IRC log for #arm-netbook on 20120413

00:25.52*** join/#arm-netbook CaCtus491 (~Kent@123-243-197-152.static.tpgi.com.au)
00:42.04*** join/#arm-netbook CaCtus491 (~Kent@123-243-197-152.static.tpgi.com.au)
01:10.53*** join/#arm-netbook eflatun (~eflatun@46.1.116.86)
01:56.33*** join/#arm-netbook CaCtus491 (~Kent@123-243-197-152.static.tpgi.com.au)
02:03.11*** join/#arm-netbook hp_|2 (~kvirc@p5B38B4D9.dip.t-dialin.net)
02:11.58*** join/#arm-netbook eflatun (~eflatun@95.173.232.76)
02:41.59*** join/#arm-netbook Bundersleague (~quassel@92.40.254.119.threembb.co.uk)
03:05.23*** join/#arm-netbook QingPei (~PPQQ@114.241.190.75)
03:23.35A0Shedslkcl:  who hoo  http://git.chromium.org/gitweb/?p=chromiumos/platform/ec.git;a=tree;f=chip;h=fab28b9e1ec6b544ed20d386984de3249a06bcb2;hb=HEAD
03:23.56A0Shedsstm32L support for EC functions from the guys at Google
05:38.37*** join/#arm-netbook hipboi (~hipboi@221.4.213.95)
06:06.40CIA-91rhombus-tech: Dmitry master * r0f88fce4beb3 /allwinner_a10/orders/fd.mdwn:
09:10.52*** join/#arm-netbook CaCtus491 (~Kent@123-243-197-152.static.tpgi.com.au)
09:23.56CIA-91rhombus-tech: Benot master * rac3eaf5ac764 /allwinner_a10/orders/15874.mdwn:
09:33.23*** join/#arm-netbook CaCtus491 (~Kent@123-243-197-152.static.tpgi.com.au)
09:51.07*** join/#arm-netbook gordan (~gordan@178.110.239.159)
10:27.07*** join/#arm-netbook gordan (~gordan@178.110.239.159)
10:54.01*** join/#arm-netbook gordan_ (~gordan@178.110.239.159)
11:14.49MaDMaLKaVso, that mali driver for 3.3 , how hard will bew to use it with our 3.3 kernel tree until lima got in a more mature state?
11:19.14*** join/#arm-netbook gordan_ (~gordan@178.110.239.159)
11:19.26mnemocandroid uses bionic as libc, so useless for any (eg|g|uc)libc
11:19.32mnemocbased system
11:19.37*** join/#arm-netbook gordan_ (~gordan@178.110.239.159)
11:22.39MaDMaLKaVmnemoc, I'm referering to the ubuntu one on latest linaro build
11:22.46MaDMaLKaVthat one should be useful, right
11:22.47MaDMaLKaV?
11:23.52mnemocno, android and ubuntu/linaro use different libc implementations so the binaries (and libraries) are not compatible between them even when the architecture is the same
11:25.15mnemocthe mali driver for android+linux3.3 is only usable on android+linux3.3
11:28.32mnemoceven if the android comes from linaro
11:28.59MaDMaLKaVso linaro builds ubuntu against bionic ?
11:29.24mnemocsure, or it would break compatiblity with the whole rest of the android world
11:29.58MaDMaLKaVI don't understand why an ubuntu build should be compatible with android at all... this linaro guys confuse me
11:30.13mnemoclinaro has two focus
11:30.44hnoMaDMaLKaV, GPU do not equal good media decoding capabilities. Some GPUs also do media decoding, but not all.
11:32.28hnoandroid is also built with softfloat so android libraries can't be used on any hardfloat system even if the libc incompatibilities are resolved somehow.
11:32.39MaDMaLKaVI see
11:32.57MaDMaLKaVbut I fail to understand how the linaro ubuntu builds works them, gonna read about to see what I'm missing of the picture
11:33.38mnemoclinaro produces different toolchains
11:33.53hnoI do not know what linaru ubuntu build includes.
11:33.55mnemocwhile improving ARM support broadly
11:34.20hnobut if linaro have an ubuntu spin with MALI support then it's built for glibc.
11:34.52hnohopefully even an hardfloat one.
