#milkymist IRC log for Monday, 2011-05-02

kristianpaulxiangfu: hi !00:56
kristianpaulhi aw :-)00:56
kristianpaulxiangfu: for the cards you sent me, if dont remenber well if you have issues trying to create files/folders on it?..00:57
awkristianpaul, hi evening!00:57
kristianpaulso far i jus tried by ftp and was no sucefullyy even whe the ftp logs dint mentioned anythin wrong00:58
awxiangfu, have u received those cables?00:59
lekernelkristianpaul: hmm, C++ :(08:51
lekernelthen we get the GCC problems or need to use 4.408:52
wolfspraullekernel: C++ for lm32 worked in 4.4 but stopped working in 4.5? and things got even worse in 4.6?08:54
wolfspraulshould xiangfu help you with any of this? are there any particular bugs, or patches you need to upstream, or some gcc/lm32 maintainer role xiangfu could play?08:55
lekernelwolfspraul: yes10:02
lekernelwolfspraul: I already have commit access to gcc, the problem is time and motivation to fix that ... and the gcc code isn't exactly nice to play with10:02
lekernelI already fixed a few things in the LM32 port, but now the 4.6 compiler doesn't work at all, and C++ is broken in 4.510:03
lekernelC++ isn't very important though10:03
wolfspraullekernel: sounds like xiangfu can help10:15
wolfspraulespecially if there are regressions, I think we should catch them asap/immediately10:15
wolfspraulthat's by far easier than letting things slip that already worked in the past, imo10:16
lekernelwell... right now we have a solution: stick to 4.5 or, in worst cases, use 4.410:16
lekernelso let's not spend our limited resources on this10:16
wolfspraullekernel: if you have patches in your gcc queue, or some gcc bugs (url to tracker?), please post so xiangfu can start to warm up on those10:16
wolfspraulyes sure understood. but like I said, I think regressions should be fixed fast, that's the easiest.10:17
lekernelalready there: http://www.milkymist.org/wiki/index.php?title=GCC_bug_list10:17
wolfsprauland gcc will keep moving forward10:17
wolfspraulok great10:17
lekernelthe 4.6 breakage has already been there for a while10:17
lekernelseptember or so10:17
kristianpaulC++ ahh, dint new it :(11:21
aw_lekernel, Do you prefer that enable SPDIF by pulling high pin44? or s/w by bit SPDF in register 5Ch should be set?12:28
lekerneldon't connect anything spdif related12:30
lekernelput the pin to the expansion header, but don't connect it12:30
aw_the SPDIF meeds digital GND not AUDIO_AGND, so I'll change two pins on J3 be as: SPDIF OUT and GND.12:30
aw_yes, I just only connect wires for them only, no further extra-circuits12:31
aw_agreed? or NO GND on J3?12:32
lekernelyou just connect the spdif stuff to J312:32
aw_J21 already have two GND pins.12:32
lekerneland we still need a GND on J312:33
aw_yes, connected.12:33
lekernelI think there are enough pins, no?12:33
aw_hum...okay..just added.12:33
lekernelyeah there are 5 grounds on J312:33
aw_#pin1 of J3 = SPDIF OUT, #pin3 of J3 = digital GND.12:34
lekernelso remove 2, and you connect the SPDIF stuff to there12:34
lekernelor 3 if you want to add digital GND as well12:34
aw_yeah. okay..done on Audio file. :-)12:35
aw_btw, another things.12:36
aw_to have a easy way that when soldering some digital grounds for user experiments/hackable purposes, I'd like to add GND test point which just open a solder layer and user can easy soldering on that bigger GND test point/soldering point. Are you okay with this?12:38
lekernelmaybe even of just a solder layer12:39
lekernelyou can use a hole with the possibility to solder a pin12:39
aw_right, just open a solder layer though.12:40
lekernelso you can attach the ground of a scope probe to it with no risk of breaking the solder or ripping the copper12:40
aw_hup? you want a hole? a thorough hole? I don't want to use a through hole. just want user don't touch parts's GND. :-)12:41
aw_sorry that I can't follow your 'hole" meanings.12:42
lekernelif you ask me, yes, i'd prefer a through hole12:42
lekernelthis way you can 1. solder a pin through it 2. attach a scope's crocodile clip to the pin, without major risk of breaking stuff12:43
