**** BEGIN LOGGING AT Mon Apr 24 03:00:02 2017 Apr 24 03:00:21 what's the mode for which the max pixel clock spec'd was made 126 MHz ? (contradicting some other sources that give a lower value) Apr 24 03:00:37 i.e. lcdc_fck = 252 MHz Apr 24 03:01:31 note btw that reducing the risk of fifo underruns should be possible using the prioritization support in the interconnect and EMIF Apr 24 03:01:57 1280x960 is the closest to that Apr 24 03:02:14 i wonder why they didn't just allow the LCD clock divider register bits to be 1 Apr 24 03:07:05 I think that would probably be highly non-straightforward, as there's stuff potentially happening on both edges of the pixel clock Apr 24 03:07:24 either that or it's a bandwidth issue Apr 24 03:07:53 yeah Apr 24 03:07:55 good point Apr 24 03:09:31 1920x1080R@50 might be valid, I'm not sure Apr 24 03:10:50 zmatt yeah the 1GB, 2GB gets up to around $5 Apr 24 03:10:56 I'm not sure how much I need Apr 24 03:11:13 ayjay_t: 2GB/chip doesn't exist Apr 24 03:11:43 according to digikey it does Apr 24 03:11:50 it'll be multi-rank Apr 24 03:12:08 which is unsupported by the am335x (I'm assuming it's for that purpose) Apr 24 03:12:24 yeah i'm thinking about it Apr 24 03:12:35 i guess that narrows down my options Apr 24 03:12:40 2GB would require two x8 chips (like the BBE), each 1GB Apr 24 03:13:03 single-rank 1GB chips are rare and expensive though Apr 24 03:13:12 oh wait so you can't do x8! Apr 24 03:13:15 whats the max then? x4?? Apr 24 03:13:28 TI doesn't support x4 chips Apr 24 03:13:47 so a 16-bit bus like the am335x means 1 x16 chip or 2 x8 chips Apr 24 03:14:26 the latter will get you additional headaches since now the cmd/addr bus is no longer point-to-point but multi-drop Apr 24 03:14:28 really so i can still do 2G just with two 128Mx8 Apr 24 03:15:23 oi i see a 2G 256M x 8 and a 2G 256M x 16 Apr 24 03:15:50 not single-rank Apr 24 03:16:17 oh wait Apr 24 03:16:27 isn't that Mbit rather than Mbyte ? Apr 24 03:16:53 fuuuuuu Apr 24 03:17:45 lol why is this whole table in bits Apr 24 03:17:47 who thinks in bits Apr 24 03:17:54 history Apr 24 03:18:23 okay so for me to get my 512Mbyte i need to get a 4Gb chip Apr 24 03:18:29 everything just got even more expensive Apr 24 03:19:23 anyone knows an app that can control the beaglebone black remotely over wifi Apr 24 03:19:26 ? Apr 24 03:19:34 ssh Apr 24 03:20:09 that's for the terminal. what about if i want to control it with my phone? Apr 24 03:20:40 have you seen the default landing page like if you navigate to the IP address with your browser? Apr 24 03:21:06 yep? Apr 24 03:22:00 it's not the easiest thing but you can modify the source and add new commands Apr 24 03:22:12 i just want something similar to blynk. but blynk doesn't have BBB as one of the supported devices Apr 24 03:23:03 is there any example out there about how i can modify the source? Apr 24 03:24:10 ayjay_t: looks like this is an example of single-rank 1GByte: https://www.micron.com/parts/dram/ddr3-sdram/mt41k512m16ha-125 Apr 24 03:24:12 it's a good idea Apr 24 03:24:44 thanks Apr 24 03:24:54 i haven't seen anyone do that Guest45641 but i'll probably take a look at it this year Apr 24 03:25:00 i know that's not saying a lot right now Apr 24 03:26:49 that took me to a 2GB zmatt Apr 24 03:26:54 oh 2Gb Apr 24 03:27:08 isn't that 256MB? Apr 24 03:27:17 huh, did I misclick.. Apr 24 03:27:25 no, it says 8Gb here Apr 24 