#qi-hardware IRC log for Wednesday, 2010-08-18

rafawpwrak: I love your work man.. wow.. if you have some wireless connectivity.. all of us will try to follow you.. at least to steal your hardware :D00:00
rafacya all00:01
wpwraknebajoth: yup. and they require a receiver of -84 dBm for this. meanwhile, some chips already do -103 dBm. if your noise floor is low enough, that could mean you get almost ten times the original distance.00:01
wpwrakrafa: sneef. and i thought already the idbg would be enough :)00:02
wpwraknebajoth: pity that i don't have my old apartment anymore, though. there i had two microwave antenna clusters right in front of my nose. the perfect environment for interference testing ;-)00:05
wpwrakof course, measuring anything with reasonable accuracy was quite hellish. instead of "add a wire and connect the scope probe" it was "make a custom coax cable and solder a connector right into the circuit"00:07
wpwrak"reasonable accuracy" = such that Vhigh-Vlow > Vpp(noise), ie., such that there's at least a reliable trigger.00:09
nebajothI'm sure one of your neighbours must have a microwave you can stand next to for a little wireless testing? :D00:15
wpwrakhmm. i could sabotage the shielding of mine a bit. should get a new one anyway.00:28
wolfspraulwpwrak: am I too cautious with saying Milkymist NanoNote needs 5 years?00:57
wolfspraulI'm all for making it faster... :-)00:58
wolfspraulI think we are already moving in that direction, it's all just a matter of executing all the required steps.00:58
qi-bot[commit] kyak: gmu-0.7.1 md5sum updated http://qi-hw.com/p/openwrt-packages/6836a7201:29
uncloudedsorry if this is a silly question: the nbd package in OpenWRT requires glibc>=2.6.  If I build nbd.ipk, will it be able to run on a NanoNote that has only uClibc installed?01:36
wpwrakwolfspraul: i think you could make a MM-NN relatively soon. MM seems to have all the features you need for this. some of them may not work perfectly yet, but they're close.01:36
wpwrakwolfspraul: basically a nice portable development platform for FPGA hackers interested in making a SoC.01:37
wpwrakwolfspraul: the advantage of making it a NanoNote would be that you don't need to engineer the case etc.01:37
wolfspraulwpwrak: ok but we are moving in that direction already, just that you jump over a number of steps.01:39
wpwrakwolfspraul: as soon as you have the ability to make your own layouts at will, which as I understand things, you don't quite have yet, a MM-NN should be fairly easy to make.01:39
wolfspraulyou mean in kicad?01:39
wolfspraulI think you know all the different things that are moving.01:39
wolfspraulso first, Milkymist One, right now only 6 boards exist, and while this was a great step forward, we are really focusing on the rc2 run now.01:40
wolfspraulif rc2 is successful - great!01:40
wpwrakwolfspraul: preferably KiCad, yes. not sure where you are with the layouts. some thing seem to be already under qi-hw control. others not.01:40
wolfspraulok I explain01:40
wolfspraulMilkymist One RC1 was done with Altium Designer and a layout outsourcing house01:41
wolfspraulthe Altium files have all been released under CC licenses01:41
wolfspraulof course also all other files, gerber, AI, etc.01:41
wpwrakso, open but hard to reuse01:41
wolfspraulRC2 will also continue to use Altium01:41
wolfspraulyes, but then there is Xue, Andres' first Milkymist derivative01:41
wpwrakok. SIE seems to be a step better, no ?01:41
wolfspraulno I think we use OrCad there :-)01:42
wpwrakor was it Xue ? i always confuse the two :)01:42
wolfspraulkicad files exist, but they are not used for production01:42
wolfspraulyes, so Xue01:42
wolfspraulXue is indeed another interesting reuse of Milkymist01:42
wolfspraulbasically we remove a lot of connectors, and add a CMOS image sensor01:42
wolfsprauland then, it's done all in kicad01:43
wolfspraulI am not sure whether Andres will really succeed with that, but he is definitely on it right now!01:43
wpwrakah yes, xue. a commit that improves the placement of a cap has a nice ring :)01:43
wolfspraulcheck this png it gives you an idea01:44
wolfspraulso yes, Xue would be a board that uses the spartan-6, and is done in kicad01:44
wolfspraulso far _all_ kicad projects have stalled or failed at some point :-)01:45
wolfspraulso there comes the point where Andres (in this case) just gives up, we take the schematics, re-enter them in orcad/altium, go to a layout house to finish the layout, and then into production01:45
