#qi-hardware IRC log for Thursday, 2011-09-01

qi-bot[commit] Werner Almesberger: mkmk-simple: exit with an error code if "cameo" fails (master) http://qi-hw.com/p/cae-tools/54101ea00:32
qi-bot[commit] Werner Almesberger: cameo/: added option "any" to "mill", to accept paths irrespective of tool size (master) http://qi-hw.com/p/cae-tools/75c3a7d00:32
qi-bot[commit] Werner Almesberger: mkmk-simple: added variable ANY (usage: ANY=any) to set "mill any" (master) http://qi-hw.com/p/cae-tools/afa746000:32
rjeffriesEvan  Prodromou of status.net and identi.ca has a new blog about open source companies. http://ope.nu/p/40 The nine revenue streams for Open Source companies01:20
jow_laptopkyak: how many programs are affected by the api change?01:22
wpwrakwolfspraul: monthly news this month ? activity is picking up again - mailing list traffic in august was at 200% of july03:08
wolfspraulyes but at the very least I need a few days now to focus only on the m1 launch03:09
wolfspraulI have to get this thing out, people are waiting...03:09
wolfspraulAnd rejon keeps me busy too03:09
rejonyah right03:10
rejonteaching you about culture03:10
rejonyour own german culture03:10
wpwrakwolfspraul: how soon will be the M1 launch ? if it's ~1 week or more from now, a bit of a teaser in the weekly news could also be good :)03:11
wpwrakrejon: wow ;-)03:11
kristianpaulm1 launch party? :-)03:11
rejoni think we will take a nice soft launch03:15
rejonand i have a plan for growing this and our dev community03:16
rejonwe need more people adding features to m103:16
rejonlooks to me like development has stopped by the main developers at present03:16
rejonfrom the commit logs03:16
wpwrakdon't forget the summer holiday effect ...03:16
wpwrakholidayS even03:16
wpwrakand of course, sebastien now has paid-for work on the side to distract him03:17
kristianpaulhum. i think sebastien is working on increasing rendering resolution too03:18
rejonkristianpaul, what are you hacking on?03:19
kristianpauldont you know?03:19
wpwrakkristianpaul: if i understand the situation correctly, he made a few changes that should help with this, but then found that it wasn't enough, the next step would be a big one (speed up the memory subsystem), and he hasn't worked up enough fury yet to attack this issue03:19
kristianpaulyes wpwrak 03:19
kristianpaulrejon: gps baseband03:19
rejonvery cool03:19
xiangfuwow: Estimated project cost: $11,719,963  <--https://www.ohloh.net/p/nanonote-firmware03:22
xiangfuEstimated project cost: $424,833 <-- https://www.ohloh.net/p/milkymist03:23
wolfspraulwe should charge that to the Linux Foundation. Isn't that how it works?03:24
kristianpaulcan you cost ben wpan too?03:24
kristianpaul(charge) lol03:24
wolfspraulrejon: yes I think the ability to mix in static images would be cool03:28
wolfspraulif high-res is too hard, maybe that's lower hanging fruit03:28
wolfspraulunfortunately the most natural way to feed it static images would be a usb stick, but that's also hard to implement03:29
rejonwolfspraul, ideally03:30
rejonyes, a folder of images03:30
rejonpoint m1 to a folder of images03:30
rejonand they can mix into the patches03:30
rejonsome advertising copmanies interested in autogeneration of ads in public spaces03:31
rejonin china there are so many03:31
Action: kristianpaul feel unware about flicernoise rendering internals..03:31
wolfspraullike I said this requires 2 todo items: ability to mix in images in the ic design & flickernoise, and then best support for usb-storage to feed them in03:32
wolfspraulboth are non-trivial, difficult to implement. not sure who is up to the task.03:32
wolfspraulkristianpaul: how would the user feed images into the box?03:32
wolfspraulpeople would want to copy them to a usb stick and plug it into m103:32
kristianpaulusing the internet? 03:32
wolfspraulah true, that's a good workaround until we have usb-storage support03:33
wolfsprauljust point it to a url03:33
wolfspraulkristianpaul: so for still images, first level of support would be download from URL only03:38
wolfsprauloh, another problem03:38
wolfspraulspace in the nor flash is limited, just a few megabytes03:39
wolfspraulso if there should be a way to cache the images, that may not be enough03:39
wolfspraulthe 2 GB memory card is present in every rc3 we ship, but more tough software work is needed to make it work03:39
rejonwolfspraul, can you and wpwrak fight about openwrt vs please?03:40
wolfspraul"fight about openwrt vs"03:40
wolfspraulso I think for a really effective implementation of static images a lot is missing right now03:40
wolfspraulusing the 2 gb card as storage03:40
wolfspraulimplementing support for still images in the ic design and flickernoise03:41
wolfsprauladding support for downloading images from a url, probably download and cache them locally03:41
wolfsprauladding support for plugging in usb sticks03:41
kristianpaulLinux :-)03:42
wolfspraulall of this together is several man years of work :-)03:42
wolfspraulrejon: what do you want to know about openwrt & oe ?03:42
wolfspraulfor my current purposes openwrt is better suited, I like it better03:42
wolfspraulthat's why we were able to make an openwrt useland for m103:43
wolfspraultry that with oe - super hard I'd say. anybody feel free to proove me wrong.03:43
kristianpaulno thank you ;)03:43
wolfspraulI'm super practical about build systems. I want them to work quickly, be small and lean.03:43
