#qi-hardware IRC log for Sunday, 2011-04-24

kristianpaulha, intel + honeycomb,  later what..00:28
wpwrakah, nice. only timer 0 and 1 seem to be used. seems that i can do a bit better than timing cache fills.00:28
wpwrakgood. now i know PCLK: it's 336 MHz/301:09
kristianpaulah, rtk hmm https://secure.wikimedia.org/wikipedia/en/wiki/Real_Time_Kinematic01:11
wpwrakinteresting. the time per GPIO change (set or clear) is 8.5 PCLK cyles.01:11
kristianpaulinteresting topic for the centimetre-level accuracy lovers ;)01:12
kristianpaulha ! http://unreasonablerocket.blogspot.com/01:15
kristianpaulhttp://www.rakon.com/Products/Public%20Documents/Specifications/GRM8652%20Hardware%20applications%20V1%200.pdf tiny GNSS front end01:27
kristianpaulbah DNA..01:27
wpwraklarsc: hmm, why do we use 32 bpp and not 16 bpp in the frame buffer ? 16 bpp wouldn't tax our meager memory bandwidth quite so badly03:10
rjeffrieswpwrak so you can display something over SPI-->UBB..>[vga connector???]04:01
wpwrakblargh. it's no fun to make a video. so photos will have to do ...04:02
rjeffriesthis is a VERY fun article about the 6502 design, and especially layout04:02
rjeffriesthey got it right the FIRST time. amazing04:03
wolfspraulI'm quite impressed by the Milkymist article on Wikipedia :-)04:05
wolfspraulit could go into much more depth in many areas, but it's a great start I think, should be safe from deletion for irrelevance or quality reasons, I would hope04:05
rjeffriesI am so old I remember when wolfspraul dismissed the idea of writing up Ben NN on wikipedia. ;)04:05
rjeffrieswe all learn and grow and change04:06
rjeffrieswpwrak I'll look for your screen photos tomorrow.04:25
rjeffrieswolfspraul agree the Milkymist article is Most Excellent. congrats!04:26
wpwrakrjeffries: you may want to take a coffee. they're just a moment away ...04:28
rjeffriesok then04:29
rjeffriesI am wondering what the FPGA chip in MM costs in the 100 qty sharism is buying04:30
qi-bot[commit] Werner Almesberger: video/video.c: pseudo-VGA output (test pattern only) http://qi-hw.com/p/ben-blinkenlights/b85ae1504:33
qi-bot[commit] Werner Almesberger: video.c: working version, with screen grabber http://qi-hw.com/p/ben-blinkenlights/18f295704:33
wolfspraulca. 39 USD, all our prices are public as well04:33
rjeffrieswolfspraul is PCB 4 layer or more, and roughly what does it cost again in these modest quatities04:47
wolfspraulUSD 24.8704:54
wolfspraul6-layer PCB04:54
rjeffriesok the reason I asked was I did not see in on RC3 bom, thx04:54
wolfspraulbecause rc3 pcb has not been made yet04:55
rjeffriesin any case from my POV at current qty 1 price your margin is pretty thin04:55
wolfspraulyou and your margins :-)04:55
wolfspraulIf there is one thing I'm not worried about, seriously, it's competition from Ron Manufacturing Enterprises, Inc.04:55
rjeffriesbuxiness guys are such a total pain in the you know what04:56
rjeffriesyes you have NO FEAR in terms of me04:56
wolfspraullooking at bom items for these small runs tells you very little. best is to try out with some runs, then you learn a lot.04:57
rjeffriesyes prices drop a lot in volume04:57
wolfspraulkeep in mind that in the end the raw materials for all this are still dirt cheap04:57
wolfspraulmostly what you pay for is people's time, knowledge, experience04:57
wolfspraulalso yield, very important. don't forget. who owns the yield?04:58
wolfspraulif you ask that, I know you are serious :-)04:58
wolfspraulpeople who ask about bom first, ...04:58
wolfspraulbut naturally 95% or more ask about bom first04:58
wolfspraulI did the same a few years back.04:58
rjeffriesbom is all I have to look at. I was only curious as to the FPGA I had no idea the price range04:58
wolfspraulgotta start somewhere04:58
rjeffriesnot as much as I thought04:59
wolfspraulproprietary IC prices are largely impacted by strength of IP, target market/customer, etc.04:59
wolfspraulsome people pay 10 USD, some 500 USD04:59
wolfspraulit all depends04:59
wolfspraulthe IC itself, if you take out all investments (with you cannot, of course), is always just a few pennies05:00
wolfspraulso xilinx 'makes' 38.50 USD on this05:01
wolfspraulin a super simplistic calculation05:01
wolfspraulprobably with high-end process nodes like 45nm in this case the contracts with Samsung also have a lot of back-and-forth payments when reaching this or that target05:02
wolfspraulthey work on 28nm now, for the -7 generation05:02
rjeffriesby the way in higher volumes yiedl belongs to the assembly house they make educated guess the first time the see a design, iuf the yield is crappy next time you pay more05:02
rjeffriesin fact it is a partnership bewteen desgn engineering and the manufactuter05:03
rjeffriesall that matters to you is it costs you $39 and change05:04
rjeffries(for thsi chip in this volume at this time)05:04
wpwrakno comments on my vga ? sneef.05:14
rjeffrieshang on05:16
rjeffrieswhere is the link wpwrak?05:17
rjeffriesoh now I see I missed it05:17
rjeffrieslike an elephant that dances. amazing period05:20
rjeffriesI will admit I would not choose to read War and Peace on that display05:21
wpwraki could be improved a little by using two more signals for additional colors. right now, it's 3 bit color :)05:22
wpwrakyeah, the horizontal resolution is faux-320. since the gpios are so slow, i only have one update cycle per pixel. in the ingenic cpus, you can't write a value directly to a port but you have to either set or clear. so i alternate set and clear commands. this means that the worst-case resolution is only 160 pixels.05:24
wpwrakand even there i had to cheat. the timing isn't quite vga. my scan lines are about 10% longer. luckily, most monitors tolerate this.05:25
rjeffriesfascinating experiment. now I am off to bed. cu05:29
wpwrakah no, only one signal left for grey levels05:32
qi-bot[commit] Werner Almesberger: Added schematic symbol for VGA DE15 connector http://qi-hw.com/p/kicad-libs/f35d48c05:40
qi-bot[commit] Werner Almesberger: video/: added Makefile and schematics http://qi-hw.com/p/ben-blinkenlights/7d95fe806:01
qi-bot[commit] Werner Almesberger: video/README: documented sources and implementation quirks http://qi-hw.com/p/ben-blinkenlights/b11dc6106:01
qi-bot[commit] Werner Almesberger: renamed "video" to ubb-vga http://qi-hw.com/p/ben-blinkenlights/3b7a7ca06:01
wolfspraulwpwrak: I see a lot of grey stripes in the video. is that on purpose or an artefact of the theora conversion? can you upload the .avi original from your camera?06:24
wpwrakwolfspraul: yeah, tat's a feature of kdenlive :-(06:34
wpwrakthe original video has copyrighted background music (radio). i'll put it at a safe place ...06:38
wpwrak(copyrighte) well, non-open06:38
wolfspraulupload it, I convert to ogg without sound, and delete the original06:41
wolfspraulhow about that?06:41
wpwrakyeah, had something similar in mind :) just via my server. the bottleneck is the upload anyway.06:45
wolfspraulnice, thanks07:34
wolfspraullooks pretty good, with the flickering light at the beginning it feels like a horror movie07:34
kyakviric: yeah, i was running mplayer in sdl. It runs OK, when decoding mpeg2 video with libmad audio (pre-encoded for 320x resolution)07:45
kyakviric: no, not mpeg2.. i encoded it with vcodec=mpeg1video:vpass=1:vbitrate=20007:46
dvdkanybody objects if i commit the new mplayer package to git already?08:47
dvdkwe'd have to do some patching to remove patented codecs, before releasing any firmware containing it.08:47
kyaki think the patented codecs should be rather disabled based on patented flag, but not removed08:53
dvdkkyak: of course.  just the current package doesn't do that yet.  so we won't be able to release without causing (patent) trouble.08:56
