#qi-hardware IRC log for Tuesday, 2011-07-19

tuxbraingood night dudes00:09
qi-botThe build was successfull, see images here: http://fidelio.qi-hardware.com/~xiangfu/compile-log/openwrt-xburst.full_system-07172011-2233/01:57
kristianpaulthis is odd, i inverted logic for the soft reset signal usign 1's complement, and now soc get in a reset loop...02:43
kristianpaulah,, lekernel is gone..03:13
kristianpaulhe, not here03:17
GitHub160[milkymist] kristianpaul pushed 1 new commit to gps-sdr-testing: http://bit.ly/mVZ0vK04:08
GitHub160[milkymist/gps-sdr-testing] missing negation in reset signal - Cristian Paul PeƱaranda Rojas04:08
kristianpaulxiangfu: you have idea how interruptions are handle by lm32?04:08
kristianpaulwith idea i mean workflow, in order properly handle it04:09
kristianpaulokay seems i have some fault with a internal reset signal with namuru, at least accum count is alive now..04:11
kristianpaulbut what happended with channel 0 carrier and code nco... lets see04:11
GitHub65[milkymist] kristianpaul pushed 1 new commit to gps-sdr-testing: http://bit.ly/n9Zu7N04:31
GitHub65[milkymist/gps-sdr-testing] Missing epoch load write registers, also added full reset values for all write registers chan0 - Cristian Paul PeƱaranda Rojas04:31
xiangfukristianpaul, don't know that. 04:34
wpwrak_tuxbrain_away: the one working on making an arduino+wpan device is "[g2]". haven't seen him here lately, though.05:48
wpwrak_kristianpaul: hmm, you invert the reset logic and then your thingy stays in reset all the time ? that actually sounds very plausible to me ;-)05:49
qi-bot[commit] Xiangfu Liu: nanonote remote -b in mc (master) http://qi-hw.com/p/gmenu2x/b3a73d206:00
qi-bot[commit] Xiangfu Liu: nanonote tetris: change the font before start (master) http://qi-hw.com/p/gmenu2x/a4a532506:07
qi-bot[commit] Xiangfu Liu: lynx: add libidn to DEPENDS (master) http://qi-hw.com/p/openwrt-packages/9544fc506:16
e1fwill the nanonote come with an arm cpu?09:43
e1fthere are several faculty members here and students who would be interested in it if it was arm09:44
wolfspraule1f: unlikely. we are building out the platform we started, or at least that's what I'm doing. of course the project can be taken in other directions but I'm not aware of anyone working on that.09:47
wolfspraulwhen it comes to new CPUs I believe in the Milkymist project - www.milkymist.org09:47
e1fwhat was the decision behind not using the arm09:48
wolfspraulit wasn't that way round, it was just that the device that seemed best as a starting point for copyleft hardware had a mips-compatible cpu09:48
e1fok. thanks for the info.09:49
wolfspraulfrom my observation so far, since we started, the number of actual buyers that bought the Ben because it was MIPS was probably equal to the number of buyers that would have bought it if it were ARM09:49
wolfspraulboth cases are rare, most people do not look at the CPU architecture first, or make that their central decision09:50
wolfspraulI'd say maybe 5% of buyers bought because it was MIPS, and I assume there would be a few more that would have bought if it would have been ARM, let's say also 5% (maybe you are in there, don't know :-))09:50
e1fin academia, esp. for teaching and research, the cpu would be the primary motivator09:51
wolfspraule1f: why do you prefer arm?09:51
wolfspraulwhat are you trying to teach or research?09:51
e1four embedded systems courses are standardizing on arm09:51
mstevense1f: so it's basically a network effect, you like arm because everyone is using arm?09:52
wolfspraulunderstood, nothing much we can do then. I spare you my comments about academia though, so happy I never went there :-)09:52
wolfspraulmstevens: yes, academia :-)09:52
wolfspraulbut I understand, arm is pretty hyped right now09:52
wolfspraulthey do good09:52
wolfspraulMilkymist uses a Mico32 core btw09:52
wolfspraulLattive Mico (lm32)09:52
wolfspraulLattice, sorry09:52
e1fmstevens: i think that would be a good summary of the situation. the industry has standardized on arm and market forces (i.e. students) want to learn it09:53
