#milkymist IRC log for Tuesday, 2011-05-03

kristianpaulone more lib http://www.hyperrealm.com/libconfig/ ;)01:59
kristianpaulvia zrafa at jlime mail list01:59
kristianpauljust in case the current config file will just going to grow, wich i think acording to xiangfu todo's list..01:59
kristianpaulxiangfu: ping02:02
xiangfukristianpaul: hi02:03
kristianpaulxiangfu: for the cards you sent me, i remenber you we're not able to write then isnt?02:04
kristianpaulor byw ftp at least..02:04
kristianpauli think this is a know bug, just wanted to confirm02:05
xiangfukristianpaul: yes. know bug. it's also mount as read only.02:05
kristianpaulah... that explain02:06
kristianpaulhow i can enable mount?02:06
kristianpaulwrite support on mount**02:06
kristianpaulbtw zrafa pointed this interesting libconfig library02:07
xiangfukristianpaul: no. there is no write code in driver02:07
xiangfustatic int memcard_disk_block_write(rtems_blkdev_request *r)02:07
xiangfu  return -RTEMS_IO_ERROR;02:07
kristianpaulergg ;)02:07
xiangfutems-milkymist.git/c/src/lib/libbsp/lm32/shared/milkymist_memcard/memcard.c LINE: 20002:07
xiangfukristianpaul: you can try to finish it :)02:08
kristianpaulhe, i wish i could..02:08
kristianpaulbut i bet this is already written may be in netbsd..02:09
xiangfukristianpaul: hmm... there is no bad block check in memcard.c02:29
lekernelkristianpaul: yeah libconfig is great, I have already compiled for lm32/rtems and it works09:01
lekernelbut I don't use it atm09:01
lekerneland no, the memory card driver isn't complete.09:02
terpstralekernel, if i might ask, now that there is (streaming) wishbone b.4, do you still see a need for the custom bus you used in the milkymist?09:52
lekernelI don't know... haven't checked wb b4 in detail09:52
lekernelthere's a new lm32 software release by lattice btw09:53
terpstrathe reason for your custom bus was for 1-1 tranfer-clock, as opposed to classic wishbone 1-2, right?09:55
terpstraor was there another reason?09:55
lekernelthe main reason is to 1. allow pipelining and 2. simplify burst (just like a SDRAM, all the transfers are bursts of the same length)09:56
terpstraahh, ok09:56
lekernelit also removes the wb_cyc signal I've never understood the purpose, too09:56
terpstrawishbone B.4 makes burst even more complicated09:56
terpstrathe CYC line is used for arbitration09:57
lekernelwhy not use STB for both?09:57
lekernelit doesn't make sense to request the  bus if you're not transferring something to it09:57
terpstrawell, i only work with streaming wishbone, and there it makes sense09:57
terpstrai am not entirely clear why you needed it for classic wishbone09:57
lekernelI don't need it, but the spec says it should be there09:57
terpstrafor streaming wishbone you need it to keep the device 'open' while waiting for your streaming ACK/ERR/RTY lines to come back in09:58
terpstrain classic wishbone you probably needed it for atomic test/set or increment09:58
terpstrafor streaming, it is much more useful in arbitration09:58
lekernelwell...  I don't need those atomic operations either09:59
lekernelso far at least :-)09:59
lekernelmaybe someday i'll have multiple lm32 cores, then we'll see :)09:59
terpstrain streaming wishbone, STB means push a (read|write) transfer into the queue and CYC means you are not done with the device yet09:59
terpstrathe lm32 doesn't have any atomic operations atm09:59
terpstrastreaming wishbone has other annoyances though10:00
terpstra... why oh why is the lm32 rpm so damn huge. :P10:00
lekernelwhich one are you using?10:01
lekernelI have lost my lattice login so I'm trying to get it from ftp :)10:01
terpstrafetching diamond_1_2-lm-92-i386-linux.rpm10:01
terpstrai can extract the lm32/rtl folder and mail it to you if you like10:01
lekernelah, that would be cool10:01
terpstrai will also see about repatching the patches on their last version10:01
terpstrabut really, a 300MB compressed archived for the 3MB uncompressed RTL. bah.10:02
lekernelyou're using exactly the same LM32 source as ours, right?10:02
terpstrayou never integrated my JTAG patches10:02
lekernelah, yes10:02
terpstraalso not the multiplier10:02
terpstra(which really is a vendor-specific thing i guess)10:03
terpstraboth of those are vendor specific things really10:03
lekernelbut except those small things, there are no other changes?10:03
lekernelok, cool :)10:03
terpstrai want to read the changelog for the new lm3210:04
terpstramaybe they added proper pipelining!10:04
terpstra(right now i have a nasty vhdl hackdpter to turn it's shitty B.3 burst mode into pipelined WB)10:04
lekernelwell... my other cores don't support it atm10:05
lekernelso, later :p10:05
terpstraall your stuff is verilog, right?10:06
terpstrathen you can't use my vhdl crossbar switch anyway ;)10:06
lekernelwell, practically I could10:06
terpstrait wouldn't be so nice10:06
terpstrait uses records10:07
lekernelbut I don't really want to mix languages since it would cause additional difficulties the day we switch to open source synthesis10:07