dvdki'll commit anyway.  let's work out the detail in time before release.08:56
dvdkconfiguring the ffmpeg that's build as part of mplayer looks a little hacky, btw08:57
kyaksure, just commit it08:59
kyakdo you mean the "current" package that you are going to commit?08:59
kyakcause the one already in git makes use of patented flag08:59
dvdkkyak: yeah 'current' is my working copy09:04
kyakyou can keep the parts in makefile where it is checking for patented flag.. or are you talking about ffmpeg mostly?09:05
dvdkkyak: the makefile is mostly untouched.  problem is the ffmpeg that's build now build by mplayer09:06
dvdkI'll commit, then have a look at whether i find a simple patch for the ffmpeg part.09:06
qi-bot[commit] David K├╝hling: mplayer: use svn versions of mplayer plus mplayer-internal ffmpeg from git http://qi-hw.com/p/openwrt-packages/6a975d209:07
dvdkhere you are.09:07
kyakperhaps we can base on the makefile of ffmpeg from openwrt and make a controllable build of ffmpeg of our own inside the mplayer build dir09:07
dvdkkyak: this is deeply intermingled, not sure that is easy.  mplayer doesn't even run ffmpeg's configure script.09:08
kyakgod... are there some configure options of mplayer that can be passed to ffmpeg during build?09:09
dvdkkyak: easiest patch would be --disable-ffmpeg ? :)09:09
dvdkkyak: mplayer configure writes ffmpeg's config.h?09:09
dvdkahh, found:09:09
dvdk--disable-decoder=DECODER disable specified FFmpeg decoder09:09
dvdk... and so on.09:09
kyakoh yes09:09
dvdk(from mplayer configure)09:09
kyakthis looks like what we need09:10
dvdk--disable-muxer : maybe we can disable the broken ffmpeg ogg demuxer?09:10
kyakdvdk: i think it's for you to decide, you are the mplayer expert :)09:10
dvdkkyak: only unwillingly )09:11
rohinterresting how much energy is put into delivering crippled sw09:11
rohjust my 2 cents (out of the perspective to have seen this all before) .. the support work would be much more bearable if you dont deliver mplayer at all instead of one which can only play half the stuff09:12
dvdkroh: we can only play quarter of the stuff anyway, due to nanonote's weak cpu :)09:13
dvdkroh: no problem to re-encode to your favourite format?09:13
rohdvdk: bwahahaha... its not 1996 anymore dude09:13
kyakin fact, we can deliver mplayer with only ONE stuff.. due to necessity to decode to Ben's screen and capabilities09:13
dvdkroh: have to do it to not get wolfgang into any trouble on nanonote shipping (openmoko phones had these problems already)09:13
dvdki.e. patent ligitation09:14
rohbesides.. in reality there is neither the screen nor the audio out good enough for 'video' or 'music' . its ok for scummvm-sound tho09:14
rohdvdk: i know. i was working at openmoko09:14
dvdkroh: it has a stereo out.  good enough isn't it?  as good as my desktop-pc09:14
rohthats why i said.. dont ship broken sw. better not ship it at all.09:14
kyakroh: in reality i watch tv shows on Ben connecting it to car's audio system, when the traffic is super hard09:14
dvdkscreen size is debatable, but werner is working on a solution already :)09:15
dvdki think an mplayer alternative would be to use a minimal player besed on liboggplay which is the demuxing/seeking framework that firefox uses.09:15
dvdkbut still more wok.09:15
kyakbesides, i don't know why roh keeps insisting that mplayer is "broken" :)09:15
dvdkkyak: look at the mplayer sourcecode, then you'll understand :)09:16
rohkyak: ever looked at the code?09:16
dvdktwo dudes, one thought.09:16
rohtoday somebody mentioned that there is a fork of mplayer now.. mplayer2. no mencoder anymore and they tried cleaning up the source.09:16
rohsounded like a project for somebody loving pain09:16
dvdkroh: saw that already. looks extremely nice.  but first things first :)09:16
kyakwhy would i look into the code? it works, and nobody has offered an alternative09:17
dvdkroh: we're all loving pain, aren't we, spending our free time in front of the pc _working_ :)09:17
rohi use mplayer code a lot. as example how real-world-fileformats look like and what fields are where. and how not to write your code if you need it secure or maintainable... nice simd in ffmpeg still...09:17
dvdkkyak: it only works sometimes (unfortunately).  but still mplayer is extremely optimized and minimalistic (no threading etc., data flow optimized for cahces), so matches nanonote quite well09:18
rohdvdk: better than spending it sitting in an office being paid for killing time NOT working.09:18
dvdkroh: cool, do you currently hire :)09:18
kyakdvdk: that's what i'm saying; the source code might be crappy, but it works better than anything else (for Ben?)09:19
rohdvdk: hehe.. no. but i know lots of people who have such jobs. and its not good for health.09:19
dvdkkyak: looks like configure enables most ffmpeg codecs by default, maybe we should rip that out and only manually enable codecs.  that keeps patents creeping in once we upgrade to newer svn revisions09:19
rohthere was a 'beware of patents' switch somewhere09:20
viricdvdk: hey, let me try your advice :)09:20
dvdkroh: not any more09:20
dvdkmorning viric09:20
viricguten tag09:20
dvdkroh: you know, mplayer/ffmpeg people already gave up on any unproductive stuff to kill their time.09:20
kyakdvdk: the whitelist approach sounds reasonable09:20
rohyou mean the flag is gone?09:20
rohnice. finally some realism *g*09:21
dvdkroh: looks like it.  mplayer release tarballs: also gone, too much work, doing release management :)09:21
rohi can understand.. who should maintain and test a code path nobody seriously uses?09:21
dvdkviric: try a few times, kyak also had a black screen on first tries.  be patient!09:21
viricI was trying to find that 'screenw' 'screenh' line you sent... I think you sent it already, but I could not find it in my mail box09:21
dvdkroh: if nobody uses it?  don't maintain it! :)  already fbdev:vidix hit that fate, just not working.09:22
kyakviric: at that time, it helped me to wait around 1 minute :)09:22
dvdkogg/ogm demuxer: similar fate.09:22
viricdvdk: well, the video *starts in black*... that makes it also a bad test :) specially when you get the floating point codecs09:22
dvdkviric: hehe.  yes nanonote put -ac tremor into the config file i guess.09:22
rohman i love this li-lo makita .... doing renovations in our new hackerspace.. and somebody brought a battery powered hammer-drill09:22
viricdvdk: and I could not see any cpu usage difference between theora and fftheora09:22
dvdkviric: maybe depending on content.  it should be.  ffmpeg renders directly into the dma buffers, whil theora has an additinoal copy.  plus many other differences (cache usage etc.)09:23
viricMPlayer interrupted by signal 4 in module: decode_audio09:23
viric- MPlayer crashed by an 'Illegal Instruction'.09:23
dvdkviric: huh?09:24
dvdkviric: tried -demuxer ogg?09:24
viricincredible. Playing it again, it works.09:24
viricI've the scaler working :) great09:24
dvdkcool to hear that.09:24
viricouch. it stopped09:25
viricTrace/breakpoint trap09:25
dvdkviric: sounds like some corruption occuring.09:25
dvdkviric: fftheora vs. libtheora does it change anything?09:25
rohi fear i cannot give back that machine *g*09:25
dvdkviric: maybe try a different mplayer version?09:25
dvdkwait, uploading the version i used here,09:26
viricI'm using mplayer from yesterday09:26
dvdkyou mean svn checkout?09:27
dvdkor the one you compiled yesterday ? :)09:27
viricI built it yesterday, from the snapshot of yesterday! :)09:27
dvdkthis is what i used:  http://mosquito.dyndns.tv/david/nanonote/MPlayer-r33304.tar.gz09:27
dvdkincludes ffmpeg09:27
viricMPlayer SVN-r33269-snapshot-4.5.1 (C) 2000-2011 MPlayer Team09:27
viricthis is what mine says.09:28
viricaccording to the svn revision, mine is slightly older :)09:28