11:36.18MaDMaLKaVthey have, or yesterday I was quite sleepy when I read that
11:36.19MaDMaLKaV:)
11:36.50MaDMaLKaVhttp://www.igloocommunity.org/support/BSP_Releases#2012.03.1
11:36.51hnoif they have then it's great. Do you have a link?
11:37.18MaDMaLKaVj1nx shared that info yesterday, but as I said I was quite sleepy, so there is serious chances I misunderstood something
11:37.54mnemocthe linaro-ubuntu team claims to enable thumb2 and hardfloat everywhere they can
11:38.47MaDMaLKaVgod, how much will I love to be able to drag'n'drop files to a VM
11:39.57mnemocMaDMaLKaV: you seem to be right, that is a real linux (not-android) BSP
11:41.40MaDMaLKaVit was j1nx who saw it , but he had connection problems because of a sandstorm
11:41.48MaDMaLKaVso, that mali driver can help us?
11:43.27mnemocyes
11:43.34MaDMaLKaVcool
12:11.22*** join/#arm-netbook eflatun (~eflatun@46.1.185.94)
12:30.34mnemochttps://wiki.linaro.org/Resources/HowTo/AddingNewHardwareSupport
12:40.22*** join/#arm-netbook QingPei (~qingpei@124.64.127.172)
14:32.26MaDMaLKaVa little offtopic if you don't care guys: I have an android tablet connected to the pc, device manager shows the adb driver ok but doing adb devices says that there is no device connected
14:32.30MaDMaLKaVany idea what can be wrong?
14:34.10specingHave you started the adb server as root?
14:38.15MaDMaLKaVnot on a unix system
14:40.02MaDMaLKaVok, let's follow first law of computer troubleshooting: reboot
14:40.06MaDMaLKaVbrb
14:40.39mnemocMaDMaLKaV: first confirm on dmesg and lsusb you actually see a usb gadjet when plugging the tablet
14:41.02mnemocnarf
14:45.24*** join/#arm-netbook MaDMaLKaV (~IceChat9@188.86.253.69)
14:45.36MaDMaLKaVback
14:45.57MaDMaLKaVdidn't work :(
14:49.19specingyou can't really do android dev on a non-unix system
14:49.48traeakhmm
14:50.01traeakthat's fine for me...does that mess up these windoze developers?
14:50.15traeakneed to give these windoze guys every excuse to ditch wp7
14:50.18traeakabomination
14:51.27MaDMaLKaVI'm not really doing android dev, and for that stuff I use a VM, I'm just setting up a recovery method with nvflash for common users, so I need stuff working on windows as it is what most of them use
14:52.01traeakahh
14:52.03traeakyeah
14:52.08specingEnjoy :D
14:52.17traeaki hate it when people only provide recovery instructions for windows
14:52.42traeaki don't pay for their crap :-p
14:53.15MaDMaLKaVit will work on linux/osx too
14:53.49traeakthe dev environ sure, i haven't tried yet but probably should
14:54.13traeakconsidering my partner in crime figured out how to calibrate a camera by displaying an image on an LCD that can be photographed
14:54.37traeakcould probably now almost write an application to be a virtual tape measure
14:54.58traeakif only those dumb phone/tablet guys would NOT put autofocus and image stabilization in their camera units
14:55.27traeakbasically those 2 features destabilize the hardware so it can't be geometrically calibrated
15:07.57*** part/#arm-netbook QingPei (~qingpei@124.64.127.172)
15:08.22*** join/#arm-netbook QingPei1 (~qingpei@124.64.127.172)
15:14.30MaDMaLKaVadb works ok in the linux VM, so probably it is a problem with the windows driver missing some vid/pid combination or something like that
15:26.38gordan_Anyone heard any news when the first EOMA modules might become available?
15:28.17gordan_Given how dismally late R-Pi has been in arriving, it would have been quite sweet to beat them to the punch. Too late now, though...
15:38.04traeakoh the rpi is finally rolling out?
15:55.59gordan_I wouldn't hold my breath for it.
15:56.33gordan_But in theory the people who ordered it 3 months ago might actually start receiving them.