wolfspraulmany people use customized buildroots03:43
wolfspraulalso fine. but openwrt has just the right level of 'platform architecture' and easy scripting / easy ways to customize03:44
kristianpaul(other crazy workaround) use build-in camera to take pics can be re-used on a patch03:44
wpwrakrejon: naw, i don't want to fight about openwrt vs. oe :) i think owrt has developed into a resource drain to the nanonote community that could have been avoided by picking a more feature-rich distribution. be it OE or whatever.03:44
wolfspraulI think openwrt is best for small systems, just a little bit more than a plain buildroot03:44
wolfspraulwpwrak: it's amazing that this didn't happen then :-)03:45
wolfspraulseems the people that are actually working with distros come to a different conclusion about workload, getting the same thing done in A vs B03:45
wolfspraulso openwrt for embedded systems - fine by me03:46
wolfspraulthat's why we have an m1 userland today, even if it's hacked together03:46
wpwrakwolfspraul: i'm not saying openwrt is a bad start. it's lean, simple, understandable. just what you want to get rolling.03:47
wolfspraulafter that I would probably look at build tools like koji, if I have a system with more than 512 mb ram or so03:47
wpwrakwolfspraul: but as the feature set increases, you get more and more into fighting the limitations of the distribution.03:47
wolfspraulI don't see that. A lot less than OE in my experience.03:47
wolfspraulopenwrt is simple, makefile based03:48
wolfspraulyou should like that ;-)03:48
kristianpaulHe want more libs ;)03:48
wolfsprauloe is using bitbake - overengineered imho03:48
wpwrakso far, most of the work that went into openwrt was just adding things that most other distros already have03:48
wolfspraulyou should just work with bitbake for a while :-)03:49
wolfspraulam I wrong saying that you never actually used it?03:49
wpwraki've used it in the early days of openmoko :)03:49
wolfspraulfor how long? and then you were smart enough to 'outsource' this to others, no? :-)03:49
wpwrakand then i decided that the only sane way of living in a place with OE was to __OUTSOURCE__ anything that had anything to do with the distro03:49
wpwrakvoila, and i lived happily ever after ;-)03:50
kristianpaulrafa around? :)03:50
wolfspraulopenwrt is good to start, we agree03:50
wolfspraulso we see it again for the m1 userland now03:50
wpwrakkristianpaul: he has just been feeding rms :)03:50
wolfspraulfor systems larger than 512mb or so, several gigabytes of storage, I'd try koji03:50
wolfspraulI mean 512mb memory03:50
wolfspraulfor smaller embedded systems, also those we build in the future, I'd still go with openwrt03:51
wolfspraulif it's so small that it cannot even run the Linux kernel, probably even openwrt is overkill, and you'd just have some manually written makefiles or buildroot03:51
wpwrakwolfspraul: i would still look for the outsourcing approach. see if any other distros already support LM32. if not, see if there's a happy maintainer community we can recruit for our purposes :)03:52
kristianpaullm32 dunno.. well rtems for now03:52
wolfspraulsure no way. we have to do it ourselves or nobody will.03:53
wpwrake.g., i would have loved to have closer ties between qi-hw and jlime. that would have given us OE on a silver plate03:53
wolfsprauljust look at the mess with gcc, llvm, dynamic linking, mmu, and so on03:53
kristianpaulkristofer from jlime seems interested in getting a M1 i remenber03:54
wpwrakyes, i think M1 needs some infrastructure work first. but that applies to any distro.03:54
wolfspraulhow active are oe and openwrt nowadays03:54
wpwraki don't keep track of them. for me, they just are :) oe should be big. so much depends on it nowadays.03:55
wolfspraullike what?03:55
kristianpaulwhy no body comment about Qt ? :)03:56
kristianpaulah, but no c++ suport with gcc-lm32...03:56
wolfspraulthere is too much work missing, really03:57
wolfspraulinstead of pointless debating I'd rather move some tech snippets forward, no matter how small03:57
wpwraki see the biggest problem of M1 sw support currently still in the very basics. the core system infrastructure. begins with the MMU, then toolchain, etc.03:58
wolfspraulm1 is a great start now and works as a video synthesizer, but as rejon said we need to find more core devs to take on hard problems03:58
wolfspraulwpwrak: yes fully agree03:58
wpwrakonce it looks like a regular linux system, many things get a lot easier03:59
wpwrakthen your main enemies come from the autocrap family, making more or less subtly wrong assumptions about your system03:59
kristianpaulyeah, but flicernoise still flickernoise.. how many people understand that code right now to add features as they wish?..04:00
wpwrakwell, how much of FN is really essential for what M1 does ? all the desktop etc. could be done in a few days with pretty much any halfway decent toolkit04:01
kristianpaulmtk yeah could be replaced..04:01
kristianpauljust redering engine i guess04:01
kristianpaulwich there is a hello world imho, published by lekernel recently http://lekernel.net/presentations/masteri2l/tp/mmtp_i2l.pdf04:02
kristianpaulmay be rejon can take a look at it and tell others?04:02
kristianpaulbut rendering is mostly sebastien master tehsis if i recall well04:03
kristianpaulmay be no lib?04:03
kristianpauldam i hate when i dont understand something well...04:03