03:27:42 hmm yeah i had to go through a list and find the one you were referencing Apr 24 03:28:17 this is *definitely* not where i expected my costs to come from Apr 24 03:28:23 first of all, because I misread the Gb as GB Apr 24 03:29:12 on the bright side, you can basically use the slowest ram you can find Apr 24 03:29:13 ugh linux can use an SD card as virtual memory too Apr 24 03:29:18 lol Apr 24 03:29:20 you don't want that Apr 24 03:29:21 maybe linux is the way to go Apr 24 03:29:35 what do you need that much ram for anyway? Apr 24 03:30:00 i mean i just assumed that if i set up 1 GB of ram i wouldn't have to worry Apr 24 03:30:15 i *want* to run chrome really well, 1 tab Apr 24 03:30:22 oh Apr 24 03:30:24 now i actually have to think about the problem XD Apr 24 03:30:32 ehm, ok, yea, well Apr 24 03:30:35 chrome Apr 24 03:30:35 ok Apr 24 03:30:44 have fun? :) Apr 24 03:31:09 oi your confidence is killing me Apr 24 03:31:12 I was gonna say I never felt limited by the 512MB of ram even in the slightest bit Apr 24 03:31:19 but then again, I've never felt the need to run chrome Apr 24 03:31:53 linux supports swap-over-NFS ? ;-) Apr 24 03:32:25 i mean, even the 512MB ram is going to end up being way more expensive than i thought Apr 24 03:32:37 I was looking at 2Gb being $5 and i was like ahhh Apr 24 03:32:47 but now i realize my $5 2Gb is actually my $5 512MB Apr 24 03:33:08 no, 2Gb = 256MB Apr 24 03:35:38 yeah /8 not four, sorry i'm a little big WTF at the moment Apr 24 03:35:48 i'm running profiles on heavy RAM eating sites to see how much i really need Apr 24 03:41:37 keep in mind that more ram probably also means more data it's trying to access, which will affect how much you get hurt by the limited memory bandwidth available Apr 24 03:42:50 one 16-bit DDR3-800 channel is not much :) Apr 24 03:43:31 yeah i'm really trying to see what reasonably i can whittle chrome down to Apr 24 03:44:30 you'd be surprised which web apps use the most RAM (*cough* gmail) Apr 24 03:44:53 like it's literally just a list of strings, and somehow, i can't even explain how it uses so much RAM Apr 24 03:45:06 not surprised there, I inspect the memory usage of tabs often enough when my laptop ends up in swap hell again Apr 24 03:45:56 so i'm down to half a GB system wide, that's heartening Apr 24 03:48:43 the browser, the gpu process, and the gmail tab = {300k, 150k, 300k}, but i believe it's counting memory twice since the GPU process and the TAB use a shared memory ring buffer to transfer openGL commands Apr 24 03:49:13 you actually got chrome to use the gpu? Apr 24 03:49:34 i think it's been doing that for a while, tbh Apr 24 03:49:42 there's a rather large team of people working on the blink renderer Apr 24 03:50:14 what environment are you running chrome in then? wayland? Apr 24 03:50:30 it uses it on the android too, except it doesn't create a separate gpu process Apr 24 03:50:37 no i'm running i3->xserver->gentoo Apr 24 03:50:59 oh I thought you meant on the beaglebone (or am335x in general) Apr 24 03:51:56 haha sorry! Apr 24 03:51:57 no! Apr 24 03:52:03 but i'm totally going to. Apr 24 03:56:18 i just downloaded the sdk XD i wonder how crazy this is going ot be Apr 24 03:56:27 the sdk? Apr 24 03:56:56 the one rcn links to in his buildscript Apr 24 03:57:06 for sgx Apr 24 03:57:11 you don't mean the graphics sdk ? Apr 24 03:57:23 (which is obsolete) Apr 24 03:58:29 