#qi-hardware IRC log for Wednesday, 2014-06-25

kyakwpwrak: where?04:15
jaime_lionanyone in here wana do a devils advocit thing and debate about the future of computer repair?06:44
wpwrakkyak: argh. i'm getting old ... here it is: www.infobae.com/2014/06/24/1575461-video-viajo-mas-80-kilometros-un-gato-la-suspension08:12
kyakwow08:31
kyaki know rats often get under the car's hood08:32
kyakpoor thing.. it looks exactly like my cat08:32
kyakdoesn't look like it's a street can either08:32
kyaks/can/cat08:32
wpwrakit's amazing that it survived that ordeal, apparently unharmed. well, i think there's a place it'll be very careful to avoid in the future :)08:34
kyakthey speak very intelligently08:36
kyaknot a swear word! :)08:36
wpwrakwow :)08:36
kyaki'm kidding, surely08:37
kyakamazing cat08:38
kyaki would keep it to myself08:39
wpwrakfor all i know, they could have been reciting dostoyevsky ;-)08:39
wpwrak(keep the cat) the way the guy was cradling the cat had a hint of possessiveness to it08:40
kyakit's a pity you don't know Russian, their language is very "vivid"08:40
wpwrakand yes, it's a cute one. and you won't find one with a more amazing story easily08:40
whitequarkkyak: ... "their"?08:41
whitequarkoh08:41
whitequarknevermind08:41
kyakwhitequark: of those guys, i mean :)08:41
kyak"C =53> <>@40 H8@5 <>59"08:41
whitequarklol08:42
kyaki mean, this video can easily be splitted into quotes08:44
wpwrak"his muzzle my wider"08:46
kyak"his mug (muzzle, snout?) is wider than mine"08:48
wpwrakhah, so that one was guessable :)08:49
astrwhat sort of leds do smartphones camera flash use?10:02
astrI'm trying to find a stereo 3.5mm jack socket that has a switch? so normally with no jack in socket it is connected to speaker amp when jack connected speaker amp disconnected10:15
astrhmm food for thought http://en.qi-hardware.com/wiki/Audio10:23
wpwrakbtw, is anyone familiar with things like jQuery and the connections between CSS and Javascript ? i'm trying to make "fancy" pop-up things with image maps12:41
wpwrakastr: (jack) jacks with switch should be easy to find. they may not switch the signal directly, but you'd want some processing anyway. e.g., turn the speaker amplifier on or off.12:42
astrwpwrak, yea, I was wonder how to switch the amp on/off when not in use to save power12:43
astrs/wonder/wondering/g12:43
astrs/was/have been/g12:44
apeleteHi larsc12:46
apeletelarsc: is it possible to know exactly how full the mmc fifo is before each transfer ?12:47
apeleteI was thinking about setting ".dst_maxburst = fifo_capacity - fifo_actual_usage"12:48
apeletefifo_capacty is 16 words12:48
apeletebut I don't know if there's a way to know fifo_actual_usage (how full it actually is)12:49
apeleteaccording to datasheet, there are only status registers for fifo_full, fifo_empty and fifo_almost_full12:50
apeleteand dma trigger level is fifo_half_capacity12:52
apeletelarsc: but I would like to set dst_maxburst so that mmc fifo bandwith usage is alway optimal, ie dst_maxburst = fifo_size - fifo_current_usage12:53
apelete(replaced 'capacity' with 'size', and 'actual' for 'current' in the first fomula)12:54
wpwrakastr: the switches usually switch to GND. so add a pull-up and you have a nice digital signal you can use for all the rest. in some designs, you'll also have multiple amplifiers, which you'd switch according to the scenario.12:57
astrwpwrak, thx12:58
larscapelete: typically half the fifo size is a good idea14:07
larscbut there might be a configurable threshold14:07
larscapelete: try to find out at which level the mmc controller triggers a dma request14:08
larscprobably its when the fifo is almost full14:08
apelete"The DMA trigger level is 8 words, that is to say, the DMA read trigger is when data words in14:23
apeleteMSC_RXFIFO is >= 8 and the DMA write trigger is when data words in MSC_TXFIFO is < 8.14:23
apeletelarsc: so mmc controller triggers a dma request when the fifo is half full/empty14:23
larscok, maxburst should be 814:24
larscbecause you know there are at least 8 samples in the fifo14:24
apeleteyes, I was hoping this can be dynamically configurable like I said before14:24
apeletebut it seems not. I'm going to settle with 8 then (it's a pity)14:25
larscit is probably ok14:26
apeletei did some benchmark yesterday with maxburst = 8, didn't notice any improvement14:26
larscthe bottleneck might be somewhere else14:28
apeleteyeah, I'll try to run perf on the code, don't know if it's possible on the ben14:29
apeletelarsc: at least we're booting 2x faster now with dma, and CPU usage is also slightly lower14:30
larscone thing you can do is process multiple transfers at once14:30
larscthat should help speeding things up14:31
apeleteyou mean configure dma to process multiple trasnfers at once ?14:31
apeletelarsc: is that multiple block read/write vs. single block read/write mode ?14:33
larscwhat you do right now is <prepare 1><transfer 1><prepare 2><transfer 2<14:36
larscwhile you could do prepare 2 and transfer one in paralell14:36
larscthe mmc framework has support for this14:37
apeletelarsc: nice. I don't remember seeing anything like that in other drivers, will take a look to see how it's done14:41
pcercueiwith chained descriptors, you could even program all transfers beforehand14:42
larscyou still need to program the transfers14:42
apeletelarsc: do you know any driver that does mutiple transfers at once ?14:48
apeletehmm, seems like mmci.c is doing it, not sure though15:00
larscany driver implementing the pre_req and post_req callbacks15:14
apeleteok, thanks15:26
--- Thu Jun 26 201400:00

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