#qi-hardware IRC log for Friday, 2011-04-29

qi-bot[commit] Werner Almesberger: ubb-vga2: new option -m to select the display mode (resolution, timing) http://qi-hw.com/p/ben-blinkenlights/71c9c7d01:14
zhichengHi guys,I'm a software engineer,I'm very interesting hardware development and MIPS Architecture.04:51
xiangfuzhicheng: great.04:56
xiangfuzhicheng: you want learn hardware design? or you want write some software(take part in u-boot kernel develop)? on MIPS arch?04:58
xiangfuzhicheng: do you have one nanonote? are you in China? (since your name is like Chinese name :)04:59
zhichengI'm an iOS App Developer :-)05:01
zhichengbut I'm very interesting hardward design.05:01
zhichengand I'm check out some MIPS architecture design,It's really really cool!05:01
zhichengI want buy an SGI Workstation,But It's not available anymore.and very expensive.05:02
xiangfuzhicheng: http://projects.qi-hardware.com/index.php/p/board-qi-avt2/source/tree/master/05:02
zhichengxiangfu: Wow, You work at Qi Hardware.05:03
xiangfuis the 'really really cool' design open? :)05:04
zhichengI mean,the MIPS arch.05:05
zhichengand open source ,too.05:05
xiangfuzhicheng: ok. I though it's a device.05:05
xiangfuzhicheng: compare to SGI Workstation, ben nanonote is cheaper :D05:06
xiangfuand also MIPS05:06
zhichengxiangfu: yeah!05:06
zhichengxiangfu: do you live in Beijing?05:06
qi-bot[commit] Xiangfu Liu: fmit fix typo http://qi-hw.com/p/openwrt-packages/1ca9a6007:58
qi-bot[commit] Xiangfu Liu: nanonote-files: config.full_system, add 4th, using upstream triggerhappy http://qi-hw.com/p/openwrt-packages/7df189007:58
qi-bot[commit] Xiangfu Liu: new package: move kmod-ks7010 from openwrt to here http://qi-hw.com/p/openwrt-packages/159a67707:58
qi-bot[commit] Xiangfu Liu: ascii-paint: using nanoote branch http://qi-hw.com/p/openwrt-packages/d89519907:58
qi-bot[commit] Mirko Vogt: add <http://downloads.qi-hardware.com/software/mirror-openwrt-sources/> http://qi-hw.com/p/openwrt-xburst/d3d6c3c09:01
qi-bot[commit] Xiangfu Liu: gitignore vim temporary files (*~) http://qi-hw.com/p/openwrt-xburst/996641d09:01
qi-bot[commit] Xiangfu Liu: optimize for ben nanonote http://qi-hw.com/p/openwrt-xburst/00b686509:01
qi-bot[commit] Xiangfu Liu: [xburst] Improve mounttime http://qi-hw.com/p/openwrt-xburst/4006bc609:01
Action: xiangfu commit very little work on rebase trunk09:01
qi-bot[commit] Xiangfu Liu: nanonote optimize http://qi-hw.com/p/openwrt-xburst/945452809:15
qi-bot[commit] Xiangfu Liu:  Add-gfortran-compiler-support-to-the-toolchain http://qi-hw.com/p/openwrt-xburst/02a7dc809:15
qi-bot[commit] Werner Almesberger: ubb-vga2: non-contiguous allocation of frame buffer memory http://qi-hw.com/p/ben-blinkenlights/623f3cc17:08
qi-bot[commit] Werner Almesberger: ubb-vga2.c (line, frame): start line timer outside the "line" function http://qi-hw.com/p/ben-blinkenlights/a63579e17:08
qi-bot[commit] Werner Almesberger: ubb-vga2.c (main): fixed check for unknown resolution and call it "mode" http://qi-hw.com/p/ben-blinkenlights/f608e5617:08
qi-bot[commit] Werner Almesberger: tstimg.c (grill): avoid fencepost errors (pixel at xres/yres) http://qi-hw.com/p/ben-blinkenlights/2cee70217:08
qi-bot[commit] Werner Almesberger: physmem.c: align memory to word and page size http://qi-hw.com/p/ben-blinkenlights/debe9cc17:08
qi-bot[commit] Werner Almesberger: renamed ubb-vga.c to ubb-vga-old.c, ubb-vga2.c to ubb-vga.c; updated Makefile http://qi-hw.com/p/ben-blinkenlights/3a8e06217:08
dvdkwpwrak: just saw your physmem.c commit.  how does this give you physical memory?  doesn't it just give you logical pages without any knowledge of phys. addresses?17:49
kyakdvdk: latest mplayer works very nice! it consumes only around 50 % CPU, soo good :)17:53
dvdkkyak: cool isn't it :)17:54
dvdkkyak: with proper 44100khz sample rate, audio quality is also very nice.17:54
dvdkkyak: 50% for which video material?17:55
kyakdvdk: yeah, audio is good! but i'm running into issues with a/v desync17:55
dvdkkyak: encoding problem?17:55
dvdkkyak: ffmpeg2theora has some problems there17:55
kyakdvdk: trying to figure out.. my ffmpeg2theora is pretty old17:55
