#milkymist IRC log for Sunday, 2012-03-04

kristianpaula computer can actually control USB power bus, right?00:16
Action: kristianpaul had similar concern for remote powercycling a UMTS dongle00:17
wpwrakat least M1r4 will be able to. not sure if all computers are that good, though. probably not.00:37
kristianpaul(not) hmm :|00:44
wolfsprauldisappointment? (reading backlog...)00:45
wolfspraulah, probably referring to USB power00:46
wolfspraulI found a neat little (tiny) OpenWrt router and I am wondering whether we can use it to give M1 Wi-Fi. It's the TP-Link TL-WR703N00:46
wolfspraulI will buy a few of these thingies and try out00:46
wolfspraul15 USD :-)00:47
kristianpaulwhere i can order? :-)00:47
wolfspraulin a lot of places already, sales may well be in the millions, it started selling about 6 months ago00:47
wolfspraulit's really tiny, also just 0.5W power supposedly00:47
kristianpaulah new product lets search a bit00:48
wolfspraulcan take power from USB, and connect to M1 over Ethernet00:48
wolfspraulactually it could also be our fast-path to more Linux USB drivers because it has a USB 2.0 Host as well00:48
wolfspraulbut one by one, first I buy one to give my M1 Wi-Fi00:49
wolfspraulI'm sure there will be problems, I shall find out soon00:49
wolfspraulthis one http://wiki.openwrt.org/toh/tp-link/tl-wr703n00:49
kristianpaulgreat combo !00:49
wpwrakFallenou: hurry ! we need to get MMU, linux, then USB-WLAN. quicker than wolfgang can sort out the quirks of the chimera he's building ! :)00:49
wolfspraulthat will work in 2 days Werner, you have no chance :-)00:50
wolfsprauland it's besides the point because it's not really a chimera but a very practial small wi-fi dongle, just connected over Ethernet which works on M1 *today*00:50
wolfsprauland... not to forget. I did an exhaustive comparison of usb-wifi dongles recently, and what I found is not pretty at all00:51
kristianpaul;-D http://en.wikipedia.org/wiki/File:Chimera_Apulia_Louvre_K362.jpg00:51
wolfsprauldrivers in kernel.org are abysmal00:51
wolfspraulthe only difference is for the wifi that comes with integrated chipsets like intel00:51
wolfsprauland also the atheros 5 and 900:51
wolfspraulbesides that you will have frequent disconnects, crashes, hangs, you-name-it00:52
wpwrakyeha, wlan and linux have a tense relationship00:52
wolfspraulra-link (acquired by mtk), atheros 6 (plus acquired by qualcomm), realtek00:52
wolfspraulour Wi-Fi strategy is 100% correct - keep this beast at arm's length00:52
wolfspraulhowever, we need to make it work for our customers00:52
wolfsprauland that WR-703N is an option I just found yesterday00:53
kristianpaulcheaper i found was 24usd :-/00:53
wolfspraullet me try out first, I am sure there are problems00:53
kristianpauloh, sure00:53
kristianpaulbut the price and china free shipping is very temtative ;-)00:53
wolfspraulI find it in the corner shop here next to batteries and cheap mice & keyboard, so you have a guess as far as the volume it must be at already00:54
wolfspraulthere is even a mainland clone, another 1 USD cheaper, but they removed the USB 2.0 host connector00:54
kristianpaularghh, sould out on DX...00:55
wolfspraulkristianpaul: hey, I didn't want to cause a buying frenzy over this, just sharing what was on my mind right now :-)00:55
wolfspraulI will buy a few and experiment00:55
kristianpaulno no, no problem at all00:55
wolfspraul*maybe* this can quickly build a wi-fi bridge for my m1, that's all00:55
kristianpaulas i said, price and features call atention00:55
wolfspraulit would make demoing and using m1 with more connectivity easier, like rss feeds, downloaded images, etc.00:56
kristianpaulM1 ships free 3G,  :-)00:56
kristianpaulbbl dinner00:56
wolfspraulbut at that pricepoint, if it really works I am not even opposed to putting one of these little thingies in the box, together with a perfectly customized "m1 wifi bridge" openwrt image on it00:56
wolfspraul"3G" is a joke, it just means you can connect a 3G USB dongle00:56
wolfspraulI have a clear idea for this, Wi-Fi bridge for m100:57
kristianpaulanyway, sorry noise, looks very interesting device for openwrt hacking anyway :-)00:57
wolfspraulonly that00:57
kristianpaulhe, sure i just have  others too ;-)00:57