wolfspraulkristianpaul: nice document04:04
kristianpaulyeah, sebastien very kind making us now about it :)04:04
kristianpaulso recalling initial talk, using internet may be a workround _if_ there is always internet acess by ethernet near to M104:06
kristianpaulor, the another idea about using video camera included on the box.. but with no decent zoom/macro i dont see it very confortable..04:07
wpwrakfor loading images ?04:08
wpwrakyou'll get full USB nearly "for free" with linux. at least the upper layer drivers. of course, as i understand it, the current USB host is only low-speed. so this would need upgrading to at least full-speed.04:09
kristianpaulno for freee04:10
wolfspraulfrom my understanding the current usb host implementation in the navre core is just hacked together to barely work with keyboard and mice04:11
kristianpaulOHCI is not complete in mm1 soc i remenber04:11
wolfspraulI think it will require substantial more work even if Linux ran on top of it04:11
kristianpaulyes, wolfspraul 04:11
wpwrak"for free" in the sense that, once you have USB working at all (host controller driver, etc.), you can talk to all the crazy little devices linux supports04:11
kristianpauloh sure :)04:11
wpwrak(hack) yeah, that's also my understanding04:12
wolfspraulkristianpaul: loading pictures over Ethernet, even if cached locally, is just a workaround, but at least it would be a start04:12
kristianpauli think that for free would aply for memcard code at least04:12
wolfspraulof course a normal user would expect to be able to plug a usb stick into the box and it should 'just work'04:12
kristianpaulwho dont those days :)04:12
wolfspraulbut even for the caching we would need to get the memcard to work, write support, etc.04:12
kristianpaulerghh, plug ?04:12
kristianpaulno more free ports !04:12
wolfspraulso adding usb-storage support may almost be less work :-) Compared to download, caching, memory card write support, etc.04:13
wolfspraulsomehow I'm hoping we can sell 50% of m1 to developers04:14
kristianpaulDocScrutinizer: http://oz9aec.net/index.php/gnu-radio/gnu-radio-blog/451-howto-receive-and-decode-noaa-apt-images-with-the-funcube-dongle-and-gqrx see what you can with satellites and 125 eur04:15
wolfspraulthat would be perfect. 50% developers, 50% users04:15
wolfspraulI mean for rc304:15
kristianpaulhow you plan call developer atention?04:16
wolfspraulno plan yet04:16
wolfspraulJon was planning to imprison them somewhere and then just force m1 upon them.04:17
wolfspraulhe is hoping to pick up some currently unemployed talent in Libya04:17
kristianpauls/pick/employ ? :-)04:18
rejonyah, they will work for cheap wolfspraul 04:18
kristianpaulDocScrutinizer: 125eur TV-like dongle :)04:19
wolfspraulI think one way to get developer attention is to go out with a really big vision of where milkymist could go04:20
wolfspraulwith every little bit that we lower barriers of entry, improve the tools, some really spectactular new products will be possible in the future04:21
wolfspraulbut that needs to come across. what products? how long will it take? how many people have to join to speed it up? etc.04:21
wolfsprauland unfortunately there is nothing really to catch up to04:21
wolfspraulI like Sebastien's idea of an autonomous flying object that can cover long distances, recharge batteries and look for resting spots on its own04:23
wolfspraulsuch a fun idea and vision, I'm thinking how I can make a little animation of it :-)04:23
wolfspraulthat thing needs an eye and/or radar to fly by vision, identify suitable resting places, etc.04:24
wolfspraulwe could probably come up with 10 or more examples of what type of products Milkymist could lead to04:24
wolfspraulin communication/radio, computer vision, medical appliances, and so on04:25
wolfspraulaudio/video equipment, embedded systems, robotics04:25
rejonwolfspraul, bradphillips from fabricatorz can make this animation04:30
rejonwolfspraul, thus is born the 2030 initiative04:31
wolfspraularchitecturally, milkymist is in a great position to make use of new chips, new hardware advances, as they come up04:31
rejonhere is what i want:04:32
wolfspraulI do believe that is true and we should talk more about it.04:32
rejon1.) be able to give presentation from ben nanonote with ubb-vga and gpdf04:32
rejon2.) give presentation from milkymist104:32
rejonin same fashion04:32
rejon3.) feed in a folder of images on m1 and be able to sell this to advertising companies04:32
wolfspraulany interesting new chip that comes out, a real hardware innovation, you have a hard time hooking it up to an unmodifiable SoC04:32
wolfspraulbut with Milkymist it's possible to hook it up faster, and come out with a product using that new chip first04:33
rejon4.) HTML5+js+css runner build of openwrt for constrained systems, that can be installed on BNN and M104:33
wolfspraulhopefully we see some opportunities and execution of that in coming years04:33
wolfspraulif I'm wrong - please correct me04:33
rejonthose 4 things i want and i see as low hanging fruit04:33
rejonas in, they are possible04:33
rejonand possible to put on point04:33
rejonyes, i know 2, 3 and 4 are all resource dependent upon 1-2 people with knowledge04:33
rejonnot impossible, but would be amazing04:34
rejonthe way lekernel added twitter support is awesome04:34