terpstra(at the bottom i instantiate the crossbar)10:07
terpstrai understand10:07
terpstrathe people here who do the 'real hardware work' all use vhdl, so i had to make stuff compatible for them10:08
terpstrathey still use gcc 4.3 in the newest lm32 package10:14
lekernelah, I thought I saw a 4.4 somewhere10:15
lekernelmaybe that was with the theobroma gcc10:15
terpstrathey've included some of the portability fixes upstream!10:15
terpstraor wait10:16
lekernelthere is also a 3.x LM32 gcc around which is super-broken10:16
terpstrawhat does '#1' do in verilog?10:16
lekernele.g. variadic functions like printf() break the stack10:16
terpstrathey added it all over the place as compared to the last lm32_top10:16
lekerneldelay (for simulation)10:16
lekernellike 'after' in vhdl10:17
lekernelthis is usually a bad idea to add it everywhere, since there are already delta delays in the simulator to perform that (though those in verilog are done in a cretinous way)10:17
terpstrawell, they add it to tons upon tons of assignments10:18
lekernelargh ...10:18
lekernelthis is no good coding practice10:18
terpstradone reading the diff10:26
terpstrathey changed nothing of consequence10:26
terpstraall the (large patch) does is: add #1 to every synchronous assignment and 2) change clogb2 to clogb2_v110:27
terpstrastill want it?10:27
terpstrasomewhat disappointing10:27
lekernelyeah and we'll have to remove those #1's in the future10:28
lekernelmaybe they laid off the original lm32 team and gave it to some other guys who just maintain it so it somewhat works ...10:29
lekernelthat sounds like "let's fix that simulation problem one of our customers has... hmm..."10:30
lekernelfunny answer from google about VP8 not being open source:10:36
lekernelAt this point we felt it would be more helpful for VP8 proliferation to concentrate our resources around developing the IP itself.10:36
lekernelI personally think that OS hardware has also less benefits than OS software as few people can access the required development tools and computational resources so they just don't have the capability of actually verifying the code they write. This is a big design (~1 M gates encoder and decoder combined) and it will take a week to run through the test suite with a single computer. A cluster of workstations and FPGA boards are in practice10:36
lekernel pretty much mandatory. If the design was considerably smaller it would certainly make more sense.10:36
wolfspraullekernel: you got an answer from Google?10:37
wolfspraulthe 'do no evil' company?10:37
lekernelmaybe I should send them a picture of me holding a board from my 384-FPGA cluster :-)10:37
wolfsprauland they flat out said source code will not be released, or what?10:38
wolfspraulI mean they made this big press PR statement about 'open' :-)10:38
lekernelwolfspraul: got two answers... first they said "I took a look at the Milkymist SoC at your website and it really looks like a great project! However, our WebM RTL is not open source although it is free, so I don't know if there is a way we could make it part of that platform. Please let me know if you have a suggestion."10:38
terpstrathey have a 1M gate video decoder??10:38
wolfspraulfree as in beer10:38
wolfspraulwell, what I told you :-)10:38
lekernelso I asked why not open source it, and they answered again10:38
rohin short 'we are shitting you and our marketing isnt part of the "no evil" doctrine'10:39
lekernelwith only 8 hours delay... progress :-)10:39
lekernelI'm not sure if they're meant to bullshit10:39
wolfspraulI am not the least surprised, I read their original announcement exactly correct, I figure.10:39
lekernelas they say: "At this point we felt it would be more helpful for VP8 proliferation to concentrate our resources around developing the IP itself. "10:39
wolfspraulVP8 proliferation is their goal10:39
rohlekernel: sure. i always assume people are that stupid before assuming they are evil ;)10:39
lekernel_maybe_ they just not want to be bothered with stupid support request from people who can't pull it off10:40
lekernelterpstra: 1M gates is encoder + decoder ... (whatever 'gate' means)10:40
wolfspraulif Samsung calls and says they want this published under GPL so they can build some chips with it, it would be GPLed 1 week later10:40
lekerneloh, certainly10:41
rohwolfspraul: well.. try it. till they find out you arent from samsung.... its in the wild10:41
rohi actually really like that idea10:41
wolfspraullekernel: you can try to sell them a few m1 as a cool development platform10:41
wolfspraulroh: any status update on the cases?10:42
rohwolfspraul: nope.10:42
wolfspraulwhat was the latest status then? all material already ordered?10:43
wolfspraulalready arrived?10:43
rohwill need to mail the laser guy again. but we god rid of the old rooms, got new ones and are still working on setting everything up again10:43
