#qi-hardware IRC log for Wednesday, 2011-07-13

qi-botThe build was successfull, see images here: http://fidelio.qi-hardware.com/~xiangfu/compile-log/openwrt-xburst.full_system-07112011-2305/03:06
stefan_schmidthi folks07:53
stefan_schmidtwpwrak: ping07:53
wpwrakstefan_schmidt: Zzz... :)09:46
stefan_schmidtwpwrak: what, you are actually sleeping? ;)09:47
wpwrakstefan_schmidt: yeah, it's this time of the year ... :)09:47
stefan_schmidtwpwrak: heh, wrong side of the earth I would say09:47
stefan_schmidtwpwrak: just some update on the tetsing I did so far09:48
stefan_schmidtArithmetic mean rountrip time: 0.000000 seconds and 151557.515625 usecs09:48
stefan_schmidtReceived 115001 bytes in 63 seconds and 947962 usec => 1798.352783 Bytes/second09:49
stefan_schmidtThat are the normal values I get09:49
wpwrakthat's actually pretty good, considering the horrible latency :)09:49
stefan_schmidtDisabling printfs in my test program as well as turning off debug in the ieee802154 stack does not make any difference09:49
stefan_schmidtOne test that is outstanding is doing them native between two laptops instaead of two kvm instances09:50
wpwraki still don't know how kyak managed to get almost 100 ms less latency on his dual-core laptop. maybe it's the dual-core (my test box is single-core)09:50
stefan_schmidtmay help to go through 2 usb stacks instead 4 :)09:50
wpwrak(stacks) ah yes :)09:50
stefan_schmidtwhat is great though is the reliability09:50
stefan_schmidtsending 1000 packets almost always gives me 1000 packtes :)09:51
stefan_schmidtthe roundtrip time is both directions of course (hence roundtrip)09:51
wpwrakmeanwhile, i've added a "hard mac" to the atusb firmware. just have to finish the counterpart on the kernel side. (basically a mix of at86rf230.c and spi_atusb.c. alas, code reuse via copy & paste)09:51
stefan_schmidtand involves copying the data to userspace and back to kernel09:52
stefan_schmidtwpwrak: (full mac) yeah seen the commit for the fw09:52
stefan_schmidtgreat to see you are tackling this09:52
wpwrakshould then be relatively straightforward to add CSMA and ACKs, too09:54
Action: stefan_schmidt grumbles about a netlink error09:54
stefan_schmidtwpwrak: don't forget resubmit in conjunction with AUTO ACK09:55
stefan_schmidtwpwrak: are you using the code form the linux driver for the firmware, the one from atmel or your own?09:55
wpwrakresubmit ?09:55
stefan_schmidtwpwrak: The chip should be able to resubmit a packet if a ack is missing within a given period of time09:56
wpwrakthe firmware is just some minimum tx/rx loop. i'll leave the configuration to the kernel.09:56
stefan_schmidtah, good09:56
wpwrakah, i don't have to resubmit. i just tell the chip to send the thing and it should take care of the rest :)09:56
stefan_schmidtso basically something that allows some "batching" of spi register reads/writes into a bigger usb URB09:57
wpwrakincluding CSMA, randomized backoff, etc.09:57
stefan_schmidtwpwrak: good :)09:57
wpwrak(batching) sort of. you tell the firmware to enter RX mode. then it takes care of handling some interrupts locally (still need to decide on a few details there). then it auto-receives frames and sends them on EP 1.09:58
wpwrakfor sending, you do a ATUSB_TX (new) request, which leaves RX mode, sends the frame, and then jumps back to RX.09:59
stefan_schmidtsounds easy enough from the kernel driver side09:59
stefan_schmidtand during probe doing the normal setup of the registers10:00
wpwrakyup10:01
stefan_schmidtsounds like a good solution to me10:05
stefan_schmidtpity that we can't re-use the 230 driver in this case. At least for ben-wpan it was possible10:05