rejonthe combination of feed support + images + m1 is an advertisers dream come true04:34
rejonautomatic commercials04:34
wolfspraulwe need to ask him for the lowest hanging fruits from his perspective. he knows best, I'm just learning, so are others (kristianpaul)04:34
rejonmaybe people outside of asia don't see this04:34
rejonthere are screens everywhere!04:34
rejonflatscreens, LED screens, screens on the headrests of taxi site passenger seats, and just empty TVs at peoples home04:35
rejonall waiting for an m1 to be plugged in04:35
rejonanother thing the m1 can be is a set top box04:35
rejonthere are loads of people installing mythTV04:35
rejonor other solutions04:35
kristianpaul(screens) hopefully not here ;)04:35
rejonm1 could be playing all these media files 04:35
rejonso many things possible04:35
wolfspraulI just have this thought you saying "empty tvs at peoples homes" = "unused ad space" :-)04:36
rejonin the elevators and in the lobbies of housing complexes here in china, all have tv ads04:37
rejonthe ad companies give a cut to the homeowners04:37
kristianpaulwolfspraul: yes, learn learn :-) video synthesis with milkymist !:)04:37
wpwrak1) for #slides == 1, that's already possible ;-) what i really need to do, though, is build the "productized" ubb-vga04:37
kristianpaullekernel: http://milkymist.org/wiki/index.php?title=Interactive_visuals_workshop_with_Milkymist_One_and_Arduino planning to do soemthing about adding video syntheis features?04:38
wpwraki'm a bit disappointed that exactly 0 people have build an ubb-vga :-(04:38
Action: kristianpaul hides04:38
wolfspraulI'm not surprised. the scene we are in, and the way we communicate, is about someone hacking something for themselves, and then showing off04:39
wolfspraulnobody wants to rebuild what someone else did04:39
wolfspraulhero stuckness essentially :-)04:39
wolfsprauleverybody wants to be a hero, everybody wants to show off what great thing THEY built04:39
wpwrakkristianpaul: you'd be amazed how good it looks. when i tried it at FISL, i got it to display flawlessly at 1024x768 on pretty much the first try on two screens (well, one screen at a booth, and on rejon's projector)04:39
wpwrakkristianpaul: the image looked much better than on my old monitor at home :)04:40
wolfspraulwe need to break out of that hero stuckness, and rejon is working on it with sharism etc.04:40
wpwrakhmm, dunno if that hero fixation is a cause or a consequence04:41
kristianpaulI like rebuild just i just  havent tought abother ubb-vga aplication besides slides, wich is not my main concern right now04:41
wolfspraulso far we know that we have the technology under control, or rather which part we have under control and which part not. That's a good starting point.04:41
kristianpaulbut i have a talk in two months so :-)04:41
rejonwpwrak, i might build a cable, wolfspraul made me buy a ubb-vga off him so i could use it to make a calbe04:42
rejonis that software in the default BNN image?04:42
wolfspraulyes rejon bought a UBB from me, but he did a lot of price haggling04:42
wpwrakwolfspraul: how many beers did he have to pay in the end ? :)04:43
wolfspraulcash + beer04:43
rejoni think i bought him beer actually04:43
wolfspraul+ taxi04:43
rejonof course, my time is not valuable to wolfspraul 04:43
wpwrakrejon: (default image) naw, it's way too experimental for that04:43
wolfspraulI'm listening. rejon - I will send the 'dudeloop' licensing fees to you from now on, whenever I use it.04:43
wpwrakrejon: the high art of manipulation :)04:43
rejonyah, if i believed in licensing i'd be rich 04:44
wpwrak(ubb-vga) it should compile nicely if you have even only the minimum toolchain set up, though. no fancy libraries required.04:46
wpwrak(or i can just make a binary)04:46
kristianpaulthe image need specia preparation?04:47
wolfspraullunch time, taking my notebook so I can work WHILE I'M EATING and make up the time I need for rejon tonight04:47
wolfspraulhow's that?04:47
kristianpaulthas you and rejon now ;)04:47
wpwrakwolfspraul: you leave the office for eating ?!?04:47
rejoni have a maid sorry04:47
rejonwolfspraul, 04:47
wpwrakwolfspraul: next thing you'll tell us that you sleep elsewhere, too04:47
rejonshe feeds me04:47
rejonwhile i wwork04:48
kristianpaulwpwrak: same i asked my self soe seconds ago ;)04:48
rejonok, i will push on those 4 things i want from the milkyqiverse04:48
rejonworking now while you guys all talk all day04:48
wolfspraulkristianpaul: we are a-qi-vers!04:49
rejonwe should rename qi-hardware #dudeloop04:49
wolfspraullunch, work, l804:49
wpwrak(ubb-vga, images) right now, i can load PBM of exactly the same resolution the screen has04:49
wpwrakrejon: what's dudeloop ?04:49
rejona loop of dudes04:50
rejona circle of dudes04:50
rejonit must be broken04:50
kristianpaulhum so not hard to make a slide04:50
rejonits a blackhole of dudeliness04:50
rejonbroken with girls04:50
rejonlinux conferences are mostly dudeloops04:50
wpwrakaah ! good idea :)04:50
wpwrakyeah, and try to pick up one of those few girls you find at the conferences :)04:52
rejonwell you know a movement is over when no girls are there04:53
wpwrakor maybe it's still starting :)04:54
kristianpauloh i can write my own libc, interesting :-)05:00
wpwraki hear the alarm bells ring at glibc headquarters :)05:01
wpwrakkinda like then linus discovered he could write his own distributed revision control system. shortly thereafter, the wind of time blew over the scorched ruins of bitkepper ...05:02