wolfspraulah yes, please try to move this forward, especially get all the material in place etc.10:43
wolfspraulI do need the cases soon, in a few weeks or so would be nice10:44
rohnope.. havent ordered the acryllic yet.. seems the other lasercutter can use bigger sheets so i would like to optimize the size of the sheets10:44
wolfspraulhow about the rest? spacers, screws, feet10:44
rohsure. just a matter of weeks anyhow. we are now working 3 weeks on the move fulltime ;)10:44
wolfspraulyes I understand10:45
wolfspraulcases are not on the critical path right now, so I'm still quiet10:45
rohneed to check. most i have in stock, some i need to order. especially if we move the top or so10:45
wolfspraulbut no need to let them become the critical path either10:45
wolfspraulmove the top?10:45
rohbecause of the dmx sockets release button10:45
rohatleast i remember some talk about it10:46
rohbbl.. need to run now10:47
wolfspraulat this moment it looks like we leave everything as is10:47
wolfspraulsome more comparisons, will let you know10:47
wolfspraulbut definitely please try to move ahead with the production run10:47
wolfspraulit's a completely firm order on my side, down-payment made, clock ticking...10:47
wolfspraullots of things moving for the run10:47
wolfspraullekernel: have you tried the obk-2010cw camera?10:51
lekernela little bit10:51
lekernelso far it looks ok, except that there is no way to mount it on a regular camera stand10:52
lekernelseems it needs to be screwed... well... I guess duct taped in most practical cases :-)10:52
wolfspraullekernel: no way? I liked the fact that it has a standard 1/4'' 20tpi threading10:53
wolfspraulwhat kind of stand do you mean?10:53
lekerneltripods for example10:53
wolfspraulyes but it has a tripod threading10:54
wolfspraulwhat does your camera look like? :-)10:54
wolfspraulit should be a bullet-style, and at the bottom of the bullet is the tripod threading10:54
lekernelah, that's tripod compatible?10:54
lekernelgood, then :)10:54
wolfspraulthe package should also have included a tiny small stand10:54
wolfspraulyes sure10:54
wolfspraulin fact that little stand is crap10:54
wolfspraulwell maybe in some near hopeless cases it may actually help, but to me the main thing is the center bullet case10:55
lekernelbut I was thinking that the camera was specific to that small stand10:55
wolfspraultripod threading at bottom of that case10:55
lekernelyeah I think we can omit that small stand10:55
wolfspraulit's a standard 1/4'' 20tpi threading10:55
wolfspraulthat was one reason why I liked this particular model, over others10:55
lekernelyeah great10:56
lekernelwell can you order without the small stand?10:56
lekernelto me it's an useless thing10:56
wolfspraulprobably yes10:56
wolfspraulwill save a few cent, and some weight too10:56
wolfspraulwe can change a lot of things, but for rc3 I want to keep it simple10:56
wolfspraulalso for bigger changes there may be higher moq (200 units), or one-time fees etc.10:57
wolfspraulespecially with schematics changes or so10:57
wolfspraulkicking out that stand should be easy10:57
lekernelif the stand didn't fall when the camera is on it, it would still have some use... but if it requires screw mounting, well...10:57
wolfspraultry the standard tripod threading10:58
wolfspraulthe stand is optional, that's the poin10:58
wolfspraulbut yes, it is very small so in most cases you need to fasten the stand somehow, most likely with screws10:58
wolfspraulI'm sure we don't want to include a tripod as well :-) (just joking)10:58
wolfspraulwell keep trying and keep me posted about obk-2010cw feedback...10:59
wolfspraulit should be more or less waterproof, btw10:59
wolfspraulbut maybe not for under-water shooting :-)10:59
awlekernel, yesterday you said that don't touch audio codec, could you check that Notes3 of figure 18 about SPDIF 48 pin? I am not sure if I understand it completely.11:02
lekernelin what document? the datasheet?11:03
awyes, thank.11:03
awpage 32.11:03
lekernelyeah do not put R111:04
lekernelconnect pin 44 to J3, done11:04
awyes, we don not need R111:04
lekerneliirc there is an internal pull down so we don't need anything else11:04
lekerneland you connect pin 48 directly to J3 as well11:05
lekernelonly two trivial modifications11:05
awpin44 to J3?11:05
lekernelpin 44 _and_ pin 4811:05
lekerneldirectly to J3 without any other parts11:06
awso pin 44 to J3, means you want enable this from outside, right?11:06
awi tied pin 44 to 3V3 by h/w it self to enable, good idea to let user to enable though. thanks.11:08
kristianpaulargg my lattice account loginn. but i get a 0 bytes rpm file... :S11:10
lekernelanything enabled is a potential source of problems, so disable it by default :)11:10
aw_some sort of a good conceipt. :)11:12
lekernelyeah p.8 of the datasheet says the spdif enable pin has an internal pull down11:12
lekernelso nothing else needs to be connected on it11:12
aw_umm..wait see page 16.11:14
