/media/r0/download/matrix.org/iAQheCVUnhXcXoLdmsKOiIrJ/message.txt >
Apr 07 18:06:27 <DhruvaGole[m]> <jkridner "Dhruva Gole: are you focusing in"> yes I am
Apr 07 18:06:50 <DhruvaGole[m]> keeping in mind the time constraint I have diverted all focus to librobotcontrol
Apr 07 18:13:04 <jkridner> DhruvaGole[m]: k. We're going to be throwing in a new SoC in the mix this summer, but I'd expect shipment to India and other factors will get in the way, but it'll give us some good focused collaboration time on the project.
Apr 07 18:13:53 <DhruvaGole[m]> great! Looking forward to it jkridner!
Apr 07 18:15:09 <lorforlinux[m]> <satacker[m] "> <@lorforlinux:matrix.org> Can "> Please go on and tell me what's the problem you facing right now?
Apr 07 18:16:24 <ArchismanDey[m]> <jkridner "Dhruva Gole: k. We're going to b"> the Beagle V?
Apr 07 18:16:34 <satacker[m]> <lorforlinux[m] "Please go on and tell me what's "> I am not sure how the instances created in /proc/devicetree/usb@address or some symlink will help it configure the gadget
Apr 07 18:17:04 <pratimu[m]> Archisman Dey: I would suggest to give passing complex data through rpmsg a bit more importance
Apr 07 18:17:15 <pratimu[m]> For example over the flexible for loop
Apr 07 18:17:59 <ArchismanDey[m]> <pratimu[m] "Archisman Dey: I would suggest t"> what do you suggest adding? There are only 3 data types in simpPRU, and I am planning to add support for sending all three, plus their arrays, so not sure what else I can add
Apr 07 18:18:10 <satacker[m]> <satacker[m] "I am not sure how the instances "> to configure a gadget I will have to add the vid,pid, and other strings including instances if its a mass storage or some other, so i imagine different nodes for it
Apr 07 18:19:35 <satacker[m]> <satacker[m] "to configure a gadget I will hav"> but how do I tell the UDC that heres your USB gadget
Apr 07 18:19:44 <cwicks[m]> <jduchniewicz1 "I am having some troubles with t"> Hello jduchniewicz are you having trouble seeing @ds2 replies in the IRC client?
Apr 07 18:20:32 <pratimu[m]> <ArchismanDey[m] "what do you suggest adding? Ther"> Archisman Dey: I meant to say that completing that milestone before the flexible for loop in the timeline would be better
Apr 07 18:20:48 <pratimu[m]> Not that more datatypes are needed
Apr 07 18:21:40 <satacker[m]> lorforlinux: To be honest my proposal feels quite shallow right now because I cannot definitely say that it works or have knowledge in depth. Barely have I experimented which is primary reason I am not getting it intuitively
Apr 07 18:22:17 <ArchismanDey[m]> <pratimu[m] "Archisman Dey: I meant to say th"> Oh okay, I also agree that it is more important, I will do that, I just added the for loop after the array since it felt more related
Apr 07 18:22:45 <lorforlinux[m]> * Late intro
Apr 07 18:22:45 <lorforlinux[m]> My name is Deepak Khatri and I am a Biotechnologist from INDIA. I am love Electronics and Computer Science more than anything. Worked on the `Cape Compatibility layer for BBAI and BBB` GSoC project last year, In free time i usually create projects involving Biology, Electronics and Code. I am the founder of Upside Down Labs where we design awesome OpenSource Low power electronics and BioMedical devices. Looking
Apr 07 18:22:45 <lorforlinux[m]> forward to help students this year on various projects :)
Apr 07 18:22:55 <satacker[m]>  * lorforlinux: To be honest my proposal (usb configfs in device tree ) feels quite shallow right now because I cannot definitely say that it works or have knowledge in depth. Barely have I experimented which is primary reason I am not getting it intuitively
Apr 07 18:24:57 <pratimu[m]> <ArchismanDey[m] "Oh okay, I also agree that it is"> Okay, I will have a more thorough look and get back :)
Apr 07 18:25:24 <ArchismanDey[m]> <pratimu[m] "Okay, I will have a more thoroug"> Thanks for the suggestion!
