#qi-hardware IRC log for Saturday, 2010-08-14

rohwolfspraul i like the open-ness00:43
rohi dont believe in fpga stuff at all. too expensive for ce-stuff. but of course a required evil (including the toolchain) on the way to custom asic(s)00:44
rohif you want  some constructive critic... less foo, more bar. small steps, but one people can buy00:45
rohfoo meaning 'endless discussions with people who have not tried $something' and bar meaning all the little cool working hacks which get done and published.00:46
rohdoes that make sense?00:49
rohwolfspraul do you know thingieverse? i'd love to have that for hardware. pcb, schem, sw/links to sw and for example some 'shop' integration for people who want to buy parts, kits or even assembled devices of that. (the shop part could be some pcb-manuf, sponsoring agreement, or make the pcbs cheaper for stuff in the shop or so.01:05
rohand i need octoparts for europe and asia!01:06
rohdoes anybody know the guys of octoparts?01:07
wolfspraulroh: ok totally agree with you, I try 'less foo more bar' a lot01:09
wolfspraulI also agree on the fpga price issue, it's a necessary evil for a while. I do like Werner's repeated attempts to install a micro-fpga somewhere, that could help moving forward step by step.01:10
wolfspraulso we see maybe we need that somewhere in fpga land, or to do sdio2whatever bridges01:10
wolfspraulI meant 'somewhere in rf land'01:10
wolfspraulnot wasting time on theoretical discussion - YES MAN I hear you :-)01:10
wolfspraulI don't know thingieverse need to check, and I don't know the octopart guys either although I like the site01:11
rohhaving thousands of chinese and other asian traders feed their 'sourcing information' into it could be interresting.01:17
rohthingieverse is like google code in very simple for physical things01:18
rohhttp://www.thingiverse.com/raumfahrtagentur for example01:18
rohthe whole physical-thing-rapid-prototyping community is even more trial and error than the sw one i think ;)01:20
wolfspraulnice, I will check it out01:25
wolfspraulneed to run for the milkymist case meeting now...01:25
wolfspraulalready late...01:25
wpwrakroh: ah, which part of ieee 802.15.4 is too complex ? i still haven't quite decided whether i like it, but at least the chips are fairly decent. more evolved than any of the sub-ghz stuff i looked at.02:03
wpwrakroh: at the same time, you can't dumb them down quite as much. but then, these features may be somewhat hypothetical anyway.02:04
wpwrakwolfspraul in absentia, roh: for the record, i'm not trying to push cpld/fpgas. i just recognize them as a necessary evil if taking certain paths. there are actually "open" cplds out there, i.e., in cypress' psoc3 chips. but they have certain drawbacks as well.02:09
rohwpwrak what features? the 'mesh' foobar?02:26
alcythis thing's real cool, for normal suers as well. even it has its own distro-hopping stage. So from openwrt->debian, now running jlime, and its brilliant.02:54
tuxbrain2rafa , kristianpaul, zear and any other jlime guru out there... how change volume on GMU?05:37
bartbesthe volume buttons don't work?05:37
bartbestry the about screen05:38
wejptuxbrain2, if the volume buttons do not work, they probably have been mapped to different key codes in jlime05:38
bartbesiirc it listed the controls05:38
zeartuxbrain2, no idea :D I don't use gmu in jlime05:38
wejpyou can change the keycode mappings for gmu without recompiling, there is a config file for that, but of course you need to know which codes they use now05:38
bartbestuxbrain2: wejp is probably right, so if you manage to find the keycodes you can05:38
bartbesaww ninjad05:38
tuxbrain2any easy way to know the keykodes?05:39
zearvolume keys in jlime are mapped as page up/page down05:39
bartbesoh that sounds useful05:39
wejpthen the keycodes should be 280 and 28105:40
tuxbrain2I will try ...05:41
wejpthe file is gmuinput.nanonote.conf05:41
tuxbrain2just to make sure actully the gmuinput.nanonote.conf have this:05:43
tuxbrain2I don't know what does 73 and 7405:43
wejpit probably doesn't do anything in your case05:43
wejpyou can ignoe those05:44
tuxbrain2can I repeat keykodes then?05:44
wejpdo not repeat them but instead better set the values for PageUp and PageDown to something else, otherwise those would be picked first if the keycode occurs, which would result in no action most likely05:45
tuxbrain2whatever it works now :)05:45
wejpor remove thoise lines, but then you need to renumber the following lines accordingly05:45
wejpah, cool :)05:45
tuxbrain2repeating keycodes05:45
tuxbrain2so Discography part of Nanowar True Edition finised :)05:46
tuxbrain2no let's pimp something els05:46
tuxbrain2damn now let's pimp somthing else05:47
bartbesugh.. the one time I compile without V=99 it errors.. of course05:59
tuxbrain2wejp: When a song doesn't have a *.txt of lyrics gmu shows the first txt it founds, in this case the lyrics of the first song, this is a desired behaviour or a bug?06:13
bartbesI believe that acts exactly like the default settings say it should06:14
wejpdesired behaviour. actually in gmu's config file you can define a list of patterns gmu should search for and this is what i had predefined06:14
wejpyou can change that to whatever you like06:14
bartbesyou can change that behaviour in ~/.gmu.conf (or whatever it was called)06:14
tuxbrain2ok :)06:14
bartbesoh heh06:15
tuxbrain2I think I will put a txt with the url of the band instead :)06:15
tuxbrain2in a 00_Readme.txt06:16
wejpthat's fine06:16
bartbesare any of you experienced in working with lua on the nn?06:21
wpwrakroh: i don't think 802.15.4 has a mesh. that's what things like xigbee add on top. ieee 802.15.4 is only phy and mac. (haven't read the standard yet, though)07:38
wpwrakroh: i mean simpler features, like clock recovery and framing. some of the sub-ghz chips let you turn off all this, and you get just a bit stream or even a continuous-time digital signal07:41
wpwrakroh: ieee 82.15.4 chips go as far as offering auto-ack and auto-arq`07:42
wpwrakwolfspraul: (cpld) i'm not promoting them. i'm just telling you they can be the consequence of your actions :-)08:22
wpwrakwolfspraul: cpld = complex programmable logic device ("complex" as opposed to the "standard logic", 4 XOR gates in a chip or so)08:23
wolfspraulsure I read it in the logs. No worries I got your point long ago.08:23
wolfspraullet's see where ben-wpan takes you, then we go from there...08:23
wolfspraulbtw - I was really looking for pervertix to add to the commitlog until I realized you made a joke08:23
wolfspraulsounds like a project name you would actually choose... :-)08:24
wpwraki also think that, given that at least half of the hw side of the project is already deep into fpgas, the pain of having a cpld - in terms of development effort - wouldn't be too great08:24
wpwrakcplds are basically small fpgas. you use the same languages and tools to program tham.08:24
wpwrak(project name) after all, i made #3 on google for "devirginator" ;-)08:25
bartbeswolfspraul: are you sure wpwrak isn't sponsored by some company specializing in logic devices? :P08:26
wpwraki have to admit that i expected the word to be more common, although not quite in IT :)08:26
wpwrakbartbes: it's actually the Reformed Vogons. they've abandoned the old way of blasting planets away and not prefer if the destructive force comes from within. after they saw the Terminator documentary, they were sold on the idea of aggressively advancing technological progress on earth, to the aforementioned end.08:27
lekernelwhat do you need a cpld for?08:28
wpwraklekernel: convert the multi-megabit Q/I bitstream from an RF frontend to something the xburst can receive in a halfway efficient way08:29
wolfspraulwpwrak: I took you out of the sige loop at some point, things are moving there now. a 4162 evb is on the way to adam, who will do a bit of study, look into the antenna etc, then send this all on to cristian paul08:29
wolfspraulI am planning a sige visit next time I am in or hear HK, and they have already confirmed08:30
lekernelfor what? wifi?08:30
wpwraklekernel: gps08:30
lekernela lot of RF modulations are super compute intensive and will blow the xburst08:30
lekernelhm, don't know so much about gps :p08:30
lekernelbut i'd still put all the gps processing into some fpga... a multi-mbit dsp stream will definitely load the cpu quite heavily08:31
lekernelbtw who's going to work on that thing?08:32
rohwpwrak what about the ti wifi stuff?08:32
wpwraklekernel: it's been claimed that gps should be within the reach of what the xburst can do. we'll see if this is true :) i don't know much about the digital signal processing side of it, but i can see the pain you'd have with the raw bit stream. particularly since the xburst can't do spi slave.08:32
rohwpwrak the one the n900 uses08:32
wpwrakroh: the one only intended for large customers ?08:32
lekernelmaybe it can do it, but if it uses 90% of the CPU time it's still not good08:32
lekernelwho's going to write all the DSP stuff?08:32
wpwrakwolfspraul: sounds good08:33
wolfspraullekernel: there seems to be a number of projects out there already, not sure how reusable etc. but they exist.08:33
rohlekernel fpga are expensive. dsp eat batteries for breakfast08:33
wpwraklekernel: wolfgnag's idea is to have a dual-core cpu next, and to use one core to "play"08:33