dvdkmine is r33304 with some git version from yesterday09:28
viricrerunning, again illegal instruction... grmbl09:28
viric(in decode_audio!)09:28
viricI'm trying fftheora.09:28
viricI don't know what it gets by default.09:28
virictrace/breakpoint trap again. In a different moment. Let's see 'theora'09:28
dvdkthere is another tremor audio codec included with mplayer (not ffmpeg?), maybe this is broken and you're using it?09:29
viricbang. illegal instruction in decode_audio too. grrr09:29
viricI'm using "-ac tremor"09:29
dvdkyeah, there are two tremors, one is even mips-optimized?09:29
dvdkneed to compare our ./configure output.09:30
viric# mplayer -ac help | grep trem09:30
virictremor      tremor    working   OggVorbis audio  [tremor]09:30
dvdkviric: try -noaudio09:30
viricI'll use '-ao null'09:30
dvdkyou may not be able to discern the two09:30
dvdkok, same output for me.09:31
viricUsing '-ao null', it decodes super-slow!09:31
dvdktry -noaudio?09:31
viricIt does not know '-noaudio'09:31
viricI'll run with gb09:32
viricOOMK. grmbl!09:33
viricmaybe with gdbserver09:33
viricI'll build a gdb for the mips..09:34
dvdkviric: does it not crash with sdl?  that it's the accel's fault.09:34
virichm I use -vo fbdev when it works normally09:34
dvdkok, so problem solved? accel is killing it?09:35
viriclet me try. -vo fbdev209:35
viricMaybe I never played the video that long.09:35
virichere it plays....09:36
viricno crash.09:36
viricno crash at all without your driver.09:37
viricI may enable core dumps...09:37
dvdkmaybe it maps the wrong memory for dma buffers?09:37
viricNo idea09:38
dvdkok, we may be able to see that by looking at /proc/<pid>/maps09:38
viricsome pictures appear...09:38
dvdkjust pause mplayer before it crashes and cat that file to stdout.09:38
viricit looks like some race condition, according to the time of crash (always different)09:38
viricmplayer catches Illegal instruction and does not emit any core dump. grmlb09:39
viricI've the maps. so?09:40
viricq http://sprunge.us/ChZE09:40
dvdkcan you upload to look at them? or mail?09:40
viricit's the last part09:40
dvdkviric: need the full file :)09:41
viric http://sprunge.us/POIR09:41
viricfull file.09:41
dvdkviric: do you have swap at nanonixos?09:42
dvdkok, same here.09:42
viricI've lots of memory used, I don't know why09:43
viricxiangfu: ni hao!09:43
xiangfuviric: hi09:43
viricxiangfu: I bought a chinese product long ago, and I found it recently. I don't know what it is. May you take a look and tell me? :)09:43
viricNo latin script in it09:43
xiangfuok take a picture of it :)09:44
dvdkviric: memory mapping looks ok, afaics, so maybe it's not my fault?09:44
xiangfutake a picture of the Chinese :)09:44
dvdkviric: one thing you still might try:09:44
dvdkviric: uncommenting these two lines in jz47xx_config_playback(), just to make sure we really don't have spurious dma operations:09:45
dvdk    unmap_unused_phys();09:45
dvdk    sdram_close();09:45
viricxiangfu: http://vicerveza.homeunix.net/~viric/tmp/prod.jpg09:46
viricdvdk: hm09:47
viricdvdk: can you post a '.patch' somewhere? :)09:47
viricor I'll have to write a 'sed' operation that does that.09:47
dvdkhmm, needs to be applied from inside the directory, maybe need to change that.  wait...09:48
viricxiangfu: I even don't know the orientation of the picture :D09:48
dvdkviric: http://mosquito.dyndns.tv/david/nanonote/jz47xx-2.patch09:49
dvdkdamn it. still wrong.  svn cannot produce quilt-conforming patches?09:49
viricno, it cannot09:49
dvdkok, 3rd try.09:49
viricthe 2nd went fine.09:50
viricapplying patch /nix/store/z9i6n580lgky42dbgwyqzczvnyhzqw2r-jz47xx-2.patch09:50
viricpatching file jz47xx_vid.c09:50
xiangfuviric: it's one kind 'Seasoning',09:50
dvdkah, nice.09:50
viricxiangfu: I sparse that over the food, and it's fine? :)09:50
xiangfuviric: it have to used during the cooking. or before you start to cooking09:52
viricdvdk: same trouble. exactly.09:52
viricxiangfu: ahh perfect.09:52
dvdkviric: i hate it.09:52
viricxiangfu: does it mention a 'rotting date'? :)09:53
dvdkviric: but maybe not my fault.  just a race due to different timing (i really hope that).09:53
dvdkviric: mayb try the same source version i'm using here (see the -r33304.tar.gz)09:53
viricdvdk: you don't see the trouble at all?09:53
viricoh, got core dumped.09:53
dvdkviric: works rock-solid here.  i admit it's pretty hacky, but dma buffers are locked in memory, shouldn't overwrite your code.09:54
viricProgram terminated with signal 5, Trace/breakpoint trap.09:54
viric#0  0x7fe076b8 in ?? ()09:54
dvdkah, wait, just looking at your maps.  where does 0x7fe076b8 reside?09:55
viricHow can I know the maps in a core dump?09:55
dvdkdunno.  but if you run in gdb, the program won't quit and you can observe /proc/<pid> after the crash.09:55
viricI can't run gdb and mplayer at once09:55
dvdknote that  0x7fe076b8 is not mapped in your example.09:56
xiangfuviric: 24 months,  "check the surface for Production date"09:56
viricxiangfu: it's rotten for sure then. :)09:56
dvdkviric: i have an idea.  your stack is much smaller than mine, did you notice?09:56
viricdvdk: it looks like a stack address, looking at the old maps09:56
viricstack size              (kbytes, -s) 819209:57
viricI don't know how big is your stack :)09:57
viricInreasing to 16MB09:58
dvdkviric: but anyway, my code onlly uses about 16kB of stack and only once durin initialization (for d-cache flushing :)09:58
viricwhy sometimes it does not 'scale'?09:59
dvdkviric: huh?09:59
viricmplayer -vo cvidix -ac tremor -screenw 320 -screenh 24009:59
dvdkviric: you have to give -fs to make it scale and -screenw -screenh to tell it your screen dimension09:59
viricwhat I am forgetting?09:59
dvdkor press 'f' during playback?  wouldn't try that, though :)09:59
viricMPlayer interrupted by signal 11 in module: video_read_frame09:59
viricit looks like 'random crashes'09:59
dvdkviric: uname -a?10:00
dvdkLinux BenNanoNote #1 PREEMPT Thu Feb 24 12:46:39 CET 2011 mips GNU/Linux10:00
viricLinux nano 2.6.36 #1 PREEMPT Thu Mar 10 12:41:34 UTC 2011 mips GNU/Linux10:00
dvdkviric: maybe here's the problem?10:00
viricdvdk: If I take out the '-fs', it works just fine. but no scaling, ofc.10:00
dvdkviric: you mean it won't crash?10:01
viricit does not crash.10:01
dvdkviric: either timing, or in full-screen it writes beyond the bottom edge of the screen10:01
dvdkhow does it look? where is the video placed?10:01
dvdkyou could try -geometry 320x240+0+0 or something instead of -fs to see when it stops working10:02
viric-geometry 320x240+0+0 hangs10:03
viricwell crashes, like with -fs10:03
dvdktry 320x120 etc.10:03
viric300x200 crashes too10:03
dvdkdoes it output some frames before crashing?10:04
viricyes for soem seconds10:04
viric256x144@0x0 works just fine10:04
dvdkviric: this would disable the scaler, try 256x10010:04
dvdkviric: maybe you just have a CPU with bad silicon ? :)10:05
dvdkmaybe dma engine messess with other cache/ram accesess? i mean there is no pattern in the crashes.10:06
dvdkenabling the scaler increases ram bandwidth, i guess (has to read more pixels for interpolation)10:06
viricthe kernel does not crash, at least.10:07
dvdkviric: it's surprisingly hard to crash :)  not even writing beyond the screen crashes it immedeately10:07
dvdkviric: should try with MPlayer-r33304.tar.gz, it could still be an mplayer bug due to different timing10:08
dvdkviric: playing from nand of from sd-card?10:08
viricfrom nand10:08
dvdksame here10:08
dvdkother than that enabling the scaler does not change *any* details how software accesses memory.  so cannot see a reason for memory corruptions.  only timing changes.10:09
viric'timing' of what?10:09
dvdkviric: scaling costs ram bandwidth, so software runs slightly slower.10:09
