**** BEGIN LOGGING AT Wed Apr 08 03:00:04 2020
Apr 08 16:29:05 <hydraulicsheep[m> Hey I’m here for the meeting. Am I in the right place?
Apr 08 16:30:26 <hydraulicsheep[m> Oh wait. I think I’m an hour off. Nvm
Apr 08 16:33:30 <jkridner[m]> I’m on my way. Who all is here?
Apr 08 16:34:27 <nwan[m]> @jkrinder hello, I'm also here
Apr 08 16:34:42 <ds2> ..
Apr 08 16:35:49 <Abhishek_> Hi , here
Apr 08 16:37:36 <pratimugale[m]> Im here
Apr 08 16:39:16 <jkridner[m]> any mentors not in the back-channel, please message me.
Apr 08 16:40:46 <hydraulicsheep[m> Here
Apr 08 16:40:57 <hydraulicsheep[m> Is this the right channel?
Apr 08 16:43:19 <hydraulicsheep[m> @jkridner?
Apr 08 16:44:44 <hydraulicsheep[m> Or is there some other channel?
Apr 08 16:45:56 <nwan[m]> hey HydraulicSheep  there is another channel for the mentors. If you are a student, this is the right channel
Apr 08 16:48:09 <hydraulicsheep[m> Ok cool
Apr 08 16:48:11 <hydraulicsheep[m> Thanks
Apr 08 16:48:30 <hydraulicsheep[m> Any other students here?
Apr 08 16:48:41 <jkridner[m]> yeah, ask any Qs you might have, especially on any required hardware.
Apr 08 16:51:15 <nwan[m]> Right now there are no further questions. Regarding hardware I already talked with Robert
Apr 08 16:51:41 <hydraulicsheep[m> I just wanted to touch base about my continued progress
Apr 08 16:52:01 <ds2> pradan?
Apr 08 16:52:08 <hydraulicsheep[m> Still busy disassembling commercial calculator models to work on interfacing keyboards
Apr 08 16:52:26 <hydraulicsheep[m> I’ve been in communication with @Pac23
Apr 08 16:53:12 <pradan[m]> Yeah actually all students are here.
Apr 08 16:53:12 <pradan[m]> @pradan:matrix.org @lorforlinux:matrix.org @ymdatta:matrix.org @vedant16:matrix.org @deepankarmaithani:matrix.org  and perhaps others too
Apr 08 16:53:49 <ds2> pradan[m]: any more experiments?
Apr 08 16:57:51 <lorforlinux[m]> hi
Apr 08 16:58:03 <cwicks[m]> Hello students - I'm an admin for BeagleBoard.org GSoC.
Apr 08 16:59:11 <lorforlinux[m]> cwicks : Heya
Apr 08 16:59:32 <nwan[m]> cwicks: hello
Apr 08 16:59:54 <cwicks[m]> This is a great time to be enhancing your elinux proposal sites
Apr 08 17:00:33 <cwicks[m]> For example any clarifications you have been working on with the mentors. (not huge differences from the GSoC proposal)
Apr 08 17:01:03 <hydraulicsheep[m> Aaah ok
Apr 08 17:01:20 <hydraulicsheep[m> I might do it tomorrow morning as it’s pretty late here
Apr 08 17:01:44 <hydraulicsheep[m> I’ll upload some photos of the disassembled calculator
Apr 08 17:01:54 <hydraulicsheep[m> And labeled diagrams of the plans
Apr 08 17:02:34 <cwicks[m]> So HydraulicSheep if you have been working with Pac23 on clarifications, enhancements, updates, etc.  that is great to add to your elinux page
Apr 08 17:03:46 <jkridner[m]> vedant16: some discusson regarding your proposals is that there should be enough already done on each to actually accomplish both (as 1 project). any thoughts on that?
Apr 08 17:03:46 <pradan[m]> @ds2 I was done with the TIDL SETUP until I felt whether I should switch the approach NOW and try the previously discussed method
Apr 08 17:04:50 <lorforlinux[m]> cwicks jkridner : For making things more manageable during coding period i would like to working on the beagle-tester. Can you give me some pointers for getting started with it?
Apr 08 17:04:56 <jkridner[m]> pradands2 : did you check out the fresh Debian 10 release with TIDL? Should have requirements to handle thet TFLite stuff.
Apr 08 17:05:26 <jkridner[m]> ds2: also, there's a new IMGTEC image for a training program they plan to launch at some point in the future. Might be worth exploring for you. Meant for SGX530. Might spill over into a bit on SGX544.