wpwrakah, just component placement for now01:45
wolfspraulnow, from either Milkymist One or Xue to a NanoNote board is still a lot of work01:46
wolfspraulthere is no LCM controller01:46
wolfspraulI don't know about heat generate by the s-6 in the small nanonote case01:46
wolfspraulpower consumption - if it works for 30 minutes I think it's a stupid idea01:47
wpwrakbtw, i wonder what the great concern about ground fill is. i needed it the first time for the wpan, and it works much better than i expected. a lot better than when i played with it a few years ago.01:47
wolfspraulit's just a bit early I think, but yes definitely this kind of idea is on the horizon. totally!01:47
wolfsprauldon't ask me.01:47
wolfspraulwe have a number of kicad projects 'live' right now01:47
wolfspraul1. the milkymist one jtag/serial reflashing cable (actually a small pcb)01:48
wolfspraul2. xue01:48
wolfspraul3. ben-wpan01:48
wpwrak(lcm) hmm, i thought one of the demos showed an lcm ? or was that just a composite video board ?01:48
wolfspraul#1 is trivial, I think we can take that into production, the entire process in kicad etc.01:48
wolfspraulthat was probably via VGA01:48
wpwrak4. idbg (just not hosted at qi-hw)01:49
wolfspraul#2 is a big challenge. Andres wants to do it, but if he gets stuck and gives up I am ready to support the project with closed tools, same as Milkymist One01:49
wolfspraul#3 - I am very curious about this, but haven't had the time yet to fully understand what this is, how we can use it, etc.01:49
wolfspraulok great, didn't realize idbg was in kicad too01:50
wolfspraulwe have kicad files for SIE, and for NanoNote, but I think they never made it into actual production01:50
wpwrakben-wpan will be an ieee 802.15.4 transceiver board for the ben. as an intermediate step, i add a USB-to-SPI converter so that i can plug the thing into any PC01:50
wolfspraulwhich doesn't mean that we won't try again, of course01:50
wpwrak(idbg) did you really expect me to use anything else ? ;-))01:50
wolfspraulso while I love the idea of Milkymist NanoNote, I think it's a bit too early.01:51
wolfspraulWe are working on it already, just indirectly and several steps away.01:51
wolfspraulfirst Milkymist One01:51
wolfspraullet's get that done right01:51
wolfspraulwith reflashing cable, with case, with great software support01:52
wpwraksure. before that, it wouldn't make sense01:52
wolfspraulit may take a while to get there01:52
wolfspraulI really don't want to jump around and leave half-finished stuff behind.01:52
wpwrakjust an idea for a direction that should be feasible soon thereafter01:52
wolfspraulso that's also true for the Ben, I will definitely continue to squeeze more out of it for a while.01:52
wolfspraulI am just starting with it actually, and your great hacking is very very helpful! (idbg, counterweight, now 802.15.4)01:53
wpwrak(ben) you mean with software or do you also envision a ben II with hardware changes ?01:53
wolfspraulyes, as direction we are 100% on the same page01:53
wolfspraulfirst software, it's obvious that it could still be much bette01:53
wolfspraulhardware, we are collecting bits and pieces01:53
wolfspraullike your scan work, your counterweight, maybe kristian paul can do more on GPS01:54
wolfspraulwe are looking into the 4760 from Ingenic01:54
wolfspraulnothing new, all the things that are in motion already01:55
wpwrakso ben II == Ya ? will ya have a cpu change and if the cpu isn't ready but some interesting hardware features are, then a ben II ?01:55
wolfspraulcannot tell right now01:55
wpwraks/will/Or will/01:55
wolfspraulthe package needs to be right, i.e. the cost to manufacture and difference from Ya has to be 'right'01:55
wolfspraulif it's only minimally different in hw, maybe continuing to improve sw is better01:55
wpwrakwell, as long as there's a steady stream of development, there are always things you can pick up and integrate :)01:56
wolfspraulif we find a large enough hw difference that is also cheap to take into production, then let's go for it, the Ya will replace the Ben01:56
wolfspraulyes correct01:56
wpwrakwhat are your thoughts on case changes ? impossible ? small things, like cutting a hole ? bigger things, like moving holes ? anything is possible (i.e., a small change will be just as expensive as a big one) ?01:58
