#qi-hardware IRC log for Sunday, 2012-08-05

wolfspraulhi qi-bot, good to have you back :-)01:00
wpwrakwhat's the trouble with the server ? aging hardware ?01:11
wolfsprauldon't know01:13
wolfspraulthere are too many pieces that I don't have full visibility into, and/or don't want to (too much effort)01:14
wolfspraulthe kernel is not 100% bugfree I would think01:14
wolfspraulthere are 2 kvm instances running01:14
wolfspraulthere's a raid-whatever mirror underneath, and lvm01:14
wolfsprauland there are 1 in 1^x hardware errors, as we all know01:14
wolfsprauldram, hdd issues, etc.01:14
wolfspraulof course I could improve all this, but why01:15
wolfspraulwe are not operating facebook or twitter with millions of users active at any moment01:15
wpwrakhmm. so you'll just migrate to a new machine and hope for the best. well, this often works :)01:15
wolfspraulin fact I have that machine already01:15
wolfspraulbut when the old one comes back up I immediately stop working on the migration :-)01:15
wolfspraulhe he01:15
wpwrakah, what happened to german discipline :)01:16
wolfspraulI use the chance of a new server also to upgrade my knowledge01:16
wolfspraulthis time for example I finally make the big ipv6 push01:16
wolfspraulI will first set it all up wtih ipv6 only, and then route ipv4 into it01:16
wpwrakhah. i'm not even sure if my isp supports it :)01:16
wolfspraulin 2012, it's about time01:16
wolfspraulsame here, but I mean on the server01:20
wpwrakwoah :) no sign of IPv6 ... but i found this: http://www.fibertelevolution.com.ar/01:22
Action: kristianpaul is waiting his isp to support ipv6 as well so i can migrate server to new home01:22
kristianpauland worldwide AAAA dns support..01:23
wpwrak30 Mbps !! and they present it as if they had just created the iphone :)01:23
kristianpaulI tough you had used to our nice adverticement methods :-)01:24
wpwrakyeah, save on tech, spend on marketing. actually a reasonable approach :)01:25
GCWDoes anyone know where I could locate newest linux source code for Ingenic jz4770 or a Programmers Manual 05:40
GCWIf so please pm me05:41
GCWor e-mail me gcwnow@gmail.com05:43
rz2kthe only one public source that I've found some time ago. http://git.varjanta.com/Ingenic-JZ4770-Android-Kernel/.git/tree06:06
kristianpaulerghh where is weather forecast when you need it..09:52
wpwrakFallenou: found it ? ;-)14:07
Fallenouwpwrak: yes =) thx for making a fool of me, I kind of deserved it !15:04
wpwraksorry, couldn't resist :) it's just so easy ...15:05
rohhey there15:07
rohbtw.. on ti chips: reading the datasheet and the register level documentation isnt enough. you HAVE to read the application notes. ALL of them and importantly: check out all the small notes. they are extremely important from time to time15:08
Fallenouand the silicon errata15:09
rohcompared to samsung or similar stuff its hard work getting these beasts to work.15:09
Fallenouthere are a lot of silicon bugs15:09
wpwrakroh: what hidden gems would one find in the app notes ?15:09
Action: Fallenou is working with omap315:09
rohon ti one can smell the shift-registers, while samsung hides em all quite well15:09
rohwpwrak: bugs, nondocumented, enforced register orders (can only be touched in that order) etc.. such stuff.15:10
rohti is hell to work with compared to other stuff.15:10
wpwrakah, nice :)15:10
Fallenouwpwrak: next time try this one : lmgtfy.com/?q=ti+am3352 ;)15:10
wpwrakFallenou: i didn't want to make it so easy for you ;-)15:10
rohanyhow. if you really want to do a product with an imagination gpu on board, then i am out.15:10
wpwrakroh: you still have the dumb frame buffer, don't you ?15:11
rohi have enough foss work in front of me and or get payed for it, i will not waste any minute of my life for such crap15:11
rohwpwrak: no.15:11
rohwpwrak: usually not.15:11
rohwpwrak: and even paying for the silicon is NOT ok.15:11
wpwrakduh. as if the "microsoft tax" had prevented us from using laptops.15:12
rohits live buying nvidia chips. its NOT giving the vendors the proper incentive.15:12
wpwrakthink of it this way: if we can make a ti-nanonote with the help of ti, it'll be pretty easy to turn this into a, say, allwinner-nanonote afterwards.15:12
rohwpwrak: it hasent. and no, i did not buy windows last few times.15:12
rohwpwrak: i think you are much too eaten by the ti-marketing-bullshit15:13