oh yeah i got it off the wiki Apr 24 03:59:37 :-/ Apr 24 03:59:48 the current sdk is the "Linux Processor SDK for AM335x", but a brief sketch on how to get the ingredients working on debian is here: https://www.mail-archive.com/beagleboard@googlegroups.com/msg43945.html Apr 24 04:00:36 I've also got qt5 eglfs_kms working (though I patched and recompiled it) Apr 24 04:00:45 haven't tried wayland yet Apr 24 04:01:23 rcn integrated the kernel patches btw Apr 24 04:01:55 i see you have the patches Apr 24 04:01:57 so a recent 4.9-bone kernel should just work Apr 24 04:02:16 yarr but i'm trying to dissect the drivers Apr 24 04:02:51 I locally built some sketchy debian packages Apr 24 04:03:04 like some userspace libraries? Apr 24 04:03:45 wait wait what you linked me to, rcn's reply, that's the link i followed to get the sdk which you say is out of date Apr 24 04:03:53 so you're saying rcn integrated your changes into the drivers/ Apr 24 04:03:54 ?* Apr 24 04:05:44 ohh, then I misunderstood which link you meant Apr 24 04:05:47 probably Apr 24 04:05:58 which link did you mean? Apr 24 04:07:11 http://software-dl.ti.com/dsps/dsps_public_sw/gfxsdk/ Apr 24 04:07:28 in his loader script from 4.4 it references that domain Apr 24 04:07:30 yeah that's ancient Apr 24 04:08:09 even if 4.9, that's in his sgx_build_modules.sh script Apr 24 04:08:19 that script hasn't been updated in ages Apr 24 04:08:39 alright well i'm going through that thread so ill dig into what you posted Apr 24 04:08:55 I'll see if I can make my work a bit friendlier to use Apr 24 04:09:08 keep in mind that though TI completely dropped support for X11 Apr 24 04:09:50 yeahh i'm going to try to mux the GPU in kernelspace Apr 24 04:09:58 mess* Apr 24 04:10:09 I have no idea what you could possibly mean by that Apr 24 04:10:15 most of the work is done in userspace Apr 24 04:10:34 the kernel driver is afaik just a shim to let userspace safely communicate with the gpu Apr 24 04:10:59 to "mux" the GPU, use wayland Apr 24 04:11:21 eh i want to Apr 24 04:11:58 i'm not doing anything productive, unfortunately, it's all for my own education Apr 24 04:12:12 although i should feel bad for that and try to contribute, i suppose Apr 24 04:12:23 I should probably try to do a better write-up of what I've managed to do so far probably Apr 24 04:12:31 sentence fail Apr 24 04:13:15 and I really need to try wayland Apr 24 04:13:18 well i can give you some feedback before you do that Apr 24 04:13:24 since i'm going to be reading through your material Apr 24 04:13:36 i'm doing it right now and taking notes so Apr 24 04:15:21 basically: if you build a 4.9-bone kernel, you can then use my build.sh in https://github.com/mvduin/omap5-sgx-ddk-linux to build the kernel module Apr 24 04:15:39 noo i don't want the easy way Apr 24 04:15:50 i want to understand what you've done, and what you're doing Apr 24 04:16:12 don't worry, you'll need to read the build script anyway since it contains hardcoded paths you need to fix Apr 24 04:16:16 :) Apr 24 04:17:15 so, the very short history is that I already knew about the git repos (omap5-sgx-ddk-linux and omap5-sgx-ddk-um-linux) for a bit longer Apr 24 04:17:26 the former for the kernel module, the latter containing the userspace libs Apr 24 04:17:52 to be able to compile the kernel module at all I had to fish some patches out of TI's kernel git repo Apr 24 04:18:19 those are the ones I