Apr 07 18:25:56 <vedant16[m]> Archisman Dey: Looks good overall, I'll give a detailed feedback in sometime, need to submit some assignments.
Apr 07 18:26:07 <vedant16[m]> I am here for any questions though. everyone
Apr 07 18:26:36 <ArchismanDey[m]> <vedant16[m] "Archisman Dey: Looks good overal"> Thanks for looking at it!
Apr 07 18:26:43 <lorforlinux[m]> <satacker[m] "I am not sure how the instances "> Okay!
Apr 07 18:27:36 <lorforlinux[m]> <satacker[m] "to configure a gadget I will hav"> USB devices are hot plug-able right?
Apr 07 18:27:58 <satacker[m]> <lorforlinux[m] "USB devices are hot plug-able ri"> yes , devices
Apr 07 18:30:32 <lorforlinux[m]> <satacker[m] "but how do I tell the UDC that h"> I believe we missing something important here, If the USB gadget is hot plug-able there must be some module in-place for the detection!
Apr 07 18:32:50 <satacker[m]> <lorforlinux[m] "I believe we missing something i"> actually there are two apis gadget vs linux host side
Apr 07 18:34:19 <satacker[m]> <satacker[m] "actually there are two apis gadg"> the one rcn-ee uses are gadgets
Apr 07 18:34:31 <lorforlinux[m]> <satacker[m] "lorforlinux: To be honest my pro"> yeah I can see that. until you get to talk to Robert on this again, things will not improve much. As he is the mentor, He can clear your doubts much better!
Apr 07 18:35:19 <satacker[m]> <lorforlinux[m] "I believe we missing something i"> are u suggesting that once the nodes are created the modules will be called by matching the compatible property or something like vid,pid
Apr 07 18:37:30 <satacker[m]> lorforlinux: does [this](https://www.kernel.org/doc/Documentation/usb/gadget_configfs.txt) look familiar w.r.t device trees
Apr 07 18:39:56 <satacker[m]> <satacker[m] "lorforlinux: does [this](https:/"> this is ```associating functions with configurations (symlinks)```
Apr 07 18:44:55 <cwicks[m]> Students - we need to assist you in developing the proposals pretty quickly. Mentors please reach out to jkridner and other mentors to help review proposals so that they are robust.  Deadline is coming up soon
Apr 07 18:45:19 <cwicks[m]> Students - we will be here every day - and as you have seen there are mentors also on IRC in India time as well as Europe time.
Apr 07 18:47:12 <satacker[m]> lorforlinux: I think I would move past it for now and would you like to have a look at [beagle-vonfig](https://elinux.org/BeagleBoard/GSoC/2021_Proposal/beagle_config)
Apr 07 18:47:45 <satacker[m]> Although Abhishek has reviewed thoroughly and I am yet to add changes I would like to have a quote
Apr 07 18:47:45 <satacker[m]>  * lorforlinux: I think I would move past it for now and would you like to have a look at [beagle-config](https://elinux.org/BeagleBoard/GSoC/2021_Proposal/beagle_config)
Apr 07 18:47:45 <satacker[m]> Although Abhishek has reviewed thoroughly and I am yet to add changes I would like to have a quote
Apr 07 18:48:07 <satacker[m]>  * lorforlinux: I think I would move past it for now and would you like to have a look at [beagle-config](https://elinux.org/BeagleBoard/GSoC/2021_Proposal/beagle_config)
Apr 07 18:48:07 <satacker[m]> Although Abhishek has reviewed thoroughly and I am yet to add changes, I would like to have a quote
Apr 07 18:48:41 <satacker[m]>  * lorforlinux: I think I would move past it for now and would you like to have a look at [beagle-config](https://elinux.org/BeagleBoard/GSoC/2021_Proposal/beagle_config)
Apr 07 18:48:41 <satacker[m]> Although Abhishek has reviewed thoroughly and I am yet to add changes, I would like to have a quote on overall idea
Apr 07 18:48:45 <cwicks[m]> Hello Omkar Bhilare Welcome!
Apr 07 18:49:33 <cwicks[m]> Omkar Bhilare: Tell us where you live, what you are interested in and if you have connected with a mentor yet.  Please include a link to your ongoing proposal.