wolfspraulwhat's a good arm platform? don't know. maybe a reverse engineered Android phone?09:53
e1fthis is one of the researchers who i was talking to who prefers arm http://www.ee.ryerson.ca/~jasmith/09:53
wolfspraul'industry has standardized on arm'09:53
wolfspraulsounds like you are teaching future lobbyists09:54
wolfspraulis this a technical study?09:54
wolfspraulI'm not going to change your standardization, just curious to understand...09:54
e1fi meant ios is arm and windows will soon(!) run on arm09:54
mstevenswasn't the openmoko arm based?09:54
wolfspraulI would never think like that, but of course, I would never be able to sit still in a college for more than a few hours...09:54
wolfspraularm sells around 2 billion licenses per year lately09:55
wolfspraulgrowing09:55
kodeinolimex makes some arm prototype things that seem pretty decent for student laborations. we have a few in the computer club09:55
wolfspraule1f: oh, watch out about 'ios is arm'09:55
kyake1f: windows already runs on mips09:55
e1farm might also have some educational licensing (i'm not sure) which makes "us" prefer arm09:55
wolfspraulI would totally disagree, that's not at all how those guys think, imho09:55
wolfspraulapple is making very strategic investments into llvm for a long time09:55
wolfspraulI'm still not trying to understand what is being taught.09:56
e1fi was trying to explain my "standardization" comment09:56
wolfspraulsorry I mean I am still not understanding what is being taught.09:56
wolfspraulCPU architecture?09:56
wolfspraulembedded systems?09:56
wolfspraulkernel development?09:56
wolfspraulcompiler architecture?09:56
e1fi said embedded systems earlier09:57
wolfspraulApple is the only company that successfully switched from one 'platform' to another one, they'll do it again :-)09:57
wolfspraulno platforms... just products...09:57
e1fthere might be other courses too09:57
kodeinwolfspraul: the m68k -> ppc switch or the ppc->x86 one? ;)09:57
wolfspraule1f: I'm sure there is nothing you as an individual can do to escape that 'industry standardization'09:58
wolfspraulI'm really just curious to think deeper about it, now that you mention it09:58
wolfspraulas a student, I would immediately resist learning this stuff09:58
wolfspraulunless I want to become a lobbyist09:58
wolfspraulwhich I hear are well paid jobs...09:58
mstevenswolfspraul: they do seem to have a decent investment in ARM09:59
wolfspraulmost important in embedded systems still seems C, kernel architecture, and lots of peripherals, I think also needs knowledge about electronic circuits at some point09:59
mstevensnot to say they wont' switch as soon as something else looks like it has an advatnage09:59
wolfspraulyou mean Apple?09:59
mstevensyes09:59
wolfspraulsure, I agree09:59
wolfspraulARM is on a roll09:59
wolfspraulIntel needs to watch out10:00
wolfspraulI think those things are all easy to agree on.10:00
wolfspraul2 billion chips a year, not bad10:00
wolfspraullicense & fabless model more powerful, more partners10:00
wolfspraulbut it's just a proprietary CPU design, that's all10:00
wolfspraulhow many lines of Verilog? let's say 10k-100k maybe10:00
wolfspraulneither the beginning nor the end of the world...10:01
wolfspraulok let's be generous, with all they do they have 500k of sources now, plus tons of tools, testing stuff, etc. a large proprietary IP enterprise.10:01
wolfspraulI just don't understand why a student needs to learn that first, since it's better to learn the underlying concepts, they can easily be applied to ARM or any other architecture10:02
wolfspraulmaybe the university can reuse some arm marketing material :-)10:02
wolfspraulbut I should hold my cynicism back, be positive! :-)10:02
wolfspraule1f: I propose an Android smartphone with good reverse engineering status on the software side10:03
wolfspraulthat'd be pretty cool for students, I think10:03
wolfspraule1f: and if you have some time and curiosity, for CPU architecture I propose you take a look at www.milkymist.org10:07
wolfspraulit's both fully open and fully functional, a combination which is rare. so I think it serves well as a study object.10:07