wpwrak(all this within the ben II / Ya timeframe)01:59
wpwrakof course, once the Ya serial number wraps from 99999999 to 00000000, the cost of a case change will probably be a very minor concern :) of course, then you may be more interested in choosing an adequate yacht :)02:00
wolfspraulcase changes: I am not actively working on any mechanical stuff around the NanoNote at the moment.02:02
wolfsprauland I won't, it simply makes no sense economically02:02
wolfspraulfirst I need to find a way to make a great case for the Milkymist One02:02
wolfspraulactually I am following with interest what your scanning activities will result in :-)02:03
wolfspraultooling for plastic parts is just really expensive02:03
wolfspraulwhy throw our limited money into that? it's hard to justify.02:03
wolfspraulso my plan there is to first learn more, find ways to do it cheaper.02:04
wpwraki think you'd have to radically simplify the case. e.g., have the widest part of any hole intersect with the tool boundary02:04
wpwrakyou can probably have two-part tools for most elements. that should help to keep costs sane.02:04
wpwrakyup, learning more is good02:05
wolfspraulI don't know, need to learn more.02:05
wolfspraulalso there is no rush, we have the Ben and it's a great platform, and we can sell it today.02:05
wolfspraulso little by little the pieces are coming together to make more powerful copyleft devices02:05
wpwrak(scanning) so far, nobody seems to actually do anything with the scans. well, except for myself. they were quite useful for the counterweight. i still had to make changes after seeing the real thing in the real case, but it helped a lot to get there.02:06
wolfspraulwithout sounding direspectful to anybody, I doubt anybody will do anything with them :-)02:06
wpwrak(rush) the things is that you can't add connectors without changing the case. and you have the imbalance problem. of course, you could just add counterweights :)02:07
wolfspraulwe are real pioneers in all this, I doubt anybody will really move from talking mode to action mode02:07
wpwrakalways the optimist :)02:07
wolfspraulbut who knows, it's out there so it's possible02:07
wolfspraulI can tell you that I am super interested.02:07
wolfspraulthose scans are the basis of any future work I will do around the mechanical02:07
wpwrakthey're at least good for sanity checking02:08
wolfspraulit's a copyleft project after all, so I am not that interested in buying great proprietary stuff and not learning anything in the process02:08
wpwrakit would probably be a good exercise to just try to remake the ben case. or at least one of the case parts.02:08
wolfspraulyes, exactly02:09
wolfspraulthose are things I imagine doing, or outsourcing to someone in China. but it won't be tomorrow, it will take a bit of time.02:09
wolfspraulit's an excellent step anyway02:09
wpwrakthe more of the process you own, the easier it is to change it, too02:09
wolfspraulbuild open foundations, then go from there step by step02:10
wpwraki just wish there was a plastic equivalent to lead/tin. those wooden molds are actually quite fun. incredibly cheap to make, and good for quite number of runs. yield is bad, because the heat distribution is all wrong, but on a good day, with the counterweight, about 80% are possible.02:16
wpwrakthe molds would even as longer if i hadn't lazily (*) skipped some basics of mold-making of course. (*) actually, mainly because heekspython doesn't have the functions i would need for giving the molds a better shape. better shape would mean non-vertical walls and rounded corners.02:18
nebajoththis paper by sebastian is super interesting02:33
wpwraknebajoth: yup. very nice work. he thought of a lot of things.02:34
wolfspraulok guys then jump into the Milkymist project! :-)02:39
nebajoththis is totally the future02:39
nebajothfpga cpus02:39
nebajoththe cpu itself becomes a spime02:40
wolfspraulyeah but we need free fpgas as well, ... :-)02:40
nebajothcursed recursive tasks02:40
wolfspraulnebajoth: do you think you can help in any way?02:40
wolfspraulI need to make a rough allocation plan for the rc2 boards02:40
nebajoth"any" is a broad word02:40
nebajothallocation plan02:41
wolfspraulwe will probably only make one run02:41
nebajothwhat is the unit cost?02:41
wolfspraulsure :-) following and giving feedback is already a start02:41
wolfspraulwell it's difficult02:41