Apr 07 18:50:49 <jduchniewicz1> <cwicks[m] "Hello jduchniewicz are you havin"> sorry did not see the message, I did not see him respond to my previous messages
Apr 07 18:50:55 <jduchniewicz1> the ones I send yesterday
Apr 07 18:51:27 <jduchniewicz1> at this moment I do not know whether I should start working on some other proposal or try to circumvent my problems with this one
Apr 07 18:52:12 <jkridner> jduchniewicz1: I don't know what the problems are, but, if you are experiencing blockers, I'd say yes.
Apr 07 18:52:30 * OmkarBhilare[m]  < https://matrix.org/_matrix/media/r0/download/matrix.org/PxUeGYJHMMxJgNZjJeqLfNHU/message.txt >
Apr 07 18:52:34 <jkridner> I believe up to 3 proposals can be submitted by a single student, though only 1 can be approved.
Apr 07 18:52:46 <jkridner> also, the 3 proposals don't need to all be to the same organization.
Apr 07 18:53:32 * OmkarBhilare[m]  < https://matrix.org/_matrix/media/r0/download/matrix.org/JqwtQnvEnVvZugjNzuXgOsCQ/message.txt >
Apr 07 18:53:58 <jduchniewicz1> Before I drop the proposal I would like to get some feedback and maybe alternative ideas from ds2
Apr 07 18:57:14 <cwicks[m]> <jduchniewicz1 "Before I drop the proposal I wou"> jduchniewicz: ds2 is replying but it seems you are not getting his replies.  Can you please advise your email address?  this may be better
Apr 07 18:58:08 <jduchniewicz1> that is weird, as I received his replies previously both on IRC and on matrix
Apr 07 18:59:50 <ds2> let see if this makes a difference
Apr 07 19:00:01 <ds2> jduchniewicz1: anything now?
Apr 07 19:00:06 <cwicks[m]> <jduchniewicz1 "that is weird, as I received his"> Some kind of weird bridge issue with IRC client I think (cwicks is NOT knowledgeable about these types of things) so email may be a good back up.
Apr 07 19:00:25 <cwicks[m]> I see you ds2
Apr 07 19:00:25 <jduchniewicz1> now I can see ds2
Apr 07 19:00:36 <ds2> sigh... matrix is doing its own thing once again
Apr 07 19:00:53 <jduchniewicz1> weird because I didn't receive notifications on the IRC native client as well
Apr 07 19:00:59 <jduchniewicz1> at least not today
Apr 07 19:01:12 <ds2> jduchiewicz1: really need to usse native IRC... I been pinging you for the last hr + been replying yesterday!
Apr 07 19:01:28 <ds2> going to be slow as I have another mtg that just started
Apr 07 19:01:49 <jduchniewicz1> ah, sorry for that in this case. Let me check logs from previous days
Apr 07 19:02:24 <jkridner> DhruvaGole[m]: "The compatibility layer comes with simple reference nodes for attaching LEDs to any gpio pin." It is really more of connecting an LED *driver* to any GPIO pin and using the LED driver instance name to map to the same header pin, despite whatever GPIO it is on.
Apr 07 19:02:42 <ds2> jduchniewicz1: what is the other HW?
Apr 07 19:03:21 <jkridner> the purpose isn't to actually connect an LED to that pin, but to have a somewhat generic driver attached that can be both named as well as driven from userspace apps. I'm not sure I think it is the best plan, but it was something reasonable that lorforlinux[m] cooked up.
Apr 07 19:04:18 <jduchniewicz1> ds2:  what HW? I am not sure to which point you are relating
Apr 07 19:04:40 <cwicks[m]> jduchniewicz: Do you have a link to an elinux.org page where you have started your proposal?  I am collecting those today to help ping the mentors to review
Apr 07 19:04:43 <ds2> jduchniewicz1: you said the newer TIDL stuff is for the new HW
Apr 07 19:04:55 <jduchniewicz1> ok, step by step
Apr 07 19:05:14 <jduchniewicz1> 1. no idea where the messages went, just checked logs from IRC client on my machine
Apr 07 19:05:23 <jduchniewicz1> 2. https://elinux.org/BeagleBoard/GSoC/2021_Proposal/YOLO_models_on_the_X15/AI
Apr 07 19:05:44 <cwicks[m]> hello @laimonade
Apr 07 19:05:44 <jkridner> DhruvaGole[m]: the cape compatibility layer provide compatibility at two-layers, at the overlay and at userspace. for librobotcontrol, the interest is at the overlay level. now, if the robotics cape overlay is assumed, that overlay depends on the compatibility at the overlay level. with the cape compatibility layer and the fact that librobotcontrol depends heavily on userspace drivers (for better or worse), an overlay may or
Apr 07 19:05:44 <jkridner> may not be required.