e1fin general, i would say faculty tend to teach the cpu with which they are familiar with10:08
wolfspraulthat's for an IC design course though, maybe too far from 'embedded systems'10:08
wolfspraulwhen you work on an 'embedded system', cpu choice is the first thing you do?10:09
wolfspraulI mean you first say "it has to be arm", and then you think about the system?10:09
wolfspraulhonestly, especially in embedded systems, that sounds very wrong to me10:09
wolfspraulbut I am not trying to change your thinking here, please dont' misunderstand10:10
wolfspraulyou are looking for an arm system, got it!10:10
wolfspraulhow about an android smartphone?10:10
wolfspraulnot sure whether that counts as 'embedded'10:10
wolfsprauldoes anybody know other cool and cheap arm boards?10:11
wolfspraulron is typically flooding this channel with them, we could search the backlog10:11
larscbeagle board, panda board, hawk board10:11
larsc...10:11
kodeinhttp://www.olimex.com/dev/sam9-L9260.html ?10:12
e1fand android would be overkill because the touch-screen is unnecessary10:12
wolfsprauldon't use it :-)10:12
kodeinthis seems to be an upgraded one; http://www.olimex.com/dev/sam9-L9261.html10:12
e1fbut it is not for me to say why he prefers arm. i was just curious over the mips choice for the nanonote10:13
wolfspraulyes, I hope we explained well. we didn't choose the CPU first.10:13
wolfspraulproduct first10:13
kodeinI seem to recall there being a few courses utilizing spim at my uni10:14
e1fby a strange coincidence i just saw this http://go.theregister.com/feed/www.reghardware.com/2011/07/19/arm_to_grab_quarter_of_laptop_market/10:44
wolfspraulwhy is that strange?10:44
e1fwe were discussing arms's market dominance10:44
wolfspraulnobody disputed that, I think they will continue to do well10:45
e1fand that articlea ppears in my rss feed10:45
wolfspraul2 billion chips (see above)10:45
wolfspraullaptop market is actually small, about 350 million/yr I think10:45
wolfspraulone quarter = 87.5 kk?10:45
wolfspraulhardly a blimp in the arm statistics, but at least those chips are expensive, that's good for arm10:45
wolfspraularm makes as little as 3-4 cents for their cheapest chips, 10-40 for the bigger ones10:46
wolfspraulcrazy numbers from an Intel perspective :-)10:46
wolfspraulthey will think it must be DOLLARS not CENTS10:46
wolfspraul:-)10:46
larscwell, there is a big difference between ARM and Intel. Intel actually builds and sells the physical chip, while ARM just sells the blueprint10:48
wolfspraulyes and no10:48
e1fbut i assume the fab is a lot cheaper than the intel cpus10:48
wolfspraulso yes, you are right, those 4-40 cents are just for the 'license'10:48
wolfspraulthen some of their licensees add IP on top10:49
wolfspraulsome don't10:49
wolfspraulit's a totally different machine/business model10:49
wolfspraulIntel is still doing manufacturing themselves, but it seems they have already decided that while this helped Intel the last 20-30 years, it will not do so the next 20-30 years10:49
wolfspraulso Intel will get out of manufacturing, and use fabs that can pool investments from many customers. maybe slowly, maybe not for their high-end chips first, but it will start.10:50
wolfspraulbut that's a huge change in Intel's business model, so I have no idea whether Intel can survive such a turnaround - they will need their cash...10:51
wolfspraulthis is not me guessing crazy stuff, but sort of reliable information from Intel Labs people that is also not a big secret10:51
kodeinthe foundry model does have its good points, yes10:51
wolfspraulthat's why Intel bought McAfee, for example. you will see more like that.10:51
larscstill, intel sells the chips at the end of the day10:52
e1fdid you read gordon moore's article in business week a few days ago about u.s. losing manufacturing industry to china?10:52
larsceven though if they were fabless10:52
wolfspraulyou mean the strength of their brand? hmm. ARM's is also strong, as we can see here :-)10:52
wolfspraulbut you are all right, this change will come slowly, it will take years10:53
wolfspraulIntel is still selling many chips > 100 USD10:53
