#milkymist IRC log for Sunday, 2011-07-24

AlarmI want to install qemu with Milkymist using the wiki: 'http://www.milkymist.org/wiki/index.php?title=Using_QEMU'07:52
AlarmI do not understand or I have to apply: 'git revert bf5547f5365c1bfeba9d5c3e87653b9803a875a5'?07:52
larschm?07:54
larscafter cloning the repo run above command07:54
AlarmI do this first: 'git clone git: / / git.qemu.org / qemu.git' ?07:57
larscyes07:57
Alarms that the qemu git clone should be in the same directory as / opt/rtems-4.11?08:02
larscdoesn't matter08:05
Alarmok08:05
Alarmvim opens when I apply the 'git revert'?08:22
larscjust safe the file and close it agaion08:25
larsc:wq08:25
AlarmI did the make after the '. / Configure08:36
AlarmI get an error 'block / raw-posix.c: 35:28: fatal error: IOKit / IOKitLib.h: No such file or directory08:36
larscIOKit sounds like  MacOS X library08:37
wpwrakwolfspraul: what kind of package do you want for the gates ? here are some choices:08:37
wpwrakhttp://www.nxp.com/documents/data_sheet/74AUP1G08.pdf08:37
wpwrakhttp://focus.ti.com/lit/ds/symlink/sn74aup1g08.pdf08:38
wpwraklekernel: (having to search the web for codecs) have you gone over to the evil side (windows) ? :)08:39
larsclekernel: what distribution are you using now? fedora?08:40
lekernelI was running Linux before computers were seriously able to play video08:40
lekernelyes, fedora08:40
wpwraklekernel: at least that's the only place where i know such adventures from (once had a new laptop that came with this junk preinstalled and wanted to try to play some video. i could not have imagined in my wildest dreams just how awful things were in that particular hell ...)08:40
lekernelanother (and more serious) annoyance on this distro is pulseaudio which has more bugs than a rainforest, and on which most of the GUI environments depend on, so you can't uninstall it cleanly08:41
Action: wpwrak tries to remember the first videos ... i don't think linux has been around back then ;-)08:41
larsci'm using debian and i never had to manually download and install codecs08:41
lekerneleven to play MP3, WMA, encrypted DVDs, etc.?08:41
wolfspraulwpwrak: the most commond smd type? can be small and no-lead etc. but most common and cheapest is what I'd pick.08:42
wpwrakwolfspraul: they're all kinda common ;-) well, let's pick one that both TI and NXP have then08:42
lekernelmy experience with Debian is I had to install extra stuff to play proprietary/patented/DRM'd formats08:42
larscat least mp3 works out of the box08:43
wolfspraulwpwrak: the cheapest :-)08:43
larsci'm not sure if i even have an wma files08:44
wpwrakmplayer on ubuntu handles pretty much everything i throw at it08:45
lekernelinteresting...08:45
Alarmlarsc: is it necessary to install MacOS X libraries?08:46
larscAlarm: do you use MacOS X?08:46
Alarmno08:46
larscwell, then not ;)08:47
larsccheck your config for some reason qemu seems to to think you have it08:47
wpwrakwolfspraul: (no-lead) contacts not accessible from the side is okay ?08:48
wolfspraulI can only answer based on my knowledge, so I say 'cheapest'.08:49
wolfspraulyes we need to integrate, but in the end I'd still just go with the cheapest08:50
wolfspraulunless we are so space constrained that even if no-lead is more expensive I'd take that08:50
wpwrakis about 3.2 x 2.5 mm occupied space for one gate okay ? (you'll need two of them)08:51
wolfspraulthat's a lot ;-)08:51
wpwraksee :)08:51
wolfspraulyes but let's look at price first. I could imagine that the most common and relatively small no-lead version is also the cheapest.08:52
wpwraklet's see what a search for cheapest turns up ...08:52
wpwrak6-XFDFN hmm ...08:53
wpwrakSOT-5 is only about 10% more expensive than 6-XFDFN08:54
Alarmlarsc: ok I understand. I typed the ". / Configure 'for MacOS X08:54
wolfspraulwpwrak: I think no-lead and bga are uncommon and expensive. guess only used when people are space constrained.09:00
wolfspraulI don't know where you look, I look at digikey and the SOT-5 come up as cheapest09:01
wolfspraul10-13 cents (in quantity)09:01
wolfspraulwhereas no-lead and bga is more like 40-50 cents09:01
wolfspraulah yes, 6-XFDFN is also there, 17 cents09:03
wolfspraulbut 6-XFDFN is smaller, right? so maybe a good choice?09:03