wolfspraulanother project claims to release sources 'soon', although I tend to leave my expectations at zero in such cases08:34
lekernelroh, if you do the IQ signal processing on a xburst, I still call that a DSP08:34
wolfspraulcristian paul will work on it08:34
wpwrakroh: it will burn battery power like there's no tomorrow for sure :)08:34
Action: roh does not want to use sdr for data communication. nice for a develboard. worthless on a real world device08:34
rohlekernel still, its stupid.08:34
lekernelsdr? worthless?08:34
wpwrakroh: i guess it's just a question of time. think of it as research.08:35
lekernelif done right SDR kicks ass08:35
rohlekernel for real world things like 'ip communication' sdr is a studid idea. yes.08:35
wpwrakroh: by the time it's actually debugged, maybe there is a chip that does sdr in a reasonably efficient way08:35
rohsdr ALWAYS eats more power than a regular approach.08:35
lekernelwhat do you call a "regular" approach?08:35
lekerneldiode and capacitor?08:36
rohdepends on the modulation.08:36
lekernelthanks to SDR you can implement complex and very efficient modulations08:36
wpwrakroh: perhaps. but the absolute value will go down. at some point you simply don't care whether you burn 1 nW or 100 nW :-)08:36
lekernelthen you need less transmitting power and in some cases less RF bandwidth08:36
lekernelboth are cool: less overall power consumption, less RF pollution08:36
rohlekernel come down from your theoretical island. in realworld that all doesnt matter.08:36
lekerneloh yeah? what about digital TV then?08:37
rohlekernel no sdr there.08:37
rohlekernel same on gsm.08:38
wpwraklekernel: good point. you can use adaptive modulation/demodulation methods. particularly demodulation could be interesting to better compensate for distortions.08:38
lekernelwell, ok, maybe a lot of GSM and digital TV chips aren't that reconfigurable that you would call them "software"08:38
lekernelbut it's still the same idea08:38
rohyes, these demods do use i and q and a dsp. but they also do clock-syncing and not use a free running adc.08:38
rohno free running adc -> no sdr.08:39
rohonly math-assisted demod of synced samples.08:39
wpwrakroh: also, don't forget that SDR for GPS+foo+bar+whatever always uses the same number of chips. with the one-chip-per-radio approach, you quickly have half a board of them.08:39
rohactually most of these demods do not even have any possibility to upload firmware (anymore). they come mask-programmed and have a but of overlay-ram to upload 'patches'08:40
rohwpwrak sdr doesnt work with the help of hw-mixer-hw and filters.08:40
wpwrakroh: well, see the wbx :)08:41
rohwpwrak so a 'allband-tranciever' is pure fantasy. the usrp (gnuradio) also needs hw-extentions for every frequency range08:41
wpwrakroh: make it a bit wider and you're there :)08:41
rohyou can buy a usrp for a few bucks, but can buy half a car's price in frontends. (and these are not even quite good ones)08:41
wpwrakroh: 50-2000 MHz doesn't sound too shabby to me08:41
wpwrakroh: naw, the usrp is still more expensive than all the useful frontends together08:43
bartbeswolfspraul: I'm currently stuck with what seems to be a lua bug, but I have a vid you might want to see08:43
qi-bot[commit] Andres Calderon: eth-phy placement http://qi-hw.com/p/xue/ef72b7508:43
rohwpwrak imho none of the original fe are worth their money. low-end-ce tuner crap.08:43
bartbeshave you seen the video of the first 'game' I was going to port?08:43
wolfspraulbartbes: always!08:43
wpwrakroh: e.g., i already have half the collection of useful frontends at home, i.e., the one for 2.4/5.x GHz :)08:44
wolfspraulno haven't seen any video from you08:44
bartbesthis one: http://vimeo.com/136711708:44
rohin the end its a question of 'power' .. i mean.. have you played with a gnuradio?08:44
wpwrakroh: ooh, the tv thingie. c'mon.08:44
rohtry some dvb-t. in realtime please. and leave enough cpupower on that core2duo for a bit of mpeg208:44
bartbeswell, here's part of it running on the nn: http://bartbes.ath.cx/public/nlove.ogv08:45
wpwrakroh: have a look at the site. it has changed in the last decade ;-)08:45
rohsame goes for wifi. ofdm in sw is simply a very bad idea.08:45
wpwrakroh: dunno about dvb-t. i see 802.11a/b/g/n, though.08:45
rohwpwrak any ofdm with some complex modulation in it.08:46
wpwrakroh: and pretty much any other protocol i find interesting08:46
rohthe point is simply: freerunning adc/dac and sdr will cost something. its experimental and not for sale if you want to earn money.08:46
wolfspraulbartbes: he he! cool!08:47
wolfspraulthe game means you just keep shooting at HK forever?08:47
bartbesoh, it's more like a vid..08:47
wolfspraulHK does not seem to be too frightened from the sea otters...08:47
rohits the question if one wants to build another usrp, or something useable to transport ip08:47
wolfspraulbartbes: sure got that already. well great!08:48
wpwrakroh: i think sdr will eventually happen also in the ce market. like nobody build boards with lots of logic chips anymore but just uses a cpld to, more often, a microcontroller08:48
wpwrakroh: it is less effcient, but it doesn't matter. and it's probably more efficient for development.08:49
rohwpwrak there is just no reason for it. sdr is more expensive than any other possibility.08:49
bartbeswolfspraul: as you may or may not be able to see (I personally can't see it all too good), the speed on the nn is a little off08:50
rohdevelopmentboards cost 20times the money of ce-electronics.08:50
wpwrakroh: today, yes. tomorrow ? see above.08:50
rohsure it helps development. for signal-engineers. not for any ce-user08:50
bartbes(but it is somewhat noticable)08:50
bartbesironically the nn runs faster ;)08:50
bartbesbut that's because due to the aforementioned lua bug it thinks time is going faster than it really is08:51
wpwrakroh: in any case, let's see what crawls out of it. i'm not sure myself whether that gps project will work.08:52
wpwrakroh: they guys playing with it seem to like it well enough, so everyone's happy :)08:53
rohwpwrak playing!=usable for ce users.08:54
wolfspraulwpwrak: am I included in this? yeah, definitely happy here!08:54
wolfsprauleven if it doesn't work, I want to really fully understand why not, otherwise how can I make the next step :-)08:54
rohwpwrak sure i find that massively interresting. but i am not ready to 'pay' multiple times the power-budget for 'gps reciever' than something with 50 hw-korrelator channels does08:55
rohwpwrak which means again 'development' .. yes. sdr is for development. not for using afterwards.08:55
rohthe idea is to learn something and design an asic from that (imho)08:56
wpwrakroh: i think that;s the end goal, yes08:56
wpwrakroh: so if you have an interim result that's a little crappy, that's okay08:56
rohmy point is simply: a device will cost n times as much and run n times less long if it uses sdr instead of hw-mod/demod. and thats not acceptable for any mobile device one wants use.08:57
rohthe moko had such problems and thats why it was worthless as day-to-day-phone.08:57
rohnot because of sdr in that case, but the end result was similar.08:58
rohpeople are not ready to accept such limitations on non-develboards08:58
lekernelroh, there's a continuum of technical solutions between doing everything in hardware and everything in software (SDR), just pick an appropriate one :)09:00
wpwrakroh: it doesnt't have to be the xburst that runs all the sdr. you can always choose the granularity of your sdr instructions. maybe throw some configurable logic into the mix.09:00
lekernelthat's what computer architecture is all about09:00
wpwraklekernel: exactly :)09:01
rohlekernel i know how sdr works and i am not against it in general. its just a totally idiotic and reality-ignoring pov to say it can be used in real-world ce applications properly. especially on mobile devices.09:01
rohwpwrak the point is: one doesnt need any sdr for what we use day-to-day on rf.09:02
lekernelthen what about SDR on a very specific processor?09:02
rohlekernel that still needs multiple times as much power09:02
wpwrakwolfspraul: (pervertix) btw, you should know that i hate those "optimized for small systems" packages - kicking them as far as i could was one of the key pleasures i had in myroot ;-)09:02
rohwe are talking dsps on hundreds of mhz, where one wouldn have a dsp at all.09:02
lekernelever heard of multimode hardware synthesis?09:03
lekernelthat's another point on the continuum I was talking about09:03
rohlekernel theory doesnt get you further here.09:04
wpwrakroh: if you want to speak a lot of rf protocols, that bom makes sdr suddenly look quite nice.09:04
rohwpwrak may be. nobody does that for a reason. the rf frontend is too different for the different protocols.09:04
wpwrakroh: just consider N chips, O(N) oscillators, mixing all those antennas, coexistence, etc. at some point in time, sdr is cheaper :)09:04
lekernelroh, I'm not the last one to complain about techno-weenie researchers and cheap talk about theory, but still I believe in SDR09:05
rohwpwrak theory. in reality you have problems like interoperability. the need to use more than one if at the same time.. etc.09:05
rohwpwrak there is no manuf atm afaik which does bt and wifi on the same radio. there are combo-chips but they have seperate rf sections and mods/demods09:06
lekerneland modifying the processor to better handle a particular task isn't theory09:06