wolfspraulARM licensees pay pennies10:53
wolfspraultotally different machine, imo10:53
wolfspraulIntel is basically doing what a whole bunch of companies in the ARM ecosystem do together10:53
wolfspraulfor my notebook I want power, as much as is reasonably priced10:54
wolfspraulso most likely I will not want those ARM notebooks10:54
wolfspraulunless Intel really screws up :-)10:54
wolfspraulbut I think the percentage of users for power notebook will go down, like the workstations did10:54
wolfspraulmost people just do a bit of browsing... skype, email, chat, etc.10:55
wolfspraullarsc: I'm sure Intel will have the most powerful chips for a long time, 10 years, or more.10:55
wolfspraulbut ARM is coming and eating into their business in a nasty way, with a new model and new ecosystem with many partners. So Intel cannot just continue like before.10:56
wolfspraulwpwrak_: [refers to discussion a while ago] the Sharism Ltd. business model is entirely different from dangerousprototypes/sparkfun/seedstudios/adafruit/arduino/etc. 100% different11:01
wolfspraulTuxbrain I don't know, he has to explain his model, I think he is moving more to a dangerousprototypes model?11:02
wolfspraulthe only other company I know that I would say has the same model as Sharism is Elphel11:02
wolfspraule1f: I did not [moore article], but maybe because I spent so much time in China I'm fairly relaxed about it11:10
wolfspraulnothing much is 'lost', imho11:11
wolfsprauldon't know where to start, really. Let me know if you want to know more. I'm not Gordon Moore though :-)11:12
wpwrak_e1f, wolfgang: for researching the internals of a CPU/SoC, M1 should indeed be perfect. similar to what happened in operating system research when linux came out. completely changed the landscape.12:48
wpwrak_e1f, wolfgang: as far as assembler programming and general architecture is concerned, ARM (which one ? nowadays, the core is much less important than the SoC around it) or MIPS (again, which one ?) hardly matters. and you rarely have the luxury of spending enough time on a single core to become a great assembler programmer there. nor are such skills much needed.12:51
wolfspraulfully agree12:51
wolfspraulbut the ARM mindset in academia will still be strong, even if it only helps to reuse ARM marketing material and tools in class 1:112:52
wolfspraulmy belief is rooted in a general suspicion of recent academic values though, so I hope Werner is right...12:53
wolfspraulWerner believes in academia, that's a good sign. Maybe they rediscover independent thought one day.12:53
wpwrak_wolfspraul: i'm a little surprised there would be an "ARM mindset" in academia. i can understand that arm is convenient (large choice of boards, large choice of books, etc.), but that should be all12:53
wolfspraulthat's what e1f said. arm = industry standard = that's the only thing faculty looks at12:54
wpwrak_if a student thinks they MUST learn arm or they'll be doomed, i think some much earlier lesson went very wrong ;-)12:54
wolfspraulor even the teacher :-)12:54
mstevenswpwrak_: I've noticed academia seems very absolutist12:54
mstevensif a tool is good, it must be the best tool, and should be used for everything12:54
wpwrak_well, very young teachers or those nicely shielded in their ivory tower could be excused ;-) although the latter sometimes hatch weirder plans :)12:55
wolfspraulI totally agree about convenience, and I'm not against ARM in any way. but it's just one proprietary architecture/business model out of many. There have been others before, and there will be others in the future.12:55
mstevensat a more technical level, I have no personal knowledge but I've heard ARM assembler is quite nice12:56
wpwrak_mstevens: be careful not to confuse individuals with academia as a whole. it's often a lot of little fiefdoms, each with its very own very strict rules ;-)12:57
wolfspraulmaybe it's even better to study a small not so ubiquitous platform, it helps focus on core principles that can then be easily applied to any other instance or currently commercially successful platform12:57