wolfsprauland powered even from a USB port, with R3 that takes 50% of the ports, with with R4 we have 4...00:57
wolfspraulso one cable into ethernet, one into USB, and there we have Wi-Fi...00:57
GitHub145[milkymist] wpwrak pushed 6 new commits to usb-debug: https://github.com/milkymist/milkymist/compare/240d96d...56e57ba02:21
GitHub145[milkymist/usb-debug] softusb: disable RX timeout message in usb_in - Werner Almesberger02:21
GitHub145[milkymist/usb-debug] softusb: don't send SOF while in full-speed bus reset - Werner Almesberger02:21
GitHub145[milkymist/usb-debug] softusb: fix two typos, one of them breaking some error detection - Werner Almesberger02:21
wolfspraulI was wondering about the lm32 compiler - we are typically compiling as a cross compiler, no? is it possible to build the lm32 toolchain natively and run it in qemu?02:26
wpwraki suppose you could do that. but why ? for distribution building ?03:32
wolfspraulwpwrak: maybe we see different or - hope dies last - less bugs in such an environment09:48
wolfspraulthinking ahead a little, and really just a little Sunday daydreaming, I am wondering whether we can remove the MIDI connectors in a future revision, and instead include one of those usb-midi cables in the box. after we get them working well I don't see easily what could justify keeping the 5-pin midi connectors.09:56
wolfspraulthe usb-midi cables could restrict us electrically or otherwise in terms of the MIDI protocol? probably not...09:57
wolfspraulanother risk is that we cannot source such cables easily anymore at some point in the future? probably also not very likely, though theoretically in the very long run the connectors are probably easier to get than such a cable09:57
wolfspraulit could free space in the side panels and on the pcb09:59
wpwrak(less bugs in a native environment) hmm, yes and no. if you get package X written by someone who doesn't care about cross-compilation, then you're better of in a native environment, that's true.12:42
wpwrakbut most things you take from a distribution or that have been vetted for distribution should already be cross-compilation clean12:43
wolfspraulyes I just wanted to ask whether someone tried that12:43
wolfspraulprobably not, although I remember reading about qemu once in a while, but less recently that more hw is easily available12:43
wpwraksebastien ran into issues with what he affectionately calls autocrap, but i don't know if these would go away by avoiding cross-compilation or whether they are simply issues of an unrecognized target (or maybe just an incomplete build environment)12:44
wpwrak(midi) you'll win with having the DIN connectors over using a USB-MIDI dongle the day you're the last one using MIDI ;-)12:45
wolfsprauldon't understand12:45
wpwraknot sure such a doomsday scenario is really relevant for milkymist's wellbeing, though ;-)12:45
wolfspraul'win' in which way, and what is the advantage over a cable?12:45
wpwrakadvantage: better integration (and imho, that's the only one)12:46
wolfspraulassuming the cable works perfectly with usb bugs fixed, what's the downside?12:46
wolfspraulit's just a cable, and you have to have a cable anyway12:46
wpwrakavailability of usb-midi cables will exist as long as there's a significant number of users of traditional midi12:46
wolfspraulif we move to dvi-i, there will be more cables, and maybe we support vga sampling with the adv7181c one day - another cable12:46
wpwrakwell, it's a "fat" cable. but yes, it's not too nasty as far as cables go12:47
wolfspraulyes, I think we can assume they will exist for a long time12:47
wolfspraulI am not suggesting to make this change now, the reasoning just popped into my mind somehow12:47
wolfsprauland I am in no way suggesting to remove midi, quite the contrary. but the cable...12:47
wolfspraulremoving the connectors would free space for interesting new things12:47
wpwrakironically, when i was gearing up for playing with midi, i found the usb-midi critter at the very first shop. that same shop (one of the largest shops in town for MIDI things, mixers, etc.) didn't have DIN-to-DIN cables12:56
wolfspraulI know :-)12:56
wpwraki only found the cable some three shops later. and the one cable may have been more expensive than the entire usb-midi dongle ;-)12:57