dvdkviric: scaler works in the background concurrently with sw.10:10
viricand that is properly semaphored' :)10:10
dvdkviric: it's producer-consumer, like a queue, so not much possibilities for mistakes.10:11
dvdkjz47xx_frame_sel() does the switching of buffers (i.e. 'semaphore')10:11
viricmaybe 2.6.36 has problems with dma...10:11
dvdkviric: if it starts to defragment physical memory, you'd be in trouble.10:11
viricwhy would it?10:12
viricmay I disable it?10:12
dvdkviric: don't know, but 2.6.38 has transparent huge page support, so might defragment.  never heared of such a thing other than that.10:12
dvdkviric: i don't even know if such a concept exists in linux kernel.10:12
dvdki only know, that if the physical->logical address mapping changes for the dma buffers, we're screwed.10:13
dvdkbut that's why i allocate them via mmap() with flag MAP_LOCKED10:13
viricmaybe larsc can suggest something10:14
dvdkbut all that is not really guaranteed, you know.  getting physical memory from user-space is just a hack after all.10:14
viricbut even without scaling, isn't the YCrCb mapping to RGB using the gpu?10:14
viriclarsc: has linux changed a lot from 2.6.32 to 2.6.36 about making a hack to mapping physical memory to process memory from userspace?10:15
dvdkviric: you're right, it's the same data-path just with less reads because no interpolation10:15
viricthere is one ram read per pixel?10:17
dvdkthis may be the culprit: ?10:17
viricper screen pixel10:17
dvdkstack protection in 2.6.36?10:17
dvdkdo you have the latest 2.6.36?10:19
viricI have the *released* 2.6.36 + patches from larsc10:20
viricI'll take, fine?10:20
viricI'll build it10:21
dvdkviric: that's beyond me.  why not use the kernel that.s released with the official openwrt/qi-hw fimrware?10:21
viricdvdk: bah, who uses openwrt?10:22
dvdkviric: just the kernel :)10:22
dvdkviric: or use the corresponding source snapshot.10:22
viricI just take the chance to get more problems10:22
viricOtherwise the nanonote would be useless10:22
dvdkviric: you got it, want to order more?10:23
dvdknananotes, bundled with problems10:23
viricno no10:23
viricAnd now... how to overwrite the kernel image without overwriting the rootfs.10:25
dvdkviric: just reflash the right part with usbboot?10:26
Jay7dvdk: wrt latest mail in ML10:26
dvdkviric: just rip the right part out of reflash_ben.sh10:26
Jay7about disabling cursor on ssh10:26
viricdvdk: usbboot had a bug about overwriting beyond the limits10:26
dvdkusbboot -c "nprog 1024 ${WORKING_DIR}/${KERNEL} 0 0 -n10:26
dvdkviric: not yet fixed?10:27
Jay7dvdk: I'm using escape sequences to disable cursor on fb in kexecboot10:27
Jay7may be this is more convient way10:27
dvdkJay7: but it's the same problem, just relying on /dev/tty, isn't it?10:28
Jay7dvdk: put sequence on stdout :)10:28
dvdkyeah, with stdout being /dev/tty10:28
Jay7        fputs("\033[?25l", f);10:28
Jay7where f is STDOUT10:28
dvdkjay7: i think KD_GRAPHICS might do more? like disabling scrolling?  also it restores the srceen when doing KD_TEXT afterwards.10:29
Jay7this will disable cursor on controlling terminal10:29
dvdkyeah, and /dev/tty is aways the controlling terminal :)10:29
Jay7when you running mplayer via ssh your pty will be controlling terminal :)10:30
viricdvdk: xiangfu was in charge. I hope he fixed it already.10:30
viric(quite long ago)10:31
dvdkjay7: that's the problem, with your code and with mine.  there is some ioctl to map framebuffer to terminal, but too much hassle...10:31
Jay7but ioctl can't disable cursor on pty as I understand10:31
Jay7so I'm using esc seq which is working there :)10:32
dvdkjay7: but that's no use anyway, since pty != framebuffer10:32
Jay7ah.. seems I understand what you mean10:32
viricdvdk: I don't use any reflash_ben :) nanonixos does not require rootfs reflashing for updating the OS ;)10:32
Jay7this will disable cursor on ssh but not on fb10:32
dvdkjay7: btw same with ioctl.10:33
Jay7yes, but ioctl will disable cursor on fb but not on ssh :)10:33
Jay7hm.. may be call both? :)10:33
dvdkjay7: i guess ioctl() is a superset of esc10:33
Jay7iirc, that ioctl is applicable only on linux console10:34
Jay7or at least tty10:34
dvdkjay7: well, but that's what i'm coding for, no video output via ssh :)10:34
Jay7I should re-read that docs10:34
Jay7libcaca ;)10:34
dvdkjay7: already supported by mplayer ? :)10:35
dvdkmplayer -vo aa10:35
Jay7I just have same problem with kexecboot10:35
Jay7e.g. you may run it with fb UI from serial console10:36
Jay7then it's hard to guess fb<->tty mapping10:37
Jay7well, afk for some hours10:42
dvdkkyak: is current multimedia/ffmpeg package  really patent free?10:45
dvdkkyak: i guess not:10:45
wolfspraulthere is no 'patent free'10:46
wolfspraulwe work this from practical threats backwards10:47
wolfspraulthat means, to my current knowledge, we should stay away from MP3 and anything from the MPEG LA10:47
dvdkFFMPEG_DECODERS=mp3, mpeg2video wmav1 h26410:47
dvdkkyak: ok, somhow ffmpeg Makefile pulls per-codec flags CONFIG_FFMPEG_ENCODER_* ? where are they described?10:48
dvdkwolfspraul: that's what i meant.10:48
dvdkah, nice, there's a Config.in  now i understand how this is supposed to work.10:48
wolfspraulour best protection is that patent lawsuits are expensive, and that there is always the (small) chance that due to a lawsuit, a patent gets declared invalid10:49
wolfspraulthat's pretty much all that is holding the parasites back10:49
wolfspraulthen of course the (still) golden 20 year rule, even if we make it a 25 year rule to be safe :-)10:50
dvdkwolfspraul: being a corporate entity really makes life difficult here.10:50
dvdkwolfspraul: i mean i can host mplayer on my dyndns server and nobody would ever come after me :)10:51
wolfspraulthe law is still impartial, in the major markets. so worst case you have to become a patent troll and make it back, even if from someone else :-)10:51
dvdkwolfspraul: that's what samsung is doing against apple currently.  aaah, i hate it.10:51
wolfspraulI cannot imagine that such hosting could cause you practical trouble, never heard of such a case at least.10:51
dvdkwolfspraul: but while they still each other, nanonote can strive for world domination.10:52
dvdkwolfspraul: I'm a not-for-profit private person, no chance under german law.10:52
wolfspraulyes sure, like I said I focus on real threats, not fear.10:53
wolfspraullet's take out mp3, mpeg4, h.26410:53
dvdkwolfspraul: and windows media, and ac3, and... only vorbis, flac etc. as far as i see., also mpeg110:54
wolfspraulOgg Theora is by now such a hopeless loose that if we are lucky we stay under radar.10:54
wolfspraulif anything the next big battle will be over WebM10:54
wolfspraullet the lawyers make their keyboards crack over that one...10:55
dvdkwolfspraul: if they create a patent pool over Theora, i'm going amok.10:55
wolfspraulwho is 'they'?10:55
wolfspraulmany people are creating patent pools10:55
dvdkwolfspraul: over theora? mpegla is trying to do that just now, afair.10:55
wolfspraulbut it will not be worth to go after Theora, because Theora users are too poor10:55
dvdkwolfspraul: but i mean it's already in firefox.10:55
wolfspraulwell let's see. and meanwhile the clock keeps ticking.10:56
dvdkwolfspraul: so that gives theora quite some market share.10:56
wolfspraulI doubt that, but I don't care either we can build our thing.10:57
wolfspraulwhat's the best guess - how many of the world's tech patent lawyers are currently full-time on the smartphone front?10:57
wolfspraulmust be thousands of them10:57
wolfspraulwriting and writing and writing. and charging someone.10:58
dvdkwolfspraul: and who writes the law?  more lawyers :)10:58
wolfspraulI feel pretty good about our patent strategy right now.10:59
dvdkcan somebody help to list the codecs that are ok to go into the firmware?10:59
