Tuesday, 2017-01-17

*** tpb has joined #timvideos00:00
*** tvCommitBot has joined #timvideos00:13
tvCommitBot[streaming-system] mithro pushed 1 new commit to master: https://git.io/vMKai00:13
tvCommitBotstreaming-system/master e67abf7 Tim 'mithro' Ansell: Kill all children!00:13
*** tvCommitBot has left #timvideos00:13
mithrojea: https://github.com/timvideos/streaming-system/blob/master/tools/preview/generate.py00:14
tpbTitle: streaming-system/generate.py at master · timvideos/streaming-system · GitHub (at github.com)00:14
*** CarlFK1 has joined #timvideos00:29
*** ChanServ sets mode: +v CarlFK100:29
*** CarlFK has quit IRC00:32
*** tvCommitBot has joined #timvideos01:11
-tvCommitBot- [lca2017-av] stefanor pushed 2 new commits to master: https://git.io/vMKob01:11
-tvCommitBot- lca2017-av/master 055af93 Stefano Rivera: Make rX actually work01:11
-tvCommitBot- lca2017-av/master 064b102 Stefano Rivera: Some new encoder T410s01:11
*** tvCommitBot has left #timvideos01:11
mithrojea: On the front page it is showing the duration of the "next talk" rather then the time it starts01:12
jeamithro: ok, I'll check that01:16
mithrojea: Same in the page01:17
mithrojea: Lunch is also showing as being 2 hours and 20 minutes...01:18
mithrojea: And the html is not rendering in the pages either01:19
*** tvCommitBot has joined #timvideos01:21
-tvCommitBot- [lca2017-av] stefanor pushed 1 new commit to master: https://git.io/vMKKR01:21
-tvCommitBot- lca2017-av/master 5322322 Stefano Rivera: More encoder boxes01:21
*** tvCommitBot has left #timvideos01:21
mithrothaytan: What is vaapih?01:22
thaytanmithro, vaapih264enc you mean?01:22
mithrothaytan: Yeah?01:23
thaytanmithro, it's the VAAPI H.264 encoder element01:23
thaytanthat does GPU encoding via VAAPI where it's supported01:23
mithrothaytan: I mean what is VAAPI :-P01:24
*** paradisaeidae_ has joined #timvideos01:25
thaytanmithro, it's an API for accessing GPU decode/encoders01:28
thaytanVA = Video Acceleration01:28
mithrothaytan: Ahh, Googling seems to tell me it's an intel graphics thingy01:29
mithroVAAPI (Video Acceleration API) is an open-source library and API specification, which provides access to graphics hardware acceleration capabilities for video processing. It consists of a main library and driver-specific acceleration backends for each supported hardware vendor.01:29
thaytanmithro, it is, but can also wrap VDPAU which is nvidia's API01:29
thaytanCarlFK1, Chris Neugebauer01:30
tumbleweedmithro: you're going to have to dig around and grab interesting video from sat/sun before space gets tight and we delete it01:54
tumbleweedI think you probably have a couple of days to do so01:54
tumbleweedwe're at: /dev/md0                  3.7T  2.1T  1.6T  57% /srv01:54
*** sb0 has quit IRC02:55
*** sb0 has joined #timvideos02:58
tumbleweedthaytan: https://anonscm.debian.org/cgit/debconf-video/ansible.git/tree/roles/voctomix/templates/videoteam-stream.j203:16
tpbTitle: debconf-video/ansible - Ansible playbooks for the DebConf Video Team's machines (at anonscm.debian.org)03:16
thaytantumbleweed, so.... turns out you need an audio track, even if it's just silence03:28
thaytanOK, so https://www.youtube.com/watch?v=WB2XlB-Zhw0 is live streaming from my laptop @ 720p with vaapih264enc03:32
thaytan25% of 1 core03:32
thaytanwhich is mostly audio enc I think03:32
tumbleweed\o/, we'll give that a shot03:32
thaytangst-launch-1.0 videotestsrc ! video/x-raw,width=1280,height=720 ! vaapih264enc keyframe-period=60 max-bframes=0 bitrate=2000 ! video/x-h264,profile=main ! flvmux streamable=true  name=mux ! queue ! rtmpsink location="rtmp://a.rtmp.youtube.com/l...."03:32
xfxfmithro: you around?03:48
xfxfstatus/status short currently doesn't give me any info the video_matrix (what is connected to what), or am I misreading?03:49
xfxfam writing something to parse it to update the UI03:49
xfxfbut i can't see what to parse :P03:49
xfxfi'm thinking of creating a new status which is meant to be entirely machine readable that has everything we care about03:50
xfxfdistinctly seperate to existing 'status short' et al which is meant for human readable logs03:50
