**** BEGIN LOGGING AT Sun Apr 11 02:59:56 2021 Apr 11 06:33:28 I'm really excited about your project. I also can relate to your feelings in the "Contingencies" section Apr 11 06:40:31 I'd love to learn what you know :) Every time I mess around with ALSA I feel like I'm wandering around a cave with a mediocre flashlight Apr 11 08:54:04 jkridner: I was a bit skeptical that drivers are to be written for zephyr rtos or in micropython. In my proposal I have written it with respect to micropython. Apr 11 11:41:01 Archisman Dey: reviewed your proposal, LGTM with a few minor comments Apr 11 11:41:11 Check the doc Apr 11 11:41:35 Checking now! Apr 11 11:45:10 Students: please do not delay in uploading your proposals in GSoC website, even if we have not been able to thoroughly reviewed them Apr 11 11:45:32 I repeat, do NOT wait until last moment Apr 11 11:46:32 If your proposal is not up on GSoC website and you have not completed the enrollment formalities, we will not be able to consider your proposal for the GSoC program no matter what Apr 11 12:57:32 Abhishek: replied to your comments on the doc and made some minor changes, have a look when you are free Apr 11 13:44:48 referring to ds2 comment, is there any mentor willing to go through my proposal of GPGPU? https://elinux.org/BeagleBoard/GSoC/2021_Proposal/GPGPU_with_GLES Apr 11 13:46:04 with the highlighting enabled in the IRC clients it will indeed tag them Apr 11 13:47:07 Oh, did not know that Apr 11 13:47:49 I will maybe finally create some good materials on the topic. Intended to do so for my thesis, but time fell short. I will dig around the plugins stuff and if it the ALSA with BELA project succeeds I may have initiative to document this thoroughly :) Apr 11 13:48:13 anyway it is a good topic for so called "information void filling" Apr 11 13:48:51 my adventures with ALSA were hindered because of the scarcity of information about this system... Apr 11 13:58:00 jduchniewicz: Can’t comment much on your proposal, I’ll let other mentors take a look on that one Apr 11 13:58:13 sure, thanks for looking anyway Apr 11 14:24:58 That’s fine. I will look at it in that regard. Apr 11 14:29:23 @Yadnik Bendale you need to look at how USB is done on the Freedom board. There is an MSP430 that connects to USB and then connects via UART and I2C to the CC1352. Apr 11 14:34:16 All hardware should be ordered by BeagleBoard.org. You should not nees to go out of pocket. If you must or must pay taxes, Beagle should reimburse you. Apr 11 14:38:22 jkridner: gentle reminder Apr 11 14:38:31 giuliomoro: okay, I am much more optimistic after reading about some previous work on creating virtual devices via plugins! It can be done by the `file` plugin in ALSA. The relevant API: https://www.alsa-project.org/alsa-doc/alsa-lib/pcm_plugins.html -> File Apr 11 14:38:55 what it does, it creates the virtual devices basing on the alsaconf file Apr 11 14:39:04 with the very arcane syntax Apr 11 14:40:45 so basically, assuming there is an FD for the BELA devices it can interact with it Apr 11 14:41:13 probably the rest of the work is to patch up libasound with the necessary RT elevation calls to Xenomai Apr 11 14:41:38 I am actually very optimistic about this being able to work! Apr 11 14:46:13 https://github.com/alsa-project/alsa-lib is updated quite often by Jaroslav Kysela and Takashi Iwai (the original creators :) Apr 11 14:47:51 giuliomoro: before I sketch anything, please check if I am not omitting anything important here (this is quite fragile topic and better double check than get into the project and bail out because it is impossible to deliver it) Apr 11 16:35:10 looking into this Apr 11 16:37:37 jduchniewicz. Bela doesn't currently expose a device file, although it could, however you'd need to interact with it with Xenomai's `cobalt_read()` and `cobalt_write()`. Looking at the source code, the `file` plugin functions call Linux's `read()` and `write()`, which is not going to work. Some of the functionality of the `hooks` plugin seem to be useful. If I understand correctly, it allows to call arbitrary Apr 11 16:37:37 functions from an arbitrary `lib` file. Great! However it does it only for open and close, so not good for us yet. Apr 11 16:37:37 If there is not an ALSA plugin readily available that accepts arbitrary functions for read and write, we could add one. I think that would stand some chance of getting mainlined as it would be a new, clean feature, backward compatible and without hacks. Apr 11 16:50:12 @Abhishek: do you know anything about using TensorFlow