16:02.37traeakit's only been ~7 weeks, not 3 months :-p
16:03.08traeaklast i saw (forget where) they were saying july or so for eoma modules?
16:03.14traeakprobably also very...ambitious
16:05.47mnemocA0Sheds expects to show them on computex.... iirc he is even getting a booth
16:06.10gordan_It was supposed to be out before xmas.
16:06.22gordan_So by that measure it's 3.5 months late
16:06.25A0Shedswe finally have all the connectors
16:06.45A0Shedsit's not a race against PI
16:06.49gordan_That's encouraging.
16:06.53mnemocverified? :)
16:07.15A0ShedsPi has nothing in common with this initiative except that maybe it has an ARM SOC
16:07.36gordan_Yes and no.
16:07.45A0ShedsPi is a stunt by some broadcom guys
16:07.48traeakvery different animals
16:07.51gordan_EOMA is a wider, more future-oriented concept
16:07.56gordan_I get that.
16:08.26gordan_But frankly, the Pi could have done with being obsoleted by a comparably priced device before it actually got into the hands of the users.
16:08.53mnemocA0Sheds: are they already on their to USA for verification?
16:08.57gordan_Not to mention that it would have gathered EOMA the sort of publicity that money can't buy.
16:09.10A0ShedsI'm finishing up the boards
16:09.25A0Shedsa10 and the AMD APU
16:09.43mnemocoh, both going to be produced at the same time?
16:09.44A0Shedsmaybe the Freescale imx6 next
16:09.52A0Shedsyeah asap
16:10.13mnemocawesome
16:10.40A0Shedsnow we have connectors for the next few years, or somebody wants usb3 or display port  :)
16:10.54mnemocwhat fusion cpu are you going to use?
16:11.11A0Shedsthe <10W versions, all pin compatible
16:11.38mnemocmy laptop is an e350, I love it
16:11.46traeakusb3 has to show up on an SOC and it'll be a couple of years before an SOC could even realistically handle the throughput
16:11.52A0Shedswhatever works out at the time of manufacture, maybe even 2 versions 2w single core and 8w 2 core
16:12.15A0Shedsmnemoc, what maodel?
16:12.19A0Shedsmodel
16:12.20traeakdoes the eoma then support hdmi ?
16:12.33MaDMaLKaVok, I finished what I had to do with that tablet. Now, it will take like two weeks to get my ttl adapter to play with the mele
16:12.37traeakor at least dvi i guess
16:12.41MaDMaLKaVwhat would I do in this time?
16:13.04mnemocA0Sheds: thinkpad x120e + amd e-350 (2x800Mhz)
16:13.32mnemocnow idea how much power it eats, but far less than the intel equivalent
16:14.51mnemocI can cross the atlantic coding non-stop on it
16:17.26mnemoccpuinfo_max_freq says it can reach 1.6Ghz on each, but everytime I look it's at 800Mhz
16:17.56A0ShedsAMD just released all the new drivers for the APU's
16:18.07gordan_echo performance > /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor and then look?
16:19.18gordan_I like my AC100. I still get 6+ hours out of it, even though the battery is getting quite old, and I have a high res screen and clocked it up to 140%, plus an RC8 USB SSD internally.
16:19.37mnemocgordan_: it was ondemand... now cpu0 is 1600 and cpu1 still 800
16:19.56gordan_So echo it to the cpu1 node, too. :)
16:20.14mnemocgood point :p
16:20.28mnemocbut well... I don't need to burn the battery faster :)
16:20.49*** join/#arm-netbook eebrah (~Ibrahim_K@212.49.88.34)
16:21.16gordan_I find that ondemand is still a lot slower than performance. Most interractive things are bursty and by the time the load goes up and it clocks up, the work is already done at slower speed.
16:21.20mnemoc2x800 is good enough for my habits
16:21.46A0Shedsimagine that in a eoma68 card
16:22.14gordan_Yeah, but I also like it in a laptop that is thinner than macbook air. :p
16:22.18A0Shedsthat you'll be able to swap into other devices
16:22.40traeaki think you'll have problems with heat
16:22.52mnemochaving cortex-a8 and x86 alternatives since day one will be great for anyone planning to make eoma-devices
16:23.09specinggordan_: theres "conservative"
16:23.11gordan_I'm not sure there's much point.