wpwrakof course, he has a history of destroying lesser projects. e.g., coherent unix never knew what hit them when linux came out ...05:03
Action: kristianpaul take a look to lekernel thesis and datasheets again05:10
qi-bot[commit] Werner Almesberger: gp2rml/: copied over from svn.openmoko.org/developers/werner/cncmap/gp2rml (master) http://qi-hw.com/p/cae-tools/fbbe45106:39
qi-bot[commit] Werner Almesberger: mkmk-simple: changed path to gp2rml from ...openmoko... to ...qi/cae-tools... (master) http://qi-hw.com/p/cae-tools/50918a206:39
wpwrakgrmbl. didn't set the board thickness. no surprise it didn't cut through the 1.6 mm board using the 0.8 mm default ...07:24
qi-bot[commit] Werner Almesberger: labsw/: rotate P5 (opto out) (master) http://qi-hw.com/p/wernermisc/8ece4f407:52
qi-bot[commit] Werner Almesberger: labsw/: changed spacers to have a 4 mm hole, with 0.2 mm of slack (master) http://qi-hw.com/p/wernermisc/a336e1d07:52
qi-bot[commit] Werner Almesberger: labsw/cam/Makefile: improved indentation; added BOARD_Z and ANY (master) http://qi-hw.com/p/wernermisc/0d6568c07:52
qi-bot[commit] Werner Almesberger: labsw/: reduced depth of "bays" for columns from 11 mm to 10.4 mm (master) http://qi-hw.com/p/wernermisc/ffaef3b07:52
qi-bot[commit] Werner Almesberger: labsw/: reduced width of "bays" for columns from 8.4 mm to 8 mm (master) http://qi-hw.com/p/wernermisc/02f7bd907:52
qi-bot[commit] Jiri Brozovsky: Console viewer for Git repositories (master) http://qi-hw.com/p/openwrt-packages/974541507:52
qi-botThe build has FAILED, see log here: http://fidelio.qi-hardware.com/~xiangfu/compile-log/openwrt-xburst.full_system-08312011-2007/08:03
wpwrak*hmm* seems that the moment of truth has come. the moment i have to settle for a 2D CAD that actually is a CAD (and not just a perl, python, etc. script)08:06
Action: wpwrak goes on teaching fped a few more tricks08:06
DocScrutinizerlekernel: nice sat pics08:16
DocScrutinizerthe method is known for long, but the dongle is really rather low-priced compared to what you had to pay in the 1980s08:20
DocScrutinizerIIRC back then they reworked fax machines to do the decoding08:22
lekernelimplementing support for still images in the ic design? wtf?09:58
lekernelthis is only a software task09:58
lekernelpictures can be uploaded with FTP09:58
lekernelrejon, commits aren't the only thing that matters in a project, popularity/publicity (or the lack thereof) are important as well10:00
rejon_lekernel, FTP is not good for normal people10:01
rejon_lekernel, sure we can send out loads of publicity10:01
rejon_but writers will write want they want10:01
rejon_not as important as quality connections and time spent with actual people10:02
wolfsprauloops, sorry, I got that wrong with 'ic design'. need to study the sources more carefully.10:02
lekerneland 99% of the codebase of milkymist is totally unknown to most people here, so have a look at the existing stuff before asking for more10:02
rejon_lekernel, we should get that 99% changed10:06
rejon_so more people can participate10:06
rejon_that would be really great for the health of the project10:06
rejon_got a hit on the betahaus event lekernel 10:07
rejon_emailing you now10:07
rejon_lekernel, were you talking to anyone at betahaus about that event?10:08
lekernelMadeleine and Eva10:11
lekernelthe next step is to write the workshop description10:11
lekernel(and have M1s ready for sale)10:12
rejon_sure, that is super close on the sale side10:14
rejon_sure lekernel 10:14
rejon_ok cool10:18
rejon_madeleine is cofounder of betahaus and eva is intern10:18
rejon_just checking, my friend jay is organizing the makerplatz during that10:18
rejon_seeing how connects before final push out10:18
rejon_also want to tie in with sharism plan10:18
rejon_ok emailed10:19
rejon_lekernel, i have three quetsions10:20
rejon_1.) how will your CERN work tie into the m110:20
rejon_how does it, so when talking with people we have an idea10:20
rejon_2.) how can SVGA/720p get onto m1 soon, and/or how much work will this take because people already asking for it constantly10:20
rejon_3.) how to get images simply into m1 so they can be mixed with10:20
Aylalekernel, you work at CERN?10:24
jiri_brozovskywprak: I'm still interested to build a copy of your ubb-vga (even got a buchn of ubbs for that). But by experience is limited and it looks like too compicated for start :-(10:26
rejon_we doing an makerlab next week in shanghai...10:32
rejon_ooh too bad, they doing an fpga workshop soon after10:33
lekernel#3 is relatively easy, just use libpng/libjpeg to decode the images, and inject into the renderer loop. the only delicate point is how you configure the images from the patches.10:50
lekernel#2 is technically difficult and will probably require doubling SDRAM frequency, which I do not feel motivated to do while the units are not being sold and the project remains so unpopular10:51
lekernel#1 it's not related other than being other open source electronics stuff and bringing new opportunities to meet people who are not afraid of advanced tech like FPGAs, contrary - i'm sorry to say - to most hackers10:53
lekernelin case you are worried about my commitment to making M1 successful, it's consulting, not employment. I'm totally free to continue working on M1.11:00