rohwpwrak simple as that. these guys arent stupid, working at atheros, broadcom, ti etc.09:06
wpwrakroh: you may not necessarily run all of the protocols at the same time. consider the ability to pick them a la carte.09:06
wpwrakroh: (wlan and bt) there may simply be no economical incentive for them them to do better09:07
rohwpwrak i still would not buy such a device if i can get an alternative one. in my experience a frontend doing discrete demod can do the same as a sdr one on less than 10% of the sdr power budget.09:07
wpwrakroh: there are a lot of areas where development stalls until someone comes with a new idea/implementation that threatens the establishment09:08
rohwpwrak oh.. there IS economic interrest there.09:08
wpwrakroh: e.g., think laptops vs. netbooks09:08
rohbt and wlan eat space. pcb space, really precious space09:08
rohwpwrak huh? nothing new there at all.09:08
rohnetbooks are just smaaÃall notebooks. not a single 'new' bit around it.09:09
rohboring old hw. same shit we had before.09:09
rohits just hyped.09:09
wpwrakroh: (wlan and bt) make one chip that's smaller than the competition. i assure you they won't rest until they have one too :)09:09
rohwpwrak you know that dsps are bigger than most discrete asic demods? (diespace) ;)09:10
rohan oamp are just a few transistors, and a few oamps and you got your demod.09:11
rohthere are cases where even running the pll for the dsp eats more power than the demod would use if done classic09:11
wpwrakroh: (laptops) well, the sub-notebook area was almost dead for several years. they all made big machines. you're right, the netbook was nothing truly new, but it changed the rules anyway. also by breaking the association between high-end CPUs == a quality product.09:11
lekernelgood luck demodulating wifi with op amps...09:11
wpwraklekernel: bah, i'd use valves ;-)09:12
wpwrakergh, tubes09:12
rohlekernel examples. not everything is ofdm09:12
wpwrakbrr. terminologies from different languages mixing.09:13
wpwrakroh: dsps are pigs, i agree09:13
rohand right now i dont see any reason to use sdr for wifi. it will eat loads more power than the full what a pc-wifi-chipset would use.09:13
wpwrakroh: a bit like fpgas09:13
rohwe are on mobile devices, so 1W is our FULL power budget.09:13
rohmeans with display and app-cpu09:13
wpwrakroh: well, remember the calypso ? there's a dsp inside too. that one wasn't so bad.09:14
qi-bot[commit] Andres Calderon: fixed xil2kc.py error http://qi-hw.com/p/xue/a8ff17c09:14
wpwraki see .fpds in xue/source/tree/master/kicad/modules/. nice ;-)09:16
qi-bot[commit] Andres Calderon: fixed xc6slx45fgg484.lib  error http://qi-hw.com/p/xue/bd8f8b909:16
wpwrakroh: anyway, i see no harm in people trying to to gps sdr on the xburst. if it works in a useful way, we all win. if it doesn't, experience will be gained and fun has been had.09:19
rohwpwrak the calypso iuuses a dsp for demod. yes. but its not sdr.09:20
rohwpwrak it uses a synced clock, so its not a free running adc. limits the amount of samples to take and do math on to an absolute minimum and removes oversampling09:21
rohand clock recovery09:21
roh(in sw)09:21
rohwpwrak do experiments for sure. but dont waste your hopes on that. neither for communication not for gps or so09:21
qi-bot[commit] Andres Calderon: fixed FPGA component bug http://qi-hw.com/p/xue/4110ee209:24
rohi really hope somebody build a free gps and or wifi chip some day. but i am not ready to wait that long.09:26
wolfspraulwpwrak: [smaller chip] totally agree. the IC sales people in china all know the die size in micron, and the company powerpoint presentation shows how much it decreases each year, etc.09:29
wolfspraulit depends on volume, but one rough number to start thinking is that the typical IC engineering will work until they have about 20% size savings, then tape-out09:29
rohwolfspraul hehe.. thats something not possible to shring endless in the rf world, where wavelength defines some structures09:29
wolfspraulin higher volume you can tape-out for lesser improvements, in lower-volume you need larger size reductions09:29
wolfspraulbut that's how the machine works, and how people operate09:30
roh20% less is quite some archievement.09:30
wolfspraulroh: first of all I'm explaining how sales people think/talk, how they present their company to customers, etc. the sales people are just that, sales people, but that's the dynamic in the industry.09:30
wolfsprauljust my feedback to wpwrak's comment about 'smaller'09:30
bartbeswolfspraul: btw, I just updated the vid to a 720p one09:31
bartbesand this one is in-focus ;)09:31
wolfsprauland the CEO of Ingenic, for example (makers of the XBurst chips in the Ben NanoNote), knows the percentage of die space for each feature.09:32
wolfspraulthat's just how their minds work09:32
wolfspraulso we look at this as 'video-out', or 'usb on-the-go', but in his mind it's 4% and 7%, or whatever :-)09:33
wolfspraulfunny, eh?09:33
rohwolfspraul we have to differenciate between digital only and mixed signal chips09:34
wpwrakCEO: we need to make it smaller", and kicks out gps :)09:34
wpwraknow, wait, this *has* happened .... ;-)09:34
wolfspraulroh: yes sure, I know little about all this, asking around, talking, learning...09:34
rohvideo-out is foobar in my eyes. waste of engineering time. ever connected you phone to some tv?09:35
wolfspraulis it in the N900? how many N900 users do you think use it?09:36
bartbesbtw, when I want to link to the ben nanonote, do I link to the wiki page?09:36
roheven if it works, it looks awful. (tv on cvbs is shitty resolution)09:36
wolfspraulbartbes: yes, wiki is good, it's the homepage09:36
rohwolfspraul ive had loads of devices with tv-out. used none of them for it ever09:36
wolfspraulthanks, links are very helpful09:36
wolfspraulyes that's you. you said that already. but what do you think, out of 100 N900 users, how many use this feature?09:37
wolfspraulI don't know. I'm like you, never used it and cannot imagine I ever will.09:37
rohotg isnt THAT helpful (nobody got a otg cable anyways). having host mode is more important. people actually use that.09:37
wolfspraulI didn't have a TV ever I think.09:37
wolfspraulit seems there definitely are people who love TV out though... guess not you and me :-)09:37
rohwolfspraul here in berlin people dont own tvs anymore. (atleast i know nobody with one).. everybody watches tv on his/her pc/notebook from dvb or via ip, if ever. cvbs is dead.09:38
wolfspraulTV out is one of the wish-list features for Ya NanoNote (the successor of the current one)09:38
alcywill Ya be priced similarly ? :)09:39
wolfspraulYa will be same as Ben, or lower09:39
rohwolfspraul some rf-data-modem is much more important than tv out09:39
wolfsprauleverything else is not hardware business. prices go down.09:39
wolfspraulalcy: yes but we have to make it there, which is not so easy. We sold about 900 Ben NanoNote today, which is a great result, but very hard to go from there to an improved device. We will try though, and every day important pieces fall in place...09:40
roheven some weight like the one werner did to keep it upright is more important than tv out ;)09:40
wolfspraulbut I guarantee you prices will go down.09:40
alcysounds good :)09:41
wolfspraulI am in China here, I feel the pressure not only from the software side (Android), or the latest Western hyped hw features, but also from the unbelievable price reductions in ce.09:41
wolfspraulI just switched my phone to a 20 USD phone.09:41
wolfspraulprices are falling like a rock I feel, and it's important to follow that momentum.09:42
rohwolfspraul *g* .. i still use my V3i. just replaced the outter glass recently09:44
rohwaiting for a small smartphone.. or one with decent battery runtime09:48
roh .oO(the web of trust is dead... https://www.eff.org/deeplinks/2010/08/open-letter-verizon )09:48
wpwrak(tv out) tv or composite video ?09:53
wpwrakroh: since when were certificates issued by "the man" every trustworthy ? :)09:54
wolfspraulwpwrak: tv I think, but not sure09:56
wpwrakspeaking of the counterweight: wolfgang, want some ? rafa will bring a few to europe, mainly for tuxbrain's heavy metal project, and now that i have molds, i can make more :)09:57
wpwrak(they're all Pb67Sn33, so not RoHS-friendly. paint varies. still haven't quite decided which one to use.)09:58
wpwrak(tv out) so that wouldn't work with, say, projectors. most of them don't have a tv tuner.09:59
rohwpwrak never.. but the whole 'web of trust' itself is failing10:07
rohhttps://www.eff.org/files/DefconSSLiverse.pdf is .... interresting10:07
rohfor those who don't believe ;)10:08
wpwrakroh: why do you call that hierarchical thing a "web" ? i never saw that use. i associate "web of trust" with, say, pgp.10:09
rohdoesnt work well either (pgp trust) .. i think we need to re-do all the trust-metrics we use today10:10
rohwpwrak https://docs.google.com/present/view?id=df9sn445_206ff3kn9gs10:11
rohTOFU/POP ;)10:11
wolfspraulwpwrak: I'm not in Europe, very rarely nowadays. David from Tuxbrain is your man.10:12
wolfspraulsure I could have one one day for myself, but it's not urgent. Maybe I can get one from David later, no need to send them around.10:12
rohwpwrak tv-out on mobile devices means 'cvbs' 'composite' 'the yellow cinch thing' on a 3.5mm jack usually10:13
rohnot rf ;)10:13
wpwrakwolfspraul: ok. i'll give rafa a few more to deposit at tuxbrain for you then :)10:14