16:24.03gordan_conservative is good for some workloads such as certain servers but it still has the same issue for desktop use.
16:24.23A0Shedstraeak, what heat problems?
16:24.37gordan_mnemoc: I'm not sure x86 has a place left in markets outside high end servers and gaming rigs.
16:24.55gordan_Shoehorning it into an EOMA seems to just be pretending that's not the case.
16:25.57mnemocgordan_: thin clients, pos, windows based embedded-wannabe controllers are still VERY x86ish
16:26.29specinggordan_: High end server? Google Tilera 100-core chips
16:26.33gordan_Why encourage delaying the inevitable. :)
16:26.42mnemocnot because they couldn't switch to linux/arm, but because they don't want to touch working products
16:27.18gordan_specing: I don't doubt ARM will get into the higher end servers, but on the unparallelizable tasks x86 has a big edge still.
16:27.31gordan_Not everything scales easily horizontally.
16:27.34traeakhttp://en.wikipedia.org/wiki/AMD_Fusion
16:27.37A0Shedsx86 is easy to use for the majority of oems duo to the software they already use
16:27.42mnemocI know a bunch of POS solution providers than don't plan to leave winxp in any nearfuture time frame
16:27.45traeakdoesn't the brazos still require a north bridge for other functions?
16:28.08gordan_There are still a lot of poorly architected, designed and implemented systems that are stuck with vertical scaling.
16:28.09mnemocjust look at the old junky visual basic (not .net) in the language rankings
16:29.28mnemochttp://www.tiobe.com/index.php/content/paperinfo/tpci/index.html <-- visual basic (not .net) is 7th
16:29.36traeakie: e350 is rated at 18W tdp, that's sans ram and north bridge
16:29.55gordan_18W is huge
16:30.17mnemochow much is intel's i5 ? :)
16:30.26traeaki5 is extremely ugly
16:30.34traeakbetter lookign at 870 or something
16:30.39traeak865 i mean
16:30.40traeakwait one
16:30.42gordan_My whole tricked out AC100 sucks about 7W including the 10" TFT panel and a USB SSD.
16:30.51A0Shedshttp://www.amd.com/us/products/notebook/apu/Pages/tablet.aspx#3
16:31.21mnemocA0Sheds: how much heat can you cope with inside a pcmcia module?
16:31.28traeakfor a while there the x130e had an intel 875 as option (like core i3 but with some stuff disabled)
16:31.31gordan_Intel N450 claims to be 7.5W TDP
16:31.59A0Shedsit's not the heat in the module, it's the enclosure of the product
16:32.13gordan_But when I got one, the mobo is sucking down 30W when booting.
16:32.40mnemoctasty
16:32.54A0Shedshttp://www.amd.com/us/products/notebook/apu/mainstream/Pages/mainstream.aspx#6
16:33.05traeaksry celeron 857
16:33.07traeak17W tdp
16:33.17gordan_Granted, some of that is due is due to the mobo manufacturer being incompetent, but 3x the expected TDP is a tad excessive.
16:33.42A0Shedsintel has misleading TDP
16:33.58A0Shedsthe CPU might be 1W bu the chipset 13W
16:34.14traeakdunno about these sandy bridges
16:34.15gordan_I was counting on the peak being 7.5W for the CPU + 2.5W for the SB + negligible overheads for the other handful of components, e.g. NIC chips.
16:34.16traeakregardless
16:34.19A0Sheds+ RAM
16:34.29traeaki would be skeered to put anything mainstream x86 on that eoma
16:34.46gordan_I was intending to run it off PoE, with expected 50% margin for error on the 15.4W class 2 limit.
16:34.53gordan_Imagine my disappointment...
16:35.13A0Shedswelcome to Intel
16:35.24gordan_I'm not convinced AMD is any better.
16:35.28gordan_If anything they are worse.
16:35.35traeakwelcome to x86, all the instruction decoder circuitry
16:35.44A0Shedswould you believe anything Micro$oft tells you?
16:35.49traeakARM is getting thaa way too
16:35.53traeakMIPS64 ftw :-p
16:35.58traeakwhich mips64 though, eh?