Lite on BeagleBone AI and why it's not supported right now? (I put a post about it in the forum and am considering tackling at is a proposal, though I'm not sure who else is working on it so far/what it's challenges are--obviously everything is coming down to time atm) Apr 11 16:52:51 I’m not very sure as I have not worked on TFlite side of things Apr 11 16:53:05 jkridner: maybe ^? Apr 11 17:31:33 There are bits out there, but rcn-er hasn’t integrated them. Apr 11 17:43:06 lpillsbury: I tried tackling YOLO deployment on BBAI but it is currently impossible because of TIDL library restrictions. TFLite is the way to go with DL on BBAI and it uses TIDL underneath Apr 11 18:04:47 jkridner: I saw thatYadnik Bendale created a proposal for MicroPython for BeagleConnect Freedom. Is it possible for me to also make a proposal for that? Apr 11 18:05:39 Yes, it is possible for multiple people to submit a proposal for the same topic Apr 11 18:06:18 giuliomoro: just looked through the source code and it seems like we could do some hacks with this API, or indeed write an entirely new plugin Apr 11 18:06:38 no idea how it would be named though :) snd_rt? Apr 11 18:06:57 or just snd_virt something Apr 11 18:07:07 whatever, not very important right now Apr 11 18:07:33 I like the idea of a new plugin and doing things without dirty hacks! Apr 11 18:08:09 Archisman Dey: k, thanks! Apr 11 18:08:56 also I would really like to have one of the mentors review my proposal on GPGPU with GLES https://elinux.org/BeagleBoard/GSoC/2021_Proposal/GPGPU_with_GLES ds2 already reviewed it and the newest version is on the Google page Apr 11 18:09:27 jduchniewicz: I took a look at your YOLO proposal, are these the problems you are talking about: The TIDL library API guidelines propose 3 approaches to utilizing the library: Apr 11 18:09:43 (taken from your proposal) Apr 11 18:10:03 How do I actually start of writing a proposal? I know its late but I still wanna give it a try! Should I make a new page on elinux? Apr 11 18:11:30 lpillsbury: nope, the problem is having unsupported layers in YOLO model Apr 11 18:12:02 Willmish: create a page on Elinux (a copy of a proposal template) Apr 11 18:12:08 Willmish: and then just edit it in the upper right corner Apr 11 18:12:34 ok, thanks! Apr 11 18:13:09 Willmish: here is the template https://elinux.org/BeagleBoard/GSoC/2021ProposalTemplate Apr 11 18:13:20 what I did, was search for the phrase "BeagleBoard/GSoC/2021 Proposal/" Apr 11 18:13:29 it then prompts you to create it Apr 11 18:13:35 and I copied the template Apr 11 18:15:13 Ill try that in a sec and start collecting some stuff there Apr 11 18:15:43 We will pick best proposals and community engagement will also be considered. Apr 11 18:16:06 Ideas are only suggestions. Apr 11 18:17:29 We won’t select 2 proposals that can’t be run independently. Apr 11 18:21:02 sorry, is rcn-er a person or community? Apr 11 18:22:51 It's Robert Nelson Apr 11 18:28:09 cool thanks Apr 11 18:32:59 What I am thinking of is a plugin that allows you to define arbitrary functions for each of the snd_pcm_* internals (so nothing at all that is Bela- or real-time specific would go in the plugin itself. The Bela parts would be implemented in a plugin configuration file). Apr 11 18:33:40 Yes, got it after reading the newer messages :) Apr 11 18:34:48 I am writing something up, will probably have it ready today or early tommorow Apr 11 18:34:54 *proposal Apr 11 18:35:10 not much time, but worth giving this idea a shot Apr 11 18:37:46 great let me know when you have it and I'll look it over Apr 11 18:38:33 It'd be good if someone else could review that as well, as it's my first year as a mentor, so the sooner it is ready, the better. Apr 11 18:42:49 sure, we got almost exactly 48 hours for this. Can do! Apr 11 19:06:54 Hey jduchniewicz are you by any chance interested in a collab? Apr 11 19:07:39 giuliomoro: I read your very detailed email and I am right away making the inclusions as suggested ASAP Apr 11 19:10:11 giuliomoro: Thanks alot, It almost feels like you handed the proposal to me on a platter :) Apr 11 19:18:24 giuliomoro: https://elinux.org/BeagleBoard/GSoC/2021_Proposal/bela_on_bbai#Proposed_method_of_Implementation_and_Timeline Apr 11 19:18:24 Have a look, I have tried to include almost everything you mentioned in the EMail. Apr 11 19:18:24 I am now submitting this proposal on GSoC incase any failure occurs at the final moment of submission. Apr 11 19:28:04 I think this is kind of the case like a FTDI USB-to-Serial Converter.I am looking