Apr 07 19:07:15 <jduchniewicz1> 3. ds2  yes, there is a release of the SDK -> PROCESSOR-SDK-DRA8X-TDA4X
Apr 07 19:07:27 <DhruvaGole[m]> Right... So my work will mostly be focused on working on userspace drivers
Apr 07 19:07:48 <jduchniewicz1> ds2:  apparently they included yolo version for reference in there and it works fine (trained in the jacinto Caffe)
Apr 07 19:07:52 <DhruvaGole[m]> * Right... So my work will mostly be focused on working on userspace drivers? Ie. The one's in librobotcontrol repo only?
Apr 07 19:07:55 <ds2> jkridner: any insight on the new TIDL SDK (see jduchniewicz1's msg)
Apr 07 19:08:28 <ds2> jduchniewicz1: any notes on procesors supported on that SDK?
Apr 07 19:08:31 <jkridner> ds2: no, nothing new. just know they are working toward supporting TFLite and not TIDL moving forward.
Apr 07 19:09:06 <jduchniewicz1> it seems like TFLite is using TIDL underneath
Apr 07 19:09:15 <ds2> TFlite is just a framework
Apr 07 19:09:21 <ds2> it can call other acceleration as needed
Apr 07 19:09:33 <ds2> tflite "interprets" your models
Apr 07 19:09:57 <jduchniewicz1> yes, put it in wrong terms - they use TIDL underneath all ML/DL components
Apr 07 19:10:41 * jduchniewicz1  < https://matrix.org/_matrix/media/r0/download/matrix.org/DfOVWWkcqQKYzATWBZWXsLJz/message.txt >
Apr 07 19:10:44 <ds2> i wonder if they really mean - TIDL won't be exposed as today but rather exposed as a Tflite backend
Apr 07 19:11:02 * jkridner wishes NishanthMenon knew more about it.... would be better resource than me.
Apr 07 19:11:12 <jkridner> he idles on #beagle
Apr 07 19:12:04 <jduchniewicz1> as far as I dug into it, TIDL is (beyond other things) a way to utilize EVEs and DSPs without low level control (anyway EVEs are an IP and seem to be not very much exposed)
Apr 07 19:13:58 <jduchniewicz1> so the most important question for me is how to:
Apr 07 19:13:58 <jduchniewicz1> 1. use the newer API on BBAI (probably not possible)
Apr 07 19:13:58 <jduchniewicz1> 2. deploy only part of the model in TIDL and accelerate the rest by other means
Apr 07 19:14:17 <jduchniewicz1> with my limited experience with TIDL it seems either all or none
Apr 07 19:15:42 <jduchniewicz1> * the last comment is about accelerating either all layers with TIDL or none
Apr 07 19:16:24 <jduchniewicz1> and it seems like TIDL at current version lacks such important features as LeakyRelu which is interlaced with conv2d in YOLOv3
Apr 07 19:18:13 <lorforlinux[m]> <satacker[m] "are u suggesting that once the n"> Yes, modules get called with the matching compatible value provided with the node.
Apr 07 19:19:49 <lorforlinux[m]> <satacker[m] "lorforlinux: does [this](https:/"> Not sure 🤔
Apr 07 19:22:21 <lorforlinux[m]> > <@satacker:matrix.org> lorforlinux: I think I would move past it for now and would you like to have a look at [beagle-config](https://elinux.org/BeagleBoard/GSoC/2021_Proposal/beagle_config)
Apr 07 19:22:22 <lorforlinux[m]> > Although Abhishek has reviewed thoroughly and I am yet to add changes, I would like to have a quote on overall idea
Apr 07 19:22:22 <lorforlinux[m]> That's would be better!