wpwraklekernel: naw, i don't think people are really afraid. most just don't have an itch that they think an FPGA will scratch. at the end of the day, that's what really motivates people.11:10
lekernelso, in that case, I can say they lack the motivation to scratch the itch of understanding the vast field of FPGA-related technologies11:11
lekernelwhich is equivalent11:11
wpwraklekernel: a few "getting started" examples (blink the LEDs, whatever) should allow people to get rolling. of course, if they could use use a free toolchain instead of the proprietary xilinx tools, that may make the whole thing more attractive. at least for people coming from the free software corner.11:11
kristianpaulwpwrak: there is a blink led example in llhdl btw .)11:12
wpwrakkristianpaul: i would have been a bit disappointed if there hadn't been one ;-)11:12
lekernelhaha indeed, and it even works11:13
wpwraklekernel: but you don't ask people to be driven by the prospect of mastering a complex and obscure technology. it's just normal that you'll have to wait fairly long for your first followers with this approach.11:14
wpwrakyou have to sneak in the FPGA world a little more subtly :)11:14
wpwraklike "oh, this would be really easy to do with an FPGA". like this [high-level picture of the concept]. here's roughly what the code would look like."11:16
kristianpaulhave some ideas about what _you_ want to do in a FPGA?11:16
wpwrakthat way, you've 1) connected it to the problem they're trying to solve, leveraging their willpower to tackle that problem. 2) you explained what can be done in terms they can readily understand without even learning anything about the arcane magic of FPGAs. 3) shown them that it's something that mere mortals (like them) can actually understand within a reasonable amount of time.11:17
wpwrak3) is important, because many people shy away from the unknown. even if they would have no trouble with it.11:18
wpwrake.g., i wrote my first compiler only after i read an article series in a computer magazine about how to write a simple compiler. that gave me a starting point.11:19
wpwrakkristianpaul: me ? my scholarly interest is in enhancing computation. basically make the compiler figure out what hardware feature/structure could solve a particular part of the computation the program requires, then synthesize it ad hoc.11:20
kristianpaulyummy :)11:22
wpwrakkristianpaul: that's of course quite basic research with a very long time horizon. maybe ~5 years before it's actually useful for more than isolated demo cases. perhaps a whole decade before it's something you'd want to consider a tool like lex or yacc.11:22
wpwrakand maybe one more decade before your compiler just does it without you even noticing :)11:22
kristianpaulgotta go read you later11:24
wpwrakover such a time span, technology will also change. fpga in ten years will be different from fpgas today. not just bigger and faster. so it's not only a question of developing a branch of science that works with today's zoo of FPGAs, but also to extract more fundamental structural elements, patterns, and trends, and to extrapolate from these.11:24
rohwpwrak: well.. if you think.. when i compare 10 years ago with now, not much changed. especially the tools seem to have stood still.11:30
rohso.. i am optimistic about a lot of things, but not any progess there.11:30
wpwrakroh: oh yes, development is super-slow at the moment. there's no synergy. and all the interesting bits and pieces are kept a secret. but when you start breaking down the walls, the critters who like to dwell in obscurity will get restless :)11:31
wpwrakroh: also, what you do see is the addition of more/larger fixed-function blocks.11:32
rohheh.. well.. the latter is quite boring11:32
rohi'd like to see some work on cpu/fpga combo chips.. they atleast have a serious chance to perform before becoming a asic11:33
wpwrakit's something you have to consider. how to incorporate them into your model of the device and how to use them efficiently.11:33
wpwrakyes, cpu+fpga combos are another dimension11:33
rohwpwrak: e.g. some 800mhz or more arm with a fpga to develop the soc around it.11:34
rohbut wasting my time with more <400mhz devices is out of the question for me.11:34
rohalteast nowadays.. was different 5 years ago, but one has to be realistic11:35
lekernelanother fact that show how much hackers suck at computer architecture is their strong belief in the MHz myth11:37
wpwraklekernel: ;-))))11:37
lekernelclock frequency is one performance factor among tens of thousands11:39
wpwrakand one where you have the fewest possibilities to improve things. well, unless you happen to have a fab in your backyard and enjoy some pretty hardcore fundamental research :)11:40
lekerneleven after years the super-conservative (real mode etc.) Intel has started making multicore chips, they still don't get it11:41
wpwrakroh: let the dinosaurs have the MHz. let us quick and smart mammals have parallelism :)11:41
wpwraklekernel: well multicore is at least one timid step in the right direction :)11:42
wpwrakjiri_brozovsky: (ubb-vga) great ! and don't worry about the complexity ! it's not all that hard. it looks much weirder than it really is :)11:50
jiri_brozovskywpwrak: I have 10 UBB boards. So hopefully I will be able to build ubb-vga at the end... 12:16
wpwrakjiri_brozovsky: yeah, you'll probably waste 1-2 before you get it right. but they're quite robust, so don't worry too much :)12:20