wolfspraulyes, that would be great10:14
wolfspraul'a few' - wow10:14
wpwrakroh: ah, ok :) yes, the yellow chinch things is exactly what i had in my mind :)10:15
tuxbrain2:) be quick wolfspraul, the BNNTSE is taking form, and it look great, and I even add a REAL TRUE PIECE OF HEAVY METAL for the first to purchase maybe you will end without it :P10:16
wpwrakwolfspraul: how many shall it be ? < 20, or rafa won't be able to have much luggage beside counterweights :)10:16
rohwpwrak woulnt 'stamping' a pice of metal out of a sheet be the cheapest way to manufacturing for a counterweight?10:17
tuxbrain2hehehe I would love see the custom agent looking at raffa luggage and the pieces of metal in the scanner...10:17
rohi always thought moulding metal is crazy expensive10:17
rohtuxbrain2 hrr..  'custom made shrapnel'10:18
rafatuxbrain2: an easy way to know keycodes on X is if you install xev and run it.10:18
wpwrakroh: for small quantities it's pretty okay. i can make the mold of wood. lasts for about 20-40 pieces before the wood gets too dry and things break off.10:18
tuxbrain2rafa thanks, it was already solved :)10:18
rafaah.. too late :D10:19
rohwpwrak uh.. wood? ok.. that doesnt burn up when molten metal flows in?10:19
Action: tuxbrain2 going again to mounting tables for the den10:20
tuxbrain2c u10:20
wpwrakroh: stamping probably needs more pressure than i can produce. the counterweight isn't from a flat sheet but has different depth at different places, following the case's shape10:20
wpwrakroh: i probably don't have anything hard enough to make a stamp. i could of course try to heat the solder before stamping. but again, heat + pressure = more difficult10:21
rafawpwrak: I need some special 100l backpack to put the counterweights?10:21
wolfspraulwpwrak: really, 'a few' like 2-3 or so is by far enough for me10:21
wolfspraulI didn't know you guys already went to mass production...10:21
wpwrakroh: the solder i use melts around 170/2xx C. wood doesn't burst into flames until reaching the high 200s. so it will just get a little black but won't burn.10:22
wpwrakrafa: 100 l would weigh about one ton ;-)10:23
rafawpwrak: I have already big problems for when I arrive at the spanish airport. South americas are not welcome, you know, and if they have something to say about many pieces of something I will be coming back to argentina in the next plane.10:23
rafaSo better if we do not give them something as reason to not let me in.10:24
tuxbrain2btw roh lekernel wpwrak , an applause for your dissertation about RF Soft vs Hard driven... Is a pleasure read such high level conversations  (so high level I think I just undertand the 10%, but Is a pleasure to struggle my mind)10:24
tuxbrain2rafa if you think you might have problems please don't carry them with you, I will import them by courier latter on10:25
wpwrakrafa: do you think immigration will search your luggage ? if yes, how should the things be packaged/documented for the least amount of trouble ?10:25
wpwrakrafa: same for customs10:25
tuxbrain2wpwrak: In spain thing are getting weird with immigration specially form Africa/South America10:26
wpwrakrafa: want some counterweights for the jlime meeting, too ?10:26
wpwraktuxbrain2: i already heard stories ...10:26
wpwraki actually wonder if the thingies could actually add credibility to the "travel to business meeting" aspect10:27
tuxbrain2or pieces of cocaine with strange form factor10:28
wolfspraulyes, chemically solved in there somehow10:28
wpwrakah yes, perhaps one or two for customs to break open could be useful, too10:29
wpwrakthe tendency is that, if you prepare for all eventualities, they won't even look at you :)10:29
rafawpwrak: I have not idea. I have read bad and good experiences.. When the Miriam colleague traveled to Spain a month ago, immigration opened hist luggage, drop the whole things inside.. and ask for some no clothes things..10:30
tuxbrain2wpwrak, seriously, thing here are very very weird10:30
rafawpwrak: there is not a manual or rules to follow, so they could ask/do whatever they want.. and it seems, from that I read recently,  no a good idea to try to avoid that10:31
tuxbrain2Thinking on it again, I preffer to use a courier instead of rafa to bring the counterheights10:31
tuxbrain2I want to see rafa in a bar not in jail10:32
wpwrakin a bar, not behind bars :)10:32
wpwrakokay, your call.10:33
rafatuxbrain2: I am okey with bringing the stuff.. Just that I was thinking in a small and tiny package.. if it is something which they will detect easily and because it is not a common piece of a tourist I do not know if they will use that to annoy.10:33
rafaHere there are a lot of travelers.. with more experiences.. but no south america travelers I guess :)10:34
tuxbrain2yes sadly the origin of the traveler influence a LOT in the customs dealing10:35
rafatuxbrain2: between bars you can pass a beer bottle to me :D10:36
tuxbrain2Pb is really easy to detect what is no easy is what is inside the Pb , that why I start to not like the idea10:36
tuxbrain2rafa is your final decition but due they just need an excuse to start annoying , my advice is better don't10:39
tuxbrain2C U10:39
wolfspraultake some, learn. otherwise all decisions are fear driven. I think this should not be a big issue.10:40
wolfspraulI am regularly traveling with tons of stuff.10:40
wolfspraulbetter not list it here since it's publicly logged, ha! :-)10:40
wolfspraulwhen I go to a 'new' country, well, I sweat a little. but after X times whatever, you realize thousands of people march through there every day...10:41
wpwrakwolfspraul: drugs, porn, weapons, diamonds, Pu, wikileaks inbox ? :)10:42
rafawolfspraul: our concern is because spain is not liking south americans right now, and they are checking the whole trip of them and checking all their bags. To give you just an example:10:42
rafawolfspraul: I need to show the passports, the whole trip paid, all the flight tickets, all the room hotels, insurance paid, the whole money in your wallet, yes, like you are reading, and I am sure that if you go to Spain you just show them your passport.10:44
rafaall the hotel rooms paid*10:44
rafayou just = you just need10:44
wolfspraulyes sure, in Western Europe you grow up with the idea that you basically can go to 90% of the world just buying an airplane ticket and flying there.10:45
wolfspraultake passport and credit card, that's enough10:45
wolfspraulof course it's not entirely true, big countries especially don't really work like that (Russia, India, China), but that's the attitude10:46
rafayes, and credit card is not helping much for south americans.. We need to show the money, still if you think that it is ridiculous10:46
wolfspraulbut this is about lead counterweights now, I still don't see how it could be a problem10:46
wolfspraulmuch much stranger things have travelled in hand- and checkin luggage I think10:46
wolfspraulI mean why should there be a problem?10:47
wpwrakwolfspraul: a while ago, all the countries you needed a visa for were considered bad places anyway10:47
wolfspraulcommercial? no10:47
wolfspraulendangered species?10:47
wolfspraulweapons of mass destruction? any weapon at all?10:47
wolfspraulcome on10:47
wolfspraulit's nothing10:47
wolfsprauldrugs, prescription medicine?10:47
wolfspraulagricultural products?10:47
rafaI think that it is something happening this year, because people who have gone to spain years ago did not have those problems like spain are doing this year with south americans10:47
wolfspraulit's nothing I think10:47
wpwrakwolfspraul: maybe. drugs, precious metals, drugs, radioactive metals, did i mention drugs ?10:47
wolfspraulif you cut up your grill into small pieces, stuff it in your check-in luggage to take to Europe, why not10:48
wolfspraulwell, who knows10:48
wolfspraulthat's why I say - first trip - take 20 of them, it's enough10:49
wolfspraulnext time, 5010:49
wolfspraulthen - a while suitcase full, 2000 :-)10:49
wolfsprauljust kidding...10:49
wolfspraulyou will enjoy the trip, I seriously cannot imagine they cause you problems. You are coming from the 'paris of the south', no?10:49
rafawolfspraul: I was asking, in the spanish consulate, this: hey, I paid the hotel here in Argentina, before the trip, like you are asking us: what if the hotel is horrible, and we want to change?.. that is not our problem was the answer :P10:50
wolfspraulyes it's embarassing, I can imagine10:51
wolfsprauljust blog about it, done10:51
wolfspraulno need to waste energy10:51
wolfspraulsend a link of your blog post to the spanish tourism office10:51
wolfspraulend of story. enjoy your trip (you still will I'm sure)10:51
wolfspraulone day they will realize and improve10:52
rafayes.. and it is because I have not traveled before as well, so we are a bit afraid with the stories from others.10:52
wolfspraulyes that won't help.10:52
wolfspraulbut I can tell you - I went to bogota!10:52
rafaand Buenos AIres!10:52
wolfsprauland if there is one place people are told horror stories about (say apart from some african places), it's bogota/colombia10:53
wolfsprauland there may be some truth to these stories, but not that much10:53
wolfspraulI would go to bogota again in an instant10:53
wolfspraula little better prepared than first time :-)10:53
rafaah.. yes.. of course.10:53
wpwrakmexico is building up quite some reputation, too ...10:53
