#qi-hardware IRC log for Monday, 2011-03-07

kristianpaulhe, lilo save my evening :-)00:45
Action: wpwrak high-fives kristianpaul 00:47
kristianpaulwpwrak: now i see why still using it :-)00:52
wpwrakkristianpaul: how did grub fail you ?01:04
kristianpaulwpwrak: easy problem about root=something_bad_set here, but was generated automatically so lilo seems not follow that path01:07
Action: kristianpaul was installing voyage linux on his soekris 550101:08
wpwrakoh dear, double-exotic ;-)01:08
wpwrakbut yes, sometimes lilo's simplicity can save your day. recently, i'm hearing an increasing number of complaints about grub.01:10
mthgrub is u-boot for non-embedded systems ;)01:12
wpwrakmth: yeah, a little boot loader that tries so hard to be a big operating system ;-)01:13
wpwrakmth: with u-boot, the fun starts when you add usb, because this gives you *true*concurrency*. all the races, but none of the locking ...01:14
mthI didn't know u-boot had USB support, but I'm not surprised :)01:16
wpwrakmth: i don't know if usb ever made it into u-boot mainline. in openmoko we had it, though01:16
wpwrakmth: and if you think of it, functionality-wise, it makes a lot of sense01:17
mththe Wii homebrew channel accepts USB uploads as well, very useful for testing new programs01:17
kristianpaulmth: You could'nt be so right, actually there is a port of grub as a replacement of pmon bootloader for lemote/yeeloong laptop01:17
wpwrakmth: the only problem is that it drives a borderline design all the way into deepest cuckoo land01:18
mthkexecboot makes more sense then01:18
Action: kristianpaul looks at Jay7 01:19
wpwraktotally. and larsc  already has the weapon of u-boot destruction ready for when Jay7 is finishes :)01:19
wpwraks/is //01:20
larsci hope so01:20
wpwrak*grin* on cue01:20
larscIt works on the JZ476001:21
larscbut I think I'm going to start from scratch and write it that way, that it can be used both for the minimal bootloader and as the usbboot firmware01:22
larscso there is less code duplication in our tools01:22
wpwrak(4760) does this imply that the 4760 wouldn't quite so much of a horrible thing to support anymore ?01:23
wpwrak(usbboot) hmm, never looked into this one so far01:23
larscwpwrak: i added basic support for it yesterday to our kernel repoi01:24
wpwrakwheee !01:24
larscso far non of the more advanced peripherals is supported01:25
wpwrakif a ya would happen with the 4760 instead of the 4740, i think that would be hugely better01:25
kristianpaulWhat device is using the 4760? or is a EVB?01:25
wpwrakrome wasn't built in a day :)01:25
larscbut there is a guy working on a jz4750 port, which is similar to the jz4760 so i'm not the only one working on getting better support01:25
larsckristianpaul: the evb01:26
wpwrak(4740 vs. 4760) my main concern at the moment is memory. without DDR, you're kinda screwed today ...01:26
wpwrak(ddr) e.g., according to some very superficial studies of mine, for the price of the ben's memory, you could get about twice the amount of SDR today, or four times DDR01:27
wpwrakand that may not even be the end of the story. with "fat" smartphones driving up the demand, quite large memories may become cheaper than smaller ones01:29
wpwrak(well, after the supply catches up :)01:29
wpwraki also like the huge number of MMC/SD/SDIO in the 476001:30
kristianpaulVideo also may allow higher resolution i guess?01:32
kristianpaulis that the dual cored xburst? or just more speed?01:32
kristianpaulhow fast?01:33
larsc600 Mhz01:33
kristianpaulthats a laptop :-)01:33
wpwrakddr is faster than sdr, so you get a lot more memory bandwidth for video01:33
larscit actually is a tripple core01:33
kristianpaulwhat?? :O01:33
larsccpu, vpu, gpu01:33
kristianpauli see01:33
larscthe gpu is 2d-only though01:34
wpwraklarsc: how useful is the vpu ?01:34
larscand i've not seen any docs for either the vpu or gpu01:34
larscwpwrak: without docs not so much01:34
wpwrakokay, that answers it ;-)01:34
kristianpaulha! :/01:34
wpwrakyeah. with what little i saw, it looked kinda semi-useless01:34
larscthe PM states though:01:35
larscRealVideo decoding up to 720P 30fps01:35
larscMPEG-2 decoding up to 720P 30fps01:35
larscMPEG-4 decoding up to 720P 30fps01:35
larscVC-1 decoding up to 720P 30fps01:35
larscH.264 decoding up to 720P 30fps01:35
larscMPEG-4 encoding up to VGA 20fps01:35
wpwrakoh, for video it's probably cool. dvdk ought to be happy :)01:35
wpwraki was thinking more of a "real" 2nd core01:35
larscV                                                                                 PU01:36
kristianpaullarsc: usb on-the-go?01:36
larscincludes an XBurst® CPU with 48kB TCSM and without CACHE, MMU, CP0, DBG and FPU.01:36
larsckristianpaul: yes01:36
kristianpaulno MMU?01:36
larscthe VPU01:36
kristianpaulah, thats second core01:36
larsckristianpaul: ftp://ftp.ingenic.cn/2soc/4760/JZ4760_pm.pdf01:37
wpwrakalright, no smp01:37
larscthis scares me: ftp://ftp.ingenic.cn/2soc/4760/JZ4760B_an01_vs4760.pdf01:39
kristianpaul apt-get install mysql-server php5-cgi php5-mysql01:40
wpwraklarsc: because figure 1 says "Jz4760E" ? ;-)01:46
kristianpaulhaha, nice catch01:46
wpwrak2005 .. kinda ancient01:47
wpwraki wonder how much of that copyright refers to products long forgotten by history01:48
wpwraksection 5 is nice :)01:49
Action: kristianpaul is stoned wich so much features01:50
mth"Change TCSM1 from 48KB to 32KB"... that's the RAM for the VPU, right? 48KB was already small...01:50
wpwraklarsc: do you deal with ingenic directly or does this go via wolfgang ?01:51
wpwrakall these changes pale in comparison to what samsung did when going from the s3c2440a to the s3c2440b01:51
wpwraki mean, no much could *possibly* change on such a minor step, right ?01:52