dvdkkyak: depending on input video material.  latest ffmpeg2theora does sync correction by default.17:55
kyaki'm thinking to build ffmpeg2thoera and libtheora from source17:55
dvdkhowever, fails on mpeg transport streams for me (i.e. raw dvb-t recording)17:55
dvdkkyak: did the same.  try to use libtheora 1.217:56
kyakStream #0.0: Video: mpeg4, yuv420p, 624x352 [PAR 1:1 DAR 39:22], 23.98 tbr, 23.98 tbn, 23.98 tbc Stream #0.1: Audio: mp3, 48000 Hz, 2 channels, s16, 128 kb/s17:56
kyakthat's the input video17:56
dvdkwhat.s the container? mpeg?17:56
kyak(latest south park episode, btw)17:56
dvdkkyak: avi shouldn't have any problems with desync, since avi comes without timestamps.17:56
dvdkhmm.  if you do mplayer -noquiet and look at the output (i.e. ssh) does it show a a-v desync?  at 50% cpu it should be around0.17:57
kyakdvdk: oh.. when i'm running this output ogv file on my PC (totem), there's no desync17:57
dvdkkyak: ok, did you test with -demuxer ogg on you pc, too?17:57
dvdkkyak: also check cpu load.  some scenes need more cpu power, causing desync.17:57
kyakdvdk: may i upload the output ogv file to you? it's aroudn 34 Mb17:58
dvdkkyak: yes, no problem.  how do we do the uploading?17:58
kyakdvdk: here :)17:59
dvdkok, downloading at 120k18:00
kyakdvdk: desync starts at around 1 minute, i'm just running mplayer ..ogv without arguments, and i don't use scrolling18:02
dvdkkyak: btw OSS audio was pretty weird, are you already running the version that does alsa exclusively?18:02
dvdkok, uploaded, testing now18:02
kyakversion "3 in period" :)18:02
kyakdvdk: hm, when i start scrolling, the sync seems to be restored18:03
dvdkok, desync is displayed in mplayer output (-noquiet): A-V: -0.37418:05
dvdkso maybe cpu problem18:05
dvdklets check18:05
dvdkno, cpu load is at 40 %18:06
kyakdvdk: when i run mplayer -demuxer ogg, it get's desynced on PC, too!18:06
dvdkmaybe the alsa patch is the culprit (changing buffer/fragment size, but was neccessary for audio to be continuous)18:06
rjeffrieswpwrak awaiting 3D in millions of color as video out on Ben NN. "Dream no little dreams."18:06
dvdktrying with -autosync 3018:07
kyakdvdk: strangely, it only happens in the beginning of the video (around first minute)18:07
dvdkmaybe a ogg framing pecularity18:08
dvdk-autosync doesn't help :(18:08
kyakit's pretty hard to determine if there is a desync on your bbv video, cause there is no speech.. But i'll try encoding with latest ffmpeg2theora and see if this goes away18:10
wpwrakdvdk: (physmem) did you see the comment saying that it's still a dummy ? :-)18:10
dvdkkyak: maybe an ogg paging problem causes that.  mplayer can buffer audio, but it cannot really buffer video.  so there are some rules about video vs. audio interleaving, maybe it fails there.18:11
wpwrakdvdk: btw, here's a nicer mechanism for mapping virt->phys: http://www.mjmwired.net/kernel/Documentation/vm/pagemap.txt18:11
dvdkwprwrak: absolutely cool.  didn't know that.18:11
dvdkwpwrak: should use that to clean up jz47xx_vid18:12
wpwrakyeah. those "search memory" heuristics are all a little evil. even though they work ...18:12
dvdkkyak: i did some encodes with ffmpeg2theora (50 minute video, 25 fps) and didn't notice any desyncs.  so maybe really ogg paging problem.  trying oggz-sort may help.18:12
dvdkoggz-sort seems to not help here :(18:14
dvdkkyak: waht to have a look at the file i tested with?  http://mosquito.dyndns.tv/david/nanonote/1nn.ogv18:16
dvdk(tell me if you download don't want to leave such stuff online for long)18:16
dvdkwpwrak: just a little slow :)18:16
dvdk(the searching)18:16
wpwrakdvdk: there it's an advantage to have so little memory ;-))18:17
kyakdvdk: downloading now18:17
kyakdvdk: "trying oggz-sort may help" - sorry, i don't understand18:17
dvdkkyak: oggz-tools contains a tool oggz-sort which does some simple ogg remuxing, sorting pages in the stream.18:18
kyakdvdk: is it also better to build oggz-tools from source?18:18
dvdkkyak: i guess your problem is that the scene at 0:01:00 has almost no motion, so you get many video frames in one ogg page.  guess mplayer has a problem with that, since it sees the video frames too late18:18
