[02:18] * Martyn swears loudly at the tegra2 board [02:18] this chip has the oddest frigging bottlenecks in performance! [02:20] So, I've got this source that hardcodes "armv4l" in ./configure, and fails to build for any other `uname -m`. Anyone have a good suggestion on the safest way to patch that? [02:20] right now, I'm trying to improve streams performance .. and do you know where I found the bottleneck? [02:21] I can't make memcpy() fast. [02:21] DMA is fast as all getgo [02:21] mem copy? Slow. Slow as dog [02:22] What! [02:23] * persia can't imagine how to implement something that would have that particular combination of performance metrics. [02:24] __aeabi_memcpy4 [02:24] is what is being used .. aligned memcpy ... [02:24] it _should_ be fast. [02:25] I'm going to try to force it to use __aeabi_memcpy instead [02:25] persia : Any ideas? [02:25] None, sorry. [02:25] I don't really understand that low-level stuff. [02:26] There is no NEON on the tegra2, so I know the GNU optimizer won't try for that [02:26] (saving a potential 50 cycle penalty) [02:29] Well .. what I think I'll do is use PLD, ldm/stm about 32bytes to match the cache line size, and keep things aligned [02:29] but it SUCKS that I need to do that. [02:29] write my own memcpy() === dyfet` is now known as dyfet [03:56] hmm, say, those Marvell 1.2GHz ARM CPUs... how do they compare in desktop usage (i.e. firefox page-load, and such) to a 1.6GHz Atom, or to a 1.6GHz Athlon XP? [04:05] You can't really compare directly. [04:06] hmm. [04:06] How would you compare them, then? [04:07] Generally, you want to find some repeatable workload, and then run it on different hardware. [04:07] Depending on the available hardware features, you'll get varying performance. [04:08] But there's a huge set of dependencies on the actual workload. Just being fast at one workload as no relevance whatsoever to being fast at another workload. [04:08] I'd find it interesting to put a radeon on an ARM processor's PCIe slot. [04:08] And lots of hardware has accellerated features for certain workloads or subsets, *especially* in the SoC realm. [04:08] I know folks have done that with some success in the past. [04:09] Speaking of ARM... I keep seeing Flash for Arm on Android... but what about every other ARM system? [04:09] MInd you, it's only interesting if you have workloads easily accelerated with a GPU... [04:09] Flash is available for just about any ARM system. [04:09] Compiz is my biggest GPU-usage thingy on any platform. [04:09] But I don't believe it's free. [04:09] Price, or source? [04:09] Both. [04:10] Dang, that sucks. [04:10] It's very definitely not free-as-in-speech, but I think it's also not free-as-in-beer. [04:10] Consider it an incentive to 1) improve gnash or 2) not use flash formats. [04:11] Right. Flash sucks -- that's not news on any platform. === XorA|gone is now known as XorA === hrw|gone is now known as hrw [10:07] morning [10:23] hey hrw [10:24] XorA: you had a problem to discuss [10:25] hrw: touchscreen with Xorg, I beleive you did some work on making them just work (tm) [10:26] what is a problem? [10:31] hrw: the fact by default Xorg thinks they are touchpads [10:31] never had such problem [10:32] or moment... [10:32] ah you never ran Angstrom without the nasty hack? [10:32] show me modalias of it? [10:32] hrw: I would if I knew how :-) [10:33] XorA: cat `find /sys/class/input/eventX -name modalias` [10:33] "input:*-e0*,3,*a0,1,*18,*" is regexp for touchscreen usually [10:34] input:b0000v0000p0000e0000-e0,1,3,k102,14A,ra0,1,10,11,18,1C,30,32,35,36,mlsfw [10:35] looks like ts for udev [10:35] tslib works with it? [10:36] hrw: yes in Angstrom [10:36] ah lack of tslib in default lucid might not help [10:42] try evdev driver in angstrom first [10:44] well thats a secondary task, I need to get back to what Im supposed to be doing :-) [10:46] tslib is available in ubuntu ... just not in any default install [10:46] ogra: is what I just said [10:46] you should be able to just install it (though it might miss config or patches for your specific board) [10:48] * XorA is always wondering if evdev should just learn about touchscreens [10:49] * ogra tries out the begale netinst image ... [10:49] tslib seems a bit pointless on modern kernels [10:49] XorA, it knows about them [10:49] XorA, but it ignores all that dont use the usbtouchscreen kernel driver [10:49] ogra: which is stupid [10:49] so make your LCD use that driver and it will work [10:49] most