Apr 07 19:23:23 <satacker[m]> <lorforlinux[m] "> <@satacker:matrix.org> lorforl"> would love suggestions from you and probably some quotes
Apr 07 19:23:43 <m_w> hello all
Apr 07 19:23:58 <m_w> is the meeting today?
Apr 07 19:24:01 <m_w> checking logs
Apr 07 19:24:07 <cwicks[m]> hello m_w !!! welcome!
Apr 07 19:24:19 <m_w> logs are broken
Apr 07 19:24:57 <m_w> cwicks[m], hey
Apr 07 19:25:12 <cwicks[m]> ah, that's not good m_w sorry
Apr 07 19:25:22 <cwicks[m]> Please meet Omkar Bhilare
Apr 07 19:25:25 <jduchniewicz1> ds2: maybe I could ditch the TIDL acceleration whatsoever as I mentioned earlier, but then how to use EVEs and other acceleration possibilities the BB platforms offer
Apr 07 19:25:27 <cwicks[m]> His proposal is here https://elinux.org/BeagleBoard/GSoC/2021_Proposal/OmkarBhilare
Apr 07 19:25:48 <cwicks[m]> We have a deadline of 4/13 to get all the proposals in
Apr 07 19:25:54 <lorforlinux[m]> <jkridner "Dhruva Gole: the cape compatibil"> Overlay is required and there will also be edits for the user-space code!
Apr 07 19:26:09 <jduchniewicz1> Alternatively I could try to come up with a plan for another proposal, for example Miropython support as it seems to be compelling to me
Apr 07 19:26:17 <jduchniewicz1> *Micropython
Apr 07 19:26:19 <vedant16[m]> Hi students, this is my proposal which got selected last year https://elinux.org/BeagleBoard/GSoC/2020Proposal/VedantParanjape2
Apr 07 19:26:29 <ds2> jkridner: what is the timeframe of the TDA4VM/BBAI?
Apr 07 19:26:48 <ds2> jduchniewicz1: there is talk about a T4DM BBAI
Apr 07 19:27:10 <DhruvaGole[m]> <lorforlinux[m] "Overlay is required and there wi"> jkridner this is why I added the overlay part as well in my proposal
Apr 07 19:27:42 <m_w> OmkarBhilare[m], you are Andrew45 on discord right?
Apr 07 19:27:54 <DhruvaGole[m]> lorforlinux if my understanding is correct, the overlays are already integrated upstream by you right?
Apr 07 19:28:03 <vedant16[m]> <m_w "Omkar Bhilare, you are Andrew45 "> yes
Apr 07 19:28:52 <ds2> jduchniewicz1: if I don't get back quick enough, can you check with jkridner? there is a new upcoming revsion of the BBAI that may address that
Apr 07 19:29:01 <cwicks[m]> Thank you vedant16 for sharing your proposal link.  lorforlinux can you also share yours?  This will definitely help students when editing theirs this year.
Apr 07 19:29:23 <ArchismanDey[m]> <cwicks[m] "Thank you vedant16 for sharing y"> All four proposals for last year available here: https://elinux.org/BeagleBoard/GSoC/2020_Projects
Apr 07 19:29:34 <jduchniewicz1> this would be quite interesting, however I don't see any information about it on google
Apr 07 19:29:41 <ds2> it is very new
Apr 07 19:29:55 <ds2> like dev in progress but sometimes they are done enough to be available for GSoC
Apr 07 19:30:04 <DhruvaGole[m]> I am very confused as to why will I need to work on DT Overlays if lorforlinux has already done the job? 😅
Apr 07 19:30:13 <vedant16[m]> <cwicks[m] "Thank you vedant16 for sharing y"> happy to help
Apr 07 19:30:47 <DhruvaGole[m]> What I understand is I need to edit only the userspace code since overlays are already integrated in the BBAI's debian image
Apr 07 19:31:53 <jduchniewicz1> ds2: assuming we are "lucky" enough to have the newer SoC architecture? we want to be able to deploy official YOLO models. I hope that this SDK does not support _only_ the Jacinto-Caffe ones as this would be terrible for potential users of this project.
