#milkymist IRC log for Sunday, 2013-04-07

--- Sun Apr 7 201300:00
_florent_hi12:00
_florent_http://www.enjoy-digital.fr/open_hardware/3/kc705_hdmi_07042013.jpg12:01
_florent_it seems I figured out how to configure the adv7511 chip... ;)12:01
Fallenou__florent_: wow :)12:14
Fallenou_very nice, I saw yesterday your github repo for the kc70512:15
Fallenou_I guess on this picture we can see the hdmi output driven by the kc705 port of milkymist-ng soc?12:15
Fallenou_expensive board though ...12:16
_florent_yes, in fact I'm using milkymist-ng frame buffer12:16
_florent_+ RGB to YCbCr conversion12:16
_florent_+ YCbCr 444 to YCbCr 42212:16
_florent_and then the ADV7511 chip12:17
_florent_sure it's expensive...!12:17
Fallenou_adv7511 takes yCbCr 422 as digital input ?12:17
_florent_it can take RGB 444 as input12:17
Fallenou_(do you have a twitter account?)12:17
_florent_but that's not possible with the kc705...12:17
Fallenou_oh, the adv chip is configured by resistors on the board?12:18
_florent_only 18 of the 36 data bit are connected12:18
Fallenou_or only a few data lines are routed12:18
Fallenou_ok12:18
_florent_no sorry I don't have a twitter account...12:18
Fallenou_ok too bad, I wanted to give you credit with your @account for the picture on twitter12:19
Fallenou_I will just quote your name instead12:19
_florent_no pb12:19
_florent_now I need a ddr controller to do interesting thing with the board12:19
Fallenou_where is the picture stored? blockram?12:20
_florent_yes!12:20
_florent_it's only a 64x64 picture12:20
Fallenou_ok12:21
Fallenou_larsc: in linux-milkymist I can see a few "extern" functions like extern DWtype __divdi3 (DWtype, DWtype);17:41
Fallenou_where are those defined?17:41
Fallenou_my guess: in the toolchain?17:41
larscshouldn't they be in arch/lm32/lib?17:43
Fallenou_well a few functions are defined over there indeed17:44
Fallenou_but not all of them17:44
larscmaybe divdi3 is never used17:44
Fallenou_I did a git grep __divdi3 for instance in linux-milkymist17:44
Fallenou_and I found no definition17:44
larscis that 64 bit divisio?17:44
Fallenou_I have no idea17:45
Fallenou_I get this while compiling NetBSD kernel17:45
Fallenou_/home/fallen/NetBSD/sys/arch/milkymist/compile/obj/GENERIC/lib/kern/libkern.o: In function `rngtest':17:45
Fallenou_(.text+0x174c): undefined reference to `__divdi3'17:45
larscI think di is short for double integer17:45
Fallenou_I am trying to sort this out17:45
Fallenou_oh, ok17:45
larscand the linux kernel does no 64bit division17:45
larscthere is a special macro to do this17:45
larscsee include/linux/math64.h17:46
Fallenou_thanks for the pointer :)17:47
wpwraknice picture. "milkymist - i am legion" ;-)17:47
Fallenou_I will try to understand what part of libkern needs __divdi3 and see if it's optional or not17:47
Fallenou_hehe yeah nice hdmi out capture :)17:48
Fallenou_and good to see that Milkymist ng SoC runs on Kintex 717:48
Fallenou_I think so far we only had numbers about the frequency it would achieve and the occupation rate of the chip17:48
Fallenou_but it never ran on actual FPGA AFAIR17:49
Fallenou_s/rate/ratio/17:49
wpwrakthat board looks odd.seems to have a LOT of regulators.17:50
Fallenou_the mezzanine pcb parts?17:53
larscwpwrak: they all have17:54
larscthe xilinx eval boards17:55
Fallenou_a high resolution picture: http://www.micron.com/~/media/Images/Content%20Images/content_image_KIntex7boardfront.jpg17:55
wpwrakseems a bit excessive. they look large. do these boards really need more than 2 high-current rails ? you always have some more odd voltages, but they tend to be low-current17:56
wpwrakeven more than i thought :)17:57
larscI don't think these boards are optimized for cost17:59
wpwrakso that's 4 rectangular modules with DC-DC converters and three smaller square modules, also DC-DC converters. .18:00
wpwrakamidst that, there could easily hide a few small LDOs18:00
wpwrakso a board with 7 high-current rails ?18:02
wpwraklarsc: yeah, but still :)18:03
Fallenouhum that's weird18:05
FallenouI get no call or reference to __divdi3 if I do objdump -D libkern.o18:05
wpwrakthe openmoko phones had about as many rail (some with the same voltage but kept separate). something like two chips took care of all that.18:05
Fallenoubut I can see it in nm , as Undefined18:05
Fallenouand I can see it in relocations while doing objdump -D -x libkern.o18:06
Fallenouhttp://pastebin.com/Te2pAkb218:07
larscwpwrak: there is basically one 10A regulator for each supply input of the fpga18:07
wpwrakoh, it gets better. the rectangular modules have a DUAL regulator each. so, 11 high-current rails. (that reg is rated for 10 A peak) (UCD7242)18:09
Fallenouoh, ok, now I can see it at 174c indeed18:09
Fallenoubut since it was "calli 0x174c" I could not grep for it18:09
Fallenouindeed there is a long long X; in rngtest() :/18:11
wpwraklarsc: yeah, someone should have told these engineers that the supplies don't need to be stable if sorted ;-)18:11
wpwrakcorrection: 10 rails, the big critter with the cap on top is just some bead/transformer18:12