reference in my mailing list post, and have been integrated into 4.9-bone (and probably 4.9-ti, didn't check) Apr 24 04:18:21 ugh my pdfs aren't rendering kern properly Apr 24 04:26:42 if you got the kernel module compiled and installed, you should be able to get the userspace libs working by installing this set of horribly kludgy debian packages I built: https://gerbil.liktaanjeneus.nl/sgx/ Apr 24 04:27:08 I will try to make nicer ones when I find some time Apr 24 04:28:57 this basically just the libs from omap5-sgx-ddk-um-linux, together with libgbm and kmscube compiled from my git repo, and some fighting with shared lib dependencies and debian dependencies Apr 24 04:29:32 it conflicts with everything mesa Apr 24 04:38:32 woh i found this gitorius thing Apr 24 04:38:37 i feel like i'm somewhere i'm not supposed to be Apr 24 04:39:11 wasn't the powervr thign closed source? is it still closed source? Apr 24 04:39:18 http://git.ti.com/graphics/ you mean? Apr 24 04:39:35 the userspace libs are closed yes Apr 24 04:39:51 which is why omap5-sgx-ddk-um-linux only contains compiled binaries Apr 24 04:39:59 wait wait Apr 24 04:40:02 the userspace libs are closed Apr 24 04:40:06 but the kernelspace libs are open Apr 24 04:40:09 THATS BACKWARDS Apr 24 04:40:12 no Apr 24 04:40:22 06:10 < zmatt> the kernel driver is afaik just a shim to let userspace safely communicate with the gpu Apr 24 04:40:28 i mean isn't it orders of magnitudes easier to reconstruct the userspace .... Apr 24 04:40:31 no Apr 24 04:40:33 okay okay Apr 24 04:40:36 everything is done in userspace Apr 24 04:40:43 you're saying it's just exposing registers or something like that, basically Apr 24 04:41:01 it's mostly just memory safety stuff, mmu setup and such Apr 24 04:41:08 yeah that sucks Apr 24 04:41:11 i got really really excited Apr 24 04:41:12 in fact that's how it works for other gpus too Apr 24 04:41:23 that's the Direct in Direct Rendering Manager Apr 24 04:41:27 LOL Apr 24 04:41:53 you don't throw all the opengl stuff and whatnot into the kernel Apr 24 04:42:04 i'm going to decompile the userspace lib, find all the SVC, and reverse engineer it Apr 24 04:43:14 if you have nothing better to do with your time the next couple of years Apr 24 04:44:02 i've got a lot of coffee Apr 24 04:44:36 i might actually compile your kernel this week Apr 24 04:46:36 jacinto6evmwut Apr 24 04:49:03 /lib is only 2080 bytes Apr 24 04:49:32 i lied Apr 24 04:52:04 my kernel? why? Apr 24 04:52:32 it's 4.9-bone + local patches that are probably not of general interest Apr 24 04:52:52 *shrug* then i wont Apr 24 04:52:55 i haven't taken a look at it Apr 24 04:53:15 Hi Apr 24 04:53:18 hi Apr 24 04:53:26 I have a question Apr 24 04:53:35 regarding the beagle bone black Apr 24 04:53:36 ayjay_t: patches of general interest (like the ones for sgx) generally end up being merged by rcn Apr 24 04:54:07 Is beagle bone black exposed the i2s lines on the expansion header? Apr 24 04:54:48 ? Apr 24 04:56:47 i believe so Apr 24 04:56:52 you shoudl google beaglebone black pinmap Apr 24 04:57:49 Thank you ayjay Apr 24 05:00:03 It seems not exposed Apr 24 05:00:08 Thank you ayjay Apr 24 05:02:10 did you find the excel sheets Apr 24 05:04:03 it has multiple i2s interfaces Apr 24 05:04:13 (exposed on the headers) Apr 24 05:05:52 thats what i thought, but i'm assuming he just found the small picture that only shows the GPIOs Apr 24 