wpwraki don't particularly care about ti marketing15:13
rohti hasent made a single foss useable design the last few years.15:13
wpwraki'm interested in someone financing qi-hw development15:13
rohtheir ┬┤documenton is sub-level compared to what i see from freescale and samsung15:13
wpwrakbecause we're all running out of reserves15:13
rohthey are much less open when it comes to details, and they are known to 'fall into your back' when you need it the least.15:14
rohah.. and their chips are known not to work without the help of their FAE and lots of pain and time.15:14
wpwrakyeah, i heard that one :)15:14
Fallenoutheir support even for a big company is shit15:14
rohas in 'sit down with their FAE to get routing changed' without any documentable reasons... sucks.15:15
Fallenoui am on the phone with a TI account manager each week15:15
rohits not that they are not trying. but their products are just crap.15:15
wpwrakmaybe the AM335x aren't all that hairy, though. after all, they're only mid-range by today's standards.15:15
Fallenouhe just buys time and say bullshit about his progress on fixing ours bugs15:15
rohwpwrak: doesnt matter. i would choose a freescale or samsung soc over them blindly.15:15
wpwrakthe cpu choice would be determined by who feeds the project :)15:16
rohFallenou: i have never seen ti accept that a bug is theirs, even when provided proof.-15:16
wpwraks/who/whom who15:16
rohwpwrak: anyhow. any ti project cannot be foss.15:16
rohatleast not with the chips they have.15:16
Fallenouroh: they accepted to say they are out of jedec spec15:16
Fallenoufor omap363015:16
rohFallenou: whohoo...15:17
Fallenouwhich implies troubles with several ram chips15:17
Fallenouat precise freq15:17
rohnah.. its so much work and pain to do a project/product. i do not need additional weight.15:17
wpwrakas far as i can tell, the LCD controller should be able to provide a dumb frame buffer. and that part is documented.15:17
Fallenouwpwrak: yes15:17
Fallenoujust drawing to framebuffer is fine15:17
rohwpwrak: doesnt matter. either you can use the accel. or you choose another chip.15:17
Fallenouyou have linux framebuffer which works15:18
Fallenoujust forget about opengl15:18
wpwrakwe don't need accel at those resolutions. hey, even my desktop doesn't have 3D accel.15:18
rohwpwrak: you need accel. even for blitting.15:18
rohwpwrak: without its painfully slow.15:18
wpwrakremember gta01. considerably weaker hardware. and graphics speed was just fine in the end.15:19
mthrz2k: whoever is maintaining that git is doing some weird things to the commits15:19
rohand yes, i also only use 2d desktops. but yes i need accel. why? for yuv conversion and bitblit with keying and alpha.15:19
mthdate and author info are lost15:19
rohwpwrak: you forget that on the moko even the screen-refresh ate 1/3rd of our total memory bandwith already due to the 'big screen'15:20
wpwraksee !15:21
wpwrakthe AM335x supports DDR3. and DMA seems pretty flexible, so even if you can't do blit with the CPU, you have a plan B15:22
rohanyhow.  making a product which is in the same market as the efika (which are there and done) will be more than hard.15:22
rohand less free as ti.15:22
roheh... i mean.. ti will be less free afaik (due to details, also imx has pvr)15:22
wpwrakfrom the discussion, i gather there are a number of undesirable properties, most of them well-known. but nothing that would make such a device un-viable.15:24
rohwpwrak: the major thing i just named: market already taken.15:25
rohyou would have to beat a 200 euro product15:25
wpwrakfor me the question is whether TI would provide the resources for such a project, accepting it to be open (according to qi-hw standards) for the rest of its structure and parts15:25
wpwrakthat's a different class of device15:26
rohthe allwinner cpu is at least interresting15:26
rohti and freescale are both extremely expensive.15:26
wpwraki wonder about the allwinner. where is the documentation ?15:27
rohme too.. i only find linux ports15:27
rohi guess its chinese15:27
rohbut atleast it has a mali-gpu .. which means its from arm and needs only one reverse-engineering, not a new one every time15:27
wpwrakheh ;-)15:28
wpwrakdoes powervr change so often ?15:28
rohwpwrak: yes and no. there are some -zig revisions15:29
rohand the company is quite extreme on legal cases.15:30