lekernelbut it's very unimportant, so if there is the slightest problem with the through hole, don't do it12:44
aw_hmm....i see know. so prefer using a GND through header pin instead of using test pointes, right?12:44
aw_no no. i frequently solder and did some experiments, so we just missed this minor things but good.12:45
aw_okay, got it.12:45
lekernelbut again: this is NOT important. so if there's any problem or delay arising from it, use just a test point, or do nothing12:45
aw_how about I just add bigger test points nearby four J12~J15? then I don't need to source just other hearder pins. :-)12:49
aw_i.e. at four corners. of board :-)12:50
lekerneldon't source header pins12:51
lekernelpeople who can use a scope can solder them themselves12:51
lekerneljust add a hole for a header pin and DNP it12:51
aw_okay...just one through hole only?12:52
aw_:-) okay.12:53
lekernelyou can also DNP J3, btw12:54
aw_no no...J3 header is good even few people use it. :-)12:55
aw_who knows that one day someone will do great works on mixing other audio sources for video effect though. :-)12:57
terpstra_the milkymist only costs 70 EUR???13:07
terpstra_that's shockingly cheap. how many have you managed to sell?13:07
lekernelnah, where did you find that?13:08
lekernelthe board is 380E13:08
lekernelwe might reach 70E at some point but we'd need _huge_ volume orders :p13:09
terpstra_the price just changed!\13:09
terpstra_it was 7013:09
terpstra_now it says 38013:09
terpstra_bait and switch!13:09
terpstra_oohhhh - it's the kit part that is 7013:09
lekernelmaybe 70 was just the case kit13:09
terpstra_how many have you sold for 380EUR? seems a bit more than most people (including me) would be willing to pay for a (private/personal) hacking/devel platform.13:10
terpstraat 70 i would've ordered one this instant ;)13:11
lekernelnot many, about 3513:11
lekernelwell that's a serious problem13:11
lekerneleither you make a hacking/devel platform, and you have low volumes, so either it's expensive or you make something ridiculously simple like arduino13:12
terpstrai am not a hardware hacker, though i've been forced to work with vhdl/verilog lately. so for me, things like the linksys routers and kindle are the stuff i buy to tweak.13:13
terpstraand those are all < 100EUR for much thicker hardware13:13
lekernel*or* you go to other markets, which is what I plan doing with milkymist, which would potentially allow larger volumes and definitely higher prices than what most hackers are often ready to pay13:13
terpstra(but no fpga)13:13
lekernelyeah but those are built in much larger volumes than the milkymist today13:14
terpstrayes, i know13:14
lekernelit's a chicken and egg problem13:14
terpstrathe FPGA part of the milkymist is what makes it both more interesting and more expensive, though, i guess13:14
lekernelnot only, that part is only about $40 (not including taxes)13:15
lekernelwe have costs everywhere13:15
terpstrano disk controller except for via usb?13:17
terpstraanyway, nifty device13:17
lekernelthere's the internal memory card13:18
lekernelyou can easily have 8GB (or so) cards those days13:18
lekernelbtw, even as a devel platform the M1 isn't that expensive. you can compare it to http://www.digilentinc.com/atlys/ for example13:20
terpstrahttp://www.altera.com/products/devkits/altera/kit-cyc3-starter.html   ... mroe RAM and bigger FPGA13:21
lekerneland less peripherals. and we're not Altera.13:21
terpstra(no A/V stuff, though, but i don't care about that part of the milkymist myself)13:21
terpstrathat link you posted is dead btw13:22
lekernelno, the server is just slow13:23
lekerneltry archive.org maybe13:23
terpstrawhat i bet you could sell quite well:13:24
terpstraa slightly more powerful milkymist that could run debian13:24
terpstrathink of all the FSF nuts who could finally have their free OS running free HW13:24
terpstrai bet you could charge 380EUR no problem then13:24
lekernelthe current one could run debian... it's mostly a software problem. but slowly.13:24
terpstrayou need an MMU13:24
terpstrano debian without a proper MMU13:24
lekernelcounts as 'software'13:25
terpstraalso no disk controller13:25
terpstraand too little RAM13:25