lekernelPin 48 may be used to output the PCM DAC playback data in SPDIF (IEC958) digital data format. In order to enable this output, bit SPDF in Register 5Ch should be set or pin 44 pulled high.11:14
lekernelyeah so it's pulled down by default, and disabled. good.11:14
aw_yeah...yes..just totally understood now. :-)11:15
aw_two ways can enable this though. :-) thanks.11:16
terpstralekernel, i found an interesting paper on how to build a smaller TLB. they propose to use a single-port RAM block for BOTH ITLB and DTLB. The trick to making it run at near full speed is that they use a register to cache the last physical address loaded from the cache, thus cutting out most hits and allowing the port to be shared.11:17
terpstrai'm not sure if a dual-port RAM block is actually more expensive than a single-port RAM block on most FPGAs, though11:17
lekernelI think it's not11:17
lekerneljust routing resources11:18
terpstraso more an ASIC improvement then11:18
wolfspraulroh: ok about rc3 cases. for dmx tx, it looks like we leave everything as-is right now11:35
wolfspraulI will give you the final confirmation in a few days11:35
wolfspraulconsidering all the different options and pros and cons, the current solution is not bad, and sellable11:35
wolfspraulfor MIC, we still plan to move the microphone to the acrylic, and let it face the acrylic. so if you could make a small hole there, and 'MIC' label, that would be great11:36
wolfspraulthe final and exact positioning will take another week or so though11:36
wolfspraulother than that, I think all is settled11:36
wolfspraulI still like philips (+) keys better than allen keys (on top and bottom side)11:37
wolfspraulbut I can live with allen11:37
wolfspraul(allen = inbus)11:37
wolfspraulwe have to be careful about the DC jack tolerances, as Adam will try to make this more precise this time11:37
wolfspraultheoretically that should make things easier, but you never know, we are moving something so we have to be careful that the cases still fit11:38
wolfspraulone idea we haven't talked about yet to improve the dmx tx push button is to make the entire top and bottom parts a few mm smaller11:39
wolfspraulthat would reduce the amount it stands 'over' the side acrylic11:39
wolfspraulyou are probably worried about mechanical stability...11:39
wolfspraulif it would be smaller, it would presumably be easier to press down that dmx tx push button...11:40
wolfspraulI think it's about 7mm right now11:40
lekernelwolfspraul: tbh both the mike and dmx push button are working today, what isn't working here is publicity11:41
wolfspraulbut even that, I think the dmx tx button problem is not as severe as to warrant a reduction of that space11:41
lekernelin the loudness of a typical concert or party the mic picks up more than enough sound11:41
lekernellast time I even hard to turn down the volume in the GUI a bit11:41
wolfspraulunderstood, but I'm not as frustrated as to stop/deny improvements when I see them and when they come easy11:42
wolfspraulalso I'm just going through some points for roh so we are moving forward on the cases...11:43
wolfspraulon publicity, I've explained my strategy, and what I will do, several times11:43
wolfspraulI will work on two angles: 1) journalists 2) actual users, word of mouth11:43
wolfspraulI'm more worried about #2 than about #1 right now11:44
wolfspraulif we have a good press release, it will be picked up, I'm sure11:44
wolfspraulbut... I sold 32 units, and I clearly don't have 32 active users right now that are dragging this unit whereever they go, preach to all their friends how great it is, etc.11:44
lekernelI was talking about that in #qi-hardware the other day ... seems to me that absolutely all projects have a great deal of drawerware11:45
lekernellook at openpandora ... they claim to have sold 4000 units and there are 100 users in their IRC channel11:46
wolfspraulin those small electronic gadgets, you typically have 50% drawerware, even for big brands11:46
wolfspraulmaybe not Apple :-)11:46
wolfspraulit's scary but that's how it is11:46
wolfspraulso the thing is bought, and maybe turned on once, then into the drawer11:46
wolfspraulyou could call this a success of capitalism, or defeat. but it's reality.11:47
wolfspraulfor the 32 units we sold, how many do we see alive?11:47
wolfspraulmaybe 5?11:47
wolfspraulJon, Kristian Paul, Lars11:48
wolfspraulI see 3 :-)11:48
wolfspraulthere are probably a few more, but a good 25 disappeared in terms of publicity11:49
lekernelwhat can we do? more users? tutorials?11:52
lekerneltutorials published in specialized magazines seem to be a good idea for me :-)11:52
wolfspraulI think some of the use-case improvements we have in mind might help, for future buyers. for example the 'update' and 'next patch' buttons.11:53
wolfsprauloh sure11:53
wolfspraulout of the hacker scene, into audio and video lovers11:53
wolfspraulso for future buyers, I think I have a solid plan, with better accessories, better out-of-box software (easier to use for beginners), etc.11:54