wolfspraulI don't want to make too many, because it's still too risky02:41
wolfspraulif there is a problem, my reputation as a manufacturer will suffer, along with the cash losses02:41
wolfspraulfirst run was 6, and for such a new design the results were very good02:42
nebajothhow different is rc2?02:42
wolfspraula safe bet would be 10-15, but we want to be more aggressive02:42
nebajothwhy do you want to be more aggressive?02:42
wolfspraulrc2 only has bugfixes from rc1, here's what we found02:43
wpwrakright now, he seems to need mainly fpga experts. and all is done with xilinx's proprietary tools. to make this a useful linux system, a mmu would be highly desirable. that's currently missing. of course, with an fpga, it's "just" a software upgrade.02:43
wolfspraulsebastien routinely tells me he does not at all only need fpga experts02:43
nebajothwhy was the mmu left out originallyy?02:43
wolfspraulin fact he would love people to jump into the user-mode app now, flickernoise02:43
nebajothI know a DJ who would be interested in it02:44
wolfsprauland there is always room to improve the toolchain, Linux kernel, distros etc.02:44
nebajothI also know someone with some fpga experience02:44
wolfspraulnebajoth: ok let me explain first about the 'allocation'02:44
wolfspraulso Sebastien wants 20 for 27c302:44
nebajothsorry, I know I ask big questions02:44
wolfspraulwe will sell them, probably at 350 USD + shipping02:44
wolfspraulit's very hard to calculate costs for such small runs, essentially all participants continue to invest money (spend more than they make), although some bean counters may do some math on the components and come up with a 'bom' of maybe 100-200 USD02:45
wolfspraulbut bottom line is the rc2 will be offered for 350 USD02:45
wpwrakwolfspraul: yes, he also wants people to work on the incompatible with the rest of the world toolkit and gui :) i'm much less convinced about that side of the story.02:46
nebajothI don't think its unreasonable to organize people ahead of time for a run this small02:46
wolfspraulwe don't want to make too few, cutting into or delaying valuable contributions, and also not too many, giving us a sales headache, especially if there are any problems with the boards (case is also missing)02:46
nebajothie: collect money02:46
wpwrakwolfspraul: well, it is compatible at the bottom - can run on SDL or X. but things on top aren't.02:46
nebajothperhaps even $40002:46
wolfspraulwpwrak: I am talking about gcc, Linux, OpenWrt02:46
nebajothfrom everyone who wants one02:46
nebajothjust put a date on it02:47
nebajothPlace your order by Aug 29, 201002:47
wpwrakwolfspraul: yup, these look more useful to me02:47
wolfspraulyes the organization starts here and now :-)02:47
wolfspraulwpwrak: Sebastien has a very VJ-driven focus, and I like his focus actually. He wants to drive the device to usability asap.02:47
wolfspraulnebajoth: I will not take pre-orders, always just creates trouble.02:48
wolfspraulaccess to such early runs is very unusual anyway. many things can go wrong, and for many of those I will take the risk.02:48
nebajothI think the VJ thing is pure genius02:48
wpwrakwolfspraul: i haven't quite figured out what his main motivation is actually. when we discussed the system, he didn't mention VJing a lot. but that doesn't have to mean much.02:49
wolfspraulso we have a rough plan of people who just 'say' they are interested, and we believe them because we know them and we know they are reliable02:49
nebajothyeah man02:49
wolfspraulI did some math on my end and maybe I need 5-10 boards for people I know.02:49
nebajothI would totally pay $350-$400 to have one of these02:49
wolfspraulI didn't know wpwrak or nebajoth should be on my list :-)02:49
nebajothsoft upgrades02:50
nebajothits an investment02:50
wpwrakwolfspraul: (mm case) seems that the case will be abit expensive, no ? also probably way overengineered for the cosy developer's lab environment ? (as opposed to the harsh on stage world)02:50
wolfspraulnebajoth: yes but it's risky, also for me. I cannot just crank up the volume because some people are excited. it's very risky. if something goes wrong I am dealing with a lot of expensive junk :-)02:50
nebajoththink how aweseome release version 68.3 is going to make my milkmist02:50
wpwrakwolfspraul: i probably shouldn't be on the list just yet. rc3 maybe :)02:51
nebajoththen take preorders :P02:51