terpstraif you had 512MB (or better 1GB) and a SATA plug then it would be able to really run a debian desktop13:25
lekerneland the MMU isn't the biggest problem, gcc, libc, linux and all those non portable GNU wares are the major pain13:25
lekernelwith a 80MHz CPU? not really I think13:25
terpstraright, forget the fpga is so weak on that board13:26
terpstramy LM32 runs at 175MHz13:26
lekerneleven 175M is slow13:26
terpstrabut not so bad13:26
terpstraarm processors at 500MHz run debian ok13:26
lekerneland if we take those FPGAs that can run at 175, it's going to cost a lot13:26
terpstrai'm not trying to push you to make the milkymist run debian. i was just envisioning a market of people who would be willing to pay a premium for an openhardware product.13:27
lekernelalso, the FSF in fact doesn't really care about open hardware13:27
terpstrayou need zealots like the people who buy apple.13:27
terpstradebian/linux people are the same only even more extreme.13:27
lekernelthe problem is they have absolutely no idea what logic synthesis is about13:28
terpstrawhy is that a problem?13:28
lekernelfor them the CPU is hardware and not their problem13:28
terpstraoh -- you mean they wouldn't buy into the free hardware angle?13:29
terpstrai don't agree. debian even ditch custom firmware from the kernel.13:29
lekernelas long as it's free from 'blobs' they're happy13:29
lekernelfirmware is 'blob'13:29
terpstrai think happy is relative13:29
lekernelnow they could even moan, rightfully I admit, about the bitstream being a 'blob'13:29
terpstrathey have the most free that they can get.13:29
terpstraif a milkymist-like product could run debian reasonably well, they'd have a new freedom option that they didn't previously13:30
lekernelwell, we can figure that out before13:30
lekernelwe can definitely ping the FSF about this13:31
terpstraand you're quite correct that they will moan about the synthesis tool not being free13:31
lekerneleven given the current state of things13:31
terpstrabut isn't that what you moan about too?13:31
terpstramaybe they would help you out! ;)13:31
lekernelsure. they would be right doing that. but that's not going to improve sales.13:31
terpstraany publicity will improve your sales.13:31
terpstrano one i have spoken to has heard of the milkymist13:31
terpstrai only found it myself when looking into opencores and researching the lm3213:32
terpstra(opencores here is open cpus, not the website)13:32
lekernelmh... don't know. maybe you're right and the FSF is worth trying...13:33
lekernelbut otoh I've contacted so many people and been turned down so many times that I have doubts about that13:33
terpstrai don't understand why you would ask the FSF?13:34
terpstrathey don't run debian13:34
terpstrapeople would only really care once/if you had debian able to boot.13:34
terpstranot many people are interested in a half-finished product13:34
lekernelthat sounds like going through a lot of pain for something very uncertan13:34
terpstraisn't that what you're already doing with the milkymist? ;)13:35
terpstraanyway, i'm not trying to pressure you to do this!13:35
terpstrai was just thinking out loud about what market would be willing to pay a premium for open hardware13:36
terpstra(b/c let's face it, there will definitely be quite a premium)13:36
terpstrathe audio/video nuts certainly like to pay too much also (gold plated usb plugs?!) so that is also a good market13:36
lekernelmy plan is not to put forward openness in most of the marketing stuff13:38
lekernelI think only a small minority of people are interested in that anyway13:38
terpstrayou are planning to make a business around the milkymist, though?13:39
lekernelsure. I rather envision a regular hardware product developed using open source practices13:40
lekernelthat being said, we could DEFINITELY benefit from more publicity in hacker/embedded circles13:40
terpstrabtw, ohwr.org is exactly aligned with your stated goals.13:41
lekernelisn't ohwr.org centered on particle physics?13:41
terpstra'regular hardware product developed with open source practices'13:41
terpstranah, that quote might as well have come from javier - they guy who founded it13:42