Apr 08 17:05:30 <hydraulicsheep[m> Thanks @cwicks
Apr 08 17:06:00 <hydraulicsheep[m> Will do. If anyone has suggestions on where to contribute relevant to my project, that would be great
Apr 08 17:06:07 <jkridner[m]> lorforlinux: As you read the setup, it is designed for working with a bar code scanner. I'd add that to the required hardware....
Apr 08 17:06:15 <hydraulicsheep[m> I’ll upload the updates tomorrow
Apr 08 17:06:28 <hydraulicsheep[m> (In 6-7 hours)
Apr 08 17:06:33 <jkridner[m]> if the code is properly integrated, the test for each cape should be triggered by scanning the barcode on that cape.
Apr 08 17:06:54 <ds2> jkridner[m]: TI image or BB image?
Apr 08 17:07:06 <jkridner[m]> ds2: bb image. bbb.io/latest
Apr 08 17:07:11 <lorforlinux[m]> <jkridner[m] "lorforlinux: As you read the set"> yes, i need a TechLab Cape too, I will update the e-Linux page accordingly.
Apr 08 17:07:24 <pradan[m]> @jkridner:matrix.org: the one Mr. Robert has been referring to?
Apr 08 17:07:42 <jkridner[m]> lorforlinux: TechLab? I didn't think the cape compatibility extended to PocketBeagle.
Apr 08 17:07:52 <jkridner[m]> maybe just for learning?
Apr 08 17:08:05 <jkridner[m]> I was planning to send one to all to be able to participate in the on-line training.
Apr 08 17:08:32 <lorforlinux[m]> <jkridner[m] "maybe just for learning?"> yes for learning.
Apr 08 17:08:46 <jkridner[m]> pradan: bbb.io/latest now has rcn-ee 's Spring Release.
Apr 08 17:08:52 <ds2> that could be interesting... or just stick with a 3.x kernel ;)
Apr 08 17:10:26 <jkridner[m]> students: remember that whatever discussion we have here does not imply you will be accepted. communication from Google is the only way to know if you've been accepted into the program.
Apr 08 17:15:08 <lorforlinux[m]> jkridner: I have done some work on the device trees now and i want to work on some other repository, can you suggest me a repository which is highly important for the cape-compatibility and can be handled without any hardware in hand?
Apr 08 17:15:16 <pradan[m]> @ds2 I will stick with the previous one... untill I use TF Lite for it
Apr 08 17:16:18 <ds2> pradan[m]: How's the local situation?
Apr 08 17:18:38 <jkridner[m]> lorforlinux: the device tree repositories, bb.org-overlays and BeagleBone-DeviceTrees, can be build-tested without hardware and you can make community requests to test forks you might create.
Apr 08 17:18:47 <jkridner[m]> er, BeagleBoard-DeviceTrees
Apr 08 17:19:18 <pradan[m]> Somewhat messed it up while setting up API and SDK for linux.
Apr 08 17:19:18 <pradan[m]> It is comparatively simpler when setting up the image directly for the BB hardware
Apr 08 17:19:18 <jkridner[m]> beagle-tester can also be build tested.
Apr 08 17:20:59 <pradan[m]> I found an alternative to FeatherCNN or let's say if I use any particular tool, how do I demonstrate its performance without the actual development board?
Apr 08 17:21:29 <lorforlinux[m]> <jkridner[m] "beagle-tester can also be build "> okay i will start working.
Apr 08 17:21:49 <ds2> pradan[m]: the board problem should sort out if the project is a go
Apr 08 17:21:51 <jkridner[m]> just out of curiousity, any other students around? vedant16 deepankarmaithani others I don't know yet?
Apr 08 17:29:41 <pradan[m]> Ok. I still don't wanna go for TIDL as relying on high-level tools is something I can't rely on ...for completion of my project 😬
Apr 08 17:32:04 <pradan[m]> I am exploring OpenCL now and think that accelerator library + opencl could be a reliable and controllable option.. even if the results tend to saturate at a point
Apr 08 17:32:38 <ds2> pradan[m]: keep in mind the opencl on the x15 is rather limited (back end is a DSP not a GPU)
Apr 08 17:34:50 <pradan[m]> I may not be exact but I think the OpenCL is responsible for backend for C66x DSPs ... and there are 2 of them
Apr 08 17:35:17 <ds2> yes, you are correct on that
Apr 08 17:37:22 <cwicks[m]> Students  and Mentors, I hope that everyone is staying healthy and safe.  While this is the end of the formal meeting time, please continue to chat and to clarify/enhance elinux docs
Apr 08 17:37:26 <pradan[m]> Even after the solution manages to improve the performance, if some of the objectives like the desired FPS>30 isn't completely achieved, is the project still considered a success?
Apr 08 17:37:39 <jkridner[m]> vaishnav98_: you around?