touchscreens not being USB [10:50] ogra: no shit [10:50] well, most new ones in the x86 world are [10:50] ogra: I have Atmel devboard which is fine with evdev driver [10:50] x86 is probably >1% of touchscreen market [10:50] and indeed development is focused in these shiny new touchscreen convertible x86 laptops :P [10:50] bug 2.0 will work with evdev driver rather then with tslib one too [10:51] hrw: in OE? [10:51] hrw: so we can steal the magic [10:51] the kernel driver needs to export some specifics (i forgot which) so evdev picks it up [10:51] XorA: at91sam9263ek [10:52] but its very unlikely usb touch will ever take off in arm world [10:52] XorA: basically write xorg.conf part for input device with driver 'evdev' [10:52] hrw: the trick is to do it without an xorg.conf [10:52] hrw: which means udev magic [10:52] XorA: and then check does it work. you can be forced to swap axes (option for evdev driver) [10:52] XorA: first test does it work at all with evdev [10:53] XorA: otherwise you will lose time [10:53] it works as a touchpad [10:53] XorA: ok [10:53] relative not absolute position? [10:53] XorA, well, its just some flags the kernel needs to set ... i'm sure you can make other drivers work with evdev, its just that none but usbtouchscreen export the right stuff yet [10:53] so relative movements and no mouse clicks [10:55] XorA: check drivers/input/touchscreen/ads7846.c - at91sam9263ek uses it [10:55] iirc [10:56] hrw: cool, will do [11:10] * ogra sits in excitement watching d-i installing on the beagle [11:11] C4? [11:12] yep [11:12] http://ports.ubuntu.com/ubuntu-ports/dists/lucid/main/installer-armel/current/images/omap/netboot/omap/ [11:13] starting with that one [11:13] if it works i'll test -server [11:13] -desktop will need another image respin until it works [11:43] ogra: hi! so i't s working better now? [11:43] ndec, it should be installable but hasnt been tested yet [11:44] i'm half way through [11:44] the bootloader stuff only entered the archive on friday and while all bits work they havent been tested inside the installer yet [11:44] ogra: does it support hub connected to OTG? [11:45] hrw, nope, amitk didnt manage to get proper testing from anyone with usable HW before kernel freeze [11:46] we'll surely push an SRU kernel poist release though, if we can get enough tests from users that have OTG cables with the shortened pins that patch might go in [11:46] hrw: do you have OTG HW? [11:47] I don't have access to the special cable nor do I have time to solder one until after release... [11:48] amitk: so far no luck with my cable [11:56] amitk: beagleboard C3 has a jumper near otg to force host anyway [11:56] C4 too ? [11:57] * ogra sees a jumper next to OTG on his C4 [11:57] hi, was anyone successful with kexec on 2.6.28-araneo? [11:59] ogra: that one [11:59] well, i might try it later then [11:59] i cant risk to damage the board before all images are 100% sure to work [11:59] hrw: good to know, now I only need a mini-to-mini cable [12:00] but i have proper mini-> normal adapters to actually do the testing if OTG switching works [12:00] amitk: you need usbA<>usbA female-female adapter [12:01] btw - how many of you go to UDS? [12:01] * lool is considering it [12:02] heh [12:02] * ogra wonders if we'll all be europeans there ... arriving by train or car this time round [12:02] given the flight situation [12:02] hrw: FYI https://wiki.ubuntu.com/UDS-M/Attendees and https://launchpad.net/sprints/uds-m/ have some lists of people too [12:03] I know [12:03] got mail from Alice about it [12:03] hrw: all of us? [12:03] (assuming the volcano is doused by some water) [12:03] amitk, well, take a ferry :P [12:04] ogra: actually not a bad idea. I've got a ferry to Rostok... [12:04] ogra: I hope for plain - 12h in 2 trains is a bit too much [12:05] amitk, yeah, rostock->bruxelles might make a nice train ride ... though all german trains are painfully full atm [12:05] hrw, especially when sitting on your suitcase for the whole ride :) [12:05] ogra: rootstock -> bruxelles!! [12:06] hahaha [12:06] Build a rootfs while taking the train [12:06] i would be feared the hang inpacts the train though :) [12:06] *fearing [12:06] *impacts [12:06] ogra: please... I once returned from London with having to stay still in overloaded night bus for 5h [12:06] sigh, i need new fingers !! [12:07] lool: to build rootfs in train power outlets are needed ;( [12:07] hrw, all german trains have them [12:07] what good is a train w/o