wolfspraulI would not want to arrive by myself at the airport though, without having a known person pick me up. not ready for that yet..10:54
wolfspraulbut anyway, I went, and some of this horror stuff turned out to be hot air10:54
wolfspraulgood! :-)10:54
wolfspraulcalling it a day, 'night10:54
Action: lekernel remembers the welcome signs at Taipei airport saying you get a death penalty if you smugle cannabis10:55
rafawolfspraul: venezuela airport scared me a bit when I arrived late one day.. but the next visits I just enjoyned Caracas :)10:55
wpwraki would not set foot on any arab soil10:56
tuxbrain2my uncle was segrsted by las FARC(now is ok don't worry) but yes some truth is there :(10:57
wpwrakthe asias are merely excessive about certain things. the arabs are outright crazy. e.g., that guy who had a poppyseed sandwich on another airport, they found some crumbs on his clothes, and jailed him.10:57
wpwraktuxbrain2: lovely :-( did he do anything considered dangerous ?10:58
tuxbrain2not just workin there10:59
wpwrak(arab) they also seem to be similarly thorough when it comes to searching for pornography. e.g., go though all your IT equipment.10:59
rafatuxbrain2: trying to prepare beta3 with the suggestions/fixes that some of you found.. Do you remember where is the gmu config file for the keys?11:00
zeari only remind this channel is being logged, not safe to talk about different enthic groups, considering they can find that logs :D11:00
wpwrakzear: well, i already decided that i won't enter these countries for their border controls, and i don't think anyone would find me insulting their border control sufficiently offensive that they would send some black ops team, so i guess the public logs are okay ;-)11:02
rafazear: if you read the whole conversation before your login.. you willl logout scared :D..11:02
zearwpwrak, who said they won't cross your border? :D11:03
wpwrakzear: for the record, i also haven't traveled to the us for quite a while for similar reasons. if a country makes it sufficiently clear that one is not welcome there, i think one should respect their wishes :)11:03
zearthey require us to have visas to usa (i'm polish)11:04
wpwrakzear: (cross border) that would be the black ops team. naw, for that, i would have to insult their religion or their leaders :)11:04
zearwe are not welcome in their country, but very welcome to help usa out in iraq and afghanistan. Weird logic :P11:05
wpwrakzear: i'm austrian, but since my passport has the wrong date (there's a window of a few months where the visa waiver doesn't apply), i would need a visa as well11:06
zearah, sucky11:07
wpwrakzear: hey, poles, blacks, indians, ... they all make great cannon fodder. global equality, finally achieved.11:07
zeartrue :D11:07
zearwell, i'm not going to usa anyway, don't want my retina in their database11:08
wpwrakzear: (visa) i had a look at the application form, and upon realizing that it contains questions i wouldn't even be sure i would know the correct answer for (e.g., in which countres have you been and when, within the last ten years ? quite a lot and i sure didn't keep a diary), i decided that any future visits to the us will have to wait until i have to renew my passport anyway11:09
zearwpwrak, how about the "are you a member of any terrorist group" question? :D11:10
zeari think i saw something like that in that form11:10
wpwrak(with all the biometric nonsense, passport renewal has become an ordeal as well. supposedly, it now takes somethine like at least 3 months to issue a new one.)11:11
wpwrakzear: (questions) ah well, i don't mind these. they're silly but i can answer them. what bothers me are questions where they're likely to know the answer better than i do.11:12
qi-bot[commit] Andres Calderon: fix http://qi-hw.com/p/xue/170ed5a11:44
lekernelwpwrak, btw you didn't answer my question: who's going to develop the DSP stuff for that GPS?12:11
larsc"the community" ;)12:30
wpwraklekernel: ah, there's a guy who'll do it for a real dsp. not sure what the path from dsp->xburst will be, though12:42
wpwraklarsc: so yes, unpaid volunteer work, as usual :)12:43
wpwraklarsc: wolfgang can still teach the chinese a few things about labour cost ;-)12:43
lekernelamazing, given how hard it is for me to get people interested in flickernoise development12:48
lekernelor, rather, do useful stuff - most of the time they're interested, but it's just hot air12:48
wpwrakgrmbl. physics isn't an exact science :-( i'm trying to detect whether my counterweights are properly sealed. my idea was to set up a salt water bath, measure the resistance, then insert the counterweight and measure again. instead of a nice constant value, i get some time-dependent curve that isn't impressed by whatever i put into the solution :-(13:02
bartbeslook at this magic: http://loveclub.pastebin.com/z6ugedbz13:09
wpwraklekernel: what is flickernoise ?14:17
lekernelend user software for milkymist14:19
lekernelI'm super busy with FPGA design and, at some point, PCB14:20
lekernel=> flickernoise didn't advance much :(14:20
lekernelhopefully I'll be done with FPGA stuff relatively soon and then I can properly take the lead on flickernoise14:21
wpwrakah, it's a bit of a niche. that may slow it. sell 900 milkymists, and you'll have your developer community :)14:21
lekernelwell it can run on x86 too14:22
lekerneland milkymist can be emulated in qemu14:22
lekernelplus selling 900 boards without software is probably tricky14:22
lekernelchicken and egg problem14:23
wpwrakbut will it be useful on a pc ? or do you need a milkymist for it to really shine ?14:23
wpwrakyeah. the ben has it a bit easier there. uses are more readily apparent.14:23
lekernelright now it's totally worthless on a PC.... then after a while it might have some uses on a PC14:23
wpwrakah, so still the same niche. how many milkymists are there in operation at the moment ?14:24
lekernelthat's again a chicken and egg problem14:24
lekernelright now only 4 are really used14:25
lekernelvery first protos14:25
lekernelnow we're planning a new batch of 2014:25
wpwrakvery small niche then :) would it be feasible to make a board that replaces the current ben's internals ?14:27
wpwrak(probably with some compromises for the large fpga)14:27
lekernelwell better just wait for the RC2 batch :)14:28
lekernelbut I'm impatient14:28
wpwrakcould you trim the milkymist down to be able to "run" in a ben, talking to keyboard, screen, and usb ?14:28
lekernelyou won't go far, the ben doesn't have graphics acceleration14:29
lekernelif it's only for running the pure software GUI part then yes14:29
lekerneljust recompile and it should work14:29
lekernelthere's a lot of crap that needs to be redone just because we're not using X14:30
lekernellike developing a file manager14:31
wpwrakthat may be a step towards spreading it then. after all, wolfgang's long-term plan/hope is to turn this into a free main cpu.14:31
wpwrakwhy not just make X work ?14:32
lekernelslow, bloated, hard to make work14:32
lekernelincomprehensible protocols14:32
lekernelsee the chapter on X in the UNIX-haters handbook14:32
lekernelthat's basically what I think about X14:33
lekerneland it's even more true in embedded systems14:33
lekernelplus on top of that you'd use toolkits like GTK or Qt, which are also slow and hard to port to RTEMS14:33
wpwrakby the time, you've reinvented it, your stuff will be just as bloated, probably more ;-)14:34
lekerneloh, sure not14:34
lekernelsingle application, single memory space, simple and pragmatic toolkit, not thousands of more or less useful abstraction layers14:35
wpwrakyou're not the first one who thinks X is too heavy. then they do their own thing, and eventually fail. also, just the basics of X don't seem to be too bad. of course, if you want the full feature set, then you get a lot.14:35
lekernelwe're using an existing toolkit called genode fx14:35
lekernelwhich isn't anywhere close to failing14:36
lekernelat least for what I want to do out of it14:36
wpwrakah well, if you find something you like better, why not. you just have to put a lot of effort into stuff you don't really care about.14:37
lekerneland btw I'm not sure using X would even solve the filemanager problem for instance14:37
lekernelunless you end up porting KDE, Gnome or XFCE which is just crazy14:37
wpwrakthe whole anti-X attempts remind me a bit of the situation about TCP. every once in a while, someone comes up with some bright idea how to make a simpler, faster, leaner TCP. it's always fun to watch how far they get before the things blows up in their face :)14:38
lekerneloh, come on14:39
lekernelthis is no comparison14:39
lekernelyou can implement TCP in 1000 lines of code14:39
lekernelif you end up with X14:39
lekernelyou have 3 toolkits to support if you want to re-use applications (gtk, qt, xlib)14:39
lekernela super complex protocol and architecture14:39
wpwrakthe days of ATM (asynchronous transfer mode) were particularly fertile for that sort of ideas. countless ideas for making TCP without TCP. none of them ever saw the light of day, beyond some timid lab experiments perhaps.14:40
lekerneland some people have succeeded at replacing X14:40
lekerneljust compare the UI responsiveness of the iPhone and of the Openmoko14:40
lekernelthis tells you basically everything about X14:40