wpwrakwell, all the pull-ups became pull-downs, kinda overnight ...01:52
larscwpwrak: because there is a whole bunch of minor changes in register layout and such in a minor chip revision.01:54
larscwpwrak: Takes the fun out of coding because you'll endup with spaghetti code01:55
larscwpwrak: i got the evb though an third party01:55
wpwraklarsc: are you sure the non-B 4760 will even exist in the wild ?01:56
larsci have one ;)01:56
wpwrakan evb ;)01:56
larsci'm pretty sure it exits in the wild01:57
wpwrakperhaps a question for wolfgang's next visit to ingenic01:59
larschm or maybe not01:59
mthlarsc: tonight we found the cause for the SD errors after resume from suspend: it turns out the cpufreq driver triggers it01:59
mththis also explains why the NanoNote does not have this problem01:59
wpwraklarsc: btw, do you deal with ingenic directly or does this go via wolfgang ? (or maybe even someone else ?)01:59
larscwpwrak: i got the evb through an third party02:00
wpwrakinteresting .. the 4740 is not recommended for new designs02:00
larscit's 5 years old02:00
wpwrakah yes, and they don't even mention the 4720 :)02:01
wpwrakgood. that settles this question. 4760b looks very likeable, even if it should have 2 out of 3 cores unusable02:02
larscwpwrak: I told wolfgang that Ingenic has a news item on their page mentioning that the linux kernel now has jz474002:02
larscwpwrak: And he answered that it might be time for a visit again02:02
wpwrakgood :)02:02
wpwraklarsc: (3rd party) so they would be sort of the first ones to have a linux-based device with the 4760(b) ?02:03
larsci've never seen the real device, but i guess yes02:05
wpwrakhmm, seems that wolfgang really ought to visit ingenic ...02:06
larscbtw. i got that board back in october02:06
larscso it's very likley that the product is already on the market02:06
wpwrakso it should be a B by now :)02:06
wpwrakhmm. you added basic support just recently ? so the product doesn't use linux ?02:07
larscit does02:09
larscbut is based on ingenics sources02:09
wpwrakah, dead-end then ;-)02:09
wpwraki'm asking because i'm wondering if there may be an opportunity to have ingenic finance the making of the ya. sort of as showcasing the 4760b. the more open the design, the more reproducibility.02:11
wpwrakkinda like ti is behind the various "1st" omap boards02:12
wpwraknot sure if ingenic think this way, though02:12
larscthey do have their evb which showcases the jz4760, while the ya would probably only expose a subset of the jz4760 functionality02:14
larscand i'm not sure if they are interested in building a community yet02:14
larscthough it is worth a try02:14
wpwrakevbs are kinda strange devices. a ya could be a lot more convincing. like companies going crazy with trying to clone the iphone/ipad, but very few hot for the (probably) underlying reference designs.02:16
wpwraksharism/qi-hw ought to have earned a few credentials by now. they must have expected it to be long dead already ;-)02:17
larscthey signed a licensing agreement with MIPS, so I guess they are targeting the non-chinese market now too.02:19
wpwraklarsc: that would be very much in our favour02:20
larscmaybe they could be convinced that it is in their interest to have something like the Ya as a reference design02:20
wpwraklarsc: yup. now, where's wolfgang when we need him ? :)02:21
larscespcially when beeing told that companies like TI and Freescale do the same02:21
larscwpwrak: probably asleep for another two hours02:21
wpwraklarsc: oh yes, sometimes reminding companies of their competitors works wonders ;-)02:21
wpwraklarsc: (wolfgang) until ~1 pm, beijing time ? did he do some particularly hard partying last night ? ;-)02:23
kristianpaulwpwrak: wolfgang have a meeting with a robotic company i remenber02:24
wpwrakprobably skynet inc.02:25
larscbtw. as far as i can see the JZ47XX SoC family chips are among the few MIPS based SoCs(If not the only ones) that are competetive in the embedded mobile market.02:26
kristianpaulwpwrak: ucrobotics.com02:26
larscso it might as well has been in MIPS Inc.02:26
wpwrakkristianpaul: cute :)02:27
kristianpaulwpwrak: yeah02:27
larscso it might as well has been in MIPS Inc.' interest to have them expand to a global market02:27
wpwraklarsc: ah yes, that would make sense. mips' role is kinda obscure anyway02:28
larscARM has become omnipresent while the same can't be said about MIPS02:29
larscthey are still used in quite a lot of routers and such02:29
wpwrakyeah, you have a point there ;-)02:29
larscbut i've yet have to see a MIPS based mobile phone02:30
wpwrakbasically arm became the intel of embedded while mips became the via02:30
qi-bot[commit] Xiangfu Liu: add reset help http://qi-hw.com/p/xburst-tools/9b840cd03:14
qi-bot[commit] Xiangfu Liu: remove duplicate code http://qi-hw.com/p/xburst-tools/759777903:14
qi-bot[commit] Jo-Philipp Wich: The cxxtools static library is not built anymore after autoreconf, do not attempt to package it http://qi-hw.com/p/openwrt-packages/61c62b603:56
qi-bot[commit] Jo-Philipp Wich: Rely on standard autoreconf facility, remove manual autoreconf all, define prereq on host guile http://qi-hw.com/p/openwrt-packages/b14002b03:56
qi-bot[commit] Xiangfu Liu: [mtd.nn] using /bin/sh http://qi-hw.com/p/openwrt-packages/b91445503:56
rjeffries_Barebones may be interesting if someone does a MIPS version07:06
wpwrakwolfspraul: have you seen the discussion with lars about the 4760/4760b in the irc log ?07:15
panda|x201wolfspraul, ah, nice talk with you but with bad coffee :-(07:22
wpwrakrjeffries: heh, "do everything just like in the kernel but avoid actually *using* the kernel at any price". sometimes you really have to wonder :)07:26
larscwpwrak: it's the kernel without userland support07:27
larscsingle-user linux07:28