dvdkffmpeg gives us:10:59
dvdkoh my god that list is too long for irc.11:00
wolfspraulwhite-list is better11:00
wolfspraulmp2 is arguably ok btw, or so I read11:00
dvdkmpeg1 and mp2, yes the xiph stuff i know.11:00
wolfspraulwhat else do we need?11:01
dvdkthe codecs used on mobile phones amrnb etc. are ok ?  dunno11:01
dvdkh261 should be ok?11:01
wolfsprauldon't know11:01
wolfspraulwait until someone wants them11:01
dvdkok, pragmatic.11:01
dvdkbut what about demuxers? no patents on them?  only allow mpeg, avi, ogg, mkv?11:02
wolfsprauldon't know. like I said I would add things case by case as needed.11:03
wolfspraulseems Ogg is a good starting point11:03
wolfspraulthen FLAC, seems popular on some p2p sites and should be patent-safe11:04
dvdkwolfspraul: ogg alone won't help much, very limited container.  mkv is open-source and used by webm, so should be ok, too11:04
wolfspraulthat's all I can tell you about11:04
wolfspraulgreat, you know this much better than me :-)11:04
dvdkof course, flac11:04
wolfspraulI can only say - whitelist, case by case upon request, and then we look into the threat situation11:04
dvdkwolfspraul: already patched configure to not include codecs by default11:04
dvdkstart with zero, then whitelist11:05
dvdkok, that's a short list of codecs to enable :)11:05
wpwrakwolfspraul: (horror movie) heh, nice idea ;-) alas, the part where the things is shown in action is very short compared to the rest. not sure if my cutting improved that, though12:16
kristianpaulwpwrak: (video)you like like make suspense !!12:47
kristianpaulthat room light switching was misterious.. actually i get cautght first by the "OPEN" green msg at the left :)12:49
kristianpaulhmm suspiciously the ben lcm get white just in the time it is showing up in the big screen, is it  light issue or?..12:58
viricBefore the dvdk code.... how was the video playing in the nanonote?13:06
viricHow could it be usable?13:06
wpwrakkristianpaul: (white screen) that's just when i turn off the lcd clock. then the display fades. i need to turn off lcd refresh because it would take away precious bus bandwidth. and it would also add noise to my timing.13:08
kristianpaulah ok13:12
viricwolfspraul: The nanonote plays VP8 video just fine!13:13
viricI've just encoded a video, and it even takes less cpu than theora decoding13:15
viricencoded in the PC, played in the nanonote.13:16
viricBut I see flicking13:18
wolfspraulviric: you mean in terms of including vp8 in the whitelisted codecs?13:19
viricwolfspraul: yes13:19
wolfspraulI think it should be fine, right? I am not aware of any agressor there, right now.13:19
viricgoogle says it's caring that there are no patent infringement in it13:19
wolfspraulthere's also dirac, which for some reason seems to get little attention but technically is not bad13:19
wolfspraulsure, let's whitelist it then :-) [vp8]13:20
viricisn't it for some peculiar bandwidth?13:20
viricvery low bandwidth, or so13:20
viricI forgot its advantadges13:20
wolfspraulno I think it's general purpose13:21
wolfsprauljust no big corporate giant behind it13:21
viricit looks like dirac requires a FPU13:22
wolfspraulno idea, it just came to mind when you mentioned vp8 as another potentially whitelisted codec13:24
wolfspraulI have never once seen dirac in use, anywhere :-)13:24
wpwrakVP8 should be about as safe a google money :)13:31
viric'google money'?13:31
wpwrakviric: google are trying to push that codec. if someone got sued over it and lost, that would be very very bad press for google.13:32
viricbtw, mplayer seeks perfectly in matroska containers, but it crashes on ogg containers13:33
viricwpwrak: I agree13:33
wolfspraulthat's my concern over VP8, that it is Google. This will be hashed out in the old boys club between Google, MPEG, Apple, Microsoft, etc.13:33
wpwrakviric: so if we'd run into problems with it, it's very likely that google would try to help. particularly given that qi-hw's business ethics are about as evil as mother theresa's, so it would look even worse.13:34
wpwrakwolfspraul: yes, it may get you drawn into a battle, that's true13:34
viricmother theresa is not a very good example I suppose13:34
wolfspraulI wouldn't expect one dime of support, or one tear, for that kind of logic.13:34
viricShe is famous for his war against gays13:34
wpwrakviric: (mother theresa) in public opinion ;-)13:35
wolfspraulthis is not how things work in the real world13:35
wolfspraulanyway, we see where VP8 goes. at this moment I agree we can consider it to be a free codec.13:35
wpwrakwolfspraul: naw, if someone got burned for using VP8, others would shun it too. that would be bad for google's investment into it and the value they expect it to have.13:36
wolfspraulfine but there are dozens of players, who knows who cares about what.13:36
wolfspraulmany of them will not give a damn about someone else's reputation13:36
wolfspraulbut we should give VP8 the benefit of the doubt, otherwise we are just in FUD land ourselves, cannot operate like that...13:37
wolfspraulhey, many may not even give a damn about their own reputation, as long as they get filthy rich13:37
wolfspraulthat wouldn't be unheard of, I think is safe to say :-)13:37
wpwrakwolfspraul: (reputation) what i mean is that qi-hw would look 100% the victim. if, say, microsoft got burnt over vp8. half the world would celebrate and the rest would say that they can very well help themselves. so there would be much less incentive for google to lend a helping hand.13:39
wolfspraulI think in politics you celebrate victory strictly only after it's really over. Until then, even 1 second before the end, expect any dirt from any direction :-) No, I wouldn't count on anything in such a scenario.13:46
wolfsprauland I wouldn't want to be in it even, that's why I still prefer peaceful and more or less forgotten (by the mainstream) Ogg Theora...13:47
wolfspraulI think several people emailed google about their vp8 ic cores - no response13:49
wpwrakwolfspraul: with ogg, you'd have the whole free software world up in arms. that may not be too bad either.13:49
wpwrakwolfspraul: but yes, staying under the rader is a good strategy as well13:49
wolfspraulstay close to home, don't venture too far13:50
wolfspraulif you really want to pick a fight with Google, Apple, etc. first thing you need to get in touch with high-profile lobbyists, high-profile lawyers, high-profile anything13:50
wolfsprauland why would they even want to talk with you? they have their own agenda. and so on...13:50
wolfspraulmaybe you could pull a groklaw and do it all grass-roots, but that kind of thing is rare, and even groklaw is shutting down now13:52
wolfspraulthat was a strange coincidence of forces to overlap from one field (free software) into another (law). but you should not count on this happening again, at least I wouldn't.13:53
wolfspraulso you may be crushed, and nobody cares13:53
wpwrakwolfspraul: all i'm saying that it would hurt google if someone lost a lawsuit over vp8. in terms of reputation ("i trusted google and lost"), in terms of discrediting their technology and their assurances ("it's not as safe as they claimed"), and perhaps also in terms of establishing a precedent. so it would only be logic if google tried to help you win that case.13:53
wolfspraulthat's happening every day, thousands of times13:53
wolfspraulonly if you connect it to a pr campaign13:54
wolfsprauland a successful one13:54
wpwrakwolfspraul: "copyleft hardware company sued over VP8" that pr writes itself ;-))13:54
wpwrakmake that "Google's VP8"13:54
wolfspraulI'm telling you it's all work.13:54
wolfspraulit's not like you or me are the loudspeaker of the world, and the other 7 billion are waking up wondering each day what we have to tell them today13:55