05:06:41 https://goo.gl/Jkcg0w Apr 24 05:06:57 (P9/P8 tabs especially) Apr 24 05:07:06 Praveen Apr 24 05:10:51 so basically TI engineers sign an NDA and then they can work on these PowerVR drivers Apr 24 05:10:55 ? Apr 24 05:11:29 something like that yes, although I'm not sure how much they work on the drivers themselves vs just the platform glue Apr 24 05:15:45 i feel like given your technology is 5 years old and being surpassed you'd FOSS it just to stay relevant Apr 24 05:15:59 but who am i Apr 24 05:16:12 there might be complicated concerns Apr 24 05:16:53 yeah there always are... Apr 24 05:17:03 I know a dev withing imgtec who's still very much in favor of open sourcing, but his suggestions do not seem to have much traction within the organisation Apr 24 05:19:44 yeah, i mean, my attitude is, if you're giving it away for free anyway Apr 24 05:20:01 and it's basically an accessory to whatever you're selling Apr 24 05:20:05 you should FOSS it Apr 24 05:20:24 because it increases the value to your user, although not to you Apr 24 05:20:34 except in this case, it never had value to you, because you don't charge for it Apr 24 05:21:54 Hi Apr 24 05:22:10 Sorry, I was lost in other work Apr 24 05:22:13 Yaa Apr 24 05:22:21 I read SRM and found the modes Apr 24 05:22:23 etc Apr 24 05:22:30 but nothing related to the i2s Apr 24 05:22:30 good good Apr 24 05:22:32 oh Apr 24 05:22:44 zmatt posted a link Apr 24 05:22:50 maybe that will help Apr 24 05:23:33 Yaa Apr 24 05:23:38 Looking into it :) Apr 24 05:25:00 LOL http://www.eetimes.com/document.asp?doc_id=1316832 Apr 24 05:26:40 what does dbm stand for anyway Apr 24 05:28:14 DBm is an abbreviation for the power ratio in decibels of the measured power referenced to one milliwatt. Apr 24 05:28:40 That's what google said :) Apr 24 05:29:38 nopeeee not that dbm Apr 24 05:29:49 thanks though Apr 24 05:30:39 Praveen: the McASP peripherals support I²S (among other things) Apr 24 05:33:14 ooh zmatt trying to disassemble these libraries is gonna be fun Apr 24 05:34:47 @zmatt, I wanna connect my board to the header using some fly wires Apr 24 05:35:14 McASP does not seem to be on the p9 and p8 Apr 24 05:35:56 Praveen, bud, you gotta read closer Apr 24 05:36:05 The asp stuff in the p8 and p9 are the things that u r talking about Apr 24 05:37:49 yes I abbreviated mcasp to asp in my spreadsheets Apr 24 05:38:18 WoW Apr 24 05:38:25 It seems I can connect to them Apr 24 05:38:50 in p8 I can see lrclk,bclk and some data lines Apr 24 05:39:08 mcasp 0 is somewhat occupied by the hdmi audio stuff, although you can disable that Apr 24 05:39:30 i'm realizing that i don't actually know what opengl does Apr 24 05:39:34 Yaa Apr 24 05:39:53 I dont want audio from the hdmi, I can connect my codec stuff here Apr 24 05:39:56 there's also a 25.576 MHz oscillator that can be enabled or disabled Apr 24 05:40:11 it's audio *to* hdmi, not *from* hdmi :) Apr 24 05:40:22 *24.576 MHz Apr 24 05:40:55 So the lines gives the i2s data or i2s data converted to hdmi ? Apr 24 05:47:07 the hdmi framer accepts i2s input Apr 24 05:49:04 Lost connection Apr 24 05:50:09 ayjay_t: I just pushed an updated version of my build script for the sgx kernel drivers, it now builds a debian package instead of a tarball Apr 24 05:50:20 no promises that I'm Doing It Right™ ;-) Apr 24 05:50:29 i'm still reading the assembly Apr 24 05:51:01 the assembly? Apr 24 05:52:19 yeah