wolfspraulso I wanted to just *ask* whether that may be an idea for R5 or not, but no urgency12:57
wolfspraulis it possible that the usb-midi cables don't support certain things?12:57
wolfspraulI will start to ask around with the pros, if I run into one.12:57
wolfspraulbut I think mostly everybody is on USB nowadays12:57
wpwrakhmm. maybe for M2. i'd love to go to a more front/back oriented form factor. not the current "octopus" arrangement12:58
wolfspraulwith DMX you find large installations, really long cables, etc.12:58
wolfspraulbut the old-style midi is definitely a legacy thing that might well be solved for good with a cable12:58
wolfspraulok, one by one12:58
wolfspraulthe din-midi vs. usb-midi question is one that stands independently12:59
wolfspraul1. are there things/features that only din-midi can do? shortcomings of the cable?12:59
wpwrakusb-midi may have some minor latency issues. but that's probably not even something we care about. i mean, we have 24 Hz granularity anyway.12:59
wolfspraul2. can we trust the cable to be sourcable for years?12:59
wolfspraul3. are we technically at the point that the usb-midi cables on m1 actually work bug-free?12:59
wolfspraulthose need to be answered calmly and professionally, in and of itself12:59
wolfspraulfor the good of our users, since as you said it starts to become really hard to even find din-midi cables...13:00
wpwraki wouldn't worry about sourcing. if it really becomes an issue, then that would mean that the old midi devices are dying out. and anyone who still has one will probably have solved the connectivity problem by then13:00
wolfsprauland if we tell our own users they should best look for a usb-midi cable, well then... :-)13:00
wpwrak3) is more difficult to answer. needs more testing.13:00
wolfspraulsure, no urgency on this13:00
wolfspraulit would free space though13:01
wolfspraulfor the next bigger revision, call it m2 or whatever, well, let's see13:01
wolfspraulsounds like by then the -7 series xilinx fpgas could be an option, ddr3, rearrangement of pins, sdi-video or other new features? touch, hdmi-in, whatever... :-)13:02
wpwrakit would also remove some "unusual" components :)13:02
wpwrakwhat's SDI-video ?13:20
wpwrakand yes, we may need more pins soonish. we've been quite generous in M1r4 ;-)13:21
wpwrakthe curse of having an abundance (of free pins, in M1rc3). you think you'll never run out of them, so you start spending ...13:21
wolfspraulsdi is a digital video standard used in pro-equipment http://en.wikipedia.org/wiki/Serial_digital_interface13:23
wolfspraulit's just one of a list of things m1/m2 could add at some point. adding it would require first learning more about lots of things there. how hard it is to support, how common, growing or shrinking, actual sales potential, and so on13:24
wpwrakoh wow, another standard. the video sector is quite productive :)13:24
wolfspraul3000 gbit/sec :-)13:25
wpwrakhuh ? i think you mean 3.000 Gbps :)13:26
wolfspraulthe video guys are really pushing the limits13:26
wolfspraul3 gbit/sec?13:26
wpwraknobody does 3 Tbps. i think not even the L1 cache in your server cpu ;-)13:27
wolfsprauldon't know, just glancing over the wiki page. anyway this requires more research first, it's just one of many things floating around together with screens, touch and others13:27
wpwrak(cache speed) yeah, L1 should be about one order of magnitude below 1 Tbps. there are fibers that carry more, though. at least in labs ;-)13:28
wpwrak3 Gbps is already quite good :)13:29
wolfspraulyes yes. misread the . and ,13:29
wolfspraulit's 3gbps of course13:29
wolfspraulnext time I look for a bunch of m1-related cables, I will try to find a vendor for a usb-midi cable as well13:31
wolfspraulby now, let me think what is on my list13:31
wolfspraulvga sub-d to rca13:31
wolfspraulthat could help if m1 supported tv-out, and also on the other side if (with adv7181c) we would support sampling vga input13:32
wolfspraulthen of course dvi-i to vga, that's a must with r413:32
wolfsprauland dvi-i to hdmi13:32
wolfsprauland then also usb2midi, since we talk to cable vendors anyway it's easy to add to the list, maybe for the future13:33
wolfspraulthe only thing that's a must for R4 is dvi-i to vga/hdmi, the rest requires software upgrades first13:33