wpwrakwell, there's not so much left of it in the end :) the core logic is quite different. besides, the driver has lots of issues :)10:08
stefan_schmidt:)10:08
stefan_schmidthmm, this small idea turned out harder that it should be10:09
stefan_schmidtGoing to teach iz a set command to allow setting the pan, addr and channel manually10:09
stefan_schmidtSetting up a network without coordinator should be possible then10:09
stefan_schmidtSomething that is needed to be "compatible" with the contiki implementation10:10
wpwrakah, no coordinator in contiki ?10:10
stefan_schmidtNeed to fight with netlink a bit it seems10:10
stefan_schmidtwpwrak: thats what I hear around me. Never used contiki myself10:10
stefan_schmidtwpwrak: But we want our DTN implementation working with one that get written fro the contiki side so we need to be compatible.10:11
stefan_schmidtMakes sense anyway I think.10:11
stefan_schmidtI'm wondering a bit on how many pitfalls I will find on the way :)10:11
wpwrakshouldn't be too horrible. after all, that's what izcoordinator has to do already :)10:14
stefan_schmidtwpwrak: indeed, getting my inspiration from there already.10:15
stefan_schmidtok, setting works now. I'm abusing the coordinator command here. Hopefully it will not have strange side effects. :)10:45
stefan_schmidtha, there we go. Communication without any coordinator involved. :)11:04
wpwrakizcoordinator joins the ranks of the jobless :)11:06
stefan_schmidtwpwrak: Aren't we all working for being jobless11:19
stefan_schmidtThe point it to get enough money before to enjoy it :)11:19
wpwrakyeah ;-) let's hope the izcoordinator could make some savings :)11:24
rjeffryawns14:26
wpwrakrjeffr: losing the eternal fight between good and boredom ?14:42
rjeffrhi werner. man, you are making some serious progress with ATusb. cool beans,14:48
wpwrakrjeffr: slowly but steadily :)14:53
rjeffrnice to see a few people are helping. interesting how subtle the issues are.14:55
wpwrakit's always the subtle ones that complicate things :)14:57
rjeffrieswpwrak have you seen a 6LoWpan implmentation on Arduino yet? would make an interesting end-point for ATben or ATusb to communicate with (sensor application)16:03
wpwrakrjeffries: you mean the xbee ? don't know what software they use, though16:24
rjeffriesthere are hardware platforms, Zigduino is nice at $70 no 6LoWpan yet however16:26
wpwrakrjeffries: ah, everybody's doing homegrown protocols then :)16:44
wpwrakrjeffries: if the avrs are big enough, they could probably run contiki, which has a 6lowpan stack. don't know how small you can make it, though.16:45
rjeffrieswpwrak I'll have a look at Contiki and ask around. thanks16:59
qi-bot[commit] Werner Almesberger: atusb/fw/mac.c: reserved code 0 and added notification of TX completion (master) http://qi-hw.com/p/ben-wpan/eeeb55918:35
qi-bot[commit] Werner Almesberger: tools/lib/atusb-common.c (atusb_tx): wait for and receive the TX completion (master) http://qi-hw.com/p/ben-wpan/8f8c41418:35
qi-bot[commit] Werner Almesberger: tools/atrf-txrx/atrf-txrx.c (main): removed endless loop in mode_hmac (master) http://qi-hw.com/p/ben-wpan/7605d8c18:35
qi-bot[commit] kyak: update asound.state (master) http://qi-hw.com/p/openwrt-packages/71066b618:43
qi-bot[commit] Werner Almesberger: spi_atusb: corrected name and typo in title comment (ben-wpan) http://qi-hw.com/p/qi-kernel/7569b8118:56
qi-bot[commit] Werner Almesberger: spi_atben, spi_atusb: trimmed "dirty" #include of at86rf230.h (ben-wpan) http://qi-hw.com/p/qi-kernel/72b376918:56
rjeffriesthis is rich: http://paul.kedrosky.com/archives/2011/07/being-a-patent-troll-now-patented.html21:05
--- Thu Jul 14 201100:00

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