wolfspraulnebajoth: maybe you think what you can contribute. If you just want to watch development 'live', install updates every few months, I think the RC2 run is too early for you.02:51
wolfspraulwpwrak: yes but I didn't even know you are remotely interested, but now I do.02:51
nebajothwell mabe02:51
wolfspraulI just need to have a vague idea who could be interested.02:51
wolfspraulI think the run will be 30-35.02:52
nebajothso why no mmu?02:52
wolfspraulthat already quite risky, but then we could have devices for a number of really great people that can help drive this project forward.02:52
wolfspraulnebajoth: simply nobody has gotten to the MMU yet02:52
wpwrakwolfspraul: i would be even extremely interested if (FPGA) synthesis was possible with free tools. we're not there yet. sebastien is contemplating to do a thesis on this, though :)02:53
wolfspraulyes I know [free tools]02:53
wpwraknebajoth: (mmu) wasn't necessary for the VJ app.02:53
nebajothfpga synthesis02:56
wolfspraulnebajoth: what you can do already today, and what would really really help, is to drive awareness to the Milkymist project02:56
nebajothprogramming the fpga I assume?02:56
wpwrakFree fpga synthesis could get very very exciting. imagine being able to load your own hw accelerators into your cpu. need a one-round-of-AES instruction ? just load it. need a really quick arccosh(x)/x ? no problem. etc.02:57
wolfspraulwrite great blog posts, talk to journalists, get slashdot/engadget/whoever to talk about it02:57
wolfspraulit's a great story, but it needs to be told right02:57
wolfspraulthe next ARM/MIPS?02:57
wolfspraula free CPU?02:57
wolfspraulwhat is this thing? why should people care?02:57
wpwraknebajoth: you write the "program" in verilog (a language a bit like C) or vhdl (a language a bit like COBOL), then feed it to a "compiler" that allocates the cells on the chip, connects then, and configures them. this is the synthesis.02:58
wolfspraulthis is a free CPU, it needs the help from a lot of free software people, because very few companies will help at the beginning, unlike proprietary CPUs where of course the proprietary companies like more 'open' software, it helps them sell their proprietary IP02:58
wolfspraulSebastien chose the LattticeMico32 core, so at least he doesn't have to start from zero02:59
nebajoththeres no open tool to do synthesis?02:59
wolfspraulbut still, we are quite lonely out there02:59
wpwrakwolfspraul: (pr) or make a MM NN. that's something whose reality they can physically perceive :)02:59
wolfspraulwell we make Milkymist One02:59
wolfspraulnebajoth: no, maybe first step the xilinx bitstream format would need to be documented03:00
wpwrak(sebastien will probably have a lot of corrections to what i'm writing here ;-)03:00
wpwrakwolfspraul: there is actually some work on that. dormant at the moment, though. lemme search for the link ...03:01
wolfspraulyes I know03:01
wolfspraulbut that works needs to be finished03:01
nebajothwhat, so we can reverse engineer it? :P03:01
nebajothare the target fpgas always xilinx?03:02
nebajothor does everybody use xilinx bitstream format?03:03
wpwrakthey support the chip MM is using. so that part may already be solved. not sure if the fpgas's elements are fully characterized, though. (timing and such)03:05
wpwraknebajoth: i'd be very surprised if there are any compatibility on such things among vendors ;-)03:06
nebajothI'm a total n00b to it03:06
nebajothxilinx fpga03:07
wpwrakof course, once you know how to synthesize for one fpga, the next won't be too hard. xilinx seem to be among the easier, do to having a very regular structure.03:07
nebajothhow much do fpgas cost?03:12
wejpdepends on how many cells they have03:15
wolfspraulexpensive, they are chips with lots of proprietary IP and under relatively little price pressure03:15
wpwrakthey're unfortunately very pricy. USD 50 for a really cheap one. the one in MM is about USD 200 in quantities of one piece from digi-key. so probably USD 100 or more for large volumes.03:15
wpwrakwolfspraul: how much do you pay for it ?03:15
wolfspraulca. 40 USD :-)03:16
wolfspraulwell I explained the whole s-6 situation earlier03:16
wolfspraulit's difficult, so also no reason to blame digikey03:16
wpwrakbut mm uses a virtex, no spartan, no ?03:17
wolfspraulah no03:17
wolfspraulvirtex is very expensive03:17
wolfspraulthere are even more proprietary IP blocks in virtex03:17
wolfspraulafaik Sebastien knows this better03:17
