#qi-hardware IRC log for Tuesday, 2012-06-19

qi-bot[commit] Adam Wang: sot.fpd: added package of SO-235 also for TI DBV(R-PDSO-G5) (master) http://qi-hw.com/p/kicad-libs/445734e02:45
qi-bot[commit] Adam Wang: generic header through hole type (master) http://qi-hw.com/p/kicad-libs/f5392ae03:23
qi-bot[commit] Adam Wang: added header.fpd: copied from Werner Almesberger (master) http://qi-hw.com/p/kicad-libs/045e84803:24
wpwrakgood. you found it :)03:25
wpwrakthat's another 80 footprints03:27
cladamwyes, but too bad that i just did another 2*row header weeks ago. :(03:30
cladamwbut still okay. :-)03:30
wpwrakwell, it's a relatively simple one :)03:31
cladamwi forgot to make test points module. seems you also didn't make before. :)03:31
cladamwTP is the last work for links. else all are done. :-)03:31
wpwraki use things from pads.fpd for TP03:32
cladamwTest Points could be many types, round, square, circle, with suitable cut-out area.03:32
cladamwoh ... that one i see. :-)03:32
cladamwbut not enough to meet future work and m1. :-)03:33
wpwrakyou can add more ;-)03:33
wpwrakbtw, there's another place to consider for scavenging: http://svn.openmoko.org/trunk/gta02-core/modules/03:34
wpwrakthese things are ancient and often with bugs, but some may still be useful03:34
cladamwaha ... nice, the server comes back. :-)03:34
wpwraka catalog is here: http://people.openmoko.org/werner/gta02-core/gta02-core-modules.pdf03:35
cladamwgood seems the diode's marker is better than ours now03:40
cladamwand there's TST module. :-)03:40
wpwrakwell, tst is a precursor of pads :)03:40
cladamwyeah .... 03:41
cladamwwell... i need to make new one. :) 03:41
wpwrakwhy not just add one in pads ?03:43
cladamwsure, i meant to add more variants in pads. :-)03:45
qi-bot[commit] Adam Wang: pads.fpd: added variant sets of C(circle), R(oround) and S(square) pads. (master) http://qi-hw.com/p/kicad-libs/3b3541606:28
DocScrutinizer05wpwrak: I strongly suggest you simulate your idea of limiting inrush current with a choke, in spice07:57
DocScrutinizer05you'll notice quite differnet effects than what you hoped for07:57
DocScrutinizer05it *could* even fry your USB plug07:59
DocScrutinizer05as in an ideal circuit the sinuoidal half wave of current will reach maximum when input and output of choke are both at 5V. Then current decays until it reaches 0 when input=5V and output=10V08:06
wpwrakDocScrutinizer: don't worry, suitable chokes are all too big anyway. i'm just solving this with procedure :) besides, i have OTA updates working now. farewell, programming cable :)12:01
ArtyomHave anyone seen/tested this toy: http://www.zedboard.org/node/8 ? 12:03
whitequarkdid they, hm, take a xilinx devboard and replace the fpga with cortex-a9?12:06
Artyomwhitequark: which devboard do you mean?12:13
rzkwhitequark: its on xilinx zynq, probably shares same design as other "default xinix devkit"-ish stuff. even digilent logo is there.12:41
rzks/xinix/xilinx12:41
whitequarkArtyom: nexys13:18
Artyomlooks similar but not exactly the same...13:28
kristianpaulArtyom, thanks for license update about kicad project !16:40
Aylamth: hi, what's the name of the upcoming feature on the kernel that stops the periodic IRQ when there's only one task running?17:34
DocScrutinizer51sth pretty unlikely to ever happen17:41
DocScrutinizer51well17:42
DocScrutinizer51no task running might be a more likely case17:42
viricwhy?17:49
viricwhat should the kernel do periodically, for one task running?17:49
DocScrutinizer51I hardly can see one task running solitary17:54
rohDocScrutinizer51: ;)17:55
whitequarkhm17:55
whitequarkwasn't that "tickless kernel"?17:55
viricnon-sleeping17:55
whitequarkand isn't it implemented eons ago?17:55
DocScrutinizer51thought so17:55
Action: roh just checked and has >200 tasks running somewhat. most sleeping till their next wakeup17:55
viricwhitequark: yes, I also thought of NO_HZ, but I don't know what's this about then17:55
whitequarkviric: link?17:56
viricwhitequark: Ayla mentioned it, not me17:56
whitequarkoops18:04
DocScrutinizer06>>"The tickless kernel feature (CONFIG_NO_HZ) enables 'on-demand' timer interrupts: if there is no timer to be expired for say 1.5 seconds when the system goes idle, then the system will stay totally idle for 1.5 seconds. This should bring cooler CPUs and power savings: on our (x86) testboxes we have measured the effective IRQ rate to go from HZ to 1-2 timer interrupts per second.<<18:46