Apr 07 19:32:36 <jduchniewicz1> ds2:  dunno how to call this TDA4VM and if this is some codename for an iteration over AM5x SoC's
Apr 07 19:32:49 <ds2> jduchniewicz1: no YOLO model are native to Jacinto-Caffe. knowing how to convert them will have to be part of the project
Apr 07 19:33:09 * OmkarBhilare[m]  < https://matrix.org/_matrix/media/r0/download/matrix.org/bHTmTAJHlKEWmLrsmZlsjwSQ/message.txt >
Apr 07 19:33:16 <jduchniewicz1> ds2:  people on e2e forums used the Jacinto-Caffe ones in SDK version 7
Apr 07 19:33:39 <jduchniewicz1> ds2:  look https://git.ti.com/cgit/jacinto-ai/pytorch-mmdetection/about/docs/det_modelzoo.md
Apr 07 19:33:46 <ds2> jduchniewicz1: SoCs have different names depending on their sales market. AM7xxx is thTDA4VM, IIRC. the AM5xxx had  many names
Apr 07 19:34:00 <jduchniewicz1> oh cool now I get it!
Apr 07 19:34:38 <ds2> jduckniewicz1: yes, I know... and tflite needs its own format
Apr 07 19:34:47 <jduchniewicz1> ds2:  the relevant post on e2e which proved it can be done https://e2e.ti.com/support/processors/f/processors-forum/965208/tidl-inference-of-yolov3-in-onnx-input-format/3567220#3567220
Apr 07 19:35:06 <ds2> onnx is the keyword...onnx is an interchange tool/format
Apr 07 19:35:19 <m_w> OmkarBhilare[m], great start on the proposal
Apr 07 19:36:28 <m_w> OmkarBhilare[m], you can ping me on discord if I am not here
Apr 07 19:36:43 <jduchniewicz1> ds2: yes, it seems like they were able to validate these models (so assuming the different implementation use the same layers, we should be fine)
Apr 07 19:37:35 <ds2> jduchniewicz1: how hard is it to try to get YOLO working on the older TIDL? (either backport or implement a hybrid TIDL/ARM setup?
Apr 07 19:37:46 <m_w> OmkarBhilare[m], I created a buildroot image for the test fixture of beaglewire if that is something you want to build upon
Apr 07 19:38:27 <ds2> jduchniewicz1: what's your alternative proposal ideas?
Apr 07 19:38:42 <m_w> OmkarBhilare[m], I think getting started needs a lot of work for the average user
Apr 07 19:39:52 <OmkarBhilare[m]> yes, I saw couple of issues related to this on beaglewire repo.
Apr 07 19:40:10 <OmkarBhilare[m]> <m_w "Omkar Bhilare, I created a build"> Where I can find this?
Apr 07 19:40:54 <jduchniewicz1> ds2: it seems like it is impossible to partial ly deploy a model with unsupported layers
Apr 07 19:42:14 <jduchniewicz1> ds2: maybe there is some way I am not seeing here but I spent considerable time on this and either you put all layers to TIDL (and split them between DSP and EVE) or you dont use it at all
Apr 07 19:42:38 <ds2> jduchniewciz1: couldn't you run layers 1-N, take that result, do layer N+1, run layers N+2-M?
Apr 07 19:42:57 <jduchniewicz1> ds2: alternatively I can do something totally different, but accelerating things on BB would be most Fun to me
Apr 07 19:43:11 <ds2> TIDL likes to see the entire model but I don't see why you can have the model just be layers 1-N
Apr 07 19:43:52 <ds2> jduchniewicz1: what about GPGPU with OpenGLES 2?
Apr 07 19:43:58 <jduchniewicz1> ds2: the problem is the relu layers which I cannot seem to deploy in TIDL and the model has a lot of them
Apr 07 19:44:00 <ds2> that's also acceleration
Apr 07 19:44:20 <jduchniewicz1> Hmm I am only very delicately versed in GL stuff
Apr 07 19:45:00 <ds2> jduchniewcicz1: RELU is to implement.... that's x = (x > 0)? x:0;
Apr 07 19:45:15 <ds2> if you can break it up into chunks, and do that on the ARM side
Apr 07 19:45:23 <ds2> makes me wonder if TFLite is doing exactly that
Apr 07 19:45:27 <jduchniewicz1> Apparently they cannot do the leakyrelu
Apr 07 19:45:35 <jduchniewicz1> And yolo does that a lot
Apr 07 19:45:43 <ds2> what is leakyrelu? any ideas?