wpwraklarsc: how much of a difference would user space notice, once the kernel has the necessary drivers in place ? should be pretty much the same, no ? except for things that explicitly touch the hardware (evil-menu2x, bit-banging, ...)07:42
larscwpwrak: what do you mean?07:43
larscdifference compared to what?07:44
larscyou mean for a jz4760 based board compared to the nanonote?07:44
wpwraklarsc: changes required in user space, compared to what we have for the ben07:44
wpwraklarsc: yes07:44
larscwpwrak: i was able to boot the nanonote userland without problems on the jz4760 evb07:45
wpwrakoooh ... sorry, i got confused about the threads ;-)07:45
larscnow it makes sense07:46
wpwrakthought you were referring to the 4760 discussion but instead you meant barebox :)07:46
wpwrakso barebox is a patched kernel ?07:46
larscit's more of a light-weight api compatible reimplementation07:47
larsci think the idea was to be able to reuse kernel drivers07:47
larscbut i must admit that i haven't actually looked at the code07:48
wpwraksounds a bit like u-boot. there are also copies of kernel code in there. of course, in heavily modified permanent forks ...07:49
larscbarebox is u-boot v207:49
wpwrakyeah. new shinier outfit, all the same design flaws ;-)07:50
wpwrakhigh time for kexecboot. let's stop all the nonsense :)07:50
larsci wonder if people realize that you can boot a linux kernel with a minimal initrd in about the same time as uboot07:50
wpwrakquite obviously they don't ;-)07:51
qi-bot[commit] Werner Almesberger: usrp/sps/20110306: examine sidebands ("spikes") and try to reduce them http://qi-hw.com/p/ben-wpan/a60fb1e07:54
wolfspraulwpwrak: no did not see the discussion.07:55
wpwrakwolfspraul: the highlights: lars is getting the 4760 to boot with "proper" linux. 4720/4740 is not recommended for new designs http://en.ingenic.cn/product.aspx?CID=1107:57
wpwrakwolfspraul: so all this suggests that the 4760 may be a better choice for a ya than the 4740, and also that this choice may be more viable than previous discussion suggested07:58
wpwrakwolfspraul: i also brought up the idea that ingenic may perhaps be interested in sponsoring (all or a part of) ya development, in order to a) mainline 4760 support, and b) have a real-life reference device07:59
wpwrakwolfspraul: ah, and 4760 has been replaced by 4760b, with many cute little changes: ftp://ftp.ingenic.cn/2soc/4760/JZ4760B_an01_vs4760.pdf08:00
wpwrakwolfspraul: maybe you want to visit ingenic some day :)08:00
wolfspraulha yes, that's a funny page (Ingenic's)08:00
wolfspraul4730 is indeed bad, but putting 4740 on the list of 'not recommended' just shows how much trouble they have getting people off of 4740.08:02
wolfspraul4750 and 4755 were never a commercial success, and they won't become one anymore this late in the game.08:02
wolfspraulI hope 4760 will be a big hit for Ingenic, because otherwise the slowly shrinking 4740 margins may squeeze them badly.08:03
wpwrakah, political EOL :)08:03
wpwrakwell, then they should like the idea of sponsoring a shiny device with an open design even better ;-)08:04
wolfspraulI highly doubt that will happen either. You don't understand how much and how long they are holding on to every last penny.08:04
wolfspraulexpecting such kind of sponsorship beyond kind words is most likely unsuccessful.08:05
wpwraki guess it depends on how desperate they are to make people "get" the 476008:05
wpwrakand of course your sweettalking skills ;-)08:06
wolfspraulno way, this is a business.08:07
wolfspraulno 'sweettalking'08:07
wolfspraulthey are not desperate, hopefully 4760 is a great product.08:07
wolfspraulyou cannot bribe your own customers...08:08
wolfspraulLars mentioned that they put some full programmers' manuals on the ftp server now?08:08
wpwrakhave you seen siemens lately ? ;-)))08:08
wolfspraulI have to check which ones are there actually, this would be great news.08:08
wpwrak_ds is there. board design guide as well08:08
wpwrakhaven't seen _pm08:08
wpwrak(business/sweettalking) marketing is also part of business ...08:10
wpwrakin a way, trying to get vendors to sponsor development work would be a much more honest model than your favoured earn-by-selling devices approach. there's a lot of free publicity and long-term value that's being created in this project that has no relation to short-term revenues. it would only be fair to remind those who derive a strong benefit from this of this fact, and to have them help with the effort08:12
wpwrakearn-by-selling has its place, but that works a lot better once you've made it past this grossly inefficient long-term investment period08:13
wolfspraulthis is called 'technical marketing' and long and aggressively being done already, with strong metrics in place to make sure there is good payback.08:15
larscwpwrak, wolfspraul: all of the _pm.pdf are on their ftp08:15
wpwrakalso, if ingenic are trying hard to get people to accept the 4760, this may be a short window of opportunity. there is nothing to be won by just waiting for it to close08:15
wpwraklarsc: where ?08:15
wolfspraulI would need to learn a lot of technical marketing tricks first before daring to go into this business - not interested.08:15
larscwpwrak: ftp://ftp.ingenic.cn/2soc/08:16
wpwrakwolfspraul: what would be the downside of asking ?08:16
wolfspraulwpwrak: relax, I have been discussing this with Ingenic many times, and many months ago, and I will continue to do so.08:16
wolfspraulbut Ingenic is a business and they have been setup to make money (sell), not to spend money.08:16
wolfspraulmaybe something can be worked out in conjunction with one of their customers.08:16
wpwrak(docs) wow. this is wonderful !08:16
wolfspraulyes, I agree. that's great news!08:16
wolfspraulwpwrak: hmm. looks like they screwed up otg on the 4760?08:22