16:36.06gordan_MS? I last used their product in anger in around 1995.
16:36.08traeakand who's making one that you can buy that has features you want?
16:36.25gordan_Nu-somethingorother
16:36.48traeak1999 i got sick of dual booting just to play games
16:36.59gordan_That looks like an awesome SoC that they have done sweet fa with for the past 18 months and now it's ready for deprecating
16:37.32gordan_traeak: This is the only reason why I am pondering getting a new x86 i7 box - working hardware PCI passthrough.
16:37.48gordan_Earlier chipsets can't handle it properly.
16:37.58traeakgordan_: high end x86 is friggin *awesome* for high perfromance multi threaded workloads
16:38.27traeakgordan_: compilation included...i do all my work on multi socket xeon systems...although i have access to some quad socket
16:38.48gordan_It'd mean I can just boot Linux, and fire up a Xen VM when I fancy a blast of some Windows-only game that won't play ball with WINE on my display setup (which is a bit unusual because T221 is 2 screens in one).
16:38.55traeakon customer just bought a pair of intel 12/24 cpus (those are crazy expensive)
16:39.06gordan_It's pointless for most tasks, though.
16:39.26gordan_If you designed your system to scale horizontally, lots of small machines is the way forward.
16:39.34gordan_More redundancy, lower overall power.
16:39.37MaDMaLKaVas i don't play, I can very well say bye to x86 right now and start using arm desktops if they were available
16:39.41traeakwell for work stuff we crunch through all that crap you see on google maps/earth microsoft virtual earth/bing, etc
16:39.58traeakso the fact we have as much throughput as we do with single boxes is pretty amazing
16:40.27gordan_I use ARM for everything except: 1) My gaming rig and 2) my NAS, and 2) only because I can't get 13+ SATA ports sensibly on an ARM box yet.
16:40.30traeakwe do with one multi socket xeon more than MS can do with a cluster whole cluster of machiens...but MS doesn't care about efficiency, they brute force everything
16:41.09gordan_I was looking at using SheevaPlugs for SSL offload for big streaming media hosting last year.
16:41.11traeaki would be more than happy with a capable arm desktop, as long as i have my ssh access
16:41.25gordan_It would have done quite well, and dirt cheap compared to the alternatives like F5.
16:42.21gordan_traeak: Try Cubox. It's more a thumb-top than a desk-top, but it's fine as long as you con't lose it down the back of the sofa (or you'll never find it again).
16:42.23traeakprobably the new "medfield" or whatever intel SOCs for portable devices would hopefully be fine with eoma
16:42.35gordan_I use my AC100 for a lot of stuff nowdays.
16:42.36traeakgordan_: i'll take a shot with the mele
16:42.42A0Shedsprobably no Intel + eoma68
16:42.54A0Shedsno coreboot code except for sandybridge
16:43.04traeakbut for a desktop i would need the DSP to work...watch videos, etc
16:43.21A0Shedshttp://blogs.coreboot.org/blog/2012/04/02/google-releases-sandybridge-support-for-coreboot/
16:43.29gordan_I only really switch to x86 for gaming or things where I need silly res (3840x2400 is more than ARM GPUs can handle).
16:44.02gordan_traeak: So nvidia products are out because the quality of their accelerated drivers for Tegra is attrocious.
16:44.12gordan_Which leaves you with very few options.
16:44.18traeaki still really need 64bit regardless...i want to cross compile
16:44.28traeakgordan_: i hope nvidia burns
16:44.30gordan_What's that got to do with 64-bitness?
16:44.46gordan_4GB of RAM not enough for you?
16:44.57traeakvirtual addressing
16:44.58traeakmmap
16:45.13gordan_You don't need that for compiling.
16:45.27traeakanyways it's no matter for now
16:45.29gordan_My build farm for RedSleeve is all Kirkwoods and Tegra2s.
16:45.37gordan_All with 512MB of RAM.
16:45.52gordan_Scale out, not up. :)
16:46.20traeakgordan_: for doing mobile stuff that would be more than fine, just thinking in terms of our current high performance stuff
16:46.32gordan_Like what?
16:46.41gordan_If you exclude gaming what actually remains?
16:47.05gordan_A9 is good enough for most things, even abominations like LibreOffice.