wpwrakwolfspraul: DSF/DRY would be suitable. they're no-lead, no lateral contacts, and are available from TI and NXP. DSF aka SOT886 aka MO-252 is the cheapest package at TI. but not stocked at digi-key from NXP. DRY aka SOT891 aka MO-287-UFAD would be.09:04
wpwrakthre09:06
wpwraklet's try this again ...09:06
wolfspraulok :-) I don't know all these package names.09:06
wpwrakthere's a hektazillion of different packages called XSON6 , so that designator doesn't help :)09:06
wolfspraulI don't think manual access is important. Less space is always nice, but I have no hard number how valuable in our case. 50% more is always OK for less space.09:06
wpwrakheh :)09:06
wolfspraulif it's 3 times more for less space, maybe we first determine how space constrained we actually are09:07
wpwrakit's about 5% from DSF to DRY09:07
wolfspraulanything < 50% difference I would pick the smallest09:07
wolfsprauleven < 100% difference probably09:07
wpwrakoh, sorry. 10.8% !09:07
wolfspraulif something is not stocked at digikey, that's a bad indicator09:09
wolfspraulif whatever is not stocked there is 30% or 50% cheaper than whatever is stocked there, I would take another look. otherwise let's prefer what is stocked on digikey (maybe I would extend that to mouse & farnell as well)09:09
wolfsprauljust my thinking...09:10
wpwraki'm lazy and try not to have to look elsewhere ;-)09:10
wolfspraulsure09:10
wolfspraulI think stocked at digikey is a good indicator, which I would only override if whatever is not stocked there is substantially cheaper09:10
wolfspraul30-50%09:10
wpwrakseems we have a winner then. SN74AUP1G08DRY aka 74AUP1G08GM09:10
wolfspraulin which case I'm sure digikey would stock it :-)09:10
wolfspraulyes, perfect09:12
wolfspraulthat's the one I found for 17 cents, only 4 cents more than the larger package (if I understand the various packages correctly)09:12
wpwraklemme complete my chart and post it on the qi-hw list ...09:15
wolfspraulfantastic!09:16
wpwrakcool. TI also have some 16BGA. called YFP, just like the 6BGA. and it creeped into the data sheet. that's what happens if you pick cryptic names ;-)09:17
wolfspraulbut I think the BGA variants are substantially more expensive, no?09:26
wolfspraulmore like 50 cents vs. 10-15 cents09:26
wolfspraulwell, I am lacking complete overview. the one you pointed our earlier looked good to me.09:27
wolfspraulpointed out09:27
wpwrakgrmbl. the DRY isn't stocked either.09:30
wpwrakrecalculating ...09:30
wpwrakwe have a match on DCK/GW. that's a leaded ~1.6x2.5 (occupied area) package09:32
wpwrakDRY can be ordered, with MOQ = 1, but 2 months lead time09:37
wpwrakthe corresponding NXP part (GM) is stocked09:37
wpwrakgot DSF/GF, it's the opposite. and MOQ is 5000 units, lead time 3 months :)09:39
wpwrakwhy don't they simply display "GO AWAY !" ;-)09:39
wolfspraulI look at the digikey overview page, the one that comes up after search. I always check "in stock" when searching. I can then see how many they have in stock, and for most of those types that's in the thousands. So there should be no issue with lead time at all.09:44
wolfspraulso for example the 1G08GM you had earlier, looks good to me, no? http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail&name=568-2552-1-ND09:45
wolfspraulI have that one at 17 cents, the GW at 13 cents09:47
wolfspraulgw url is http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail&name=568-2553-1-ND09:47
wpwrakyes, the GM is good. my search criteria are: parts exists from TI, compatible part exists from NXP, TI part is stocked, NXP part is stocked.09:48
wpwrak(stocked) all at digi-key09:48
wpwrakthe next level would be statistical availability at dk/farnell/mouser/arrow09:49
Alarmqemu-system-lm32 -M milkymist -kernel flickernoise10:24
Alarmoss: Could not initialize ADC10:24
Alarmoss: Failed to open `/dev/dsp'10:24
AlarmI have no folder 'dsp' in 'dev'?10:25
lekernelthis error is not critical10:27
Alarm Failed to create voice `mm_ac97.out'10:39
Alarm> flickernoise: No such file or directory10:39
Alarm> qemu: could not load kernel 'flickernoise'10:39
larscthat on the other hand is critical10:40
lekernelAlarm, http://www.milkymist.org/updates/current/10:46
GitHub123[milkymist] sbourdeauducq pushed 1 new commit to master: https://github.com/milkymist/milkymist/commit/d7b64286d4a1d1a170ee9c5a290a71e3bf51018110:57