wolfspraulso the 4760b fixes that08:22
wolfspraulI just quickly glanced over the 4760b pdf.08:23
wolfspraulyou need to read this the Chinese way. those docs are super euphemistic.08:23
wolfspraulbasically every new 'design' is like a big gamble08:23
wolfspraulbecause they work with totally under-qualified employees08:23
wolfspraulwill it work? which bugs do we have?08:23
wolfspraulonly after some time you see whether this thing actually works, and how many features have to be 'removed' :-)08:23
larscjust like software08:24
wolfspraulof course their customers know this as well, and work with similar staff. quickly turn aroudn the product, throw to the market, see whether it sells (i.e. whether it works)08:24
wolfspraulI mean if a non-working product would sell they would of course love this business too, but it's rather unlikely...08:24
wolfspraulsince I started working with Ingenic, all their staff has changed, except for the top level management which is far removed from technical reality.08:25
wpwrakyeah, revisions are usually mainly admissions of prior screw-ups. nothing new there :)08:25
wpwrakoh dear08:25
wpwrakdo you still consider them a viable source then ?08:26
Action: wpwrak wonders if this could be used to our advantage when discussing sponsorship. "given that all the engineering has changed [which i haven't told so many people yet] we would like to help strengthen the design-credibility of the 4760" ... sometimes, a little polite blackmail can go a long way ;-)08:29
wpwrakand even more so if it's actually true :)08:30
wolfspraulwpwrak: of course all is fine with them. If the pm's are open now that's great.08:32
wolfspraulthey are a chip supplier, that's all. I need to check how good the 4760 really is, and whether we take unnecessary design risks or whether it's safe.08:32
wolfspraulabout 'sponsorship', I already said - if a lot of smiles and nice words is good, that's possible08:33
wolfspraulalso some free lunches08:33
wolfsprauland declarations of eternal friendship, new era of whatever08:33
wolfspraulbut rest assured they will hold onto every penny of cold hard cash like their life is on the line08:33
wpwrakso you're saying they don't believe technical marketing is worth the effort08:34
wolfspraulChinese companies are not that advanced yet, I think.08:34
wolfspraulplus the best company to propose technical marketing to them would be a successful existing technical marketing company.08:34
wolfspraulI don't even know by what metrics they work, report their results, etc.08:34
wolfspraulI'm just a product guy.08:34
wpwrakwell, maybe you can share some reactions with them. ask a few people what they think of the 4760b. mention how much changed since the 4760. mention that it's a new team, not the one that made the - proven - 4740 and friends08:35
wolfspraulI just need to ask how it's selling, and how 4740 is selling.08:35
wolfspraulI've heard that supposedly there is a 'xburst android phone already shipping in Europe'08:36
wolfspraulthat sounds like the uber-version of wishful thinking, all combined into the least amount of words08:36
wpwraksure, you may not be the best advisor on this. but then, you're someone who does have something that could help, namely a product where everyone can look under the hood as much and as long as they want08:36
wolfspraulbut I can look into it, say by trying to buy one of those 'android xburst phones already selling in europe'08:36
wpwrakyeah, a bit of extra information can't hurt08:37
wolfspraulwow nice, just checking on ingenic ftp server08:44
wolfspraulthe pms are indeed all there08:44
wolfspraulI'd say that alone puts them into a rare category, no? I haven't checked what Marvell, Freescale, Qualcomm, TI and the others are doing, but full documentation without NDA is at good as it gets. Probably redistribution rights are still missing, simply because they just cannot understand what that thing could possible be.08:46
wolfsprauleven in several meetings, I was unable to convince them that this strange additional paragraph I ask them to add isn't some kind of trap that they just don't understand :-)08:46
wolfspraulso whatever, screw the redistribution rights08:46
wpwrakwe have dsv to work around these ;-)08:47
wolfspraulI get a phenomenal 2 KiB / second from their FTP server, that hasn't changed a bit in the last 2 years.08:47
wolfspraulcan wget do a recursive retrieve on a whole ftp tree?08:48
wolfspraulI think I will backup everything just in case.08:48
wpwrakwget -r seems to work just fine ;-)08:49
wpwrakBOOKSHELF will be next08:49
wpwrakand indeed, this little move makes them quite a lot better for our purposes08:50
wpwrakwolfspraul: you should send them flowers or whatever is appropriate :)08:51
wpwrak2.14 kB/s here08:52
wpwrakinteresting. there's now a jz4720_pm08:53
wolfspraulI'd say 50% or more of the reason they open the pms is because of Lars work and the kernel.org inclusion08:55
wolfspraulthe other 50% is the MIPS license they finally had to take, and start paying money to MIPS (in other words, their 'patent-free' strategy failed)08:55
wpwrakso they no longer have to be afraid08:55
wolfspraulso they think08:56
wpwrakanother item for your next visit: find out what made them change their mind08:56
wolfspraulthay haven't heard about freelancing patent trolls yet, and I will not enlighten them08:56
wpwrak!censor enlighten08:57
wpwrak(i hope qi-bot got the hint ;-)08:58
wpwrakthe deal with mips may also provide some protection in this regard08:59
wolfspraulnot sure the patent trolls care, and it would be hard for MIPS to find a reason for them to stop08:59
wolfspraulI will let a wget -r run over the 2soc tree now... (or the next days, it seems :-))09:00