power outlets... [12:07] yeah [12:08] some even have WLAN but not all of them yet [12:08] amitk: come to Poland - here trains even have some power outlets but I would not put anything other then multimeter there [12:09] ogra: on ICEs you dont have power everywhere ... just by coincident from my experience on some seats [12:10] asac, all ICEs have power between the seats [12:10] i havent seen one without ever [12:10] i have ;) [12:10] so everybody cross your fingers ... [12:11] beagle d-i install finished without errors ... pray it boots ! [12:13] hmm, it boots but / is mounted ro [12:14] sigh [12:15] ext4 fs errors [12:18] ogra@beagle:~$ lsb_release -a [12:18] No LSB modules are available. [12:18] Distributor ID: Ubuntu [12:18] Description: Ubuntu lucid (development branch) [12:18] Release: 10.04 [12:18] Codename: lucid [12:18] wohoo [12:19] though i had to do two manual fsck's [12:19] not sure thats the kernel or my USB key [12:19] * ogra repeats the install and installs to SD this time [12:20] amitk: sorry to disturb you directly - is kexec on 2.6.28-araneo broken somehow? [12:21] amitk: or - do you know, who is aware of that? [12:22] im trying to get the mcspi4 controller working on my beagleboard using the kernel sources here: https://code.launchpad.net/~beagleboard-kernel/2.6-stable/ but, i dont know which files to edit, as there looks to be many ways to implement spi. should i just set the pinmux (arch/arm/mach-omap2/mux.c), or should i setup spidev (arch/arm/mach-omap2/board-omap3beagle.c)? [12:25] ndec, so http://ports.ubuntu.com/ubuntu-ports/dists/lucid/main/installer-armel/current/images/omap/netboot/omap/ definately works ... (you need a NIC thats supported by the ubuntu kernel, i use an asix here ... create a vfat on SD, put the three files on it and just boot into the installer) [12:25] ogra: thx. this is cool! [12:25] install took about 1h here [12:26] ogra: install on SD ? [12:26] but pulls everything from network so speed varies based on net connection [12:26] my test install was to USB key, but SD is supposed to work too, i'm testing that now [12:26] Sleep_Walker: I am not working on araneo, please talk to AceLan in #ubuntu-kernel [12:26] ogra: where on SD. you need to create a separate partition first? [12:27] ndec, no, the netboot installer lives in ram after boot, you can just wipe the media you just booted from ;) [12:27] it actually offers you the mmc in the partitioner [12:27] ogra: this is great. where do you put the uImage? [12:28] ndec, you mean after install ? [12:28] ogra: yes. so that it finds it after reboot [12:28] ndec, it turned out to be easier to overwrite NAND for now, so initrd and kernel live in NAND [12:28] amitk: thank you! [12:29] ogra: K. i see. so when kernel is updated it goes to NAND as well. is it possible to keep a FAT boot partition for kernel ? [12:29] ndec, we surely need to revisit that for 10.n+ but time was very short to get anything going [12:30] ogra: yes, for sure. we might have boards without NAND as well! [12:30] ndec, as i said we have to come up with something, /boot on fat isnt possible in itself since dpkg tries to create hardlinks during package upgrades for the vmlinuz file, what we will likely do is to have a /boot/uboot partition that we treat like flash ors something similar in later releases [12:31] ndec, or better something thats mounted during flash-kernel phase or some such, i discussed several models with lool last week, for beagle C4 the NAND variant is the best one for now [12:32] s/mounted/mounted temporary/ [12:32] and especially for the short time i had to create an image it was the only achievalbe thing [12:32] *achievable [12:33] all future images can change and there is a lot room for improvement :) [12:44] ogra: ok. we can talk on that after 10.04 [12:45] ndec, yes, we have to [12:46] its on my list for things we will have to improve [12:46] ogra: and we want to as well ;-) [12:46] yeah :) [12:47] * ogra goes for a coffebreak, install to SD is a lot slower [13:43] amitk: I have OTG cables that are known to work [13:45] XorA: I'll push out a kernel with OTG enabled for your testing pleasure (probably tomorrow after some release-related work) [13:46] * XorA has zero omap3 devices with EHCI ports [13:47] XorA: no C2-4 beagleboard even? [13:47] hrw: C3 EHCI is broken :-( [13:47] ok [13:47] hrw: it was working for about an hour, then it just died and works no longer [13:56] welcome in cruel omap3 world [13:56] * hrw -> out === hrw