wpwrakthere's a lot of stuff that only uses xlib or the older libs (xaw, for example). of course, all the new things prefer more abstraction.14:41
lekernelxaw? hahahaha14:41
lekernelcouldn't you think of something more ugly?14:41
wpwrakit tells you a lot about picking a graphics accelerator that doens't have negative acceleration ;-)14:41
lekernelxaw is 1980s graphics14:41
wpwrakwell, i grant you that :)14:41
wpwrakxaw3d perhaps :)14:42
lekernelstill ugly14:42
wpwrakyou're a hard man to please :)14:42
lekernellooks like winshit 9514:42
lekernelin worse14:42
lekerneland slower14:42
wpwrakregarding tcp, if may look simple to implement, but understanding the dynamics is pretty hard. but then, there's a good reason for every little bit you find in there.14:44
wpwrakyou can of course make a very basic implementation of tcp, but that won't perform very well14:44
lekernelthat's how genode fx looks like14:45
lekernelreasonably good looking, fast, lightweight and easy to port14:45
wpwraklooks pretty nice14:45
lekernelone drawback: very little application support, and many widgets are missing14:45
lekernelquite nothing like the xaw crap14:46
lekernelthis thing was developed by hardcore amiga demosceners, not bearded unix hippies14:47
wpwrakwell, if pretty buttons is all you desire, you could probably do this on top of xlib in a day, too :) the advantage would be that you'd be compatible with all the other applications that you don't care enough about to reimplement, but that are still desirable to have.14:48
wpwrakor just bite the bullent, make an X server, then keep on hacking on your own thing. that way, you can increase the user base of milkymist a bit.14:48
lekernelor just have the balls to kick X out just like Apple did14:49
lekernelX is a piece of junk14:49
lekernelface it14:49
wpwrakah, your zeal remembered me a bit of the amiga days :)14:49
lekernelMacintosh users hate X applications14:49
lekerneland I understand their arguments14:50
wpwrakx does the job and there's a ton of stuff i'd like if i didn't have x.14:50
wpwrakquite a typo :)14:50
lekernelX does the job? again, compare the UI of Openmoko and iPhone :)14:51
wpwrakcan the genode fx api be used to run things on top of X ? even if slowly ?14:51
lekerneland how nice scrolling works on the two14:51
lekernelwhen I tested it you have something like a 0.5s delay between your finger and the screen on the moko14:52
lekernelit's awful14:52
wpwrakthe moko gui's problems have many sources. x is the very least of the worries. trust me on that. i've been there :(14:53
lekernelthen why does every handheld using X seems to have the same problems?14:53
lekerneland those who don't use X, like android and iphone, don't have the problem either?14:53
lekernelsounds like a big coincidence14:53
wpwrakmaybe it just is14:53
lekernelespecially when I look at the output of "top" showing the X server in a PC14:54
lekerneland how the X junky protocol works14:54
lekernel(or does not)14:54
wpwrakx on top usually means some flash nonsense looping in your web browser :-(14:55
lekernelbtw: yes, you can somehow run genode fx apps on X... right now all in the same X windows14:55
lekernelI don't think it's very difficult to map each genode fx window to a new X window14:56
wpwrak(genode on x) great. that will make it a bit easier for your community.14:56
lekernelactually it has a SDL port14:56
lekernelso you can run on X, and directly on the Linux framebuffer, etc.14:56
wpwrakgood. so people don't have to give up their regular environment for development.14:57
lekerneloh, definitely not14:57
lekernelthere aren't many technical barriers to start flickernoise development14:58
lekerneljust install sdl headers, check out the sources, compile genode fx then flickernoise, and you're done14:59
lekernelthe problem is right now there's little actual use reinventing many wheels on the PC platform14:59
lekernelso this is probably de-motivational (imo)14:59
lekernelbut having to choose between rewriting a file manager and a X server + porting GTK/Qt/Xlib, I pick the first15:00
wpwrakyup, definitely. projects that don't scratch a person's itch are unlikely to get that person involved.15:01
wpwrakwould porting xlib and such really so hard ?15:01
lekernelI don't want X, period :)15:01
wpwrakalso, the x server doesn't seem to be too much of a problem. we had one pretty quickly for the glamo. also, you can start with just a dumb frame buffer and add acceleration later. that way, you have a usable result every once in a while.15:02
lekernelxlib may not be the hardest part (and still, I'm not so sure given the horrible build scripts of the Xorg distribution) but writing the X server will definitely be a headache15:02
lekerneloh, and my board as 32MB flash15:02
lekernelGenode FX is 100K approximately15:02
wpwrakah well, if milkymist is successful, then someone will one day put X on it. then you can compare the two :)15:03
lekernelQt+GTK+Xlib+X isn't 100K15:03
lekernelit's more 200MB15:03
lekernelfor doing about the same thing, but in a slower, more CPU intensive and less responsive way15:04
wpwraki think the actual display driver can be pretty small. i don't think the one for the glamo had much more than 1000 lines.15:04
lekerneland btw my system runs RTEMS and does not have a MMU15:04
lekernelI'm not sure Xorg would work in these conditions15:04
wpwrakrtems ?15:04
lekernelyeah, a real time embedded OS15:05
wpwraknot sure whether it needs an mmu15:05
lekernelPOSIX compliant, single process15:05
lekernelmulti thread15:05
wpwrakah, okay. hmm. single process would be an issue.15:05
lekernelkicked uClinux away after banging my head on my desk about the subtle and undocumented GNUtarded interaction between Binutils and the nommu FDPIC executable loader15:06
wpwrakah ? what's the problem there ?15:07
lekernelmaking the oh-so-awfully-coded Binutils generate properly relocatable binaries, then making uClibc and the kernel relocate them15:08
lekernelthis was an unneccessary problem anyway, since I only need one process after all15:08
lekernelso RTEMS is perfect15:08
lekerneland it's also easier to write RTEMS drivers than to write drivers for the Linux kernel15:09
lekernelthe all-function-pointer approach makes it horribly hard to debug and understand, if they want to do object programming they should use C++... but, well, that's a different troll :)15:10
wpwrakso you have your own compiler as well ?15:10
lekernelit's a modified GCC15:10
lekernelwell now it's been merged in 4.5+15:10
wpwrakphew :)15:10
lekernelbut with a broken linux executable generator15:11
lekernelanyway, it gets the job done on rtems15:11
wpwrakthen add an mmu and port linux. you'll learn to like the function pointers :)15:12
lekernelfuck that15:12
lekernelagain it's easier to write a file manager for genode fx15:12
lekerneland the mmu is not needed for what i'm doing15:13
lekernelneither is Linux15:13
lekernelkeep it simple15:13
lekerneland Linux is slower than RTEMS, too15:14
wpwrakkepping it compatible may be simpler in the long run :)15:14
lekernelRTEMS is also a pretty compliant and widespread OS (ok, not as much as Linux)...15:14
lekernela lot of Linux libs compile on it, as long as they don't use the GNU Autocrap system15:14
wpwrakheh, for once, that's a hatred we share15:15
Action: wpwrak enjoys writing makefiles15:16
lekernelsame here :)15:16
lekernelbut well15:16
lekernelif someone wants linux and mmu15:16
lekernelthey just make it, heh :)15:16
lekernelthen I have additional developers :p15:16
lekernelI can't do everything on my own15:16
lekernelRTEMS + genode fx is the fastest and technically better route to what I want to do15:17
lekerneland btw it already works today, on the hardware15:18
lekernelwith vga out, but no input devices yet... USB (for mice and keyboards) is horribly complicated too15:18
wpwrakwhat's the actual objective by the way ? it seems rather unusual to design your own cpu just to make a vj station.15:18
lekernelthat's why I have no time for things like a MMU15:19
lekernelwell, once I have that device out, then maybe this can start MMU projects15:19
wpwrakusb can be a bit of a challenge, yes. to which level have you made it so far ?15:19
lekernelthe low level part (serial transmissions) works, in both low and full speed15:20
lekernelthe protocol is handled by an AVR softcore, the design of which is done and allows some transfers to be made15:20
lekernelbut still has some bugs... i'll work on exhaustive test benches pretty soon15:21
lekernelthen one would need to implement OHCI as a C program for the AVR15:21
wpwraksounds quite good then. have you found a way to sniff the wire ?15:21
lekernelyeah, I bought a USB analyzer15:21
lekernelcouldn't do without15:21
wpwrakyup, that helps15:22
lekernelunsurprisingly enough the opencores USB serial engine was full of bugs15:22
lekernelso I had to redesign that as well15:22
wpwraki debugged my little usb stack on the silabs mcu with a scope. that doesn't go quite as far as an analyzer, though.15:22
lekernelthe amount of junk at opencores never ceases to amaze me15:22
lekernelso, well, right now the USB core reads the device descriptor of full and low speed devices and printf's it out15:24
lekernelsome other C programs crash or otherwise fail because of AVR softcore bugs15:24
lekernelonce those bugs are fixed, the last step is to write that ohci software implementation15:24
wpwrak"avr softcore" = V-USB ?15:25
lekernelno, I designed my own softcore which is put inside the fpgza15:25