out for driver code for the same to take some reference from it. Apr 11 19:29:37 If such drivers are implemented I think it will ease out the proceess on the Freedom board. Apr 11 19:29:47 * If such drivers are implemented I think it will ease out the process on the Freedom board. Apr 11 19:38:41 jkridner: I also came across the MSP430 dev pack which contains necessary source code and sample applications required for developing a USB-based MSP430 projects.https://www.ti.com/tool/MSP430USBDEVPACK Apr 11 19:52:19 Dhruva Gole: I sent you some more feedback via email. If any more experienced mentor would like to help with reviewing this https://elinux.org/BeagleBoard/GSoC/2021_Proposal/bela_on_bbai , that'd be helpful. Apart from the more in-depth technical bits, is the overall shape OK? Apr 11 20:04:08 jkridner:The Descriptor Tool generates the USB interfaces needed that then write an MSP430 application that communicates with the host through these interfaces. Apr 11 20:20:38 Hi Abhishek , Apr 11 20:20:38 https://elinux.org/BeagleBoard/GSoC/2021_Proposal/OmkarBhilare Apr 11 20:20:38 Can you check this proposal in your free time? Michael once said that you were also interested in beaglwire. I was gonna actually focus on new ecp5 cape design but Michael said it will be too intensive so I focused on the current beaglewire software support and wrote proposal for that. Apr 11 20:21:02 * OmkarBhilare[m] < https://matrix.org/_matrix/media/r0/download/matrix.org/oXGlxtYwBXZTePOopxmnsJZI/message.txt > Apr 11 20:21:10 * OmkarBhilare[m] < https://matrix.org/_matrix/media/r0/download/matrix.org/YjgvTfDjnsEnrHbCPXPJUGNb/message.txt > Apr 11 20:21:16 * OmkarBhilare[m] < https://matrix.org/_matrix/media/r0/download/matrix.org/zaMRpYsljHUkUgsNtoDSiuof/message.txt > Apr 11 20:21:33 * OmkarBhilare[m] < https://matrix.org/_matrix/media/r0/download/matrix.org/kiBODcXqHChPbxlafSgvRhZM/message.txt > Apr 11 20:22:00 * OmkarBhilare[m] < https://matrix.org/_matrix/media/r0/download/matrix.org/jriHpoYUzqUjxnuqMKKgQLot/message.txt > Apr 11 21:14:35 Yes, I'm on it... Apr 11 23:05:58 Hello students and mentors!! Apr 11 23:24:34 Hi Yadnik Bendale have you started your draft yet on the official GSoC site? Apr 11 23:25:54 lpillsbury: Have you also started a draft on the official GSoC site? Apr 11 23:26:26 No I have not done it on the GSoC website.I was waiting for inputs from jkridner.Once I get it I'll surely do it. Apr 11 23:26:47 Welcome Willmish Apr 11 23:27:06 cwicks: I will start the process and fill out all the details till then on the GSoC site. Apr 11 23:27:27 Hi! Apr 11 23:27:36 Glad to be here Apr 11 23:29:09 Dhruva Gole: Can you help me understand something about the GSoC system from the student side? Are you able to move a proposal from final back to edit to make updates til the deadline? Apr 11 23:52:22 get at least a draft in. Apr 11 23:52:43 don’t miss submitting, even without more feedback Apr 11 23:52:48 Hi cwicks, I started one for Bela Audio, but I think that I want to focus on making Tensorflow Lite work with BBAI (so should start a newly titled proposal). I have posted on our forum asking @RobertNelson for more info to make sure this is a manageable/relevant project scope since @jkridner suggested that Robert is leading the effort. I think I need to start thinking Apr 11 23:52:49 through the approach more on my own regardless of whether I hear back from him given the deadline and that the point is learning on my own 😀. I've also reached out to a few others who were discussing the issues on the Tensorflow Lite Google groups Apr 11 23:53:02 must be final by deadline. no exceptions. Apr 11 23:53:25 to confirm, deadline is 11am Tuesday? (what timezone?) Apr 11 23:54:47 Check the site. Apr 11 23:56:28 Yeah, but you’ll need to think about how to pass the messages and errors. Apr 12 00:09:06 FYI for everyone: it just says 11:00 Pacific Daylight Time doesn't say am or pm so let's assume am Apr 12 00:10:42 Have you pinged Robert Nelson on the IRC? Apr 12 00:11:23 Students -we like to get things in a day early at a minimum Apr 12 00:11:52 yes Apr 12 00:12:26 he responds best to email Apr 12 00:12:38 OK thanks will do :) Apr 12 02:48:58 StevenSchrstedt[: I haven't heard from you in days... having said that... the contents is about right. prehaps a little bit more with a basics of GPGPU type sample... something that illustrates how to access the data.. or even stuff on what isn't good due to the limitations of GLSL - things like textures being read OR write **** ENDING LOGGING AT Mon Apr 12 02:59:56 2021