xfxfah, sorry, it does, but only if something valid is connected - i.e. input0 > output1 without input0 being connected won't show anything in status03:56
jeaxfxf: i have considered adding in a json logging mode, just for that purpose04:08
xfxfi'm not convinced mutliline json is the right solution here considering we can't guarantee we'll get it intact (other things coming from the multiplexed connection), but happy to be convinced otherwise04:15
mithrocr1901_modern: ping?04:31
cr1901_modernmithro: pong. I'm stumped, so I'm trying to revive the lm32 testbench and look at iverilog output. Current goal is to mimic an SPI flash in software04:32
mithrocr1901_modern: Do you have working romboot?04:33
mithrocr1901_modern: I just got the same failure when going back to romboot04:33
cr1901_modernmithro: Yes. I also have a stashed away failing romboot04:33
cr1901_modernnow that I think about it04:33
cr1901_modernMind sending me the .v file so I can do a diff?04:33
mithrocr1901_modern: Do you have a non-failing romboot?04:34
* cr1901_modern checks what commit he's on04:35
mithroIt seems to be04:36
cr1901_modernmithro: Give me two minutes please, sorry. Looks like I went back to master since I last flashed04:37
mithro            integrated_sram_size=0x4000,04:37
cr1901_modernmithro: Yes, that was indeed different between failing and non-failing boot04:37
mithrocr1901_modern: Did you try spiflash booting with that set back?04:41
cr1901_modernmithro: No. Will try in a minute04:42
*** sb0 has quit IRC04:46
mithrocr1901_modern: I'm testing here now04:47
mithroIt works!04:50
cr1901_modernUn-freaking-believable. I tried everything BUT that04:52
cr1901_modernmithro: Actually, that's not true...04:54
cr1901_modernmithro: I looked at the stack usage previously, and concluded it wasn't possible to exhaust the stack before the errors started appearing04:55
cr1901_modernEven deep within the UART code.04:55
cr1901_modernSo I suspect something more egregious was occurring than simple stack exhaustion.04:55
*** paradisaeidae_ has quit IRC04:55
mithrocr1901_modern: flashboot is failing though...05:07
cr1901_modernmithro: Okay, so what DOES "work" then?05:07
cr1901_modernerm, remove quotations05:07
mithrocr1901_modern: The bios shows BIOS> prompt05:08
cr1901_modernhmmm... mithro, tbh. I'm not sure I ever stopped flashboot now that I think about it05:10
cr1901_modernso it goes to BIOS05:10
cr1901_modernmithro: Actually, thinking of that more, that shouldn't matter, b/c the BIOS crashed before boot began05:16
cr1901_modernIf I'm honest, I have no idea why extending the amount of SRAM fixed even part of the problems. Need to write up a document of what worked vs what didn't05:18
cr1901_modernmithro: Also, amazingly, litex can't be found on my path anymore when I try to build it using Atom keybindings. Worked fine before the third-party stuff. Also need to fix that05:18
mithrocr1901_modern: So - I want to know why changing the sram breaks the bios05:29
mithrocr1901_modern: If you make the sram too small, it linker complains about it being too small05:29
cr1901_modernright, that's expect05:29
mithrocr1901_modern: So it confuses me that changing it would fix something05:29
cr1901_modernI'm writing a document of what's worked/what hasn't05:30
mithrocr1901_modern: So - _florent_ merged that wip branch for the spiflash - does that work for the spiflash?05:30
cr1901_modernmithro: I don't know, I currently can't build bitstreams b/c going to the third-party directory approach broke my build :( (I only pull when I have to)05:31
mithrocr1901_modern: The idea with the third_party+conda is that you always have the exact same version of the libraries as everyone else05:32
cr1901_modernWell, the build broke b/c litex can't be found05:34
cr1901_modernit's prob some windows-related bullshit05:34
mithroxfxf: Your switcher should use ncurses! https://github.com/groklearning/microbit-simulated-dal/raw/master/docs/curses-gui.png05:34
xfxfi'm trying to build a generic Hdmi2Usb object which keeps state05:36
xfxfso hopefully other versions not too hard05:37