larscFallenou: can't you just copy the function from libgcc?18:15
larscwpwrak: we actually had power issues with some more low cost eval board.18:17
larscwhen running a real operating system the system would crash under high load18:18
larscthey tried to blame it on our software18:18
larschttp://billauer.co.il/blog/wp-content/uploads/2012/09/zedboard-front.jpg18:19
larscnot so many regulators anymore18:19
FallenouI guess I understood the contrary of what it meant over there: https://github.com/fallen/NetBSD/blob/master/sys/lib/libkern/Makefile.libkern#L52https://github.com/fallen/NetBSD/blob/master/sys/lib/libkern/Makefile.libkern#L5218:48
Fallenouarg18:48
Fallenouhttps://github.com/fallen/NetBSD/blob/master/sys/lib/libkern/Makefile.libkern#L5218:48
FallenouI think I should just remove the exception on lm32, I guess those .c files add basic definitions for quad operations18:49
Action: Fallenou must have a twisted mind18:49
larscyea, all other architectures in that list are 64bit18:51
FallenouI should have smelt the trap18:52
Fallenougreat, no more libgcc annoying problems :)18:59
Fallenouthat's a good thing done!18:59
Fallenoucrap now I must implement mutex_enter() and such19:06
wpwraklarsc: heh, the opposite extreme ;-)19:08
wpwraklekernel: heya, welcome back ! :)20:13
lekernelhi20:15
wpwraklekernel: DHL seem to be working hard on establishing a new record for slowness. as far as i can tell, the board isn't even in argentina yet :-(20:15
Action: lekernel had a nice week end20:16
lekernelwpwrak, it's not DHL - it's regular mail... DHL only takes documents20:16
lekernelmeh, a few days before I could track it on http://www.correoargentino.com.ar - now nothing20:17
wpwraklooong easter weekend :)20:17
lekernel(it said it was still in germany, though)20:17
wpwrakyeah. correoargentino.com.ar sometimes fails. but it tends to come back after a few hours. could also be that they're doing maintenance during the weekend.20:18
wpwrakwhat normally happens is that you get an entry when it arrives in the country and than another one when it goes to the post office. then it gets assigned a local number for transport inside argentina. so you can track it pretty much all the way, but the number changes in the middle.20:20
wpwrakand didn't DHL buy "regular mail" in germany ? of course, that doesn't make regular mail courier in germany, even if it says "DHL".20:21
larsc'regular mail' bought DHL couple of years back20:28
wpwrakoh, it's them who bought DHL ? interesting20:45
larscyea, it's owned by Deutsche Post20:49
GitHub193[NetBSD] fallen pushed 2 new commits to master: http://git.io/N1CF8w21:06
GitHub193NetBSD/master 461f181 Yann Sionneau: Fix mutex related link issue...21:06
GitHub193NetBSD/master 9d7eeea Yann Sionneau: Fix libcc related link issue, at last...21:06
Fallenou23:06 < GitHub193> [NetBSD] fallen pushed 2 new commits to master: http://git.io/N1CF8w                                                stekern21:06
Fallenouoops21:07
wpwraktracking works again ... but it's still somewhere between .de and .ar23:25
kristianpaulwhat are you waiting btw?23:28
kristianpauloh the HDMI exp?23:28
kristianpaulstill waiting..23:28
wpwrakyup. when it arrives, i'll have a look at the I2C mystery.23:29
wpwraksebastien sent it wednesday before easter. they said it'll arrive within six days. well, maybe they meant business days.23:31
kristianpaulDHL is usually quite fast on my experience23:31
wpwrakwell, it's the postal service. not even sure now if it'll have the DHL brand on it.23:33
wpwrakbut it's indeed a bit surprising that it would hardly move during the holidays. even to make out of the postal office in berlin took some 3-4 days.23:34
wpwrakthe entry for the 3rd day says "Insert item into bag". the one for the 4th day "Send item abroad"23:36
wpwrakthat was a week ago. since then, silence.23:36
wpwrakwell, i haven't lost any parcel shipments yet. so i'm not worried this far.23:37
kristianpaulwell a letter from the UK (cheap royal mail) took like a month a half to arrive here..23:38
kristianpaulme either just lost the patience hehe23:39
wpwrakheh :)23:39
wpwrakwell, i once had a paper sent, i think it was by UPS, to the UK. believe it or not, it got held up at UK customs and they didn't notify anyone.23:40
wpwrakback then, conference papers still had to be submitted on paper. even to networking conferences :)23:40
wpwrakactually, i think there were two or three different papers in that envelope. of course, we missed the deadline that way. they still let us in, though, but i think we didn't make it into the printed proceedings.23:42
kristianpaulhmmm shame on UPS23:45
wpwrakyeah, earned them a place of honor on my blacklist :)23:46
wpwrakcould of course also be that the people from the conference ignored/lost the notification23:48
wpwrakbut then, most couriers start calling if something it stuck for too long. at least they did back then.23:49
kristianpaulhmm i have never been called23:53
wpwrakyou probably don't let things sit around for weeks :)23:57
kristianpaulhehe23:57
--- Mon Apr 8 201300:00

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