rohthe mali stuff will be less stressfull from that corner afaik15:30
wpwrakhunting reverse engineers ?15:30
wpwrakwell, eventually someone will crack the secret ... :)15:31
wpwrakit's not as if people couldn't figure out how to work together on things that must avoid the light of day :)15:32
rohwpwrak: anyhow.. why license a gpu which is closed and even pay for it?15:32
rohthats not sensible behaviour15:32
wpwrakmy reasoning is based on TI financing the project. if they don't, it's obviously a no-go.15:33
rohwpwrak: you know that intel had the sgx in one of their products?15:34
rohwhich they dropped completely because of that15:34
wpwrakif they finance it, this would allow us to make that ben successor we've always talked about but never started with.15:34
wpwrakmaybe intel are smarter than ti, at least in that regard ? :)15:34
rohthe poulsbo chipset15:35
wpwrakalso, if ti finance such a project, that would completely offset the extra cost of the gpu. that is, unless we sell millions of units.15:35
rohintel seems to have a 'need to write open drivers or release datasheets' paradigm, which they couldnt do with sgx, so they remove the product in the process15:35
wpwrakin which case, we'd of course have the means to change the design to use whatever chip we truly like.15:36
wpwrakyes, intel are very friendly15:36
wpwraktoo bad they're stuck with x8615:36
rohwell.. sometimes friendly, sometimes just good-willed15:36
rohfor some stuff there is no documentation, just drivers (even open ones)15:37
whitequarknowadays if I would ever want to buy a notebook, that'll be an Ultrabook15:37
whitequarkbecause it awesomely works with Linux out of the box.15:37
wpwrakmaybe they just didn't get around to writing proper docs15:37
whitequarknot to mention that they tend to generally have quite good hardware.15:37
Action: roh buys 5 year old thinkpads. because they work great and are only a few hundred euros for hw which doesnt get built that way anymore15:38
whitequarkwell, I think I'll get myself an Asus UX31E with i715:38
whitequarkI need quite a lot of CPU power for my work15:38
larscroh: R61/T61?15:38
wpwrakti should also feel some pressure on the gpu. the android customer base if getting increasingly unhappy with the update situation and binary-only drivers don't exactly help. (of course, there are many more factors)15:38
rohlarsc: t60p etc.15:38
whitequarkand Ivy Bridge graphics is finally something that can be called a GPU. well, this is offtopic anyway.15:39
rohno R series (plastic crap) and no 61 series (nvidia)15:39
rohmagnesium frame and intel/amd chipsets15:39
rohand screens with resolution for men without glasses.15:40
wpwraknetbooks are also nice. disposable computers. and with enough cost pressure to eliminate any fancy proprietary stuff quite rapidly ;-)15:40
larsci had tought I had hd a T61, but since it had a intel graphics card it must have been T6015:40
whitequarkwpwrak: AC100 was a... nice try.15:40
whitequarkit was cheap and it also had Nvidia Tegra15:40
whitequarknuff said15:40
whitequark(translates to: "No fucking drivers! Throw out that crap.")15:41
rohwhitequark: i bet that all of that cheap netbook crap is nonworking defective stuff in a short time.15:41
larscit's still running, but the screen hinges have broken multiple times15:41
rohwhitequark: just wait 3-4 years15:41
whitequarkroh: exactly15:41
whitequarkI had a Toshiba netbook, and southbridge died after half a year15:42
rohi still use the same machine i worked on at openmoko15:42
whitequarknow I have a Samsung one (not me actually, given it to a friend), and it works for 2.5 years already15:42
whitequarknot sure how long it'll last15:42
whitequarkroh: well, for me, upgrade from 1+ year old Core2 to 0 year old i7 does matter much, so I upgrade frequently.15:42
rohand the hammerhead gps story healed me when it comes to binary drivers15:43
rohwhitequark: whatfor?15:43
whitequarkroh: 1. compiling stuff 2. writing very high-level software in Ruby.15:43
rohwhitequark: i only find io too slow (only up to some 30-70mbyte/sec) .. cpu... well...15:43
rohok.. forget it. i only do C and python15:44
whitequarkroh: try writing LLVM in python and you'll understand what I mean.15:44
whitequarkthat's basically what I did in Ruby.15:44
rohstill bad idea to use ruby for that15:44
whitequarknot at all15:44