Apr 08 17:37:45 <ds2> pradan[m]: IMO - I'd say so
Apr 08 17:37:45 <cwicks[m]> this room is always open for students to come and engage.  We encourage you to.
Apr 08 17:38:51 <Pac23> hydraulicsheep great to know that you have been working on this,looking forward to your diagrams
Apr 08 17:39:11 <jkridner[m]> ds2: What if it is demonstrated that the hardware cannot reasonably handle it as far as access is given to program it today?
Apr 08 17:40:11 <jkridner[m]> ds2: ie., can the deliverable be a somewhat definitive profile of the performance bottleneck illustrating where additional performance is needed out of the TI accelerator offering?
Apr 08 17:41:13 <ds2> jkridner[m]: I think even that is okay as long as that is discussed with the mentoring team... what is not okay is - nothing happens, no communications what soever and a eval comes up
Apr 08 17:41:31 <ds2> the AI hard is somewhat of an unknown... inference on embedded device is so new
Apr 08 17:42:14 <ds2> answers to - how does the AI stack up against the jetson for AI purposes, or against the movidius would be useful
Apr 08 17:43:04 <ds2> I know YOLO can run at about 10sec/frame... how much better can we do... 30fps would be nice
Apr 08 17:46:14 <ds2> pradan, jkridner: speed is just one aspect of it... how to do some of this is another aspect.... the TIDL docs are horrible (IMO) - I am hoping blog posts from this project can serve as an ad-hoc guide for others doing this stuff
Apr 08 17:46:30 <ds2> kind of like BeagleLogic is to the PRU
Apr 08 17:46:45 <ds2> pradan[m]: does what I am saying make sense?
Apr 08 17:47:57 <pradan[m]> Yes its crystal clear...
Apr 08 17:52:25 <pradan[m]> > answers to - how does the AI stack up against the jetson for AI purposes, or against the movidius would be useful
Apr 08 17:52:25 <pradan[m]> This seems a good turn... I am familiar with their Model optimization toolkit : OpenVino.. and will try bringing in the hardware acceleration part
Apr 08 17:53:07 <pradan[m]> The rest, is just another method of model pruning
Apr 08 17:53:20 <ds2> pradan[m]: very important to remember - GSoC period is short...
Apr 08 17:56:50 <pradan[m]> Yes . I remember... I'll try my best till there's no time
Apr 08 18:01:53 <vaishnav98_[m]> jkridner : Hi
Apr 08 18:02:18 <jkridner[m]> vaishnav98_: I pinged you on Slack as well. you let me know which is better.
Apr 08 18:02:32 <jkridner[m]> actually, probably Slack to get more of the relavent folks.
Apr 08 18:03:28 <vaishnav98_[m]> jkridner : I will be on slack , for some reason I didn't receive the notification
Apr 08 18:43:20 <vedant16[m]> <jkridner[m] "vedant16: some discusson regardi"> Sorry for the late replies. If REPL is implemented on top of Botspeaks, it is quite possible to do that. I could remove the task of porting remoteproc to C. I am doing courses on compilers. I can comeup with a plan
Apr 08 18:45:18 <vedant16[m]> <jkridner[m] "just out of curiousity, any othe"> All there
Apr 08 18:46:09 <jkridner[m]> vedant16: are you working on a unified proposal on eLinux?
Apr 08 18:49:01 <vedant16[m]> Not as of now. I will start tomorrow.
Apr 08 18:50:13 <vedant16[m]> jkridner: I had a question regarding PRU's design, why wasn't a chip more like atmega used? I mean wrt to memory and ram sizes
Apr 08 18:52:01 <vedant16[m]> Should create a new elinux page or edit the one's i submitted?
Apr 08 19:12:31 <jkridner[m]> vedant16: either way. any potential mentor have an opinion?
Apr 08 19:12:57 <jkridner[m]> vedant16: the architecture of PRU is to have super-low latency....
Apr 08 19:13:17 <jkridner[m]> it isn't a separate chip, it is a separate CPU on the same chip.
Apr 08 19:13:27 <jkridner[m]> well, 2 CPUs.
Apr 08 19:14:06 <jkridner[m]> anyway, they have single cycle access to pins, which you can't do with an ATMega. they are 32-bit and 200MHz.
Apr 08 19:45:36 <vedant16[m]> > anyway, they have single cycle access to pins, which you can't do with an ATMega. they are 32-bit and 200MHz.
Apr 08 19:45:36 <vedant16[m]> Ohh, i see.
Apr 08 19:46:45 <vedant16[m]> > vedant16: either way. any potential mentor have an opinion?
Apr 08 19:46:45 <vedant16[m]> But will it be accepted by Google? Since it will be a 4th proposal technically?
**** ENDING LOGGING AT Thu Apr 09 02:59:57 2020