wolfspraulwhether we can kick the rc2 customers alive I don't know :-)11:54
wolfspraulfor journalists - well, I just hate to create fake news. I'm not good at it, I abhor it. so I won't do it. Maybe someone else is better at this, then they should go for it.11:54
lekernelguyzmo: so what do you think about publishing that Lua tutorial?11:55
wolfspraulso without fake news, the main thing I'm going for is a proper launch press release, once we have neatly made rc3 in stock and ready for shipping11:55
lekernelwolfspraul: it doesn't have to be fake. we can contact journalists at each software release, for example.11:56
lekerneland tell them a nice story11:56
lekernelabout how great those new features are11:56
lekernelit sure takes a lot of energy and time, but I think there is no way around that11:56
wolfspraulyes and no. yes for building relationships with journalists, and keeping them posted.11:57
wolfspraulactually I do that - one reason I spend some time with montly qi news is that I then send that url to some journalists to keep them fed11:57
wolfspraulanother one is that a good story is not necessarily the same thing as a bunch of great facts11:57
lekerneluhm well... tbh this lacks the 'nice story' aspect11:57
wolfspraulyes, correct11:58
wolfspraulbut that's a competition of its own11:58
lekernelit's good as an internal newsletter for people who are already interested in our community11:58
lekernelbut it's terrible for outsiders11:58
wolfspraulso if you climb onto a Berlin highrise, and threaten to jump down unless slashdot (ahem) carries the MIlkymist news, and we shoot a video of that, then, as stupid as this is, that may be covered somewhere11:58
wolfspraulunfortunately there are tons of marketing/pr agencies flooding the world with hyped shit all the time11:59
wolfspraulbut the stories are still 'good', as in Hollywood good11:59
wolfspraulfor a journalist it's hard to wade through this and make good decisions11:59
wolfspraulthe qi community news is a 'copyleft hardware' thing, specialized12:00
wolfspraulfor Milkymist we need an entirely different approach, that goes over the video/synthesizer effects, etc.12:00
wolfspraulalso to make it big fast, we need a strong marketing partner, or a whole company that resells the product, as Roland or whatever12:01
lekernel_sorry, got disconnected after 'for a journalist it's hard to wade through this and make good decisions'12:02
wolfspraulthe qi community news is a 'copyleft hardware' thing, specialized12:02
wolfspraulfor Milkymist we need an entirely different approach, that goes over the video/synthesizer effects, etc.12:02
wolfspraulthat was all12:02
wolfspraullekernel: let's try to support Jon well. if he can really use m1 during the keynote or between presentations, that could make a difference12:22
wolfspraulthe ways we can support this are:12:22
wolfsprauleasy way for him to have some background picture, well maybe he can just point the camera to a printout :-)12:22
wolfspraulrss feature would be nice, but not a must imo if it's not in-time12:23
lekernelshould be in time12:23
wolfspraulmaybe there should be a way to continously display 'powered by Milkymist', even during rendering12:23
wolfspraulis that possible?12:23
wpwrakwolfspraul: maybe try to see if the monthly news have one item worth brushing up and bringing to the press ? it would lose the "instantaneousness", but at least it would align with a process we already have12:23
lekernelI already got atom and rss feeds from twitter downloaded and parsed by my m112:24
wolfspraulnot sure he would like that, but little details like this can matter if many people just pass by and briefly look over their shoulders12:24
lekerneland it displays the latest tweets from a particular hashtag on the serial port atm12:24
lekernelyeah this will come with RSS as well .... it will display a background message if there was no recent entry12:24
wolfspraulwe can also support Jon with a press release, even if it sounds bloated. but if someone can exercise it through it may create some press coverage, or at least some journalists have heard the name12:24
lekernelah yes12:25
wolfspraulbut for a press release, well, it's a lot of work :-) the details have to be right12:25
wpwrakwolfspraul: (powered by milkymist) for the presentation ? maybe better to just have one slide saying that this is presented on the mm112:25
lekernelwolfspraul: what exactly is Jon planning to do with the m1 during the presentation?12:25
wolfspraulwhat is the headline? "Secretive Milkymist video synthesizer shown to larger audience at LGM for the first time" :-)12:25
wolfspraulyes, see. I don't know.12:26
wolfspraulI don't even know when the exact dates are, I know nothing.12:26
wolfspraulif we do a press release, at least the facts have to be right12:26
lekernel"Milkymist video synthesizer enters the realm of the Internet of Things"12:26