wpwrakvga-in also needs some signal mixing, to squeeze vga into 3 signals. not sure if there are ready-made cables for this. maybe yes.13:34
wolfspraulI hope one day we can still support tv-out13:35
wpwrakif it will still matter that day :)13:36
wolfspraulbecause by watching a number of projectors, I think VGA is disappearing faster then TV-IN13:36
wolfspraulyes, it will. tv-in/out will stay longer than vga I think, much longer maybe. that's just my impression afer following and watching this more closely for a while now13:36
wpwrakit may become more and more crappy over time, though13:37
wolfsprauleven the most high-end HD-whatnot screens have RCA jacks, and I cannot believe how many installed high-end screens I find where people use exactly those jacks to connect stuff13:37
wolfspraulbut vga, compared to RCA, seems on the fast track to computer junkyard13:37
wpwrakvga has the disadvantage of taking up a bit of space13:38
wpwrakRCA is nice and simple, as long as you don't care about bandwidth :)13:38
wolfspraulso I think as long as we have an expensive DAC in our product, we should make use of that DAC to support tv-out13:38
wolfspraulI know, just saying from observation of projectors and screens.13:38
wolfspraulvga is on the way out, rca is not13:39
ThihiIsn't a realistic modern tv-out hdmi, not rca?13:39
wolfspraulyes, sure13:39
wolfspraulthat's a separate line of thought though, and we are adding digital video-out in R413:39
wolfspraulm1 has 90% of what it needs to support tv-out, so as long as it has a DAC (and can thus support either vga or tv-out), it would be great to support tv-out13:40
wolfspraula nice feature to add one day...13:40
wolfspraulnot urgent either though, but before removing the dav and analog video entirely one day, we should first add tv-out since that may create enough value to justify keeping the dac13:41
wolfspraulthis is not about now, I mean some time in the future13:41
wolfspraulThihi: hi there btw, and thanks for your feedback a little while back!13:41
wolfspraulthat is really very appreciated, and I take it serious too13:42
wolfspraulalso I think in m1 we have to be careful to not fiddle around the hardware too much and too fast if sw cannot move along and use all that is already possible with the hw13:43
wolfspraulit needs to be a healthy give and take, with focus on community and contributor growth13:43
wpwraksw lagging also causes holes in hw design testing13:45
wolfspraulyet not adding new hw features means that no new opportunities to contribute emerge13:46
wolfspraulit's a subtle balance13:46
wolfspraulgood thing the product itself starts to work better, and can guide us through the process13:46
wpwrakhmm yes, seems that we could support composite out just by soft- and gateware (plus a fancy adapter cable)13:52
wolfspraulnot fancy, easy to source and in fact we bought a few already and sent them here and there13:53
wolfspraulthis used to be a feature of some graphics cards13:53
wolfspraulthe cable goes from vga to rca and s-video I think, there seem to be some differences in how they wire up ddc, I think13:54
wolfspraulI mean different versions of this kind of cable exist13:54
Thihiwolfspraul, np. I just boxed the M1 btw. yesterday. I'm taking it to the office some day next week, and then Rasmus can send it back to you. Do we have a return address? You could mail me one at tapiob@mikrobitti.fi13:55
wolfspraulI have no pressing need to get it back, but of course if you want to get rid of it I have a return address in Germany for you.13:56
wolfsprauldid you get the replacement power supply adapter I sent at some point?13:56
wpwrak(not so fancy cable) oh, cool.13:56
wolfspraulso yes, I can email you a return address13:56
ThihiYou don't need it? No other reporters looking for a review kit? :)13:56
wolfspraulare you still generally interested in the product? I mean I could send it to you again let's say in 6 months or so if/when it's a lot better :-)13:57
wolfspraulif you like...13:57
ThihiWell, sure, why not.13:57
wolfspraulit's up to you13:57
wolfspraulwere you able to use it for any reporting this time?13:57
ThihiWell, there is a lot of potential, and I'd love to see it later on, too, so yeah, let's do that13:57