wolfspraulI would never believe in vp8 hoping that if there was a problem, Google would come help me.13:55
wpwrakwolfspraul: no, but everyone who is contemplating the use of VP8 because of its supposed patent-free state would want to listen in this case13:55
wolfspraulonly that I say "nice story, google pr team" :-)13:56
wolfspraulyes, but getting that message out is hard work in itself13:56
wolfspraulI'm not saying it's impossible, I understand you.13:56
wolfspraulbut it will not happen by itself, and there are many strategies.13:56
wolfspraulgoogle may also distance themselves from you13:56
wolfspraulif you even get them to issue 1 press release including your name13:56
wolfspraul"this case is about XYZ, not about VP8."13:57
wolfspraulwhat do you do then?13:57
wolfspraulthat press release will be picked up globally13:57
wolfspraulyou counter it saying "no!"13:57
wolfspraulha ha13:57
wolfspraulI want to see how that continues :-)13:57
wpwrakwolfspraul: you should probably contact them before making a press release :)13:57
wolfspraulno I meant them, I'm just randomly suggesting another approach for Google, the "that's not our stuff" approach.13:58
wolfspraulthey won't even do you the favor of listing your name in their statement13:58
wpwrakwolfspraul: sure, they can do that. but it may make them look even worse.13:58
wolfspraulbut their loudspeaker first of all is 1000 times louder than yours13:58
wolfspraulthey will first try the easy thing to get over this13:58
wolfspraulthey are fighting with stuff every day, they also need to keep their desks clean13:58
wpwrakwolfspraul: do you think they have lawsuits over VP8 going on at the moment ?13:59
wolfspraulhmm. pure speculation.13:59
wolfspraulyou might think that some lawyers somewhere are looking at this, for sure.13:59
wolfspraulmaybe some are collecting evidence, preparing a case14:00
wolfspraulthere is a lot you can do before 'lawsuit', i.e. going to a public court14:00
wolfspraulI'm sure Google's legal time has ongoing VP8 related work14:00
wolfspraulI trust Google here actually.14:01
wolfspraulthey want this thing to 'just work' so they can spread their ads.14:01
wolfspraulwill Apple/Microsoft really go all-in over this? no idea. just random speculation.14:01
wolfspraulright now it seems the smartphones wars keep them busy14:01
wpwrakyeah. a good way to occupy the giants :)14:04
wolfspraulI am sure there are private attempts under-way for VP8 licensing arrangements.14:04
wolfspraulthat's pretty much a given14:04
wolfspraulthe MPEG LA will 'suggest' this or that to Google14:04
wolfspraulI think that we can be sure of.14:04
wolfspraulmostly they want money, not go to court with all sorts of pesky rules. SO those licensing negotiations may drag on for years.14:05
viricIsn't it simply nice for quite all the parties (apart from the patent owners that want revenue)?14:06
wolfspraulas a side effect of the licensing negotiations they can start to create documents/paper-trail for going to court later, but that's several steps later due to the uncertainties once they do that14:06
viricVP8 decoderes in hardware, in software, all for free for everyone.14:06
wolfspraulviric: keep in mind how the 'patent protection' works14:07
wolfspraulI love to explain it to you.14:07
wolfspraulit's a way for all of us to become rich14:07
wolfspraulwe can sell the same thing many times14:07
wolfspraulnobody 'owns' anything14:07
wolfspraulit's a form of communism actually14:07
wolfspraulso it works like this14:07
wolfspraulon your website, you call your friends to submit 'patents' to you that make VP8 a 'safe' technology14:07
wolfspraulwhen you have a whole stack of such patents, you offer all VP8 users the safety of your protection by virtue of a payment into your bank account14:08
wolfspraulthat's all14:08
wolfspraulfor those that don't understand the beauty of your protection, your stick are the public courts14:08
wolfspraulyou can do this14:08
viricDo you think companies are not starting to put vp8 on their devices afraid of something?14:09
wolfsprauland if Werner likes your business model, he can start too, asking his friends to submit, creating another 'safety' pool, and charging the same people you are charging already14:09
wolfspraulyou can even sell your customer list to him!14:09
wpwrakwolfspraul: you think that's the plan behind OIN and such ?14:09
wolfspraulthere's more and more money everywhere, can you see it?14:09
wolfspraulno more house flipping14:09
wolfspraullet's get into patents now14:09
viricI've not understood clearly 'my role', 'my website', and all that in the explanation... I may need more words14:09
viricI don't understand 'submit patents' either14:10
wolfspraulok I was just kidding anyway14:10
wolfspraulthat 'submit' can be left very undefined14:10
wolfspraulyou can run the whole scheme on pure fear14:10
wolfspraulthe power of words14:10
wolfspraulwhen you write to your customers, you don't need to tell them exactly which patents you own14:10
wolfspraulyou just need a good name and logo14:10
wolfspraulviric: which country do you live in?14:10
viricSo, I would be selling some assurance that the payer will not be infringing anything14:10
viricthen I got it14:11
wolfspraulthat they are not infringing on the intellectual property of you and your friends14:11
viricwolfspraul: I live in Catalonia, but most governments in the world won't recognize it as a "country" by internationally agreed legal terms.14:12
viricwolfspraul: and thus I'm counted as a Spanish citizen14:12
viricregardless of my opinion.14:12
wpwrakviric: no, you can't sell an assurance against infringing on anything. you can sell an assurance that you won't go after them, though. whether they accept or not depends on how dangerous they think you are. see SCO and those (few) licenses they sold.14:12
viricwpwrak: ah clear. ok.14:12
wolfspraulyes, my wording may not have been very precise.14:13
wolfsprauland SCO probably shows you the limits of the model, I agree with Werner.14:13
wolfspraulI don't understand OIN, no idea.14:13
wolfspraullet's hope Google can keep VP8 free14:14
wolfsprauleven if they drag out any public lawsuits for a few years, with endless licensing negotiations - fine by me14:14
viricand Google advertised itself as assuming responsabilities if VP8 infringes any patent, proposing its own lawyer forces to protect VP8 users?14:14
wolfspraulGoogle cannot do that14:14
wolfspraulGoogle has no power over the patent system.14:14
viriclawyer forces14:15
wolfspraulGoogle will never ever lend out their high-paid legal team to support some Joe Doe in a patent case.14:15
viricSo if someone gets sued for VP8 usage infringing patents, google lawyers will come as help14:15
wolfspraulnot in a million years14:15
viricIs google using vp8 anywherE?14:16
viricI remember youtube used h264 in its html5...14:16
wolfspraulyes I think they are migrating quite strongly, no?14:16
viricI don't remember youtube switching from h264 to vp814:16
viricmaybe it happened14:16
viricI don't know14:16
viricThey are closing google video, that's true. I received a letter as a google video uploader about that14:17
viricThe first brave user of vp8 should be google itself. But I don't know how brave it is by now :)14:17
wolfspraulit's a very respected codec, remember it was bought from a company with quite some history14:18
wolfspraulit's a solid option, for sure14:18
viricI mean brave related to patents14:18
wolfspraulyeah well. that's outside of Google's control mostly.14:18
wolfspraulthey are an attractive target because they have money.14:18
viricBut it would look strange, if Google said VP8 is patent free and Google itself was not using it14:18