wpwrakwolfspraul: a like "video synthesizer" much better than "vj station"12:27
wolfspraulyou are not the only one, we will switch to that12:27
wpwrakexcellent ! :)12:27
lekernelyeah, let's not sound too sophisticated12:27
wolfspraulpress release soudns easy, but then actually it's a lot of work to do one12:28
wolfsprauland the usual quote, etc.12:28
wolfsprauldetails details details12:28
wpwraklekernel: "video synth" also sounds less confined. "vj station" suggests that, if you're not a vj, it's not for you.12:28
wolfspraulok let's start :-) when is that lgm thing actually? :-)12:28
wpwrakwolfspraul: (quote) you could interview lekernel ;-)12:28
wolfspraulyes sure, he is the founder that's always good12:30
wpwrakwolfspraul: if you look at them, quite a lot of those quotes do actually come from people deeply involved in whatever the news item is. so they're not really sort of an external opinion or assessment, but just a soundbite12:30
wolfspraulmay 10-1312:31
wolfspraulso it starts next tuesday12:32
wolfspraulwhen will rejon actually turn on and use/display m1 to the widest audience he has there?12:32
wolfspraulwe should do the press release on that same day12:32
wolfspraulrejon: can you post some factoids here? maybe we can make the release like some "during his keynote at xx am/pm today, Jon Phillips ..."12:34
wolfspraulmaybe we can create a nice screenshot or two in advance? not sure how we can get the LGM logo/text into the display then12:35
wolfspraulfirst step - let's see when rejon comes back and what he says :-)12:35
wolfspraulI am willing to spend a day or two to support this release, in terms of writing it, getting the pieces together, and then shopping it to journalists, whether it's picked up or not.12:35
wpwrakwolfspraul: ask him to ask someone to snap a picture ?12:36
wolfspraulwe need to create the screenshots before the actual use, otherwise timing won't work12:36
wolfspraulI'd rather not get a crappy picture in there, and delay the news for it even.12:36
wolfspraulI need feedback from rejon now, I'm sure we get it when he gets up in Missouri :-)12:37
wolfspraulof course if we make a release we need to promote lgm itself as well, both inside the release and also at the bottom with a paragraph, what LGM is etc.12:38
wolfspraulhow many people go there? a few hundred? more/less? no idea :-)12:38
wpwraki think a picture showing him in the room with the mm1 in view would certainly add value to a news item. give it credibility.12:38
wolfspraulyes, but how to pull it off? :-) good quality pic, fast.12:39
wpwrakdoes he have people there whom he knows a bit ? if yes, ask them to make a few pictures. explain the objectives.12:40
wpwrakif he doesn't know anyone, ask the organizers if they can help12:40
wolfspraulI'm sure he knows everybody, but I'm realistic. once in the chaos of such a conference, it's hard to pull off things in real-time.12:41
wpwrakprepare some non-real-time pics as plan b, in case nothing good comes out12:41
wolfspraulyes :-)12:41
wpwrakhe should contact them beforhand. e.g., to make sure they bring a camera ;-)12:41
wpwrakotherwise you'd get some dutiful soul who will snap some phone cam shots, but that would be about it ;-)12:42
wpwrakof course, if you're going for "the blair witch project" aesthetics, ... :)12:43
wolfspraulI think they are quite good about documenting, tweeting, flickering, etc.12:47
wpwrakyou mean the conference organizers ?12:48
wolfspraulyes that whole event, everybody12:48
wpwrakdunno if they pay attention to producing "press" pictures.12:50
wolfsprauldoesn't matter if we have a good quality pic, and even officially cc-by/public domain etc. we can use it12:56
lekernelhi xiangfu13:02
lekernelI got libcurl to work on the M1 so you can use it later for the web update feature13:02
wolfspraullekernel: is it possible to display a static text during rendering?13:07
wolfspraulfor promotion purposes, like 'powered by Milkymist'13:07
lekernelatm there are OSC messages to display text, but it fades out after a few seconds13:07
lekernelnot hard to disable the fade out though13:07
lekernelit'll come as well when I add the RSS stuff13:08
lekernelworst case we quickhack the code if we can't get that on time13:08
lekernelhttp://www.highbeam.com/doc/1G1-169406902.html maybe that's another reason why they won't release the code13:12
lekernelthe VP8 codec might have common parts with stuff they licensed to anti open source companies like ARM13:13
guyzmolekernel - I'm not sure the focus on lua is really a good idea, but we're gonna discuss that with john to see what can we write about the MM1, and sure make an article in make, open Si, or electkor13:16
guyzmoanyway, it'd be nice to test the MM1 with lots of projectors to show what it can do13:17
wpwraklekernel: arm probably love open source - as long as it sells chips with their IP:)13:18
wolfspraulwpwrak: yes, that's exactly 100% why TI loves open source as well.13:23