GitHub123[milkymist/master] TMU prefetch: true dual port RAM module - Sebastien Bourdeauducq10:57
kristianpaul(codecs) well, in debian i always install vlc and i'm almost done for video playing, besides, i dont use the dvd unit too much13:12
GitHub61[milkymist] sbourdeauducq pushed 1 new commit to master: https://github.com/milkymist/milkymist/commit/926a3bd0621cb4c090d55bd7fca47551a941bf5813:15
GitHub61[milkymist/master] TMU prefetch: quad port RAM module - Sebastien Bourdeauducq13:15
wpwrakbeer supplies refilled :) now back to those gates ...14:53
larscsounds as if you kidnapped bill gates' family14:54
wpwraknow there's an idea :)14:56
wpwraklekernel: for the reset logic, what's the minimum voltage at which the gates should be fully operational ? possile choices are: 0.8 V, 1.4 V, 1.65 V, 2 V15:00
lekernelmh, i'm not totally sure15:00
lekernel0.8V is the safest of those choices ;)15:01
wpwraklekernel: coward ;-))15:02
lekernelhe, pick your battles15:03
lekernelthis problem has used more than enough time already15:03
wpwrak74AUP it is then15:03
wpwrakyeah, true enough15:03
wpwrakhow do you want the gate at INIT_B - push/pull with series resistor or pull-up with open drain ?15:06
lekernelpull-up with open drain is what is closest to the xilinx reference design15:06
wpwrakokay. want a series resistor too ? in case INIT_B is driven high by the FPGA and the gate pulls down ? (not sure if this is likely to happen, considering that this means that PROGRAM_B_2 is low too)15:08
lekernelin theory, INIT_B is never driven high by the FPGA15:08
lekernelplus FPGA IOBs are fairly resistant to short circuits15:09
lekernelwe can't include such safeguards everywhere, if we did, we'd have to put resistors e.g. on the video input pixel bus15:09
wpwrakperfect. one component less :)15:09
lekernelthen the PCB layout would become a mess and signal integrity would probably suffer as well15:10
wpwrakyeah. just want to tune this to your paranoia level. every engineer has different preferences :)15:11
wpwraksome may add overvoltage protection, a fuse, and a bead ;-)  (well, i'm exaggerating slightly here :)15:12
lekernelbeads turned out to be a bad idea, i'll 0-ohm them by default in my next design15:14
wpwrakwasn't that just because you got a bad batch ?15:15
lekernelwe had a) voltage drop issues because of improper sourcing b) time wasted to tracking down video input instability problems15:15
lekernelall attributed to beads15:15
lekernelso, all beads 0 ohm, then go to the EMC lab with a bag of beads to install them after a problem is actually detected15:17
wpwrak(video instability) you mean the ground issue ?15:18
lekernelyes15:18
wpwrakwell, that bead had no reason to be there. it actually did its work perfectly - kept all the transients from crossing via the ground plane. so instead they went through the chip ;-)15:19
wpwrakoh nice. there's a circuit symbol for open drain. very handy :) even better, both TI and NXP use it. toshiba apparently too. ST only reluctantly. fairchild don't seem to like it15:22
wpwrakthat's an almost suspiciously high level of agreement :)15:23
wpwrakof course, it would have hurt just too much to also agree on which input of an AND gate they call A and which B ...15:27
lekernel?15:28
lekernelit doesn't matter15:28
wpwrakyes and no. it doesn't matter for the function of the AND gate, but if you want to use the name to refer to the net or the pin, you may get some confusion15:29
Alarmow to reflash the card with flterm?16:23
kristianpaulflterm is not for reflash16:26
kristianpaulurjtag it is16:27
kristianpaulwith help from fjmem of course16:28
lekernel_Alarm, what do you want to reflash?16:32
AlarmWhen I got the card I was told that the apps were not up to date. I guess it flickernoise?16:36
lekernel_what version are you running atm?16:36
lekernel_0.1?16:36
lekernel_you can check http://lekernel.net/blog/2010/12/upgrading-from-flickernoise-0-1-without-a-jtag-cable/ with the images at www.milkymist.org/updates/current/16:37
lekernel_or, if you have the jtag adapter, use urjtag and the same images16:38
Alarmyes 0.116:39
lekernel_rhaa...16:39
lekernel_I see. then yes, upgrade. it should fix many bugs too :)16:40