cr1901_modernmithro: It still may come down to simulating a minimal component SoC... just so I can examine RAM reads/writes in real time. I'm not sure litescope is gonna help me that much unless I extend it for fine control for gating the CPU clock05:39
mithroxfxf: Sure05:39
_florent_mithro: I'm working on the SPI Flash with the minispartan605:39
mithro_florent_: minispartan6 or the mimasv2?05:40
_florent_mithro: minispartan6, if it works for it, it will work for the mimasv305:40
_florent_mithro: because the minispartan6 is easier to work with05:40
mithro_florent_: So, it seems like spiflash is working for the mimasv205:40
_florent_mithro: what are you testing?05:41
mithro_florent_: When I was at your place early december I was using the wb bridge to read from the spiflash05:41
cr1901_modernMay I interject?05:42
_florent_mithro: ok but that's not enough05:42
mithro_florent_: Bit banging writing didn't work05:42
_florent_mithro: I'm writing a test that will run on the lm3205:42
*** hyades has joined #timvideos05:42
cr1901_modernWho is doing *what* right now?05:42
cr1901_modernmithro wrote an SPI flash fix. That worked with the mimas scope target.05:43
_florent_mithro: it's possible the wb bridge test doesn't catch all possible issues (ie that's not a real time test)05:43
mithro_florent_: I think I proved the BIOS was booting from spiflash on Jim's MimasV2 about an hour ago05:43
_florent_mithro: ah, wasn't aware05:43
cr1901_modernmithro also PR'ed a *different* SPI flash fix to _florent_'s LiteX repo05:43
cr1901_modernTaking a look over the PR, I don't think it's going to work as-is05:44
cr1901_modernBut admittedly, I forgot that mithro made that pull request05:44
cr1901_modernb/c I was using the hot-patch that mithro wrote initially05:44
_florent_mithro: ok so it's working?05:44
cr1901_modern_florent_: ALL of my tests have been done with this commit: https://github.com/mithro/litex/commit/1aabe4aa2c526554f120aeec46adc544904f6a8c05:46
tpbTitle: Temp: Trying to get the SPI flash driver working in 1x mode. · mithro/[email protected] · GitHub (at github.com)05:46
_florent_mithro: I was trying to help, but if it's working I'll do something else :)05:46
cr1901_modernthis is NOT the same as the pull request you received05:46
mithro_florent_: Unsure, lets sync up about this after the lightning talks05:46
* cr1901_modern is being cautious about introducing a single change each recompile05:47
_florent_mithro: ok05:47
mithro_florent_: Out of interest, do you know to put a SD card into "spi mode"? I'm interested in trying to get an SD card working on the mimasv2 for loading the firmware05:49
* cr1901_modern is in commit hell atm, please leave a message at the beep05:51
cr1901_modernmithro: were you using https://github.com/mithro/litex/commit/1aabe4aa2c526554f120aeec46adc544904f6a8c when testing the flash?05:51
tpbTitle: Temp: Trying to get the SPI flash driver working in 1x mode. · mithro/[email protected] · GitHub (at github.com)05:51
mithrocr1901_modern: Yes06:05
cr1901_modernmithro: I think for now, I'm going to use my copy of litex/or patch the third-party litex using this commit. So when I report anything, it will be done using that commit06:09
cr1901_modernmithro: For now, I'm going to figure out why the SRAM extension made things succeed. B/c I agree w/ you; I don't think it should've worked.06:10
mithrocr1901_modern: Will have more info in a couple of hours06:10
cr1901_modernIt may also shed light on why the flash isn't working06:10
cr1901_modernmithro: Ack, I'll just do my own thing for now. Anything else you need me to do?06:10
mithrocr1901_modern: The SRAM thing is a good thing to look at06:11
cr1901_modernAlright, let me know more info in a few hours. If I don't respond immediately, I fell asleep06:11
cr1901_modernBut I'll read the scrollback when I wake up06:12
*** tvCommitBot has joined #timvideos06:24
tvCommitBot[streaming-system] joeladdison pushed 2 new commits to master: https://git.io/vMKb906:24
tvCommitBotstreaming-system/master 2e1767b Joel Addison: Handle talk with no start or end time06:24
tvCommitBotstreaming-system/master 00d6f69 Joel Addison: Fix up timing information displayed about talks...06:24