wolfspraulor is it just too exotic right now?13:57
ThihiYeah, I wrote a half page article about it.13:58
wolfspraulyes that's probably better than it languishing around at your place for months and months13:58
wolfspraulis there a URL to that article?13:58
ThihiNope, it's in printed media.13:58
ThihiAnd in finnish.13:58
wolfspraulcan you include 2 or 3 color photocopies of the article in the shipment back?13:59
ThihiBut I can give you the gist of it later on, and maybe provide some quotes. Of course I can give you the whole text, but since it's finnish I don't think you need it?13:59
ThihiWell, sure.13:59
wolfspraulgoogle translate works better and better13:59
ThihiI think I can send you the magazine even. But it's not out quite yet - in a few weeks.13:59
wolfspraulyes, please do so [color photocopies]13:59
wolfspraulyes, magazine even better14:00
wolfspraullet's wait for that14:00
wolfsprauland thanks a lot for you taking the time to do all this, I hope you feel it was worth it14:00
wolfspraulthe milkymist soc and m1 are very fragile ideas nowadays, they need a lot of nurturing to become big and strong14:00
ThihiNp. :)14:00
wolfspraulmaybe now that you saw m1 you know what I mean :-)14:01
wolfspraulat least I think you would agree with me that there really is a very unique and different idea in this, so with a little luck and caring, it may grow strong14:01
wolfspraulwe shall find out :-)14:01
wolfspraulit's great that you follow it!14:02
ThihiAnyway, breakfast (yeah, it's four a clock in the afternoon, so sue me, it's sunday) is waiting for me. So talk to you later.14:02
wolfsprauli'm off to sleep soon, 10 pm here and alarm at 5.30, urgh14:03
kristianpaulalarm :-)14:04
kristianpaulflying early ? :)14:04
wolfspraulno flying, I improve my co2 footprint :-)14:05
wolfspraulI getup every day at 5.30 now since it's easiest with the kindergarten schedule of my daughter14:06
wolfspraulkicks me in shape by 7 AM, not bad actually14:06
kristianpauloh, i see :)14:06
wolfspraulbut as soon as I can I will go back to the more natural sleep from 4 AM to 1 PM14:06
kristianpaulwakup early is really refreshing, here is around 20°C at that time... 1 PM is sunlight hell of 32°C :-|14:20
ThihiUgh. I wish for a perpetual 17 to 25 C.14:21
ThihiUnfortunately, since I'm finnish, that's two months of the year ;P14:22
Fallenou01:54 < wpwrak> Fallenou: hurry ! we need to get MMU, linux, then USB-WLAN. quicker than wolfgang can sort out the quirks of the chimera he's building ! :) < hehe working on it :) I do the best I can with the limited time I have :x I just wrote 16 test cases for memory loads, will make it run today !15:23
wpwrakwhee ! :)15:27
kristianpaulargh, how i detect  read cycle... i had done this so far https://gist.github.com/fdf5c2b4779b07a9b91d16:19
kristianpaulproblem is when i add the  dummy_flag_clear <= read_flag_wire; to the same case statment for read.. well it always clear the reg..16:21
kristianpaulhmm lets try add wb_we_i the other always block..16:22
GitHub65[migen] sbourdeauducq pushed 1 new commit to master: https://github.com/milkymist/migen/commit/29859acc34d181442050adf98514021fbad752f218:23
GitHub65[migen/master] sim: two way IPC working - Sebastien Bourdeauducq18:23
GitHub135[milkymist] wpwrak pushed 1 new commit to hid: https://github.com/milkymist/milkymist/commit/7696bf902dc8569ad3a774d6b77e6d35bc8e7be518:54
GitHub135[milkymist/hid] Merge branch 'hid' of github.com:milkymist/milkymist into hid - Werner Almesberger18:54
GitHub99[migen] sbourdeauducq pushed 1 new commit to master: https://github.com/milkymist/migen/commit/c0b0161ec9f99e41003f8073fd0de45297268e4b20:33
GitHub99[migen/master] sim: compile VPI module - Sebastien Bourdeauducq20:33
GitHub163[migen] sbourdeauducq pushed 1 new commit to master: https://github.com/milkymist/migen/commit/2cd71e4b5ea88c75aa247b499cac7654c6e56c0921:39
GitHub163[migen/master] sim: signal reads working - Sebastien Bourdeauducq21:39
GitHub163[migen] sbourdeauducq pushed 1 new commit to master: https://github.com/milkymist/migen/commit/9bbec278c6c294bca5fa4a541c68a7d23bc8bdad22:03
GitHub163[migen/master] sim: cleanups - Sebastien Bourdeauducq22:03
--- Mon Mar 5 201200:00

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