dvdkkyak: don't think so, pretty old and reliable stuff, just use ubuntu or whatever packages18:19
dvdkkyak: but oggz-sort doesn't help anyway18:19
kyakheh :)18:19
dvdka new ffmpeg2theora *might* do the job.  or we'd have to fix the mplayer demuxer (yuck)18:19
kyakok.. i'm about to try to build ffmpeg2theora from source18:20
dvdkkyak: how's the download?18:20
kyak13% done18:20
dvdktell me when it's done, taking it offline then.  how many people are leaching here ? :)18:21
kyaksure, i'll tell you :)18:21
kyakbtw, there is a binary of ffmpeg2theora18:21
kyaki think i'll start with that.. guess libtheora is linked in statically18:22
kyakok, encoding with the latest ffmpeg2theora now18:24
kyakdvdk: seems i will finished encoding faster than downloading from you :)18:28
dvdkkyak: you're getting almost all my bandwidth :r18:28
kyakdvdk: your internet connection doesn't look very good :)18:30
kyakdvdk: nope, it's still desynced at around 1:00 -\ I'm glad it is reproduced on PC18:33
kyakdvdk: 53% :)18:34
kyakdvdk: "A-V:" reaches up to 5 seconds on PC18:37
kyakdvdk: i wonder why scrolling makes it go away? and why won't mplayer autocompensate?18:38
dvdkkyak: i guess demuxer problem.18:39
dvdksorry have to go now.  maybe again tomorrow.18:39
dvdk(leaving download up, until apache bandwidth goes down)18:39
kyaketa is 6 minutes18:39
wpwrakgrmbl. the bloody dma doesn't want to start. it seems that it did for the first 2-3 tries, but now nothing :-(19:09
kyakmirko: ping19:09
kyakmirko: when you have some time, could you have a look at https://dev.openwrt.org/ticket/9332, seems there is some issue with the latest qt4 changeset19:13
Action: kyak gone19:15
whitequarkwpwrak: hey21:26
whitequarkremember I was talking about crappy bluetooth headset?21:27
whitequarkI've played 4000 Hz through it. in normal mode, the recorded audio has peak at 400421:27
whitequarkand in fucked up mode the peak is at 4047 Hz21:27
whitequark~1% is _very_ detectible by ear.21:27
whitequarkah yes, the worst part is that it's sliding21:28
wpwrakwhitequark: (1%) interesting. have you also tried with another headset ?22:04
whitequarkwpwrak: I only have one22:08
whitequarkI'll soon write something that would be a blog post if only I had a blog22:08
whitequarkwith images!22:08
whitequarkwpwrak: ah no, I was wrong22:11
whitequarkthat's not 1%22:11
whitequarkit is sliding fast from one frequency to other22:12
wpwrakah, that's why you hear it22:12
whitequarkthat's _very_ weird and unpleasant sense22:17
whitequarkit somehow makes you think about smashing headsets with a hammer :)22:18
wpwrakwhitequark: have you used the headset with other sound sources ? maybe it's not the headset but the source22:37
whitequarkwpwrak: what can possibly go wrong with just a simple and stupid bluetooth dongle?22:40
wpwrakwhat can possibly go wrong with just a simple and stupid bluetooth headset ? :-)22:41
whitequarkwpwrak: check my article: <p>For the curious ones, here is the raw data (in22:43
whitequark<a href="http://audacity.sf.net">Audacity</a> format): <a href="raw.tbz2">raw.tbz2</a>22:43
whitequark(33M compressed,around 110M unpacked)</p>22:43
whitequarkwrong paste :/22:43
whitequarkhere's the right one: http://whitequark.org/bluecrap/22:43
wpwrakwhitequark: well, your sample size is one. a bit small for such sweeping conclusions ;-)23:29
Action: wpwrak curses the voodooness of DMA23:54
kristianpaulwpwrak: still getting ramdom DMA response?23:55
wpwrakthe response if quite predictable now: i simply doesn't even start23:55
wpwrakthe odd thing is that the MMC controller sometimes throws a timeout, and that depends on the weirdest circumstances23:56
kristianpaulmay be try to check if you are  a SD-lile compliance device ;-)23:57
wpwrakoh, it's much more subtle than this: if i disable DMA in the MMC controller and don't even start the DMA controller, and the right set of voodoo is applied, i get the timeout23:59
--- Sat Apr 30 201100:00

Generated by irclog2html.py 2.9.2 by Marius Gedminas - find it at mg.pov.lt!