DocScrutinizer06so AIUI this means there shouldn't be any such thing like a periodic IRQ since ~200618:49
Aylawhitequark: I don't know how to explain it, mainly because I don't really understand what it is21:44
Aylamth said that the "tickless" feature is not exactly what he thought it was, only a part of it21:45
Aylaah, found it22:00
Aylahttps://lkml.org/lkml/2011/8/15/24522:01
whitequarkAyla: yeah, got it. an interesting idea22:39
whitequarkhttps://github.com/fweisbec/linux-dynticks/wiki/TODO22:40
whitequarklooks like it won't get to mainline any soon22:40
Aylayes... but now I realize that it's probably not useful at all if you have only one CPU22:40
whitequarkmobile devices start to explore the smp world22:41
whitequarkand that's where power consumption matters22:41
mthwhitequark: current tickless only disables the timer interrupt if the system is idle23:09
mthwhile it could also be disabled if there is only one task that wants to run23:09
Aylabut then how to count the jiffies?23:09
mthif you can read a timer register, you could use that23:10
mthso make it count, just don't make it fire often23:10
Aylathen it'd be specific to the hardware23:12
Aylaas a generic MIPS processor doesn't have a timer register (it does?)23:12
mthit would only work on certain hardware, but for embedded systems I don't see that as a problem23:14
mthif your hw supports it, you enable it, otherwise you don't23:14
Aylaok23:14
Aylaand does the jz4740 have that?23:15
mthiirc you can read the timer as it's counting, but let me check23:15
mthit can be read,  but it's only 16 bits23:16
mthyou can set a divider for the clock input of the timer though23:17
mthRTC clock is 32768 Hz, so if you prescale that by a factor 64 you get 512 Hz and the timer overflows after more than 2 minutes23:20
mthmaybe that's why HZ=256 is useful23:21
mtheven at prescale 1 though, you've got 2 seconds until the timer overflows23:22
mthso if you set the timer interrupt to once per second and read the timer counter inbetween, you still save a huge number of interrupts23:22
AylaRTC clock is fixed?23:23
Aylait seems to depend on the PLL23:23
mthno, RTC clock does not23:23
mthit cannot, since it has to run even during sleep and the PLL is off then23:23
Aylaok23:25
mthwhat I don't know is how much impact such a change would have on the kernel code23:26
mthbut in theory it could work23:26
mthI guess jiffies would have to be read with a function rather than from a variable (it is a variable now, right?)23:27
mthbut a function can be inlined if this feature is off in the config, so it's no extra overhead then23:27
Aylait's a variable, yes23:28
AylaI'm surprised, the ondemand governor works quite well now23:43
DocScrutinizer06whitequark: (smp) http://www.stericsson.com/products/L8540.jsp23:51
DocScrutinizer06my daily job ;-D23:51
DocScrutinizer06mth: the case when exactly one task is "R" is really rare23:52
Aylamth, larsc_: the USB driver really doesn't like CPU frequency changes23:54
DocScrutinizer06lol, sure23:54
DocScrutinizer06esp when it's changing master clock23:54
AylaI can't keep a telnet session open for a long time with the ondemand governor23:54
Aylano, that's a bug23:55
Aylait doesn't crash on PC, it shouldn't crash on dingoo/nanonote23:55
DocScrutinizer06the clock generator is definitely a special unique critter on every platform/SoC23:56
DocScrutinizer06ideally you can change CPU clock without any impact on UART et al clocks23:56
mthwhy would there be more than one running task when a typical framebuffer application is running?23:56
DocScrutinizer06on S3C2410 this wasn't the case23:56
mthif I run htop in the Dingoo while in the menu, I see only htop itself as running23:57
mtheven the menu is sleeping until I press a key23:57
DocScrutinizer06still I bet X is running too23:57
mththere is no X23:57
mththis is not a desktop, it's very much stripped down23:58
DocScrutinizer06well, just saying one R task is relatively rare23:58
DocScrutinizer06and then I fail to see why there's any IRQ firing in that case23:58
DocScrutinizer06but probably I'm just too tired now23:59
DocScrutinizer06inline with that: n8 fellas23:59
mththe IRQ fires because that's how the kernel is written23:59
mthit doesn't really have to, in theory23:59
mthnite23:59
DocScrutinizer06it soesn't, since no-hz AIUI23:59
--- Wed Jun 20 201200:00

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