*** tvCommitBot has left #timvideos06:24
*** cr1901_modern has quit IRC06:25
*** cr1901_modern has joined #timvideos06:28
_florent_mithro: I was thinking it would be simple to get SD card working in SPI mode but it seems it's not the case06:40
_florent_mithro: http://elm-chan.org/docs/mmc/mmc_e.html06:40
tpbTitle: How to Use MMC/SDC (at elm-chan.org)06:40
cr1901_modern_florent_: SD cards are a PITA. Even porting an SD library to a microcontroller isn't an exceptionally fun thing to do06:42
_florent_mithro: I'm going out, we'll see later for the SPI flash06:42
_florent_cr1901_modern: Ah ok, not sure I ever used one directly06:42
cr1901_modern_florent_: Certainly it can be done, and 4 years ago is a long time ;)06:43
cr1901_modernI'd prob have more luck now06:43
*** hyades has quit IRC07:51
*** Guest5668 has quit IRC08:45
*** jea has quit IRC08:46
*** miselin has joined #timvideos08:46
*** jea has joined #timvideos08:47
*** CarlFK1 has quit IRC08:57
*** miselin has quit IRC09:28
*** miselin has joined #timvideos09:29
*** miselin has joined #timvideos09:29
*** hyadez has quit IRC09:51
*** hyadez has joined #timvideos10:05
*** [d__d] has quit IRC10:26
*** [d__d] has joined #timvideos10:28
*** CarlFK has joined #timvideos12:08
*** ChanServ sets mode: +v CarlFK12:08
*** CarlFK has quit IRC12:09
*** CarlFK has joined #timvideos12:10
*** ChanServ sets mode: +v CarlFK12:10
*** sb0 has joined #timvideos12:32
CarlFKxfxf: http://veyepar.nextdayvideo.com/main/C/lca/S/lca2017/E/12052/12:42
tpbTitle: veyepar: Lightweight inventory management with Pallet Jack (at veyepar.nextdayvideo.com)12:42
xfxfCarlFK: http://veyepar.nextdayvideo.com/main/C/lca/S/lca2017/L/plenary/E/12165/#ep-sm12:46
tpbTitle: veyepar: Turtles all the way down - Thin LVM + KVM tips and Tricks (at veyepar.nextdayvideo.com)12:46
xfxfand a few talks after (click next)12:46
*** rohitksingh has joined #timvideos13:43
*** hyadez has quit IRC15:11
*** hyadez has joined #timvideos15:15
*** nrossi has quit IRC16:30
*** rohitksingh has quit IRC18:34
CarlFKwhy are all the encoders going so slow?!!!20:03
*** CarlFK has quit IRC21:10
*** CarlFK has joined #timvideos21:39
*** ChanServ sets mode: +v CarlFK21:39
*** CarlFK has quit IRC22:04
*** CarlFK has joined #timvideos22:05
*** ChanServ sets mode: +v CarlFK22:05
tumbleweedcr1901_modern: mount /srv/gw1/video -o remount,async22:17
tumbleweedCarlFK: I mean22:17
xfxfCarlFK: http://www.9news.com.au/national/2017/01/17/10/57/melbourne-woman-ripped-to-pieces-by-kangaroo-during-morning-run22:46
tpbTitle: Melbourne body-builder ‘ripped to pieces’ by kangaroo during morning run - 9news.com.au (at www.9news.com.au)22:46
*** tvCommitBot has joined #timvideos22:48
-tvCommitBot- [lca2017-av] stefanor pushed 2 new commits to master: https://git.io/vMPqf22:48
-tvCommitBot- lca2017-av/master 14ee932 Stefano Rivera: VA-API by default (except rX, which has no intel GPU22:48
-tvCommitBot- lca2017-av/master 57c8783 Stefano Rivera: Add big enc6 box22:48
*** tvCommitBot has left #timvideos22:48
mithro_florent_: I'm more confident that I'm getting the bios from flash23:09
jeathaytan: we are trying out the vaapi streaming now. sometimes we get this message from youtube:23:20
jeaPlease use a keyframe frequency of four seconds or less. Currently, keyframes are not being sent often enough, which will cause buffering. The current keyframe frequency is 14.9 seconds. Note that ingestion errors can cause incorrect GOP (group of pictures) sizes.23:20
jeait is still working fine from what I can see, but it sounds like something to look into23:21
mithro_florent_: But I've broken the DDR23:23
micolousjea: are you getting ringing on the audio in wlg2?23:51
micolousjea: re: keyframes, I've seen that too before, I ignored it23:52
*** tvCommitBot has joined #timvideos23:54
-tvCommitBot- [lca2017-av] stefanor pushed 1 new commit to master: https://git.io/vMPcg23:54
-tvCommitBot- lca2017-av/master 60e948c Stefano Rivera: rXgrab{2,3} are now enc boxes23:54
*** tvCommitBot has left #timvideos23:54

Generated by irclog2html.py 2.13.1 by Marius Gedminas - find it at mg.pov.lt!