lekerneland talks to the usb serial engine, also put into the fpga15:25
lekernelI can't handle the serial transmission on the AVR too because I need full-speed too15:25
lekerneland the AVR is too slow for that15:25
lekernelso I assisted it with a small "UART"15:26
wpwrakmakes sense15:26
wpwrakyou seem to have a lot of space in that fpga15:26
lekernelno, I just don't use bloated Opencores designs15:26
wpwraki mean, you have lots of features in there. how crowded is it by now ?15:27
lekernelright now I'm only using 70% of it with a SoC that has almost everything planned (merely only bugfixing is needed now)15:27
wpwrakbtw, nobody ever answered my question about the cpld spi conversion :-(15:28
lekernelthis leaves plenty of space for things like a MMU, if people are interested in those things15:28
wpwrak(on the developer now discussion list)15:28
wpwrakkewl. maybe even some cache ;-)15:28
wpwrakor some DSP instructions, to annoy roh a little :)15:29
lekernelthere is already 2 levels of cache15:29
lekernelbut growing them bigger causes timing problems15:29
wpwrak(2 levels) not bad !15:29
wpwrakwb ?15:29
lekernelor you'd have to pipeline them on more stages... and/or add a 3rd level... but this means more design work15:30
lekernelI'm using 3 different buses15:30
wpwrakno, writeback.15:30
lekernelL1 is write through, L2 is write back15:30
lekerneland critical word first, too15:31
wpwrak3rd level seems a bit excessive for such a "small" cpu.15:31
lekernelI don't know15:31
lekernelthe way to know is to sniff the internal buses15:31
lekernelalready did that on the DRAM bus btw :)15:31
lekerneland there's more than a CPU15:32
lekernelmany peripherals have DMA15:32
lekernelthere's even a graphics pipeline which has its own cache directly connected to the DRAM15:32
wpwrakcan peripherals use the cache(s) ?15:33
wpwrake.g., you set up some packet in cache, then fire off a peripheral. never needs to touch dram.15:35
lekernelmost DMAs (ie those which transfer small data with eg the CPU) are done through the L215:35
wpwrakvery good15:36
lekernelDRAM-direct DMAs are done for bulky transfers like VGA screen refreshes15:36
wpwrakyup, doesn't make sense for everything15:36
lekernelthey're also L2-coherent, so the CPU doesn't need to care about the cache (L1 is write through) when writing to the framebuffer15:37
wpwrakperfect. one worry and annoyance less.15:38
wpwrakoh, you even have an fpu. the days of arm are counted :)15:41
lekernelit's not a traditional FPU, which would have been too slow15:42
lekernelit's a floating point VLIW coprocessor15:42
lekernelputs out about 75MFlops at 100MHz on real cases using the compiler described in the thesis15:43
lekernelit's not fully IEEE 754 compliant either15:44
lekernelI only use it for graphics transformations basically, so it's not a big issue if it's not super precise15:44
lekerneland it saves hardware resources and especially design time15:45
lekernelwhich is even more precious than hardware resources15:45
wpwrakvery nice. would ieee 754 be possible ?15:45
lekernelbut it's like the MMU: not needed => kick it away15:45
wpwrakkewl. that would also help.15:45
lekernelit's such a big project you can't spend time on unneeded details15:46
wpwrakwell, to take over the main cpu, you need these things :)15:46
lekernelv1.0 out first, then optimizations15:46
wpwrakyup. one step at a time :)15:46
lekernelplus the v1.0 being out may help a lot attracting more developers15:46
wpwraki still wonder if it wouldn't make sense to put this into a more common form factor. e.g., inside a ben.15:47
wpwrakthat may help to attract people more quickly15:47
lekernelthere are so many chicken-and-egg problems in that project that you need to stick to the "v1.0 out first" paradigm15:48
wpwrakof course, you may prefer not to have too many folks rush in15:48
lekerneldepends which folks15:48
lekernelif it's the annoying luser type, then no15:48
wpwrakv1.0 seems to be soonish, no ?15:48
lekernelif it's talented FPGA or software developers, they're most welcome15:48
wpwrakyou always get a bit of a mix15:48
wpwrakand it takes time to organize the crowd, no matter how good they are15:49
lekernelbut I think it's easier to ignore lusers than to motivate talented people, so more folks should be cool :)15:49
lekernelv1.0 should be soonish for the hardware, but not the software15:52
lekernelflickernoise development will probably take at least an additional 6 months or so15:52
lekernelbefore it can be really end-user15:52
wpwrakbtw, what's the state of the art on open fpga synthesis ? i find the dependency on proprietary tools rather galling. a while ago, nothing better than some ancient PAL had a publicly documented fuse map. now, there are the psoc3 with a bit of a cpld inside, which come with full documentation. still nothing big, though.15:52
lekernelthere's nothing15:52
lekernelwell there's ulogic (reverse engineering effort)15:52
lekernelof the bitstream format15:52
lekernelbut it's not the hard part15:52
wpwrakit's the enabler for the rest15:53
lekernelthe hard part is the HDL->netlist "conversion"15:53
lekernelfor which all docs are available, ironically15:53
lekernelthe closed bitstream is an excuse imo15:53
wpwrakisn't there some synthesis tool for asics ? from somewhere in france15:53
lekernelyou look a lot more smarter by saying "oh, we're poor developers who don't have fpga specs, we can't do anything!!! bad, evil manufacturers!!!"15:54
lekernelthan by saying "I have no fucking idea how to write a logic synthesis program"15:54
lekernelyeah, there's Alliance15:54
lekernelthey royally fucked up on this one15:54
wpwrak(excuse) i don't know. it;s something that always kept me from touching cplds or fpgas15:54
lekernelit's poorly coded, and, above all, defective by design15:55
wpwraki wouldn't mind having a poor synthesis tool, as long as it does give me the freedom15:55
lekernelthe approach to synthesis they have is never going to yield good results (ie comparable to those of the proprietary tools)15:55
lekernelplus a lot of features are missing15:55
wpwrak(poor tool) to get started, of course. things like milkymist are a different matter.15:55
lekernelwhich makes it totally counterproductive and incompatible with other (and better) EDA methodologies15:55
wpwrakah, pity15:56
wpwraki think an open bitstream could provide a better incentive for people to do some work in these areas. if you already know that you'll be blocked by the bitstream, that's not nice.15:57
lekernelit's not even hard to reverse engineer15:57
lekernelxilinx provides a GUI to manipulate individual FPGA cells15:58
lekernelthen it's only a matter of knowing which button flips which bits15:58
wpwrakis the behaviour of the fuses documented ? (without NDA)15:58
wpwrakthat sounds good15:58
wpwrakflip a bit, diff the streams :)15:59
lekerneland for the timing analyzer, timing tables from the binary database are nicely printed out in human readable form by a tool called partgen (iirc)15:59
lekernelwhat fuses? FPGAs are SRAM15:59
lekerneloh and there's also a lot of bitstream format documented by xilinx too16:00
lekernelframe format, CRCs, so you won't even to reverse engineer this hard part16:00
lekernelseriously people who use the closed bitstream format excuse are just incompetent and/or slackers16:01
wpwrakulogic looks quite encouraging16:01
wpwrakdon't you still call it the "fuse map" on fpgas ?16:01
lekernelhuh? no, never heard that word16:01
lekernelthat's for PALs, no?16:02
lekernelFPGAs have a very different architecture than PALs16:02
wpwrakthe bitstream documentation only tells you how to transmit the stuff, not what's inside.16:02
lekernelyeah, and for what's inside you just have to use the xilinx editor and see what changes what :)16:03
wpwraki think i've seen it at least for cplds. not sure about fpgas.16:03
lekernelfpga's are 100% SRAM, there's no fuse or flash whatsoever16:03
wpwrakfpgas are a bit too big and pricy for my taste :) at least for "playing"16:03
lekerneleven for parts that support encryption, you need a backup battery16:03
lekernelon the PCB16:04
tuxbrain2rafa: /home/root/music/gmuinput.nanonote.conf16:04
wpwrakyes, i know that fpgas have no persistent state, while cplds do. but don't they have more or less the same architecture for the rest ?16:05
lekernelthere's nothing in common between the cpld and fpga architectures16:07
wpwrakah, interesting16:08
lekernelbtw fpga's aren't so pricey, the one i'm using in milkymist is $50 in single quantities16:08
lekerneland 2 years ago avnet released a $49 devkit16:08
wpwraki call that costly :) for playing, i want to be able to go through a lot of chips. put them in different circuits. get a feeling for them.16:09
wpwrakmaybe make ten units of something useful.16:09
wpwrakthey're also difficult to hand-solder :)16:09
lekerneljust dare doing it :16:16
lekernelif you're patient too ...otherwise you can also make a pcb and use the solder wick method (which I prefer because it's fast)16:17
lekernela lot of fpga's are still available in qfp that you can solder with these techniques16:17
wpwrak;-)) naw, i make my own pcbs. up to qfn is okay. bga is where i admit defeat.16:17
lekernelyeah, diy bga isn't easy16:18
wpwrak(haven't actually tried, though. i do have a toaster oven in my lab :)16:18
lekernelI've seen some people doing it but it's tricky16:18
lekerneland you need a multilayer pcb in many cases16:18