wpwrakrejon__: i think your spiritual guidance is required in the logo thread12:33
wpwrakrejon__: you need to declare a winner for the basic shape. or propose changes. all we strive for is to approximate the splendor of your original design, yet it would be blasphemous to even think we could decide among ourselves which is best :)12:35
wpwrak(let's see if groveling gets this done :)12:35
tick_ratewpwrak: Just joined, but have been lurking for the last year. I built the ubb-vga weeks ago on a breadboard. Just diplayed a 640x480 shot on a dell.13:05
wpwraktick_rate: wheee !!!! great !13:08
tick_rateI also have the atben to atben working. (yep I have 2 bens) I tried messing with the atusb code and try to help, but its not going well.13:09
wpwrak(atben to atben) nice !13:12
wpwrakwhere did you run into problems with the atusb code ?13:13
qi-bot[commit] Werner Almesberger: labsw/: oops, forgot to commit base.fig (master) http://qi-hw.com/p/wernermisc/2488cc313:17
qi-bot[commit] Werner Almesberger: renamed labsw/cam/ to labsw/pcb/ (master) http://qi-hw.com/p/wernermisc/90a945313:17
qi-bot[commit] Werner Almesberger: labsw/me/front.fpd: first CAD version of the front panel (master) http://qi-hw.com/p/wernermisc/b1cd1d413:17
qi-bot[commit] Werner Almesberger: labsw/: moved base and front Xfig drawings to draft/ (master) http://qi-hw.com/p/wernermisc/e3322a513:17
qi-bot[commit] Werner Almesberger: labsw/: increased drill hole in 200 mil headers from 1.0 mm to 1.1 mm (master) http://qi-hw.com/p/wernermisc/e9be8ab13:26
qi-bot[commit] Werner Almesberger: labsw/: renamed me/ (Mechanical Engineering) to mech/ (master) http://qi-hw.com/p/wernermisc/db7b3c413:26
qi-bot[commit] Werner Almesberger: labsw/: changed 90 deg front corner to 45 deg; added date, author, etc. text (master) http://qi-hw.com/p/wernermisc/ab8986313:41
tick_rateinserting the device works fine, and I can ping (same problem with DUPS). When I remove it it locks the whole PC.13:48
Action: wpwrak notices approvingly the dwindling stock of pertinax boards. soon, i'll be FR4 only. well, may have to restock on single-sided ones. they're quite handy for mounting things for the mill.13:49
wpwraktick_rate: oh, i actually never tried to hot-unplug ;-))13:49
tick_ratei tried looking in to the code (even bought a kernel book). I guess its beyond me at the moment.13:49
wpwrakbut yes, i think i can see a number of things that would go terribly wrong ... :)13:49
wpwrakhehe :) yeah, atusb isn't the friendliest code ... does a few dirty things. the new version will be nicer - i'll get rid of the spi layer and just be an ieee 802.15.4 device. that way, some of the more awkward bits can go.13:51
rohlekernel: i am well aware of parallelism vs mhz. but as long as we dont have the whole sw world switch to sw arch which can really gain from that we still need fast cpus13:55
tick_ratethis weekend, i hope to set up the ben-ubb->arduino. That'll be cool. Has anyone messed with serial wombat and ben?13:57
wpwraktick_rate: probably not ... looks as if it's been obsoleted by arduino in recent years14:04
wpwraklekernel: obsoleted by arduino, that's a what i call a cruel fate ! ;-)14:05
wpwraklekernel: that "serial wombat": http://www.serialwombat.com/14:05
lekernelthere's a whole bunch of crap (the basic stamp for example, if you remember it) that I'm happy seeing being obsoleted by arduino :)14:06
tick_ratei had one lying around. (build a ship in a bottle) nothing useful, just to say 'Look'. then move on.14:06
lekerneland the arduino programming environment is better than mplab in about every way14:07
wpwraklekernel: ah yes, the basic stamp ... i don't actually know much about it, but the whole idea reeks of evil14:07
lekernelit was all over some French electronics magazines in the late 90s14:08
wpwrak"tool index 11 too large (> 10)" grmbl. now what's that supposed to mean. why of why do my programs have to throw so cryptic error messages ?14:09
lekernelthis kind of junk http://www.elmelectronics.com/ebench.html#ELM460 (and all other ELM Electronics products) is also wonderfully obsoleted by arduino14:10
wpwrakah, clever kicad issued the same tool multiple times. very smart.14:11
wpwraklekernel: phew. good riddance ;-)14:11
wpwraklekernel: did they make actual chips ? or are those SMT circuits on some DIL base ?14:13
tick_ratelekernel: been thru those, yep arduino eclipses that stuff.14:13
lekernelwpwrak, sure not, all the ELM stuff is pre-programmed PIC microcontrollers14:16
lekernelthey just erase the marking and put their sticker on it14:16
kristianpaul(MHz myth) oh yeah.. people always laught on me when i said i want to focus all my work and learning about FPGA, if no hihg Mhz nothing great can be achieved quoting then..14:18
lekernelkristianpaul, it just becomes more difficult. you can still quote math operations per second in many cases...14:19
kristianpaul(basic stamp) he, i rememnber thats when i decided learn some C and drop microchip asm ;)14:20
lekernelbut in some cases (e.g. on a system using speculative execution, or with a memory-bound process) this metric becomes insufficient too14:20
tick_ratehas anyone looked in to this: http://pocketworkstation.org/fbvnc.html for the ben. I think it was for the W100 display only on the Z. It worked great on the 5500 and C860. 14:40
wpwraklekernel: (ELM = re-branded PIC) oh, clever :)14:50