wolfspraulmaybe I can get permission to mirror it at downloads.qi-hardware.com09:01
wpwrakwell, mips could challenge their patents. look at google fighting patents [potentially] used against android09:01
wpwrakoh yes. intermittent 2 kB/s be gone :)09:01
wpwrakbut let's finish the wget -r first, then the dsv setup, ...09:03
wolfsprauldid David already succeed in getting XBurst hardware acceleration into mplayer?09:06
wolfspraulhis mail sounds terrific to me...09:06
wpwrakhe seems to be partially there. yes, amazing stuff09:06
wpwrakall the people clamoring for their iWhatever-grade multimedia nanonote may still get their wish :)09:07
wpwrakmy wget is struggling. connection stalls all the time :-(09:11
wolfspraulactually I may demo milkymist one to Ingenic next time09:12
wolfspraulwe have a whole list as usual, and surely 4760 is there as well. but I will demo m1 and see what they think.09:13
xiangfuwpwrak: mine is 116Kb/s.09:13
wpwrakxiangfu: so you're the one snatching all the bandwidth !!! :)09:13
wpwrakwolfspraul: sounds cool. you never know what may interest them. when do you think of meeting them ?09:14
xiangfuwpwrak: then I can try to send them to you. if we can put those file in our server??09:15
wolfspraulno fixed plan, maybe in a few weeks. I have some urgent things to settle first, like an upgrade of our payment gateway.09:16
wpwrakxiangfu: we can't put them there officially ...09:18
xiangfuwpwrak: ok. you can try to download in my blog server. if my blog is faster then ingenic's server.09:19
wolfspraulxiangfu: unless they added a redistribution clause, let's wait until we get an official 'ok' in a meeting with Liu Qiang.09:19
xiangfuwolfspraul: ok.09:19
wolfspraulit will be hard to even explain this question, but I gladly go through this inter-cultural experience :-)09:20
wpwrakwolfspraul: the technical argument alone may be sufficient :)09:22
wolfspraulcorrect. I will offer to donate bandwidth to them.09:31
wolfspraulthey count every penny (literally), and the ftp server is so slow because it's the cheapest they could find.09:31
wolfspraulI am not kidding.09:31
wolfspraulif you think about that for a while you realize that your idea of 'sponsorship' (where you actually think about thousands or tens of thousands of USD at least), may not work all that well :-)09:32
xMffmaybe they should offer to send CDs for a fee09:32
wolfsprauldon't give them that idea09:32
xMffI won't :)09:32
wolfspraulthey will hire some people and maybe can make a few cents on it.09:32
xMffits a pita for casual tinkerers09:32
xiangfuwpwrak: Downloaded: 79 files, 203M in 24m 45s (140 KB/s): http://www.openmobilefree.net/other/Ingenic_Docs/ftp.ingenic.cn/2soc/09:35
wpwrakwolfspraul: ever heard "pennywise smart, poundwise foolish" ? ;-) you should thank them for saving all the pennies so that they can pay you many pounds :)09:37
wpwrakxiangfu: thanks ! i like the bandwidth of your server ;-)09:38
wolfspraulnice try. they have much more money than me, and I am not just trying to get some of that? None of this will work.09:38
wolfspraulah yes, xiangfu just did the Chinese solution09:38
wolfsprauljust finished downloading the 4760 pm, and yes, it looks like it's all complete and all public09:42
wolfspraul917 pages, soon they are in the 1k+ pages datasheet club09:42
wpwrakwolfspraul: (none of this will work) and only certain way of failing is to never try :)09:45
wpwrakftp.ingenic.cn actually picked up quite a bit of speed now. 80 kB/s !09:48
wpwrak100 kB/s.09:48
El-Aurianhey, hows everyone doing?10:55
El-Auriani'm a linux user, and i've got my hp printer to work, using foo2hp... but i'm having trouble getting the printer to work on a windows computer10:56
El-Auriani'm wondering if its possible to use the linux driver... since i think its a PostScript driver, on a windows machine?10:57
El-Auriani don't think PostScript is a compiled binary, but rather an answer file, so i hoped it was possible... does anyone have any experience with this?10:57
jow_laptopEl-Aurian: wrong channel, sorry10:57
El-Aurianjow_laptop: kk, which channel should i ask on?10:58
jow_laptopEl-Aurian: afair there is a general #hardware channel10:58
El-Aurianjow_laptop: yeh, its ##hardware, thanks :)10:59
kyakjow_laptop: there is a problem. libopcodes.so is not built anymore by binutils (only libopcodes.la). It is a part of this problem: https://dev.openwrt.org/ticket/8603. It was addressed in gcc-mips by cp $(STAGING_DIR)/usr/lib/libopcodes*.so $(1)/usr/lib11:22
kyakbut now even this ugly hack won't work11:23
kyaksomeone suggested to install libopcodes.so in objdump package (https://dev.openwrt.org/attachment/ticket/8603/binutils_libopcode.patch), but this patch wasn't accepted (though it wouldn't work now anyway)11:25
jow_laptopkyak: not even libopcodes.a ?11:26
kyakjow_laptop: there is libopcodes.a11:28
kyaki'm trying now if --enable-shared would work11:28
jow_laptopkyak: there's some libtool heisenbug that, depending on the used autoreconf version (wtf?), randomly disabled or enables shared/static library building11:29
kyakheh. does it mean we should always provide --enable-shared or --enable-static?11:30
jow_laptopno, thats the default11:31
jow_laptopat least its opposed to be11:31
jow_laptopif the wheather conditions are okay, and there's no solar maximum etc.11:31
kyakjow_laptop: ok, the --enable-shared made libopcodes.so build11:32
jow_laptopkyak: ok good, anything to fix on the openwrt side?11:34
kyakjow_laptop: just a minute, i'm going to put together the --enable-shared and installation of libopcodes11:35
dvdkdon't migrate to the jz4760 yet, the jz47xx_vid might need a complete rewrite (different IPU used, afaics) :)11:44
kyakdvdk: tested your driver for mplayer, it worked fine! However, the video was still very slow (no different from -vo sdl)11:50