rohbesides... its a notebook. if you need power.. thats what rackmounted hw is for15:44
whitequarkdeveloper cycles are way more important than processor cycles15:44
wpwrak(powervr vs. gnu) well, that's the gnu side (rightly) being worried. doesn't say anything about imagination having issues any threats15:44
whitequarkroh: it's noninteractive software. the fact that I can write a sophisticated SSA transformation in less than 200 LOC _does_ matter a lot15:45
whitequarkthe fact that it'll execute for minutes does not, mostly15:45
whitequarkI even have map/reduce for the cases where you have >1 core/machine.15:45
whitequarkroh: (rackmounted hw) not if you travel much. I do, and I work from... strange places sometimes15:46
rohand there is no internet there?15:46
whitequarksometimes there isn't (~20kbps UMTS with 1000+ms latency)15:47
whitequarkit's as much as you can get anywhere outside of Moscow if you don't have a dedicated (LAN/ADSL) line15:47
wpwrakwhen it's not snowing15:47
rohyour bandwith is snow-dependant?15:48
whitequarkroh: I dunno how 2g/3g works in your country, but here you're lucky if you have 3g _in a center of a big city_, and even there it's crap.15:49
whitequarkand yes, it's snow-dependent15:49
whitequarkand luck-dependent15:49
roh;) pun aside.. i am used to quiet computers... even that c2d is quite loud15:49
whitequarka few days ago 3g just stopped working in the place where I live15:49
whitequarkmaybe some piece of shit at the tower died, I dunno15:49
whitequarkhappens all the time.15:49
roh .oO(but putin loves you all so much) *scnr*15:50
whitequarker, not only 3g but 2g internet too.15:50
whitequarkif I could get out of Russia right at this precise moment, I won't think even once.15:50
rohwhats the issue? passport?15:51
wpwrakprobably also depends on whether the girl at the switchboard had the right quantity of vodka for breakfast. too much or too little and she'll misroute the packets.15:51
whitequarkroh: I'm 19 with no university degree, which means if I lose my current job I'll probably have a hard time finding another one, and yes, there are some problems with passport due to obscure details of russian law.15:52
whitequarkthe company I'm currently working for is absolutely fine with me working from any other point on earth, which is good.15:53
rohwpwrak: http://e2e.ti.com/support/dsp/omap_applications_processors/f/42/t/89852.aspx15:53
rohwpwrak: in short: ti fucked up their own hawkboard evb design and need to respin it completely (routing issues or so between soc and ram) and recommends other evalboards in the meantime15:54
wpwrak"recommended layout recommendations" ;-))15:54
rohwhitequark: well.. sounds good and bad15:54
wpwrakif the project was sponsored by ti, i wouldn't worry all that much about this sort of design weaknesses. even if it should turn out in the end that the chip is broken beyond repair, we'd still have gathered valuable experience in other areas.15:56
wpwrakit all depends on the angle from which you view the issues15:57
whitequarkroh: I hope to resolve the latter in a year or so, and a diploma can be bought. (I'm not going to waste 4+ years on 80% shit I'll never actually use and 20% shit I could figure out myself.)15:57
rohwpwrak: sure. i get your point. but to be fair.. isnt it much more fun working on sth which works for sure in the end?16:09
rohwpwrak: have you seen the stuff from busware?16:10
rohhttp://busware.de/ .. he does small mcu stuff and some freescale designs for where a 32bit cpu is needed16:11
wpwraknothing "works for sure" ;-)16:12
wpwrak"busware.de ist bei Facebook" eek !16:12
larsceverybody is bei facebook16:13
Action: wpwrak isn't16:14
larscyour so not going with the time ;)16:17
Action: roh would rather like to do multiple smaller scale projects with higher chance of success than doing another big 'failed' one16:17
rohlarsc: nah. that verdict is still out.16:18
wpwraki'm already in the post-facebook future ;-)16:18
rohwpwrak: hrhr16:18
rohi think we can sit that one out. live icq or win9x16:18
wpwrakroh: agreed on trying smaller projects. that's what i do at the moment :)16:18
wpwrakbut at the same time, if there's an opportunity to make a ben successor, i think we should take it16:19
rohwpwrak: what are you hacking on atm? and sure. but maybe not with something that pushes so many people away16:19