i'm starting to pick apart the userspace libraries Apr 24 05:55:07 i'm hoping it takes about 2 weeks Apr 24 05:55:20 lol Apr 24 05:58:42 i mean i'm gonna skip most of the linux specific stuff, like i think, dbm, and the wayland stuff Apr 24 05:58:53 and i'm just interested in egl, vg, and gles Apr 24 06:00:33 ayjay_t... I don't mean to be discouraging, but... Apr 24 06:01:09 i mean it's just for learning Apr 24 06:03:54 you're looking at 1.3 MB of compiled code for the core parts of the library (libGLESv2 + libglslcompiler + libusc) Apr 24 06:05:03 even a few KB of code can easily take weeks Apr 24 06:05:49 no way! Apr 24 06:05:50 reverse-engineering the sgx gles driver is probably many man-years of work Apr 24 06:06:02 how fast do you think someone reads code? Apr 24 06:06:07 lines per minute? Apr 24 06:06:10 i say 20 Apr 24 06:06:15 read or understand? Apr 24 06:06:45 read for comprehension Apr 24 06:06:53 * zmatt shrugs Apr 24 06:06:55 good luck Apr 24 06:07:43 right now i'm actually just trying to figure out what opengl does Apr 24 06:07:46 I still plan to dig further into the platform glue logic, and even that seems like a fairly big task, even though it's tiny in comparison Apr 24 06:07:49 i mean, i know it does graphics and parallel processing Apr 24 06:08:00 then there are probably much better resources to learn that Apr 24 06:08:26 yeah i'm wokring on the khronos site too Apr 24 06:08:26 understanding modern opengl deeply is a prerequisite for trying to figure out the sgx driver, not a motivation Apr 24 06:08:53 LOL i wouldn't call it modern, its a few versions behind XD Apr 24 06:09:03 note btw that GLESv1 and v2 are very different Apr 24 06:09:44 well, one major version behind... Apr 24 06:10:14 i don't know what version of EGL to look at tho Apr 24 06:10:55 1.4 i gues Apr 24 06:12:02 egl is basically just platform glue though Apr 24 06:13:44 thats basically what khronos says in more words Apr 24 06:14:52 anyway, stuff to do... Apr 24 16:50:06 hello all, I have a Intel Curie board that Iam trying to flash with this https://github.com/CurieBSP/main , unfortunately After burning the bootloader and the sample program as shown in the steps, Iam not able to flash any application again,as the curie now is not getting detected through USB..Can anybody help? Apr 24 16:52:22 try some iRadium? Apr 24 18:33:18 i have a beaglebone like board, (3352), only has 256MB that runs WinCE, not porting Linux, what GUI option do I have? need run vlc, light-weight browser, a fluxbox/openbox Apr 24 18:34:10 i feel x-windows is too heavy here, so the options left are QT? SDL? some kind of framebuffer? Apr 24 19:03:12 am3352 with 256MB ram... heh Apr 24 19:04:39 I mean, it's no doubt possible :) Apr 24 19:38:03 zmatt: possible with what option....is the question, android/x11 is definitely out, tiny-X is no longer active, wayland is heavy? Apr 24 19:41:15 only QT / SDL ? note 3352 has no GPU Apr 24 20:44:17 Hello? Apr 24 20:44:24 I have a question Apr 24 20:45:07 What's the maximum size sd card I can use for BBB? Apr 24 21:11:26 Hello Apr 24 21:12:04 Anyone there? Apr 25 02:30:28 hmm, I wouldn't try to insert anything bigger than a microSD into the BBB ;) Apr 25 02:33:40 heh Apr 25 02:38:40 that explains everything! Apr 25 02:42:45 (slightly late reaction to < Patrick___> What's the maximum size sd card I can use for BBB? ) **** ENDING LOGGING AT Tue Apr 25 03:00:02 2017