nebajothrc1 has a virtex?03:17
wpwrakXC4VLX25, according to sebastien's thesis03:18
wpwrakin a 363BGA, that one is USD 211.2 at digi-key (1 unit)03:18
wpwrakno volume discount listed03:19
wejpthe new firmware image for the ben looks pretty good, but gmu has been built without FLAC support. this should be changed for the final release. several people already asked about flac support on the ben03:19
wolfspraulthe beauty of copyleft hardware: http://en.qi-hardware.com/wiki/Milkymist_One_RC1_BOM03:21
wolfspraulmilkymist one has a 484 pin package03:21
wolfspraulES = engineering sample03:22
wolfspraulthat will change for rc203:22
wolfspraulI think digikey lists it for ca. 50 USD?03:22
wolfsprauldon't know03:22
wpwrakah, so he switched fgpas. nice.03:23
wpwrakusd 53.94 at digi-key. and they have 546 in stock :)03:23
wolfspraulwejp: do you have commit rights on the projects server? can you help us there? or maybe we write it up in the issue tracker or wishlist on the wiki03:23
wejpwolfspraul, no i don't have commit rights, but sure i can help you there and actually it is very easy03:24
wolfspraulwow, great03:24
wolfsprauldo you have an account on projects?03:24
wolfspraulI just found one :-)03:26
wpwrakbig chip. 23x23mm. there's one with 324 balls that's smaller. only 15x15.03:26
wolfspraulwejp: there is a 'wejp' user on projects03:26
wejpbasically all that needs to be done is add "decoders/flac.so" to the DECODERS_TO_BUILD= line in nanonote.mk, then recompile Gmu03:26
wejpwolfspraul, oh right, sorry03:26
wejpits been a while since i created that account03:26
wolfspraulwejp: I added you as member to both openwrt-xburst and openwrt-packages03:28
wolfspraulso you should be able to commit, and thanks a lot for any fix you can contribute there!03:28
wolfspraulit's really a whole lot of work across the distro/OS, so any bit helps03:29
wejpsure :)03:29
wolfspraulI like FLAC too, let me see what Wikipedia says about the patent situation03:29
wolfspraulI think it's OK, at least there is nobody aggressive enforcing anything...03:29
wejpyes, should definately be okay03:30
wejpthey were actually trying not to use any patented stuff when developing FLAC03:30
wejpjust like with vorbis03:31
wolfspraulyes, I see it03:31
wejpand both vorbis and flac are maintained by xiph.org03:31
wpwrakhmm they did the latticemico32 port (lm32-nommu) in 2008 but it's still not in mainline.03:35
wolfspraulnot much manpower behind all this03:35
wpwrakif the code isn't horrible, it shouldn't need much03:36
wpwrakparticularly not from the makers03:37
lekernelwpwrak, the code is horrible04:34
lekerneli'd even say it's a bunch of crap04:35
lekernelif you're looking for a lm32 linux to work on, I suggest you use http://github.com/tmatsuya/linux-2.604:36
lekernelthere's still a lot of trash in this, but we fixed and improved a couple of things04:36
bartbesbtw, I'd like to state that the nn holds up nicely under 'heavy' load04:44
nebajothyeah, she's a sturdy little beast04:48
bartbesI have had no problems drawing basic images over SDL04:48
bartbesit started getting slower (obviously) when I started rotozoomzing everything every frame04:48
bartbesbut tbh it doesn't take a lot more for slowdown to be noticable on my computer04:49
nebajothwhat are you working on?04:49
nebajothjust stress testing it?04:49
bartbes(because the rotozooming is *amazingly* slow, I should really buffer that)04:49
nebajoththat's a super cool web design too04:51
bartbesyes the project starters were very cool guys ;)04:52
nebajothepic volley04:54
nebajothbartbes vs qubodup04:54
nebajothI heard this match is OFF THE HOOK04:54
bartbesI think it's the only recorded match04:54
nebajothfire the lazors!04:55
bartbeshave you seen my vid?04:55
nebajothit seems you might have many04:55
bartbesI have only uploaded 1 myself04:56
nebajothMY BRAIN04:56
nebajothSPACE OTTERS04:56
bartbesbtw, that version ran a little too fast due to a bug I had atm04:57
bartbesat *that* moment, I should say04:57
nebajothgood problem to have04:57
bartbesyeah, it is ;)04:57
lekernelwpwrak, btw my thesis was written before the milkymist one, while i was still using a virtex4 ml401 board to develop the system05:02
lekernelthen I ported it to the milkymist one (and spartan6) in about 10 days, which is a good sign of portability of the design :)05:03
lekernelwpwrak, btw if you don't believe in the toolkit i'm using, check this:05:05