lekernelactually, some programmable chip companies (including xilinx) let you change the markings on the chips, add your logo, etc. most probably subject to a large MOQ, though.14:54
qi-bot[commit] Werner Almesberger: gp2rml/gp2rml.c: allow unit of Z clearance (mm) to be specified explictly (master) http://qi-hw.com/p/cae-tools/00166f315:04
qi-bot[commit] Werner Almesberger: mkmk-simple: new option CLEARANCE (default: 2mm) (master) http://qi-hw.com/p/cae-tools/5651a4015:04
qi-bot[commit] Werner Almesberger: mkmk-simple: exit with an error code if "gp2rml" fails (master) http://qi-hw.com/p/cae-tools/98b0e9615:04
qi-bot[commit] Werner Almesberger: cameo/excellon.c: increased MAX_TOOL from 10 to 20 (master) http://qi-hw.com/p/cae-tools/41bffa815:04
qi-bot[commit] Werner Almesberger: labsw/pcb/Makefile: increased clearance from 2 mm to 3 mm (master) http://qi-hw.com/p/wernermisc/413780315:04
wpwraklekernel: and for smaller numbers, they just file the markings off. of drop a bit of epoxy on the whole thing15:05
qi-botThe build has FAILED, see log here: http://fidelio.qi-hardware.com/~xiangfu/compile-log/openwrt-xburst.full_system-09012011-1213/20:39
wpwrakmaybe one could write an automatic bug fixer. serialize the whole source, change a random byte, determine how far the build goes, apply a genetic algorithm with the time-to-error as the "fitness" factor. repeat until it all compiles cleanly :)20:57
jow_laptopsounds like autoconf debugging21:09
DocScrutinizersounds like core wars v3.021:26
wpwrakoh dear ;-)21:27
DocScrutinizerhttps://estore.ti.com/MSP-WDS430BT2000D-Bluetooth-Wearable-Watch-development-system-with-Digital-display-P2447.aspx   http://www.metawatch.org/21:28
DocScrutinizerfor brain food21:29
mthwpwrak: an automated "git bisect" on a failed build could work21:54
mthalthough with openwrt it's probably clear most of the time since it's highly modular21:54
mthand there won't be many commits on the same area typically21:55
kristianpaulDocScrutinizer: that watch is expensive, i prefer wear my nanonote ;)22:18
viricDocScrutinizer: until yesterday, it was half price I think22:19
viricDocScrutinizer: a friend of mine has one.22:19
DocScrutinizererrr, it's not even selling yet22:20
kristianpaulshame this version dobnt have 433Mhz tranceiver, i think early ones had it..22:20
DocScrutinizerthat's another one22:20
DocScrutinizerwhich is like 50 bucks iirc22:20
viricah, bluetooth? no no. I meant one not bluetooth22:21
DocScrutinizeralso wasn't a water resistant model, up to 30m22:21
kristianpauland BIG? :_)22:21
viriceZ430 https://estore.ti.com/eZ430-Chronos-433-eZ430-Chronos-Wireless-Watch-Development-Tool-P1734.aspx22:21
kristianpaulyup, that one22:22
viricand someone ported gcc, he said22:22
DocScrutinizerviric: beat me on it22:22
viricI don't know what that means :)22:23
DocScrutinizeryou were faster than me22:23
viricI wouldn't know what to do with it though22:24
DocScrutinizermeh, the ez430 has a 96(?) segment display22:25
DocScrutinizerand a platic housing with a steel bottom that's screwed on with 4 philips head screws - or sth like that22:26
viricyou can encode numbers up to 2^96 values22:26
DocScrutinizermeta watch is a *real* watch22:27
viricthat friend tells me that the belt causes sweat. And it cannot be replaced by another you like more.22:27
DocScrutinizerwell worh the price tag22:27
viric'meta watch'?22:27
DocScrutinizerMSP-WDS430BT2000D-Bluetooth-Wearable-Watch-development-system-with-Digital-display-P2447.aspx 22:27
viric 52>;NF8O!22:28
DocScrutinizernfc :-D22:28
viricIs it Texas Instruments using Communist propaganda kind of advertisement?22:29
viricThere is quite an upside-down of values in this world22:30
DocScrutinizerthat's Fossil I guess, not TI22:33
viricclose enough though22:34
DocScrutinizeranyway the electronic hardware design is rather intriguing22:37
DocScrutinizerhttp://www.metawatch.org/metawatch_system_overview.pdf  http://www.metawatch.org/metawatch_host_message_interface.pdf22:38
kristianpaulalso  i wonder where are those bluetooth acess point for getting update weather forecast..22:38
kristianpaulhum, looks like a accesory/companion for a mobilephone..22:40
kristianpaulbah 22:41
kristianpauli dont own fancy phones..22:42
viricor fancy watches22:42
kristianpaulbut is wearable22:45
DocScrutinizerit's running freeRTOS or sth like that22:47
DocScrutinizerand can run your custom code as one task besides the core tasks22:48
DocScrutinizerI only wish they had thrown in some xxMB of flash22:50
DocScrutinizere.g. to record some hours of data from BT-cardio-monitor and the builton accelerometer22:51
DocScrutinizerit still makes a great watch22:52
DocScrutinizereven without tweaking it22:53
DocScrutinizerand just the builtin vibra-alarm used as an extension to the phone's ringtone and vibra is priceless22:54
kristianpaul(alarm) if is loud i buy it! :)23:01
wpwrakDocScrutinizer: this one is a lot more subtle: https://estore.ti.com/MSP-WDS430BT1000AD-Bluetooth-Wearable-Watch-development-system-with-Analog-Digital-display-P2446.aspx#23:02
wpwrakbut i have to wonder what madmad designed those buttons23:03
DocScrutinizerhmm the analog-digital version has oled which is supposed to be normally-off23:28
DocScrutinizerand the amount of info you could cram there is not really making it look worthwhile23:28
--- Fri Sep 2 201100:00

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