rohi would for example like to see some joining forces with other foss hardware initiatives... maybe we can build alliances rather than compete on 'less foss' soc products16:20
wpwraki don't think too many people would object to a ti-based nanonote if it works well and if we communicate the limitations properly16:20
wpwrakroh: at the moment i'm trying to bring my led toy into the modern age :)16:21
wpwrakalliances would be good. but the issue on our side is that we're on a catastrophically unsustainable path. there's little point in forging alliances if you may not make it to the end of the month.16:22
rohlike the pi or these guys http://rhombus-tech.net/16:23
rohwpwrak: true. is it that grim already?16:24
wpwrakpretty much, yes16:24
rohi thought that qi was one of the posterchilds when it came to 'process foss-i-fication' ... means documentation on all areas.. most others focus only on the sw or what comes after sth. is done and working hw is on the table.16:25
larscroh: the point is you can't get it all at once16:26
larscit's a process16:26
larscgetting there means taking one step at a time16:26
wpwrakexactly !16:27
larscwhile the SoC may not be open this might be a chance to gain more experience in other areas16:27
larsclike creating the schematics16:27
larscand getting the product into fabrication16:28
wpwrakthe soc is reasonably open. i mean, it's considerably better than, say, the ingenic we have today.16:28
larscwpwrak: I mean 'open' as in you don't have access to the hdl16:28
larscs/open/not open/16:29
rohlarsc: sure. but shouldnt it be much easier to use a chip where you get support then?16:29
wpwrakroh: didn't you just say the ti chips are all about getting support ? ;-)))16:29
rohand where you know in advance that there are no 'extra difficulties' with a design16:29
wpwrakthere are always "extra difficulties" :)16:30
rohwpwrak: sure. then avoid as many as possible and dont steer into them full frontal16:30
wpwrakheh :)16:30
larscroh: given the alternative such a chip would be preferable. but there is no alternative16:30
larsccause nobody is going to pay for it16:30
rohsure, freescale is expensive, but then there are tqfp chips there, which you can even solder by hand.. its all a pro and con16:30
rohi am not saying that ti builds bad chips. i say they build complex beasts which have a high ratio of needed investment and time and volume to sales price which is resulting in higher financial risks than simpler designs, where also a lower volume on sales can make it work16:32
rohas in: lets NOT build mobile devices for a change16:33
wpwrakyou can't turn back time :)16:34
roheverything from notebooks, over netbooks, pda and smartphones and inbetween is like the 'kings class' of devices. high computing power, small volume.. very high integration.. and also a very competive market. not the best field to innovate in with modified goals16:34
rohwpwrak: its not about time. there are different devices all around you. loads of them.16:35
wpwrakthe nanonote is just in a niche. a niche where i think it has potential.16:35
rohwpwrak: try getting away from the few sample points we have right now.16:36
rohespecially the ones which havent worked that well economically16:36
wpwraki mean, even the ben with all its weaknesses sold some 1300 units. just imagine how much more attractive a device without all those issues would be.16:36
wpwrakplus, we could make a new design such that it can reach into other fields. e.g., the ben isn't very good at acting as a "terminal" (control interface to other equipment) but that could be fixed with relatively little effort.16:37
wpwrakand i'm not saying the nanonote is the only thing worth doing. far from it.16:38
rohi think we just need to open our minds a bit to other fields of devices and think for a while and then we will find more feasible fields to apply foss hw in small ways16:38
rohwpwrak: i am also not saying we should abandon ben or nanonote at all. just maybe delay it for the right time and chipset (and the money to invest)16:39
wpwraki've heard the military is a pretty good field. alas, you run into cultural difficulties with globally distributed projects ...16:39
wpwrakanother, even better, would be health. wouldn't know how to get a foot in there, though.16:39
rohi know somebody in the health field doing foss work16:40