lekernelwhy do you think there should be a premium for open source hardware, btw?13:47
terpstraif you use and fpga and sell low volumes ...13:47
terpstraanyway, i'm off13:47
terpstra(i don't think it's open hardware per se, but a small enthusist project like you have now that costs)13:48
lekernelyeah sure13:59
lekernelbut hopefully this is a temporary state13:59
lekernelit's going to be hard to get past it, but I hope it will happen14:01
lekernelas a matter of fact, the milkymist product is soon reaching a state of usability and quality where we can begin trying hard to contact retail chains, businessmen, etc.14:03
lestathi there15:10
kristianpaulhi lestat17:56
lestathello kristianpaul17:57
kristianpaulhe, rms is requesting for a Free Software GSM phone at osmocom baseband mailist18:26
lekerneloh, miracle18:27
lekernelGoogle answered18:27
lekernelThanks for contacting us through the WebM project and sorry for the long wait.18:27
lekernelI 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.18:27
kristianpaulha.. i was waiting for such us answer18:28
lekernelyeah, but I wonder why they do that18:29
kristianpauljust a dirty marketing strategy i guess18:30
kristianpaulor just realize that IP have so much value that cant be open/libre18:30
lekernelthey claim it is free of charge18:31
kristianpaulor they jus waiting emails from companies that will surelly pay for  the licesing of this IP and not mails from open source projects18:31
lekerneland that you get the source (but with NDA I'd guess)18:31
kristianpaulhmm well yes18:31
lekernelso why restrict it?18:32
kristianpaulbecause the fear to leaks?18:32
lekernelwhat's 'leak' ?18:32
kristianpaulinformation leak?18:32
lekernelwhat I mean is: why not allow people to redistribute that source, if they're not making any money from it18:33
kristianpaul(money) no yet ;)18:33
kristianpaulthe lied, thats all18:34
lekerneldon't know18:35
lekerneli'll ask them this question18:35
lekernelthen maybe in another 6 months I'll get the answer :-)18:35
larschm... provide an interface to the webm rtl and tell everybody to request the source themself18:35
kristianpaulthey will be happy if you answer with a technical solution to use the free/gratis bitstream ;)18:40
lekernelnah, i'll clearly mention this isn't an option18:41
lekernelif they don't answer or tell me to go to hell, starting the development of an independent WebM encoder because Google isn't open sourcing it definitely is PR-worthy and troll-prone material, so either case we'll gain something :p18:44
kristianpauli agree18:46
lekernel"it would be the first mobile phone that he himself would be able to use, given that there is no proprietary software on the baseband anymore"18:59
lekernelyeah well18:59
lekernelwe do more than software18:59
lekerneland anyway open source hardware exposes flaws in the FSF talking, so ....19:00
kristianpaulyeah, i agree with Sylvain Munaut about DSP part, of course FSF will no care about that..19:01
lekernelplus those projects are a gift that the mobile phone manufacturers do not deserve19:02
kristianpauland for a first phone why no? mm1 is not a DSP focus yet..19:02
kristianpaulbut a phone is an interesting turn-key problem to be solve using free software ;)19:03
kristianpaulalso VJs of course ;)19:03
lekernelmobile phones are a very competitive business...19:07
lekernelso after you get that phone to work technically (openmoko didn't even get there), how do you sell it?19:08
kristianpaulopenmoko was to ambitious project19:11
kristianpaulof course the needed to make atractive for potencial buyers i gues..19:11
kristianpaulis like if you're trying to make milkymist a full featured computer, something of course is not and not ready for19:12
lekernelif we make a full featured computer, what is the selling point? freedom?19:13
kristianpaulthats a very small target surelly :-)19:14
lekernelhow many people care that the masks of their CPU are open? guess not more than one in five million?19:15
lekernelnow that minority is interesting, and that's why we should have OSH CPUs19:16
lekernelbut OSH is NOT a selling point19:16