wolfspraulopen source helps keep their partners margins in check, and them with the fattest margins and thus most innovation potential13:24
xiangfulekernel: thanks. I try that a little. not make it work. I will try libcurl http://www.milkymist.org/wiki/index.php?title=Flickernoise_build_instructions#libcurl13:25
lekernelxiangfu: can you get the keymap done first? should be easy and then you will not hear about it again13:30
xiangfulekernel: ok13:33
lekernelwolfspraul: also, something I noticed with the camera is that when the picture is all black, it automatically adjust brightness/contrast and this results in a lot of CCD noise at the output13:39
lekernelI don't know if we can deal with that somehow, will check13:40
wolfspraulthat would be hard to change most likely, it's probably a feature of the Sony CCD or immediately surrounding circuitry, I would think13:41
wolfspraulI guess it still tries to show something, so brightness goes up automatically?13:41
lekernelbut maybe this can be dealt with by adjusting contrast/brightness in M113:44
lekernelshould try13:44
kristianpauli agree with guyzmo comments about lua14:47
lekernelkristianpaul: send me a message on twitter with #milkymist hashtag14:49
lekerneli'm testing the twit wall atm :p14:50
kristianpaulsent from identi.ca it should forward on twitter as well14:54
lekernelyay works14:55
kristianpaulnice so now we can control patches with twits and dents :D14:55
lekernelkristianpaul: http://www.milkymist.org/03-05-11_1649.jpg14:56
kristianpaulnice :-)14:57
kristianpaulfree advertisement !14:58
kristianpaulthat will be the next product ;)14:58
kristianpaulwolfspraul:  is this the new battery pack for the coming mm1 ? http://en.qi-hardware.com/wiki/File%3ADc168_pic3.jpg15:05
wolfspraulno just a crazy Chinese hack I found15:06
lekernelso they just charge the 12V li-ion batteries by applying a continuous 12.6V to them?15:12
wolfsprauldid you look at the picture closely?15:21
wolfspraulit's three 3.7V lithium batteries15:21
wpwrakwhat does the contraption connect to ?15:22
wolfspraulthe idea is to connect a 12V dc supply15:22
wolfspraulit will all 'sort of' work, I guess15:22
wolfspraulthe outside is labeled as 1800mAh15:23
wolfspraulso yeah, it _should_ be 12.615:23
wolfspraulfor charging15:23
wolfspraulbut it is 12.0, if the supply is accurate15:23
wpwraki think we could do better: bundle 40 batteries, add a diode, and charge from 110 V mains. US only ;-)15:23
wolfsprauland then to draw power, I guess the camera has to be robust, i.e. also work with 11.5 or so :-)15:23
wolfspraulnot even the slightest attempt at managing charging cycles either15:24
wolfspraulat least in China anybody can become electrical engineer15:25
wolfsprauljust plug some wires together and turn on and see whether it works :-)15:25
wpwrakbattery life is overrated anyway :)15:25
CIA-48flickernoise: Sebastien Bourdeauducq master * r657cda0 / (src/renderer.c src/rsswall.c src/rsswall.h): RSS wall: print messages on patch - http://bit.ly/jVCAJZ15:33
lekernelxiangfu: btw, the interrogation mark key doesn't work on the german keyboard16:06
lekerneli'll file a bug against that... but I think it will go with the rewrite of the keyboard code with the different keymaps16:07
xiangfulekernel: ok. thanks16:52
rejonwpwrak http://libregraphicsmeeting.org/ i am one of organizer17:16
rejonyes, all sounds good lekernel17:16
rejoni'm tempted to go to sanjose tonite to meet up with ifixit and jeri ellsworth17:16
lekernelrejon: kewl17:17
lekernelbtw i'm putting your identi.ca by default on the RSS wall feature17:18
lekernelor, ATOM for that purpose17:18
lekernelit's working btw17:18
lekerneljust need to polish a few edges17:18
mumptaiterpstra, you are knowledgeable about the different versions of the zpu?17:52
CIA-48flickernoise: Sebastien Bourdeauducq master * re1eb6cd / src/rsswall.c : RSS wall: configuration dialog box - http://bit.ly/m83Ltz18:14
carlobarhi lekernel, i have been doing some modifications to the HPDMC controller, there were some errors with a buffer, and thats why the controller sometimes worked and sometimes didnt... now it doesnt work at all :-D.... i dont know if the problem is on read/write or both.. when you developed it, how do you did tests? is there any way to watch signals on real time? thanks18:21
lekerneluse the video out18:22
lekernelit's much easier to guess DRAM problems when you can visualize what they're doing on a framebuffer18:23
carlobarhow must i do that? do you used that in tests?18:23
lekernelbtw, what do you need HPDMC for?18:24
carlobarwrite and read data sent by the processor18:25
lekernelI mean, what application?18:26
lekerneldo you have the source code online somewhere?18:26
carlobarhaha, run the kernel and some testbenches18:26
lekernelok, but what is your project about?18:27
carlobari have to run some applications and measure power consumption and execution time18:27
lekernelwhat for?18:27
lekernelcompany work?18:27
lekernelschool project?18:27