lekernelanyway, there are interesting fpgas in more friendly qfp16:19
wpwrakyup. that's part of the problem. also debugging a badly soldered bga should be a pain.16:19
lekernelthere's xray control for that16:19
lekernelcould be fun to make a diy xray machine too :)16:19
lekernelmaking xrays isn't actually difficult: old CRT high voltage power supply (on which the AC output is accessible), voltage doublers, and valve16:20
wpwrakor just hang out near a sports arena and ask people who break something and need to get an xray to take a board with them ;-)16:20
lekernelwhat's difficult is safety and image resolution16:20
wpwrakyes :)16:20
wpwrakmaybe once your life expectancy is only a few years and any cancer you may add won't kill you before the rest :)16:21
lekernelwell at some point I might dare doing it :)16:21
lekernelbut with a proper dosimeter16:21
lekerneland lots of lead16:22
wpwrakafter you finish v2.0 with the mmu please :)16:22
wpwrak(milkymist) pretty impressive project for "just" an MSc. are you now continuing this towards a PhD ?16:25
lekernelI don't actually know what i'll be doing as PhD, and if I'll do a PhD at all16:25
lekernelpossible subjects I'm thinking about are developing an open source synthesis tool16:26
wpwrakphd is fun. you get to play around, travel, and earn money :)16:26
wpwrakthat would be excellent ! :)16:26
lekerneland research on automated FPGA bitstream reverse engineering to find out if companies use GPL'd cores without releasing the sources16:26
lekernelthe problem is both projects are difficult and will slow down MM development16:27
wpwrakuse MM as your test case for synthesis16:27
lekernelthat's was one of the initial goals of MM actually16:27
wpwrakperfect then16:27
lekernelget me into fpga development and provide a test case/verilog subset for a synthesis tool16:28
wpwrakit's quite normal for a phd to involve "non-essential" projects. in your case, mm would even fit16:29
lekernelmaybe mm soc 2.0 as phd ;)16:29
lekernelbut it includes a variety of subjects16:29
wpwrakvery broad may be a problem16:30
lekernelalso i need to find a university with a relaxed copyright policy, which isn't easy in every country it seems16:30
lekernelthe nice thing with KTH is you retain copyright on all your work, so I could easily do MM as thesis without much impact on the project; it's not the case everywhere16:31
wpwraki had no problems with that at epfl. for papers, we used technical reports as a work-around. anything else was gpl or such anyway.16:32
wpwrakso even if the university has the copyright, it doesn't matter much.16:33
wpwrakthe only real limitation this creates is that can't just give a project to harald, for his gpl-violations ammunition stash16:35
lekernelI'd have a little preference for the logic synthesis tool, it's clearer in my mind what needs to be done and I'm more certain to succeed16:36
lekernelso I have better control on timing and could work on that while mitigating the impact on MM development16:36
wpwraksounds great. it's also something that's quite obviously missing in the world.16:37
lekernelotoh enforcing GPL in FPGA design too is much needed too imo16:37
lekerneland also gives nice ammunition to Harald16:37
wpwrakyou can do that later :)16:37
lekernelI'm already suspecting two companies of using my code in proprietary FPGA designs16:37
wpwrakwhat are the hints ?16:38
rohi wonder how much commercial stuff gets stolen, if already so much opensource hdl designs seem to be (ab)used16:38
wpwrak(that is, if you don't mind putting suggestions for avoiding your attention on this publicly logged channel)16:39
wpwrakroh: the best theft prevention still seems to be to encourage people to take it :)16:39
wpwrakroh: that makes them suspicious and fearful16:40
wpwrakroh: e.g., LILO was ported only once, to BeeOS (i think they renamed it to "BILO"), even though i put it under a BSD license16:41
wpwrakor BeOS16:41
wpwrakroh: then, afaik, nothing ever came out of tcp connection passing. maybe they think i've patented it :)16:45
lekernelwpwrak, well, basically asking if they can use it in their proprietary stuff, getting an answer saying they have to pay me or opensource the complete design, and then no contact16:50
wpwraklekernel: and soon thereafter, there's a product with an fpga. i see. would indeed nice to test this in court.16:51
lekernelwith all the "FPGA bitstream formats are impenetrable" bullshit around, there isn't a lot of incentive for them to pay16:52
lekernelif there's a phd report + a tool that proves the opposite, this gives you quite a few tools to fight back16:52
rohfor sure16:53
wpwraklekernel: plus some interesting violations to open :)16:54
wpwraklekernel: or, considering that they may have mixed it with closed IP that's not theirs to release, make you a rich man :)16:55
wpwrakroh: the ssl observatory slides are nice. not surprising, though. make something sufficiently complex and obscure and people will screw up.16:57
lekerneloh, you wrote LILO?16:59
lekernelnice :)16:59
lekernelit was my first linux bootloader back in 199916:59
wpwrakwhee ! that was already towards the end of its days of glory17:03
wpwrakregarding the MM developer community: i wonder if people might be fooled by the project being presented as a VJ station. you don't quite expect to find a novel SoC there.17:08
lekernelwell I put the SoC forward in all communication material17:09
lekerneland the project is well represented on opencores too17:10
lekerneleven if most of opencores is trash, that's where people go to when looking for open source hdl17:10
wpwrakokay, so fpga people will find it then. but how about software developers ?17:11
lekernelwhat are the communities of embedded developers you know of?17:11
wpwrakno idea :) they're kinda everywhere17:12
wpwrakthere are a few well-known projects, but that doesn't mean that anyone interested in such things will ever visit one of them for more than a quick glance17:13
wpwrakinterested as a developer17:13
wpwrakhas slashdot reported on milkymist yet ? that tends to spread the message17:17
lekernelI tried posting a news a year ago but it got rejected17:19
lekernelI similarly got kicked out of the google summer of code this year17:19
wpwrakblargh :-(17:19
lekernelotoh I got posted on make magazine and made a couple of nice conferences (26C3, LSM)17:20
lekerneland others17:20
wpwrak26c3 gets good coverage. but it's easy to disappear in the sheer amount of things that get reported.17:21
lekernelthe 26C3 presentation sucked a bit, it was the first time I was in front of such a big audience and I was stressed out17:21
lekerneland I'm not sure everyone got what open source SoCs were about, people were actually pretty unfamiliar with the topic17:21
lekernelthe LSM presentation was much better17:21
lekernelbut fewer people and coverage17:21
lekernelI'll submit a software development workshop on MM platform for the 27C3 I think :)17:22
lekernelthen i'll be sure everyone will get it's a SoC17:23
wpwrakLSM ... Libre Software Meeting, kinda regional ?17:24
wpwrak(sw workshop) ;-)17:25
lekernelnot regional, but french (even though they advertise it as worldwide)17:25
wpwrakokay, quelques regions ;-)17:26
wpwrakinteresting. a tank of salt water with two electrodes measuring resistance (with a multimeter) is an excellent vibration sensor17:34
wpwrakand at the same time, it's perfectly useless for measuring the insertion of a conductive object. that much about my clever idea for detecting defects in the coating of the counterweight :-(17:36
lekerneliirc there's also a nice AC component you can get on the electrodes17:38
lekernelin the audio range17:38
lekerneli've seen an artist using it to make weird music :)17:38
lekernelnot sure it happens with all metals and salts, but it definitely does with some17:39
lekernelprobably depends on current, voltage and other factors too17:39
wpwrakkewl. i have to try that some day :)17:45
wpwrakmaybe you can even tune it by shaking the water :)17:45
wpwrak"shake it, baby !"17:46
wpwrakhmm, ulogic looks abandoned. no changes in the last 3 years :-(18:05
lekernelyeah, I know the guys behind it18:07
lekernelthey're a bit of the "overwhelmed by job" type :(18:07
wpwrakyou mean "job" as in "real work" ?18:13
lekerneljob as in "doing engineering work for a big company you have no control over", not sure if that's what you call "real work"18:14
wpwrakyou get money but it sucks. sounds like "real work" :)18:14
wpwrakwell, once you've taken over arm's empire, you can hire them to finish their work :)18:15
qi-bot[commit] Mirko Vogt: [mandoc]: mark as broken, as it does not compile http://qi-hw.com/p/openwrt-packages/273bf8020:17
qi-bot[commit] Mirko Vogt: Merge branch 'master' of projects.qi-hardware.com:openwrt-packages http://qi-hw.com/p/openwrt-packages/272a62820:17
qi-bot[commit] Werner Almesberger: cam/: scripts for cutting the PCB (just the board - traces are etched) http://qi-hw.com/p/ben-wpan/2cb3f1821:39
qi-bot[commit] Werner Almesberger: Typo causing silent failure. $z would be Perl, but this is Python. http://qi-hw.com/p/ben-wpan/9bf969e22:59
qi-bot[commit] Werner Almesberger: Typo causing silent failure: forgot the semicolon, so "doit" thought these http://qi-hw.com/p/ben-wpan/12ebe2c22:59
wpwrakgrr. i reverted that one.23:01
--- Sun Aug 15 201000:00

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