lekernellet's say we are contributing to the development of an alternate toolkit that solves many of the problems X has especially on embedded systems05:06
lekernelbut if you like X that much, just port it to the milkymist one05:07
lekernelthen you'd perhaps have to add a MMU, port linux, fight with GNU toolchain and autocrap problems, etc. etc.05:07
lekerneland in the end it will be fucking slower from a end user perspective05:08
wpwraklekernel: (ugly lm32) ah, pity. is takeshi working on getting the port into linux mainline ?10:50
wpwraklekernel: (virtex->spartan) yeah, it's nice to see that this was fairly painless. and it's also a much more affordable chip :)10:51
wpwraklekernel: (autocrap) there we agree ;-)10:53
kristianpaulno more talks about craping software wpwrak  with lekernel is enought ;)11:02
larscwell, actually usually 'autocrap' solves more problem then it causes11:04
wpwraklarsc: also considering that most of the pre-posix systems it was originally designed have long moved to the graveyards of history ? :)11:08
larscwpwrak: yes11:22
larscespecially when it comes to crosscompling11:28
wpwraklarsc: hmm, wouldn't a little HOWTO solve this, too ? of course, it may be harder to make people read that one than to cook up something that works with their current version of autotools ...11:30
larscwpwrak: sure11:31
zearhey guys, who's responsible for the qi hardware wiki?12:54
zearsome of jlime dev members have doubts about something, so that's why i think i'm gonna ask the wiki maintainers this question:12:55
zearis it ok to add an article about Jlime in the qi hardware wiki? Jlime is a distro for the nanonote, so i guess it is ok12:55
zearbut since some had doubts, i'm asking you guys ;)12:55
zythzear, well, regestry and add self )12:56
zearzyth, oh i have an account on the wiki for a long time now12:56
zearjust some jlime team members raised a discussion that jlime contains non-free codecs (ex. mp3) and qi-hardware might not be legally authorised to put any information about jlime12:57
zearbut i doubt this is true12:57
zearwiki is a community based work after all12:58
zythoh, yeah12:58
zearbut it turns out nobody knows about jlime, and features it offers (x11, mplayer, wiki reader, dictionary)12:59
zearso i guess mentioning about jlime might help raise the sells of the nn13:00
zythask mirko13:00
zythnobody seems alive here )13:05
zearyea ;)13:05
wpwrakphew. quite the ordeal to solder this board13:13
wpwrakbah. now that i'm done with the soldering, the sun comes out from behind the clouds.13:33
wpwraktuxbrain2: regarding the SIE logo ... see the right side of http://www.almesberger.net/misc/ben/wpan/first-pcb.jpg13:37
lekernelwpwrak, no, takeshi is just playing around basically14:00
lekernelhe's a researcher, not a pragmatist14:00
lekernelso, if someone could take over, cleanup the code, fix the bugs and get it into mainline that would be nice14:01
lekernelpersonally I gave up, too much work and too little advantages compared to rtems14:01
Action: wpwrak wonders if lars' fingers are twitching. he's done a lot of upstreaming lately.14:07
larscdoes upstream actually support non mmu archs?14:09
wpwraklarsc: yes, e.g., m68knommu14:09
wpwraklarsc: it also shows nicely how one can branch a nommu off a mmu arch (m68k in this case)14:09
wpwraklarsc: there may be more. don't quite know all the exotic ones :)14:10
lekernelbtw the uClibc/binutils support is pretty shitty too14:14
lekernel(for the executable loader)14:14
wpwraklekernel: how fast do you think you could scale that core ? e.g., compared to the 4740, for some random workload14:14
lekernelwhat's 4740?14:15
wpwraklekernel: the whole "let's optimize for small" area seems to suffer that problem a bit. uclibc, busybox, dropbear, the various "reduced" shells, etc.14:16
wpwrakthe ben's cpu14:16
lekernelwell if you use the whole gnu libc it's going to suck too14:17
lekernelok, well you have the mibench results in my thesis14:17
lekerneljust run mibench on the nanonote14:17
lekernelthat's the only real way to know14:17
lekerneland fyi the lm32 should run at 800MHz in 90nm14:17
wpwraklekernel: so you expect it to scale without problems ?14:19
lekernelwhat do you mean, scale?14:23
wpwraklekernel: no delay mismatches that are insignificant at low speeds but that matter at high speeds, etc. not sure what kind of problems exactly to expect14:25