rohbut much more low-level.. no mcus' but mechanical hacks16:40
rohfor developing countries... low-tech.. hi-tech isnt serviceable there16:40
rohgot asked if i can help building a machine which helps dispensing a fluid onto tests-paper and then cut that into stripes16:41
wpwrakseems that integration is a big issue in health. they have the equipment but connecting it to a computer costs an arm and a leg.16:41
roha machine from locally available/scrap parts which can be serviced whereever16:41
rohwpwrak: e.g. they got a ultrasound machine from siemens and it broke down and nobody could fix ir16:42
wpwrakmakerbot to the rescue ! :)16:42
rohor they got a machine and it had a color-crt and then there was world cup and it got disassembled and nobody could reassmeble it16:42
wpwrakof course, it's difficult to make money with those clients/customers who are extremely poor and who need individualized service16:43
rohor, machines get broken due to flakey power supply from the grid. not everything works well in dusty environments or high air humidity16:43
wpwrakof course, fixing that may quickly mean a complete redesign16:45
rohexactly. sometimes also a completely differnt attempt at the problem16:45
wpwrakthere's one opportunity, though: a lot of medical equipment is priced with us lawsuit risk in mind. if you make it for countries where a hot cup of coffee doens't get you sued for the equivalent of the gross domestic product of a small industry nation, you should be able to sell things at a considerably better price point.16:47
rohi think the selling works differently in that field16:49
rohthat one guy i know gets paid from aid organisations.16:49
rohso its more a 'developing stuff' than selling stuff what gets paid, and not from customers16:49
wpwrakalright, that may be a more viable approach for one-off tasks16:50
wpwrakto make money from the customer (e.g., hospitals), you'd have to solve a problem that appears at many places16:51
rohi think that 'selling the product' will be not the major focus.. rather 'providing a solution'16:54
rohthe ideal open source product can be easily made everywhere by lots of people. so one needs to find a way living from designing it, not making it.16:55
wpwrakyup. it's basically consulting17:00
kristianpaul:-|, UPS start failing in the worst season of the year...22:44
wolfspraulkristianpaul: good morning. how so, what happened?23:11
kristianpaulserver got corrupted its CF..23:16
kristianpaulwell, and dunno how my ecrypfs home resisted several hard reboots today..23:18
kristianpaulwolfspraul: what TI SoC do you have in mind? what do you think about the ASIC + FPGA combo?23:46
kristianpaulwpwrak i meant*23:46
wpwrakthe TI inquiry was about AM335x (x = 2, 4, ...)23:47
wpwrakthat looks like a reasonable chip. besides the evil GPU, it seems to have nice properties. (judging by the documentation. of course, there could be surprises.)23:48
kristianpaulyeap but you agree on that?23:48
wpwrakis would fit the nanonote class of devices. it's a chip to get work done, not one for record benchnarks.23:49
wpwrakasic+fpga would make a lot of sense if we need peripherals ti don't provide. as far as i can tell, the chip has everything we should need.23:50
kristianpaulabout Off the shell hardware, tplink routers like mr11u seems another interesting alternative for a no care work with linux solution23:50
kristianpaulperhaps GPU/VGA can move to a small fgpa :)23:50
wpwrakasic+fpga or asic+cpld could also make sense for flexibility. e.g., an 8:10 card slot on steroids. would depend on the cost calculation if we could make an FPGA fit in the equation.23:51
wpwrakin general, it would be nice to incorporate an fpga to keep that knowledge alive in the community23:51
wpwrak(gpu in fpga) naw, then the fpga needs to talk to the memory bus. that's too messy.23:52
kristianpaulwhat you think fpga can do then?23:52
kristianpaulor taking bits from milkymist... what reamains if you already have a soc..23:53
wpwrakit could be a general-purpose controller that sits on the 8:10 card interface. so if you want to do something fancy there, such as vga out or whatever, it could take care of that23:53
wpwrakbut as such a "controller for hacking", it would have to be cheap. a few dollars, not more.23:54
--- Mon Aug 6 201200:00

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