wpwrakwolfspraul: why should google not offer legal advise ? what they can't do is offer to just compensate for all damages, otherwise you'd have trolls going after even the least small shop, hoping for big goolg emoney14:18
viricGRrr how to measure the battery left in the nanontoe?14:19
wolfspraulwpwrak: he, correct. I see now you start to think around some corners too :-)14:19
wolfspraulI have no idea, it's all speculation.14:19
wolfspraulI would never for one second hope that Google helps me with _ANYTHING_ if I have a problem.14:19
wolfspraulif you want to believe that, go try and we find out... Maybe I'm wrong.14:19
wpwrakwolfspraul: well, it's always safer to not count on getting help14:20
wolfsprauland like you also noticed now - they have to be careful.14:20
wolfspraulthere are lots of setups14:20
viricGrr I don't have /sys/class/power_supply/battery in
wpwrakwolfspraul: but in this sort of case, it seems that it would be to google's disadvantage if you lost a lawsuit over their codec14:20
wolfspraulyou create a small company, I sue you. google helps you. After some more twists we both get rich from google money :-)14:20
wolfspraulif you think that kind of thing cannot happen, well...14:21
wolfspraulif there even were a 0.001% chance that it might work, someone will try14:21
wpwrakwolfspraul: yes, of course. or you sell a ben with all codecs under the sun, including vp8. then sisvel knocks on your door for mp3 and you run to google crying about vp8.14:21
wolfspraulI want to stay out of all this.14:22
wolfspraulI want to make products that work, build a customer base, grow it. Have real users. etc.14:22
wpwrakwolfspraul: i note with approval that you've been thinking of ways to improve the financial situation of sharim ;-)14:22
wpwrakwolfspraul: create a sister company that takes care of the legal issues, suism.cc ;-)14:23
wpwrakthe first product would not be the ben nanonote but the qizha patent pool :)14:26
wpwrakhey, who understands chinese IP laws anyway ? a golden opportunity for a reign of FUD ;-)14:26
rozzintrollism.cc. I might actually like to have that one....14:52
wpwrakjumo.cc ?14:54
wpwrak(according to google translate, "ju mo" is chinese for "troll")14:54
wpwrakdarn. already taken.14:54
wpwrakhmm, i don't think they taste so good. i'd just throw them away, with the other kitchen garbage :)14:56
wpwraklet the rats and stray dogs have some nice meal, too14:57
dvdkviric: done some more benchmarking?  how much is vorbis decoding costing us?15:55
viricdvdk: ~50%, audio+video for bbb.ogv15:55
viricdvdk: ~40% for vorbis+vp8+mkv15:56
viricfor the same size. 256x14415:56
dvdkviric: and the audio part?  we can speed audio up a little, albeit at cost of lower quality15:56
dvdkviric: already mailed, asking how you encoded  the video.15:56
viricah wait, the vorbis+vp8+mkv test has 64kb/s audio15:56
viricnot fair, I imagine. I think bbb.ogv has a higher rte15:57
dvdkno bbb.ogv has quality=0 i.e. 55-65kb/s15:57
viricah ok15:57
dvdkviric: encode by quality, not bitrate.  bitrate constraints are bad for ogg and theora15:57
dvdkhow did you encode?  standard ffmpeg?15:58
viricso, basically... 'ffmpeg -vcodec libvpx -acodec libvorbis -ss 256x144'15:58
viricThe rest, filenames.15:58
viricI even did not specify any quality :D15:58
viricThe original video is x264, 640x360 15fps 500kbps, 64kbps audio15:59
viricaudio in AAC15:59
viricdvdk: I thought '-v X' was for setting the quality, and I later saw it was the verbosity level. hehe16:00
dvdkok, so i need libvpx or is that ffmpeg-internal?16:00
dvdkwhat size has the resulting file?16:01
kyakdvdk: (ffmpeg makefile) yes, they use that nice Config.in trick16:01
dvdktried to encode for full-screen 320x176 or sth?16:01
viric1h30m, 130MB16:01
dvdkkyak: teh new mplayer makefile won't be that nice, though :)  well not in the first version, anyways16:01
viricah no wait. This is my last test at 320x180 size.16:01
viricdvdk: btw, there is flickering16:02
dvdkviric: about 2MB/minute, that's about as good as ffmpeg2theora -v816:02
dvdkviric: from the accel or from the encode?16:02
kyakdvdk: it's nice enough as long as it works ;)16:02
viricI can't say.16:02
dvdkviric: there may be minimal frame tearing due to video refresh rate != lcd refresh rate.  but it's hardly noticeable.16:03
viricThe video I encode has a scene at the beginning that shows the flickering very clearly16:03
dvdkviric: only on NN or also on pc?16:03
viriconly on NN16:04
viricI may send you the first seconds16:04
dvdkviric: not even sure my mplayer build can decode it, but good for testing16:06
kristianpaulah, CGA is not QVGA..16:06
kristianpaulyeah README already explain why..16:07
viricIt's about a documentary on Kissinger :)16:07
dvdkviric: it says 'video codec: ffodivx ffmpeg mpeg-4'16:07
viricdvdk: :D right. I thought ffmpeg would 'copy'16:07
dvdk-acodec copy -vcodec copy ?16:08
viricI set that, and it outputs divx16:08
viricah now16:08
viricdvdk: redownload16:08
viric(I needed '-vcodec copy -acodec copy' *before* the -i file)16:09
dvdkstill the same16:10
dvdkok.  ffvp816:10
viricdvdk: I noticed the horizontal artifacts at some lines16:11
viricAnd on sharp edges, I can also notice (unfortunately) the peculiarity of the nanonote LCD16:12
kristianpaullinphone-nox a voip client for the nanonote?16:16
LunaVorax_miniTalking about the packages, why is cdparanoia available for the nanonote. Is a CD-Drive extension planned ? ;P16:18
viricLunaVorax_mini: you can use it through network block devices (nbd) :)16:19
dvdkviric: this looks like frame tearing16:19
dvdkviric: when i press pause, i see no artifacts16:19
viricdvdk: about what?16:19
dvdkviric: looks like vp8 needs so much ram bandwidth, that screen redraws are lowed down?16:20
dvdkviric: your example video16:20
LunaVorax_minioh really viric :o I didn't knew that16:20
viricI used nbd long ago, when I used to have only one machine in the LAN with a DVD drive, to play DVD videos :)16:21
viricdvdk: I've not played that 320x180 video in the nanonote. I only played a 256x144 version of it16:21
viricdvdk: maybe 320x180 is too much for the nn?16:22
dvdkviric: looks ok, just need some time at the start of the video to get to speed (tlb misses etc.)16:22
viricdvdk: in general, the vp8 file I tried used less CPU (according to 'top') than the theora.16:22
LunaVorax_minidvdk, according to what I read, vp8 seriously need to be fixed anyway16:22
viricLunaVorax_mini: fixing what?16:22
LunaVorax_miniI'm not a coder, go read this article to have detail I wouldn't be able to give you http://x264dev.multimedia.cx/archives/37716:25
wpwrakLunaVorax_mini: have you seen, we now have video out ! ;-)16:27
LunaVorax_miniwhat what what what ?16:30
LunaVorax_miniI need to see that !16:30
wpwrakLunaVorax_mini: ah, you don't read the mailing list ? :)16:32
LunaVorax_miniI'm sorry not yet16:32
LunaVorax_miniNot YET16:32
wpwrakLunaVorax_mini: video: http://en.qi-hardware.com/w/images/7/76/Ben_UBB_VGA.ogv16:32
wpwrakLunaVorax_mini: more information: http://lists.en.qi-hardware.com/pipermail/discussion/2011-April/007817.html16:33
LunaVorax_miniHolly crud ! That's awesome16:36
LunaVorax_miniI didn't knew at all that the minisd card slot was offering so much possibilities :o16:36
LunaVorax_miniThat's insanely awesome16:36
wpwrakLunaVorax_mini: in fact, we can even do a little better. i now have doubled the number of updates (at half the pixel clock) and it more or less works at least on one of my monitors16:37
wpwrakLunaVorax_mini: (possibilities) never underestimate the power of bitbanging ;-)16:38
LunaVorax_miniThat's fantastic wpwrak16:38
kyaknow, that UBB-VGA output by wpwrak fits together well with job being done by dvdk :)16:40
LunaVorax_miniThat's not true 320x240, I understood it right ?16:40