lekernelpersonal research?18:28
kristianpaulinteresting this topic of transputer18:28
kristianpaulmay be he just want to verify your thesis results :-)18:28
carlobaruniversity project... the project is about a comparation between two processors running uclinux18:29
kristianpaulabout lm32 vs micoblaze18:29
kristianpaulyay, i knew it !18:29
carlobarlm32 vs leon318:29
lekernelcarlobar: ok. first, please publish your code. HPDMC is GPL.18:29
kristianpaulwell, almost ;)18:29
kristianpaulleon3, nice :-)18:29
lekernelcarlobar: and same for LEON318:29
carlobarok, ill do it... github is fine right?18:30
lekernelsites like github makes it very easy for you18:30
kristianpaulwell make code publically avaliable is an option, but came on, sharing is good :-)18:30
carlobari was planning do that when it were working18:31
kristianpaulbad idea, i mean, is better publish as you go18:31
kristianpaulso ask for support get interested18:32
kristianpaulcarlobar: will be nice you publish the bistream loader for you board18:32
lekernelcarlobar: btw, maybe your uni could buy a M1 for you, so you don't have to go through SDRAM problems which aren't the topic of your work I guess18:32
carlobarmaybe.. but sure it will take a lot of time18:39
kristianpaulcarlobar: may be if you publish the tests/procedures needed for your benchmark other people with a mm1 can try it out :-)18:41
kristianpaulsorry i'm not in bogota but you can came to buga if you need the mm1 for a hacking day ;)18:43
kristianpaulor you just pass me the binary to load, etc..18:44
kristianpaulanyway, if you like work this way, just in case :-)18:44
kristianpaulremenber you can use qemu for initial test, and i later can reply that on my board18:44
carlobarohh thanks :), i will take that into account18:49
lekernelof course libcurl timeouts are broken on rtems. grmbl. having them work first time would have been too good to be true.18:58
CIA-48flickernoise: Sebastien Bourdeauducq master * r8d3ffcc / src/rsswall.c : RSS wall: use configuration values - http://bit.ly/iHpKiO19:04
CIA-48flickernoise: Sebastien Bourdeauducq master * r1c8d4a5 / src/rsswall.c : RSS wall: add timeouts - http://bit.ly/jI97Rn19:04
kristianpaulwhat? the minimac2 rxb0 have two clks and one of then is the phy_rx_clk..19:12
kristianpauli tought it wasnt posible, or not recomended to do19:12
kristianpaulwell, not bad, the nibble is coming from the PHY and i this will not hurt once the data get in the ram19:20
kristianpaul *i* *guess*19:21
lekernelkristianpaul: what?19:30
Action: kristianpaul hides19:32
kristianpauli just took a quick look to the minimac2 core, i surelly missed some details ;)19:32
lekernelwell I don't know. maybe you have spotted some bugs19:33
lekernelbut yeah the minimac2 has 3 clock domains19:33
lekernelthose should be handled properly though19:33
kristianpauli'm checking sync part right now :-)19:35
carlobarin that case do i have to create a new repo, or do fork??20:05
kristianpaulis better to track you using github that way20:06
kristianpaultrack your changes i mean, not you actually ;)20:08
lekernelkristianpaul: can you post a long tweet to test the line break? :)20:34
lekernelah, it's working already20:35
carlobarhere are my repos: https://github.com/carlobar ... i dont know if thats ok.. i had a mess here, so if something is wrong please tell me20:43
CIA-48flickernoise: Sebastien Bourdeauducq master * r6c20769 / (src/font.c src/font.h src/osd.c): OSD: multiline text - http://bit.ly/lXXfKh20:50
carlobari have to go... i'll come back in some hours...20:59
lekernelcarlobar: cool, thanks for uploading the source21:03
CIA-48flickernoise: Sebastien Bourdeauducq master * r62900f3 / src/cp.c : fix performance start button - http://bit.ly/kgYRO121:28
lekernelfor some reason the time it takes to compress a PNG screenshot has dropped to a few seconds. I like when problems fix themselves like that with magic.21:33
lekerneldecoding large wallpapers is still slow, though21:33
azonenberg... wow division is slow lol22:20
azonenbergI just wrote a suite of unit tests for my ALU22:21
azonenbergin the time it takes to do one division it can test every other instruction on the CPU twice22:21
wpwrakazonenberg: in the good old days of CISC, this wouldn't have happened. e.g., the VAX POLYH instruction. calculates an arbitrary-length polynomial with 128-bit floating-point math.22:25
CIA-48flickernoise: Sebastien Bourdeauducq master * ra7caf21 / src/rsswall.c : RSS wall: do not display empty idle message - http://bit.ly/m1pXiD22:26
lekernel<andrey_> Using different clock edges helps to make less spikes on the ground/power lines, using clk/2 - helps to reduce power and increases the amount of logic I can put between the registers without additional enable signals and/or double cycle timing constraints23:53
lekernelok well23:53
lekernelI guess one more crap FPGA design23:53
--- Wed May 4 201100:00

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