Apr 07 19:47:06 <jduchniewicz1> It is relu but slightly different :) https://www.researchgate.net/profile/Stefano-Romanazzi/publication/325226633/figure/fig9/AS:627667623768071@1526659031098/Plot-of-the-LeakyReLU-function.png
Apr 07 19:47:58 <ds2> jduchniewicz1: how are you in parallel programming (generic, OpenCL, CUDA, etc)? and how are you at understanding convolution, matrix multiple, etc?
Apr 07 19:49:08 <m_w> OmkarBhilare[m], it is only on my laptop as of now
Apr 07 19:49:38 <m_w> OmkarBhilare[m], if you want to leverage it, let me know and I can post to to github
Apr 07 19:49:42 <ds2> leaky doesn't look too bad to implement... some nasty NEON should handle it
Apr 07 19:50:49 <ds2> jduchiewicz1: the GPGPU stuff with OpenGLES isn't that OpenGL centric - it is basically - create a OpenGLES context w/shader hooks (generic code)... then writing the shader code to implement the math; the other glue is reading/writing textures
Apr 07 19:51:20 <ds2> i have code for the first generic stuff... if you want to tackle the writing the shader code for the algs part.
Apr 07 19:51:38 <ds2> not that much an alg person hence I been stalled on that
Apr 07 19:53:06 <ds2> think it is getting late for you... feel free to email
Apr 07 19:54:39 <jduchniewicz1> I will think about it further. Answering for now: leakyrelu is after every layer of conv2d so I have some problems visualizing how it would look like -> collecting accelerated data from TIDL and then running ARM code for leakyrelu?
Apr 07 19:55:20 <jduchniewicz1> As for the CUDA and convolutions, I have some experience, with convolutions on the HDL side mostly and CUDA with matrix multiplications
Apr 07 19:56:32 <jduchniewicz1> I will try to think on this “per layer” acceleration as this may solve the problem if TIDL allows for such things
Apr 07 19:57:41 <ds2> jduchniewicz1: if you are interested in writing up and debugging sample code for the GPGPU stuff - ping me.. as I said, it is mostly writing the alg side; I can help with the OpenGLES end
Apr 07 19:57:56 <ds2> anyways, I'll let you sleep
Apr 07 20:05:40 <lorforlinux[m]> <DhruvaGole[m] "lorforlinux if my understanding "> Major overlays are there in place, it will be a breeze for you to create the overlay for Robotics cape!
Apr 07 20:07:08 <DhruvaGole[m]> Okay, so overall does my proposal make sense now? I am thinking of submission
Apr 07 20:07:54 <DhruvaGole[m]> This will be the last meet before the deadline so please let me know of any obvious mistakes/ improvements/ or if my proposal is okayish?
Apr 07 20:08:36 <DhruvaGole[m]> Link : https://elinux.org/BeagleBoard/GSoC/2021_Proposal/DhruvaGole
Apr 07 20:50:23 <jkridner[m]> <ds2 "jkridner: what is the timeframe "> Theory is samples in early June and production early Sept or so, ish.
Apr 07 20:53:48 <jkridner> vedant16[m] proposal drafts and template are at https://elinux.org/index.php?title=Category:GSoCProposal2021
Apr 07 20:54:33 <vedant16[m]> <jkridner "vedant16 proposal drafts and tem"> Cool
Apr 07 20:59:45 <satacker[m]> Mine wasn't showing so i added ```[[Category:GSoCProposal2021]]``` incase somebody needs help
Apr 07 21:00:12 <satacker[m]>  * Mine wasn't showing so i added `[[Category:GSoCProposal2021]]`
Apr 07 21:00:12 <satacker[m]> And now it shows in the list.
Apr 07 21:00:12 <satacker[m]> ( incase somebody needed help)
Apr 07 21:02:39 <vedant16[m]> cwicks: can you add this link to about me section of this channel ?
Apr 07 22:55:58 <jkridner[m]1> I'll make a shortlink.
Apr 07 22:59:45 <jkridner> vedant16[m]: cwicks[m] done.
**** ENDING LOGGING AT Thu Apr 08 02:59:57 2021