kyakdvdk: and fbdev output always has these horizontal lines on top and bottom of screen.11:51
kyaki.e. i can see the console in background11:51
dvdkkyak: nice to hear it works11:53
dvdkabout your complaints: just details easiy fix.11:53
dvdks/easiy/easy to/11:53
dvdkkyak: when the scaler works we can use slightly smaller video and zoom up to full-screen.11:54
dvdkthis should get us proper 30FPS video11:54
kyakslightly smaller, why?11:54
dvdktheora full-screen decode at 30FPS is too slow for me currently11:55
dvdkbut 240x180 or something should work.  and scaling is practically for free11:55
kyakah, ok11:56
dvdkmaybe ffmpeg-theora is going to be fast enough, though11:56
dvdkalso, re-encoding with 15FPS is an option11:57
kyaki think it will be noticeable?11:57
kyakjow_laptop: http://pastebin.mandriva.com/2205912:17
kyakjow_laptop: apparently, installation of libbfd*.so is also necessary.. Otherwise binutils complain that it missing (guess it is due to --enable-shared)12:18
kyakjow_laptop: thanks a lot for your support :) moving on, jfbterm is the next failure12:18
qi-bot[commit] kyak: jfbterm: fix build http://qi-hw.com/p/openwrt-packages/97ed0cb12:41
kuribasHi, if the nanonote processor supports USB 1.1 host, why doesn't the nanonote support it?13:09
dvdkkuribas: because nobody cared to implement it?  maybe no 5V USB bus-power available to create a host port?13:10
kuribasImplement in hardware you mean?13:11
wpwrakkuribas: the main obstacle is that the USB host signals are not "exported" from the chip (the chip is bonded directly to the board)13:12
kuribasI see.13:12
wpwrakkuribas: why that happened, i have no idea. maybe it was an oversight. maybe there were technical reasons.13:12
kuribasHost mode is something I would find very useful.13:13
wpwrakdvdk: and yes, you'd need 5 V, too. if you feel sneaky, you could just run a cable from USB device VBUS to USB host VBUS ;-)13:13
wpwrakkuribas: you're not alone ;-)13:13
kuribasI would like to connect a USB audio interface to the nanonote, and have a portable audio recorder!13:14
kuribasThe zipit has a hack to make usb host mode possible.13:15
wpwrakkuribas: you could try to remove the black stuff that covers the chip and add your own bond wires. you may need a few bens before you get this right, but i'm sure wolfgang won't mind selling you a bunch of ten-packs or so :)13:17
kuribasNo thanks...13:18
kuribasI wanted to make an open source recording solution (open hardware), but it shouldn't be hack.13:18
dvdkkuribas: people are working at adding hardware add-on cards to the Ben's micro-sd slot.  adding a A/D converter plus buffer might not be too dificult if you have some hw experience13:25
dvdksee here: http://en.qi-hardware.com/wiki/UBB13:25
kuribasInteresting, but it is still kind of a hack.13:33
kuribasIf I create a usb interface, it would be silly to make a connector that would only work for one device.13:34
kuribasAn audio interface I mean.13:34
kuribasIt's interesting for one off projects though.13:36
dvdkkuribas: an off-the-shelf usb audio capturing device might use more power than the complete Nanonote itself.13:39
dvdkkuribas: btw what's wrong wtih the ben's built-in microphone ;)13:39
kuribasdvdk: Yeah, I am considering an external power supply.13:44
kuribasIt would be for a professional audio applications.13:44
kuribasThere are some nice atmel chips that support SD cards, high speed USB 2.0, FAT, SSC, LCD... Maybe it would be easier no make a simple linux recording device.13:46
kuribasRather than trying to connect it...13:47
kuribasIt would also reduce usb overhead.13:48
kyakjow_laptop: the next failure is kexec-tools (from openwrt base packages). The error is strange: ./configure: line 5923: syntax error: unexpected end of file. Since there is a newer kexec-tools in openwrt-trunk, and it builds fine (i just checked), could it be backported to backfire?14:05
kyakjow_laptop: i think it makes more sense then poking the older kexec-tools...14:06
dvdkkyak (jz47xx_vid is as slow as sdl): maybe you should redirect debug output to dev/null to speed it up a bit14:37
dvdkcurrently it's extremely verbose14:37
qi-bot[commit] Xiangfu Liu: config.full_system: add openssh-sftp-server http://qi-hw.com/p/openwrt-xburst/99830f715:01
qi-bot[commit] Xiangfu Liu: update opkg.conf, don't using /tmp keep packages information http://qi-hw.com/p/openwrt-xburst/ca5134615:01
jow_laptopkyak: sorry, was afk - I'll take a look now15:54
kyakjow_laptop: no problem!15:56
jow_laptopkyak: oh btw, the mandriva pastebin is not so good15:57
jow_laptopit eats the @@ hunk headers15:57
kyakdvdk: it didn't really help -\ i think i'll just wait for the driver to become more stable :)15:57
kyakjow_laptop: hmm,, you are right. Will change the provide in wgetpaste :)15:59
kyakjow_laptop: http://dpaste.com/481542/ :)16:01
jow_laptopkyak: and the kexec-tools in trunk/package work fine?16:01
kyakjow_laptop: yep, it works fine16:03
kyak(well, it builds fine, but still some problems with passing cmdline)16:03
jow_laptopok I'll backport it16:05
kyakfast :) thank you!16:07
qi-bot[commit] kyak: gcc-mips: fix build http://qi-hw.com/p/openwrt-packages/a0b91c616:08
kyakxiangfu: ping16:09
xiangfukyak: Hi16:12
kyakxiangfu: hey :) could you please merge the latest backfire? It would fix kexec-tools compilation issue..16:13
xiangfukyak: sure.16:14
xiangfukyak: I will work on that tomorrow morning. now is 00:12 here. I don't want mess up the git. :)  I am working on build the RTEMS crosschain. try to write a Makefile. automatic a little. :)16:15
kyakxiangfu: sure, thanks :)16:16
kristianpaulxiangfu: RTEMS nice !16:16
kristianpaulMakefile, awesome !!!16:16
xiangfukristianpaul: just start a little : http://www.openmobilefree.net/other/downloads/tmp/Makefile16:17