qi-bot[commit] Werner Almesberger: Brown paper bag time: the VDD net was connected to VBUS, not VDD/VIO. http://qi-hw.com/p/ben-wpan/524665f14:47
lekernelit's a fully synchronous design, so no such problems in theory15:18
wpwraklekernel: kewl. now all wolfgang has to do is find a fab :)15:35
qi-bot[commit] Werner Almesberger: Net name correction, layout update. http://qi-hw.com/p/ben-wpan/c9e61ee16:22
Textmoderafa: I converted the vid files with the command you linked. they don't work.16:46
rafaTextmode: can you give me some of those files?17:04
bartbesyes Textmode, give him the files17:08
Action: bartbes puts on his jedi mask17:08
bartbesYou will give rafa the files.17:08
Action: bartbes removes jedi mask17:08
Textmodejedi have masks?17:09
bartbesI have a mask that makes me a jedi17:09
bartbesI guess..17:10
bartbesdon't you try to make sense out of my horrible jokes17:10
wpwrakrafa, may the Files be with you :)17:11
Textmoderafa: can you handle lzma files?17:11
rafaTextmode: yeah I guess ;)17:13
rafawpwrak: haha .. that would be better17:13
Textmoderafa: you got that?17:16
rafait is retrying to talk with the host.. wget gives me something like "expired the connection time"17:17
rafaseveral times17:17
rafawithout luck17:17
rafasomething is not okey with that website17:17
rafaTextmode: nop.. wget can not get the file17:18
rafawhat if you use some crapupload?/megaupload?rapidshare?17:19
tuxbrain2wpwrak: we must sue you? or you to me? :P17:33
Action: Textmode huggles rafa17:43
rafaTextmode: hey, I got the file.. let me check.. it gives me error as well.17:54
Textmodeso at least its not just me.17:56
wpwraktuxbrain2: naw, you have to sue TI :) it's their design.18:07
Textmoderafa: ?19:21
wpwrakoh. the ben's memory bus is only 16 bits wide. hadn't noticed that before.20:39
qi-bot[commit] Juan64Bits: Somthing routing http://qi-hw.com/p/xue/048e8b421:16
kristianpaultuxbrain2: :)21:19
kristianpaulJlime !!!21:19
kristianpaulAnd YES it runs X on the nanonote :)21:19
tuxbrain2TRUE X yeah!21:19
tuxbrain2time to sleep :) n821:20
kristianpauli love this http://en.qi-hardware.com/wiki/File:Frontmockup.jpg21:20
kristianpaultuxbrain2: good sleep :)21:21
tuxbrain2hehe this is finally discarted , too much colors sticker too much expensive :P21:21
wpwrakrafa: btw, might be interesting to see how fast this runs with the jlime toolchain: http://pastebin.ca/191977721:22
wpwrakrafa: memory bandwidth is 256/run_time21:23
wpwrakgrr. 256 MB / run_time21:31
wpwrakwolfspraul: btw, did you get any reaction on the HTML file in repository viewing issue ?21:48
wolfspraulwpwrak: yes, negative :-)21:59
wolfspraulone second...21:59
wolfspraulI just got it a few hours ago22:00
wpwrakcowards :)22:03
wolfspraulthey should just have a HTML 'sanitizer' library in PHP that projects like indefero could quickly pump html through to only leave the wanted features in it22:04
Snicker843i want to buy a new computer, but i cant decide if Socket 1366 or 1156 for an i7 ... because 1366 is limited to 1066 mhz RAM, and for the Socket 1156 is no i7 with 3 ghz+ available.22:15
wpwrakhmm. the milkymist should already run circles around the 4720 in terms of memory bandwidth.22:18
Snicker843i dont understandm sorry22:18
qi-bot[commit] Werner Almesberger: ben-dispbase-back-100um is done. Pushing data files ... http://qi-hw.com/p/ben-scans/7d7fb5722:19
wpwrakSnicker843: i think you're not where you think you are :-) this isn't a general hardware discussion channel. we have rather peculiar topics at times :)22:20
Snicker843sorry then22:20
Snicker843im not familiar with irc22:20
Snicker843just searching for help22:20
wolfspraulwpwrak: peculiar :-)22:26
qi-bot[commit] Werner Almesberger: Index page update for ben-dispbase-back-100um completion. http://qi-hw.com/p/ben-scans/32b314822:28
qi-bot[commit] Andres Calderon: DFN10 footprint added http://qi-hw.com/p/xue/76c952622:37
wpwrakwolfspraul: ah, may I ask you for another USB product ID ? for ben-wpan. 1540 would be good. 154 for the end of 802.15.4, 0 for the first set of devices for that protocol.22:57
wpwrak0x1540, to be precise22:58
wolfspraulsure just do it23:01
wolfspraulwe have 65000 of them :-)23:01
wpwrakthanks ! i hope you're keeping track of them :)23:01
qi-bot[commit] Andres Calderon: 3.3v dc-dc added http://qi-hw.com/p/xue/e2867d723:10
wolfspraulfirst step - it's logged :-)23:11
--- Thu Aug 19 201000:00

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