wpwrakkyak: alas, you can't combine the two. each already eats most of the memory nbandwidth16:43
wpwrakLunaVorax_mini: it's somewhere between 160-320x240, with line doubling16:44
kyakwpwrak: hm.. are you saying there is no practical application for that thing ? -\16:44
wpwrakkyak: you could use it for slide shows and similar applications16:45
kyakstill, it is very good16:47
LunaVorax_miniNo transition but is the Ya Nanonote still planned16:49
LunaVorax_miniAnd therefore planned for 2012 ?16:49
wpwrakLunaVorax_mini: i think the ya will be planned as soon as someone provides funding :)16:51
LunaVorax_miniHow much do you need ?16:52
kyak(in most cases, this question is followed by money ;)16:54
wpwrakLunaVorax_mini: i guess about 500 kUSD may be sufficient. my lowest estimate would be around 300 kUSD, but that's probably too low16:55
LunaVorax_miniHum ok16:56
LunaVorax_miniYou need 1000 fanboys like me then16:56
wpwrakLunaVorax_mini: that would be development and a small (3 kunits) initial production run16:56
LunaVorax_miniOk, I see16:56
wpwrakLunaVorax_mini: also assumes very few people on the payroll and them working with compensation close to subsistence level16:58
wpwrakin other words, the community needs to participate as well. otherwise R&D cost goes up quite a bit.16:58
LunaVorax_miniOk I see16:59
LunaVorax_miniI think it's time to save for a donation17:00
LunaVorax_miniCount on me I'll do so asap17:00
wpwrakLunaVorax_mini: cool, thanks ! i don't know what wolfgang thinks of donations as a means of financing, but if it comes to that, any contribution would be welcome17:03
wpwrakLunaVorax_mini: btw, with community participation, i meant sharing the design work. we did part of this in the gta02-core project and it worked quite nicely for a while. the problem we had there was that we depended on openmoko.com to supply components. and this promise was never fulfilled.17:05
wpwrakkristianpaul: hmm, if you want a ben-phone, you could just add a GSM module ...17:06
kristianpaulwpwrak: i have one17:07
kristianpaul(GSM Module)17:07
kristianpaulbut too big too muchs wires, is a sim54817:07
kristianpaulalso the board dint router voice part..17:07
LunaVorax_miniwpwrak, in some way, donations is the only efficient way for me to contribute to what you are doing as my programming and electronic design knowledge are very limited for now17:08
LunaVorax_miniSo far the only "usefull" thing I've done was to show the Ben to people at knowledge and impress them about "how small it is" ;)17:09
LunaVorax_miniFunny to see that in some people's head "computer" = "big"17:09
LunaVorax_miniI may be na17:12
LunaVorax_mininaive but17:12
LunaVorax_miniIsn't the FSF interested about giving you some fund ?17:12
LunaVorax_miniSupposing that the FSF have funds to give, I don't know how they work exactly17:13
kristianpaulwpwrak: wich GSM Module you recomend? :-)17:14
wpwrakLunaVorax_mini: hmm, i don't think the FSF sponsors projects directly17:14
LunaVorax_miniOk wpwrak17:14
wpwrakkristianpaul: if you want to go fancy, there's a company called "option" that make nice little UMTS modules. but it may be hard to get documentation.17:14
wpwrakkristianpaul: otherwise, lemme dig ...17:15
dvdkbtw anybody knows the patent status of AC3?  this is utilized in mplayer, as an audio-filter (not codec) (getting linker errors if I disable it).17:19
wpwrakkristianpaul: this one is nice and small: http://www.telit.com/en/products/gsm-gprs.php?p_id=12&p_ac=show&p=4717:20
wpwrakkristianpaul: there are also other, slightly larger models, in the same family17:21
wpwrakkristianpaul: note the large quantity of documentation at the bottom of the page :)17:21
kristianpaulwow is amll17:21
kristianpaul(doc) oh yes17:22
wpwrakkristianpaul: if you dig a little in gta02-core, you'll also find untested schematics and there should be a footprint for the ge864 somewhere as well17:23
wpwrakhere it is: svn.openmoko.org/developers/werner/gta02-core/ge864/ge864.fpd17:24
wpwrakalso completely untested, of course ;-)17:25
kristianpaulanyway, may be the 3g/edge/gsm part will be just enought17:26
qwebirc81180hi guys, i wonder if there is the ability of booting the ben nn by 'alarm'. (bios-clock?) so the device could p.e. boot every day @ 05:00 and wake me up by music :D17:27
qwebirc81180would be great for using as pda17:28
qwebirc81180reminders are appliance #1 for pda's17:28
wpwrakqwebirc81180: good question :) i don't know if the hardware and the kernel support it, but i can at least not find anything quickly that would indicate it's impossible17:34
wpwrakqwebirc81180: i.e., you may want to experiment a bit :)17:34
qwebirc81180the bad is: I'm an advanced user, able to start a compile-run, but unable to really change sourcecode :\17:35
wpwrakqwebirc81180: here is a program that exercises the kernel's wakeup api: http://svn.openmoko.org/developers/werner/wkalrm/17:35
wpwrak(designed for a different architecture, but it should also work on the ben)17:36
qwebirc81180but in my opinion, this could massively improve the useability of the ben nn17:36
wpwrakqwebirc81180: change "unable" to "not yet able" ;-)17:36
qwebirc81180hehe, freerunner-code on ben?? :D17:36
wpwrakit's all linux ;-)17:37
qwebirc81180wpwrak: right ;) but i'm not sure if it's a good idea to start learning programming with age 53 ;)17:37
qwebirc81180as for linux: I'm with it since 0.99pl6 (SLS-Linux) :D17:38
qwebirc81180but only as user and 'make'er ;)17:38
qwebirc81180oh, btw, I should use my 'real' nick ;)17:39
Fusinis there a RTC in Ben, or does it use only software clocks?17:41
viricthere is17:41
LunaVorax_miniHum, reading the Guo Nanonote page. Are you really sure that it'll have the same design in 5years ?17:43
LunaVorax_miniEarly question may be a stupid question :)17:44
wpwrakFusin: (53) bah, that's no age ;-) and the hardest part of doing anything is convincing yourself that you can do it ...18:41
Fusinmy problem is: i already run for following evolution in hard and software technology18:42
wpwrakLunaVorax_mini: i wouldn't consider anything beyond a ya as anything than a small quantity of vapour released in the hard vacuum of outer space ;-)18:42
Fusinhow could i learn to master them?18:42
wpwrakFusin: mastery take s a while :) start small. pick objectives that are a little beyond your knowledge, so that they don't overwhelm you but still give you something new.18:44
Fusinben is new and freerunner too (linux outside big boxes is really exiting)18:45
Fusinaltough to tell the truth: for daily use i just ordered an android Phone and an android pad :D18:45
Fusinfor the 'real' use18:45
Fusinben and freerunner are atm toys to play with18:46
Fusinand for showing collegues what linux can do18:46
wpwrakFusin: yeah. FR is getting a bit long in its teeth, too. well, same thing for the ben. there are many incremental improvements that could/should be made.18:47
viricFusin: what *linux* can do? With an Android you can show what it can do too18:54
Fusinright, but android is more a gui than the os itself18:55
viricFusin: and linux is only the kernel18:55
viricMaybe something more like 'what companies + volunteers  publishing free software and hardware' can do18:56
Fusinnot enough evangelists ;)18:58
mstevensFusin: I want to like android but I can't get on with it19:05
FusinI don't like it, I just want to use it, because it works and does what it's suposed to do. no more no less19:06
Fusinand it's not owned by gates nor jobs ;)19:08
wpwrakFusin: (ping) you need to specify the destination address :)22:48
Fusinwithout destination we accept ALL pongs ;)22:49
wpwrakFusin: still, where do you send the ping packet ? :)22:51
Fusinwhole world :P22:51
wpwrakaii .. instant internet meltdown22:51
--- Mon Apr 25 201100:00

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