Action: kristianpaul had re-compiled rtems severals times last weeks by hand...16:17
Action: xiangfu have no idea should I setup a repo for this makefile.16:18
xiangfukristianpaul: first compile toolchain. them rtms.  :)16:18
xiangfus/them rtms/then rtems16:21
kristianpaulmake vs bash ;-)16:24
kristianpaulI was thiking in a bash script but this makefile looks interesting16:24
kristianpaulAlso if allow to re-compile just some parts of rtems, ie when changing a driver source.16:25
kristianpaulI guess rtems handle that in the boostraping16:26
wpwrakhmm ... how to call that UART board ... UART is unpleasantly generic. NanoUART ? NXU (Nanonote eXternal Uart) ? NNXU (NanoNote eXternal Uart) ? NXUART ? NNXUART ? NX-Serial ? NNX-Serial ?16:48
wpwrakhehe, just typo'ed NNXUNART :) uarts are a bad habit in a way indeed ;-)16:50
wolfsprauli'd say nxu or nxuart, maybe nxuart16:50
wpwrakanyone else ?16:54
wpwrakalright, now we just need one dissenter to show that the vote wasn't rigged16:55
xMffwhat about (NanoNote|Ben) Serial Extension?16:58
xMffNSX or BSX16:58
wpwrakNSX is a car. BS-X ? ;-) also, i'd hope that all the 8:10 cards that work in the ben will also work in future nanonotes16:59
qi-bot[commit] Werner Almesberger: BOOKSHELF.ingenic: Jz4720/Jz4740/Jz4760 manuals http://qi-hw.com/p/wernermisc/68e84b317:31
rjeffrieswpwrak nxuart is cool, nxs or nxu also fine17:40
wpwrak3:1 seems that we have a winner :)17:42
rjeffriesso much easier in irc than say, in Libia17:44
rjeffriesso happy i did not typo as labia17:44
wpwrakoh, in Libya, voting is probably also easy, as long as you know what's good for you :)17:46
kyakwpwrak: what is nanonote uart?17:57
wpwrakthis: http://downloads.qi-hardware.com/people/werner/uart/pix/17:58
kyakso it's basically what we were discussing the other day, UBB+hardware IC that implements UART?18:00
kyakcool :)18:01
wpwraki posted about it on the mailing list already one month ago (feb 6), subject "a bevy of boards"18:01
wpwraknobody reads my walls of text. sneef :(18:01
kyakmy bad, i'm reading the mailing list with "one eye"18:03
kyakas opposed to IRC18:03
kristianpauli read it wpwrak18:04
kristianpaulbut i'm not interested in uart for now18:05
wpwrakkristianpaul: ah, i have at least one true fan who follows my writings. thanks :-)18:05
wpwrakkristianpaul: (interested) also for me, the host interface is a lot more interesting than the actual uart ;-)18:06
wpwraklfuse set to 0x60. and nxuart no longer responds. yes ! just as it should ;) now let's add that clock output ...18:11
rjeffrieswpwrak I read every word, every slight nuance of your screeds on mail list and here18:13
wpwrakrjeffries: aha, that would make you a member of the secret police then :)18:14
rjeffrieswpwrak what feedback if ANY DO YOU HAVE FROM EARLY USERS OF ATBEN ATusb ETC18:15
rjeffriesWPWRAK YOU CAN RUN, BUT YOU CAN not HIDE. We are everywhere and nowhere.18:15
rjeffriessorry of caps lock18:15
wpwrakso far, i'm the only early user. there's one early carrier and one early storer, though, but i haven't heard from these yet. in fact, the less i hear from them, the better ;-)18:16
wpwrakrjeffries: for your .xmodmap: remove Lock    = Caps_Lock18:16
rjeffrieswhat data transfer if any owrks between atben and atUSB?18:17
wpwrak(early carrier) i.e., fedex. now somewhere between memphis and anchorage. the early storer would be correo argentino, who apparently still have the parcel sitting in the post office, a few blocks away18:17
wpwraki have some experimental communication programs, but nothing really useful18:18
rjeffriesas a matter of suriosity what portions of communication stack do you have so far? I assume very minimal18:19
wpwrakof the "real" stack ? nothing18:20
wpwrakmy tools are for design verification18:20
rjeffriesfair enough18:21
rjeffriesI m wondering what quantity wolfspraul may produce in first run of those two boards. likewise the uart 8:10 for Ben (which may be the bigger seller near term)18:22
wpwrakthere are a few bits and pieces of an ieee 802.15.4 and ipv6 stack (6lowpan) that will be reusable18:22
wpwraki don't think wolfgang plans to make any nxuart18:23
rjeffriesoh really?  because?18:23
wpwrakfor wpan, he mentioned something like 100 kits (atben plus atusb)18:23
wpwrakno use case18:24
rjeffriesn=100 sounds about right or maybe 100 atBen and more atUSB18:24
rjeffriesno use case? wtf??18:24
wpwrakbut maybe tuxbrain is interested ? after the ubb success story, that would be a nice progression18:24
wpwrakor you could make them. have to find a less expensive pcb source, though :)18:24
rjeffriesI gave up on that first outfit, and frankly do not wish to compete with Tuxbrain or tuxbrain_away18:25
wpwraki doesn't have to be competition - it could be a complement18:26
rjeffriesa less elegant way to get serial functionality would be a little pcb that uses UBB as connection to Ben, then a piece of flat ribbon cable to a small pcb18:27
rjeffrieswould heve more room that way18:27
wpwrakproducing things takes time and some investment. everyone has only so much of each. so if wolfgang isn't interested and tuxbrain is busy, there would be no conflict if you picked up that production18:27
rjeffriesone thing I do not understand (among many...) supposedly there are as many as 400 Ben NN owners in USA18:28
rjeffriesI see ZERO feedback on the list or in irc from USA besides my own ranting and raving. makes me wonder wtf?18:29
rjeffriesare all of those Bens "drawer computers" ???18:29
wpwrak(ubb to cable) kinda like the femtoduino. why not. there are many possible ways to approach this.18:29
Action: rjeffries googles "femtoduino"18:30
wpwrakrjeffries: it may be a somewhat cultural thing. we also saw this in openmoko. about as many people in the us and in germany, but you saw comparably little activity from the ones in the us18:31