mstevenswpwrak_: I'm handwaving a bit, but see eg the way most of academic CS had a java love fest in the late 90s12:57
wpwrak_mstevens: ARM, MIPS, LM32, they're all okay. learn one and you know the rest. heck, even x86 will do.12:57
wolfspraulbut I may be naive, if people only want to spend the minimum amount of hours, and make the maximum return on investment, maybe they want the straightest path to employment only12:58
wpwrak_mstevens: academia has its fashions and fads too, that is true :)12:58
wolfspraulgod yeah, the Java thing was bad :-)12:58
wolfspraulis that over now?12:58
mstevenswpwrak_: my degree didn't actually do any assembler, we did CPU design and C, but nothing inbetween12:58
wolfspraulnow it's Ruby on Rails?12:58
mstevenswolfspraul: I dunno, I left :)12:58
wpwrak_mstevens: funny to skip assembler but do cpu design ;-)13:01
mstevenswpwrak_: I actually regret missing the more mathematical stuff like formal methods13:04
wpwrak_mstevens: would you have appreciated it while studying ? :)13:04
mstevenswpwrak_: well I kinda regretted not doing straight computer science at the time13:06
wpwrak_mstevens: one problem with the theoretical stuff is that it doesn't age well (unless you use it regularly)13:07
mstevenswpwrak_: I found the same effect with the electronics!"13:08
wpwrak_heh :)13:19
wpwrak_today is tuesday ? wow. could have sworn it was wednesday. guess i'll need a resync soon :)13:20
mstevensIn theory my degree was half electronic engineering, in reality I remember nothing13:22
wpwrak_wolfspraul: regarding the selection process, i agree. there are so many constraints that it simply doesn't make sense to make the CPU architecture one of them. except for things like whether it's supported by gcc.13:22
wpwrak_there should have been a good amount of math in the EE13:23
mstevensbut not the cool math13:23
wpwrak_;-)13:24
Action: mstevens regrets not doing things like lambda calculus and more algorithms13:25
wolfspraulwpwrak_: correct. in embedded systems the cpu architecture is just not so central a decision, other constraints often outweigh it. You may still end up with 50% arm or whatever, but you don't solve your equation in that order.13:26
wpwrak_mstevens: you can always get a good book and learn from there. now you'll even have the proper motivation for it ;)13:30
rozzinmstevens: So, do lambda-calculus :)13:32
mstevensrozzin: I will one day, my current obsession is french13:32
rozzinmstevens: Ben NanoNote comes with Guile ;)13:32
rozzinhttp://lambdadance.spacebar.org/13:33
rozzinOh, Epiphany--so broken I cannot see the animations.13:33
mstevensje voudrais apprendre francais13:34
mstevensonly I'm still pretty crap13:34
rozzinmstevens: The best college-educated hackers I've known have tended to be `electrical engineers' by schooling.13:38
kristianpaulwpwrak_: (reset) sure,well the problem was my own reset logic is not negated so i did some small portinng for some namuru modules harcoded to altera stuf,, anyway,the thing was i forgot namutu logic for reset is negate, but thats fixed now i think14:37
valhallaI couldn't find on google a nanonote reseller in italy because there isn't one, right? (I know that tuxbrain sends to italy, I'm just writing a list for the local LUG)14:55
wolfspraulyes I think Tuxbrain in Barcelona is closest14:56
valhallathanks14:58
tuxbrain_awayHi valhala , if is a group sales, send me an email at david@tuxbrain.com and tellme how many, to try to arrange you the price acordinly.15:08
valhallasorry, no, I haven't found other people who are going to buy one :( 15:09
valhallaI'm just collecting a list of local sellers of hardware that is related to the foss world, for my LUG, and it would have been nice to smuggle some free hardware in there :D15:10
wolfspraulwow so cool, I think Niels is working on a NanoMap update for the Ben!18:34
wolfspraulhttp://en.qi-hardware.com/wiki/Special:NewFiles18:34
wolfspraulthat's a good way to end the day, quite a bit of headache on all fronts today :-)18:35
wolfsprauln818:35
kristianpaul'/win 2719:36
--- Wed Jul 20 201100:00

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