kristianpaulyou got jtag/serial adapter with your board?16:40
Alarmyes16:40
kristianpaulgood16:40
kristianpaulthere is a script that make all process automatically16:41
Alarm:)16:41
lekernel_iirc that script uses the snapshots, no?16:41
kristianpaulah...16:41
lekernel_better download the files from http://www.milkymist.org/updates/current/16:41
lekernel_and run jtag -n reflash_all.batch16:42
kristianpauland be aware in fedora of issues with libusb..16:42
lekernel_you'll also need fjmem.bit from http://www.milkymist.org/msd/16:42
lekernel_phew. hopefully there will be no more 0.1 boards soon...16:43
kristianpaulhaha ;)16:43
kristianpaulthe'll pop up.. one by one16:43
wpwrakcould it be more straightforward to make upgrade via uSD the default process ? that should have a lot fewer "unusual" dependencies16:44
lekernel_*SD is pesky, and the latest version supports automatic updates from the web16:45
wpwraklekernel_: "automatic" if you've set up all the things needed for it :) think of computer illiterate users16:45
kristianpaul*SD is also no the best place to reach by hjand16:46
lekernel_connect ethernet cable, then it does dhcp, and press a button?16:46
wpwrakkristianpaul: i've heard that at least the mid-air collision between jtag board and uSD is gone now :)16:46
lekernel_doesn't look too hard16:46
wpwrakkristianpaul: i had a good laugh when i saw that conflict in rejon's M1 ;-)16:46
kristianpaulwpwrak: (collition) yeah, i forgot16:47
wpwraklekernel_: hmm. i wonder whether those wifi cable modems still have ether ports or if they've saved that dollar16:48
lekernel_adding wifi is out of the question anyway16:49
wpwrakkristianpaul: the uSD holder in rejon's M1 was also a bit on the shitty side in terms of mechanical properties. but maybe the current ones are different16:49
kristianpaulshitty all the way...16:50
wpwraklekernel_: that's what i mean - what do people do whose internet access is via wifi ?16:50
lekernel_proprietary modules + USB peskiness + driver issues + regulatory problems = major time wastage16:50
wpwraklekernel_: again, i'm not suggesting you should add wifi support. just think of users who only have wifi. how common are they ?16:51
wpwrak(only wifi) that is, without extra non-trivial setup16:52
lekernel_I wrote a wifi stack and driver as a consultant years ago. months later, my apartment was filled with routers with which this or that WPA/WEP/whatever feature did not work. this will not happen again.16:54
kristianpauli guess very common, also look how apple aded non-pc required feature for coming ios5,16:55
wpwraklekernel_: ah, you had the inside view of wifi hell too. good ;-)16:56
kristianpaulopencores.org is down?17:05
kristianpaulbah... just the day i need to found a forum/ML to ask for wishbone support..17:07
kristianpaullekernel_: but after all p54 solved all wifi issues, or prism54 project was not related to what you just said?17:11
lekernel_that's a different project17:11
lekernel_also prism54 used the freebsd/linux wifi stack17:11
kristianpaulah, "Lightweight Wifi" :-)17:18
wpwraklekernel_: hmm. the gates in small packages are somewhat difficult to source (you can get them, but not as easily as i'd like to). you don't happen to by any chance to have a good amount of space there, say, for a pair of SOT-23 ? :)17:42
wpwrakstock at farnell is amazing. 5 whole units of one of these chips. why do they even bother ? ;-)17:43
mwallelarsc: lekernel_: is there a milkymist-uclibc repository?18:32
mwallelarsc: where are you setting _dtb_start?18:33
kristianpaulmwalle: i tought it was you having the uclibc repo?18:35
mwallekristianpaul: yeah, maybe theres some offical one :)18:35
mwallelekernel_: could you add a uclibc-lm32 and a elf2flt-lm32 repo to the mm github?18:36
kristianpaulmwalle:  yes it should be on the organization page in github18:37
lekernel_mwalle, done19:01
mwallelekernel_: thx19:01
lekernel_wpwrak, sot23 should be fine imo19:02
lekernel_the m1 pcb isn't that dense19:02
mwallelarsc: ah ok generic stuff19:06
lekernelAlarm, have you managed to upgrade your board?19:08
wpwraklekernel: oh, great. that will make things easier.19:11
Alarmot yet I have not finished loading the update.19:18
kristianpaullekernel: therically what will happen to milkymist conbus/lm32 if in a read cycle after a asserted cyc a slave reply with an extra ack?19:36