kristianpaulsure not, but for at least those buyers will not complaing too much about missing features ;)19:18
kristianpaula year ago i wrote rms about mm1 but *i* think he dont like the idea of uclinux.. also if it cant run emacs well :p19:18
lekernelno, they will just buy intel or ARM :-)19:18
lekerneland? no answer from rms?19:19
kristianpaulnot after i told him about it just can run uclinux19:20
lekernelbut did he answer in the first place?19:20
lekernelinteresting. what did he say?19:20
kristianpaullet me find the mail, was time ago..19:21
kristianpaul"What job does this hardware do?"19:22
kristianpaul"ISTR that uclinux is meant for small machines and does not give all                                                          "19:23
kristianpaulIs that true?""19:23
kristianpaul"Are you saying this computer lacks a mmu?"19:23
kristianpauland i said yes,i was the last mail..19:23
kristianpaulnow i realize the mmu is very important for software developer :-)19:24
kristianpaul(huh?) ?19:24
lekerneldidn't understand the second line19:24
lekernel"ISTR that uclinux is meant for small machines and does not give all"19:24
kristianpaulah, sorry i forgot to append and the end a comma followed by Is "that true?"19:25
lekernelwell, we can add a mmu, then people will start complaining about clock frequency, which we can't really solve without making an asic19:26
lekernelwhich in turn brings it lots of problems related to price and volume, and as I said before, freedom is no selling point19:27
lekernelso you should answer that to people like rms next time :-)19:28
lekernelbtw, there's opensparc, which is free and made into an ASIC19:29
lekerneldoes it sell a lot? is it popular? no.19:29
lekernelthe opensparc community is even smaller than ours19:29
lekernelso that's also a good example you can point to him19:29
kristianpaul(smaller) oh really?19:29
lekernelyeah probably19:30
lekernellast time I checked it was mostly students who were there to get their degree and then leave19:30
lekernelbut opensparc has that famous MMU and is even 64 bits :p19:31
lekernellol http://www.facebook.com/SynopsysUniversityProgram19:49
lekernelyay, got libcurl to work :-)21:09
lekernelmy m1 just displayed the source of milkymist.org21:10
lekernellibcurl is a bit bloated (nearly 200k) but well...21:10
lekernelit gets the job done and I guess potential other developers would appreciate to have this well known lib there :-)21:11
kristianpaulnow firefox..21:12
lekernelxml parser now21:13
lekernelfor RSS feeds21:13
kristianpaulxml :S21:13
lekernelyeah well21:13
CIA-48flickernoise: Sebastien Bourdeauducq master * r05dfccf / (src/Makefile src/cp.c src/main.c src/rsswall.c src/rsswall.h): libcurl demo - http://bit.ly/m1jGqG21:14
lekernelm1 needs to be web 2.0 compliant. it's teh hype.21:14
lekernelwhat about expat?21:18
lekernelunrelated: http://ebookee.org/Scientific-American-s-quot-The-Amateur-Scientist-quot-The-Complete-20th-Century-Collection-Version-2-0-1928-2001-_595588.html <= good stuff in there21:20
kristianpaulnice, busybox have inteesting options for nommu system plus static binary support21:29
lekerneljuliusb_: btw, have you heard of any open source FPGA video encoder that fucking works?21:31
lekernelpreferably not H.264 as it's a good target for patent trolls21:31
lekernelyeah... expat looks ok. one more lib for kristianpaul to compile :-)21:50
kristianpaulhmm rtems dont have byteswap.h..21:55
kristianpaullekernel: (libs) you already know my pain :-)21:55
roh lekernel wasnt there a hdl theora encoder?22:05
lekernelin elphel... but it looks suspicious22:05
lekernelyay, got expat to work and the m1 parses the twitter feeds now22:06
roh;) nevermind.. on my way to bed. just nagging *g*22:07
kristianpaulwhat is the git url for getting lm32 ulibc?22:09
lekernelhttp://git.serverraum.org/?p=mw/uclibc-lm32.git;a=summary i'd guess22:10
CIA-48flickernoise: Sebastien Bourdeauducq master * rbea814c / (src/Makefile src/main.c src/rsswall.c): Twitter demo - http://bit.ly/lx4GWD22:22
juliusb_lekernel: no proper, working open source video encoders as far as I'm aware22:30
--- Tue May 3 201100:00

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