16:47.31traeakgordan_: work stuff is what i mean, all our current code base .... the IO stuff especially is all 64bit optimized
16:47.42traeakgordan_: but NOT x86 optimized
16:48.14gordan_People said the same thing of ZFS, and that's working on ARM now, and it took next to no changes at all.
16:48.36gordan_I might actually see if I can cram a pile of Cubox-es into a 1U case with PoE.
16:48.44traeakhmm...
16:48.46gordan_Would tidy up my build farm massively.
16:49.31MaDMaLKaVgordan, a cubox is twice the price of the mele a1000, is also twice the performance ?
16:49.36A0Shedsi don't anything nvidia making it into anything eoma68 soon  :)
16:49.45traeakone of our products (we sold that one off a few years ago) was to photograph a kitchen, upload the images to a laptop and then use those images to create a program for a CNC machine to cut a granite countertop
16:49.58traeakthat could be retooled to basically cut out the PC now
16:50.18gordan_MaDMaLKaV: Not sure about performance yet, haven't played with my Armadas yet. But yes, I'd expect it to be faster.
16:50.23gordan_It's got double the RAM for a start.
16:50.56MaDMaLKaVI mean: for a farm, it will be better one cubox than two meles ?
16:51.13gordan_I suspect the extra RAM would win.
16:51.19gordan_But I'll reserve judgement on that.
16:51.46gordan_Marvell ARM performance is unpredictable without trying
16:51.55gordan_They aren't just straight ARM cores.
16:52.30gordan_There's still a LOT of hangover from when Intel owned the business and did a lot of proprietary stuff with ARMs.
16:52.54gordan_That's why Kirkwoods are way faster than an ARMv5 should be.
16:53.38gordan_I have a couple of Compulab SBC-A510 boards (ATX lovelyness but damn expensive) that I've not had time to play with for 2 months now.
16:53.57gordan_Once I get around to it I'll let you know how they compare. :)
16:54.54gordan_In the meantime I must have a go at getting recent tbb to build on ARM. Need it for hip-hop because PHP is quite slow on a Kirkwood.
16:55.14gordan_And since my sites run on Kirkwoods, I could do with a 30x speed boost. :)
16:55.45gordan_aways for now
16:55.56MaDMaLKaVgordan, you on the USA? there are some hp thin clients that use same armada soc on ebay that would save you some shipping / customs money compared to cubox
16:56.56MaDMaLKaVtoo late
17:17.51*** part/#arm-netbook QingPei1 (~qingpei@124.64.127.172)
18:45.43*** join/#arm-netbook eebrah (~Ibrahim_K@212.49.88.35)
19:22.43lkclA0Sheds: whey, good to see the ec2 commit!
20:17.06CIA-91rhombus-tech: Bente master * rff2d65009931 /allwinner_a10/orders/a10-preorder.mdwn:
20:20.11mnemoclkcl: hi, now that the connectors are ready and the designs mostly finished, when do you think you'll have an estimated sale price for the alpha modules and production ones?
20:26.11traeakand just as impoortant: how much for the interface boards?
21:41.22CIA-91rhombus-tech: Grant master * r9965f9bec365 /allwinner_a10/orders/chebs.mdwn:
22:08.47flaushyhej. Did you get applications for gsoc openec32?
22:35.43*** join/#arm-netbook mikey_w (~mike@c-71-63-115-202.hsd1.va.comcast.net)
22:36.00*** join/#arm-netbook mike_ (~mike@c-71-63-115-202.hsd1.va.comcast.net)
22:50.58*** join/#arm-netbook mikey_w (~mike@c-71-63-115-202.hsd1.va.comcast.net)
22:51.19*** join/#arm-netbook mike__ (~mike@c-71-63-115-202.hsd1.va.comcast.net)
23:11.06*** join/#arm-netbook uwe__ (~uwe_@dslb-088-066-186-052.pools.arcor-ip.net)
23:15.02*** join/#arm-netbook CIA-158 (~CIA@cia.atheme.org)
23:16.28*** join/#arm-netbook flaushy_ (~flaushy@euve10332.vserver.de)

Generated by irclog2html.pl Modified by Tim Riker to work with infobot.