wpwraklekernel: which approach do you like better - left or right ? http://downloads.qi-hardware.com/people/werner/m1/rst/rst.ps20:52
wpwraklekernel: (and please ignore the twisted arc in the AND gate. thought that bug was dead already :-( )20:52
lekerneli'm not sure the reset ic should drive program_b20:54
lekernelr60 on the left is unnecessary20:54
lekernelotherwise, I prefer the left one20:55
lekernelbut I hope those gates really would work for this purpose ...20:55
lekernelconnecting program_b makes some sense though - pulsing it low to clear the FPGA would also reset the flash20:56
wpwrakhmm, you want a gate between reset and PROGRAM_B_2 pull-up as well ?20:56
lekernelno, but for delaying configuration program_b has no effect20:56
wpwrakah, but you still need to reset, don't you ?20:56
lekernelfrom my understanding it's a falling edge sensitive signal that clears an already configured fpga20:56
wpwrakso the real reset happens elsewhere ?20:57
lekernelthe fpga configuration is delayed by holding init_b low when it's still unconfigured20:57
wpwrakyes, but what initiates the configuration ?20:58
lekernelI don't know why they have not combined those two signals into one ...20:58
lekernelthe FPGA, as soon as init_b goes high20:58
wpwrakwhy make it simple if you can make it mysterious ? ;-)20:58
wpwrakoh, so INIT_B is reset and delay ?20:59
lekernelinit_b is delay, program_b is reset20:59
lekernelbut the fpga does not need a reset pulse to start - applying voltage is sufficient20:59
lekernelaccording to the xilinx datasheet, it begins configuration after all 3 supply voltages have reached minimum levels for 5 ms21:00
wpwraki see. and there's no other reset source around there either. so shall i disconnect PROGRAM_B_2 ?21:00
lekernelas I said above - PROGRAM_B could be useful for manually clearing the fpga. in that case, resetting the flash at the same time makes sense.21:00
wpwrakso that would be via TP36 ?21:01
lekernelyes21:01
lekernelalso, it currently works with PROGRAM_B connected and I do not want to spend more time experimenting with this stuff21:02
wpwrakok. so i just remove R60 ? does C238 stay ?21:02
wpwrakhehe ;-)21:02
kristianpaullekernel: no need to anwer, it crashes :-)21:02
lekernelC238? no, it shouldn't stay21:03
wpwrakperfect :)21:03
wpwrakL48N is always defined then, after reset Z plus pull-up, later either that, or push or pull ?21:05
lekernelwhat's L48N?21:12
wpwrakbtw, the one on the right side would have the advantage of needing only one type of gate21:12
lekernelhow much are those gates btw?21:13
wpwrak25-29 US-cents each if you buy 10021:18
GitHub9[milkymist] sbourdeauducq pushed 1 new commit to master: https://github.com/milkymist/milkymist/commit/70d9ba910c0637466a77050c4f90da77222ecb9521:28
GitHub9[milkymist/master] TMU prefetch: cache data memory. Prefetch complete but untested. - Sebastien Bourdeauducq21:28
GitHub107[milkymist] sbourdeauducq pushed 1 new commit to master: https://github.com/milkymist/milkymist/commit/44b92e56a920987e5db594cf05d6ff05ac363c0121:41
GitHub107[milkymist/master] TMU prefetch: fixed compilation errors - Sebastien Bourdeauducq21:41
GitHub77[milkymist] sbourdeauducq pushed 3 new commits to master: https://github.com/milkymist/milkymist/compare/44b92e5...3ad5eb922:13
GitHub77[milkymist/master] TMU: check fdest cache valid bit - Sebastien Bourdeauducq22:13
GitHub77[milkymist/master] TMU: check for address generator busy - Sebastien Bourdeauducq22:13
GitHub77[milkymist/master] TMU prefetch: fix Xst warnings - Sebastien Bourdeauducq22:13
wpwrak(l48n) that's the fpga pin that drives FLASH_RESET_N. full name is IO_L48N_M1DQ9_1. i abbreviated it to IO_L48N, because i've seen similar names in the neighbourhood22:30
GitHub120[linux-milkymist] mwalle pushed 5 new commits to master: https://github.com/milkymist/linux-milkymist/compare/5b92aa2...aaa25ab23:12
GitHub120[linux-milkymist/master] lm32: lm32 has no dma capabilities - Michael Walle23:12
GitHub120[linux-milkymist/master] lm32: support for simpleImage - Michael Walle23:12
GitHub120[linux-milkymist/master] lm32: move dtb section definition - Michael Walle23:12
--- Mon Jul 25 201100:00

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