rjeffriesOMG http://www.varesano.net/projects/hardware/Femtoduino18:31
wpwrakrjeffries: not entirely sure why either. it's not that free software or diy hardware wouldn't happen over there :-)18:31
wpwrakcute, isn't it ? :)18:31
rjeffriesalso n=400 is a LOT of people you'd expect a few are active18:32
wpwrakwell, jane may be in the us18:33
rjeffriesfemtoduino hits all the right freedom notes, license etc, used KiCad18:34
rjeffrieswonder if anybody is selling it18:34
wpwrakyeah, femtoduino looks quite nice. i particularly appreciate the emphasis on smallness :)18:35
rjeffrieshave you compared his design with your uart design?18:39
wpwrak(compared) not really. they're a little different, of course. also, my chip is smaller because i expect it to do very very little.18:43
rjeffriesI wonder if femtoduino can be powered from Ben18:54
qi-bot[commit] kyak: config.full_system: busybox enhancements http://qi-hw.com/p/openwrt-xburst/c3cafbf18:54
rjeffriesone nice thing about femtoduino is it should appeal to a wider market where Ben owners are a subset18:54
wpwrak(powered) possibly even through the regulator. else, connect after it. (for the 3.3 v version)18:57
qi-bot[commit] kyak: config.full_system: more busybox options http://qi-hw.com/p/openwrt-xburst/d6a32b519:18
wpwrakheh, i'm so cool. clock output worked on the very first try ;-)19:53
wpwrakhmm. now i just have to figure out why ... :)20:03
rjeffrieswpwrak the femtoduino guy tells me: I get all my PCBs from http://dorkbotpdx.org/wiki/pcb_order .. cheap, great quality, ROHS compliant and made in USA!\20:12
wpwrakrjeffries: yeah, no setup fee is great for prototypes. doesn't say if he can do 0.8 mm boards, though20:17
qi-bot[commit] Werner Almesberger: usb/cam/Makefile: added dependency in Makefile itself; local parameter update http://qi-hw.com/p/ben-blinkenlights/a05155e20:25
qi-bot[commit] Werner Almesberger: uart/avrdude: renamed to ./avrdude (i.e., moved to the top-level) http://qi-hw.com/p/ben-blinkenlights/271a50f20:25
qi-bot[commit] Werner Almesberger: uart/: great renaming to nxuart, including references in Makefile http://qi-hw.com/p/ben-blinkenlights/5b14b5020:25
qi-bot[commit] Werner Almesberger: nxuart/: changed title in nxuart.sch and nxuart.brd http://qi-hw.com/p/ben-blinkenlights/4fa0a3420:25
qi-bot[commit] Werner Almesberger: avrdude/patches: renamed "uart" to "nxuart" http://qi-hw.com/p/ben-blinkenlights/de826dd20:25
qi-bot[commit] Werner Almesberger: nxuart/fw/: renamed "uart" to "nxuart" as well http://qi-hw.com/p/ben-blinkenlights/fc3245920:25
qi-bot[commit] Werner Almesberger: avrdude/patches/nanonote.patch: added clock output with -x clk=#MHz http://qi-hw.com/p/ben-blinkenlights/0c2b98220:25
qi-bot[commit] Werner Almesberger: nxuart/fw/Makefile (prog): supply an 8 MHz clock while programming http://qi-hw.com/p/ben-blinkenlights/a8c10a020:25
rjeffriesI doubt he can handle the thinner PCBs. does atUSB require 0.8 mm? also if we have UBBs via tuxbrain, then attaching femtoduino via s short cable will be ok/fine20:39
rjeffrieschallenge if design stays as is will be how to attach ribbon cable to femtoduino I guess20:41
wpwrakatusb and atben are both 0.8 mm, yes. didn't want to introduce yet another variable.20:46
dvdkkyak: investigating the gforth build problem you reported.20:50
dvdkvery strange new behaviour introduced by last openwrt changes20:50
dvdkafter ./configure, somewhere before make, somebody calls ./config.status regenerating a file from .in that i previously patched20:51
dvdksolution is probably to patch the .in file instead20:51
xMffdvdk: the joy of autoconf ;)20:53
xMffdvdk: patching generated files is a no-go by now, I briefly looked at the gforth issue but couldn't figure it out20:54
Action: dvdk loves autoconf/automake and would never complain and call it autocrap or autofail :)20:54
dvdkxMff: not patching per se, it justs overwrites it via echo from the configure rule.20:54
xMffdvdk: hm20:55
dvdkgforth is probably the most ugly makefile in openwrt-xburst.git20:56
dvdkbut the upstream makefile isn't much better anyways :)20:56
dvdkok, seems to build now.20:56
qi-bot[commit] David Kühling: gforth: fix build problem introduced by new openwrt autoconf support http://qi-hw.com/p/openwrt-packages/51b2d6d20:59
xMffah, heh21:00
dvdkbtw just realizing that libtheora is compiled with -Os.  this may not be good for performance.21:05
xMffyeah, iirc it yeels about it too21:05
xMffbut that -Os is from the global arch flags afaik21:06
xMfffor the NN you could think about switching to -O2 in general21:06
xMffby supplying CONFIG_TARGET_OPTIMIZATION="..."21:08
dvdkxMff: i already override -Os to O2 for the gforth engine (which otherwise looses its dynamic code copying/generation capabilities)21:12
dvdkmaybe we should patch libtheora Makefile only?21:12
dvdk-Os is otherwise certainly a good idea for most of the code, given the tiny cache and the slow TLB lookups of Xburst?21:13
xMffprobably yes21:13
dvdkarggh, libtheora not part of openwrt-packages.git21:14
dvdkok, gotta go21:14
kyakxMff: did you have a chance to look into flite's Makefile? Here's the version from xiangfu, slightly changed: http://dpaste.com/481700/. It uses the latest version of flite (1.4), alsa-lib and builds the static binary (i'm not sure libflite is used by someone)21:20
xMffkyak: not not yet21:23
qi-bot[commit] David Kühling: libgii: remove libtool/host dependency http://qi-hw.com/p/openwrt-packages/a5e3b8522:57
qi-bot[commit] David Kühling: gnuplot-gfx: libtool and other fixes to make it compile again http://qi-hw.com/p/openwrt-packages/5751eb122:57
--- Tue Mar 8 201100:00

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