is now known as hrw|gone [14:00] https://wiki.ubuntu.com/ARM/BeagleNetInstall [14:00] everybody: feel free to enhance :) [14:06] I just finished following this with pretty good success (so far) http://elinux.org/BeagleBoardUbuntu [14:07] https://wiki.ubuntu.com/ARM/BeagleNetInstall seems like a better place to have good docs though [14:07] well, one is community built, one is using an official ubuntu image [14:07] I understand that [14:08] https://wiki.ubuntu.com/ARM/BeagleNetInstall is for the image we will release with ubuntu arm this month [14:08] just thinking might be a good place to help with adding to the official docs initially [14:08] cool [14:09] http://elinux.org/BeagleBoardUbuntu is definately more flexible and uses its own kernels which might have additional features over the ubuntu kernel though [14:10] we will also have a netbook and a server installation image [14:10] cool [14:14] I'll be following the arm progress for sure [14:15] my beagleboard needs an OS and I'd prefer to keep with something familiar ;) [14:15] thats why we build the images :) [14:15] indeed [14:20] yay [14:20] so install to SD works as well [14:21] amitk, i see an opps, seems OTG related [14:21] ogra@beagleC4:~$ lsb_release -a [14:21] No LSB modules are available. [14:21] Distributor ID: Ubuntu [14:21] Description: Ubuntu lucid (development branch) [14:21] Release: 10.04 [14:21] Codename: lucid [14:22] :D [14:22] ogra@beagleC4:~$ mount |grep 'on / ' [14:22] /dev/mmcblk0p1 on / type ext4 (rw,errors=remount-ro) [14:22] \o/ [14:24] sweet [14:27] * ogra files bug 566639 [14:27] Launchpad bug 566639 in apt-setup (Ubuntu) "omap install ends up with security.ubuntu.com urls in sources.list after install (affects: 1)" [Undecided,New] https://launchpad.net/bugs/566639 [14:29] plars, GrueMaster ^^^ can you check if thats also the case under imx51/dove ? [14:36] ogra: the imx51 install I did on Friday did not have security.ubuntu.com [14:37] plars, cool, can you comment on the bug that you dont see it on imx51 ? [14:37] amitk, bug 566645 for you [14:37] Launchpad bug 566645 in linux-ti-omap (Ubuntu) "doing a netinstall to SD card results in OTG related oops on first boot (affects: 1)" [Undecided,New] https://launchpad.net/bugs/566645 [14:37] ogra: will do [14:38] thanks :) [14:38] plars, also https://wiki.ubuntu.com/ARM/BeagleNetInstall ... happy installing :) [14:38] ogra: woo! [14:40] * ogra goes to test the server image now [14:40] -netbook still needs a respin with the latest ubiquity === hrw|gone is now known as hrw [14:44] re [14:46] 7Mbps usb ethernet suxx === jldugger is now known as pwnguin [17:10] there an easy way to configure the auto mount to not open a window on the desktop every time it automounts a device? [17:19] prpplague: yes, though there's no user-friendly menu config option for it. Instead, run gconf-editor, and disable apps -> nautilus -> preferences -> media_automount_open [17:19] ugh [17:20] indeed there is a config option for it [17:20] its in the nautilus preferences [17:20] Really? Oh well, do that then :) [17:20] dmart: thanks [17:20] * prpplague tries that [17:20] just open a nautilus window ... edit->preferences ... [17:20] in the devices tab [17:21] ogra: hmm, i don't see a devices tab [17:21] might be called differently, german installation here :) [17:22] prpplague, you are on lucid, right ? [17:22] There's Media -> "Browse media when inserted" which is probably what prpplague wants [17:22] media, yeah [17:22] ahh found it [17:22] I often want to disable automount completely, which you can only do through gconf-editor :/ [17:22] * prpplague tests [17:22] there is also "dont perform any action ..." here [17:22] dmart: yea i hate auto mount [17:22] * ogra loves it [17:23] i rely on it for all my SD stuff [17:23] Depends what you're trying to do :) [17:23] lets you get very lazy though [17:23] and eject doesn't seem to work as expected in ubuntu [17:23] ?? [17:23] it removes the device [17:23] its a kernel feature [17:24] there is a separate unmount function in the context menu if you want to keep the device node of an USB device [17:25] ogra: yea, normally i would expect that if you have /dev/sdb1 and /dev/sdb2 mount, you could issue the command: eject /dev/sdb and it would unmount /dev/sdb1 and /dev/sdb2 and eject the removable device [17:25] ah [17:25] ogra: on ubuntu it fails [17:25] Was eject ever that clever? [17:25] well, there were some upstream kernel changes in that area, not sure thats ubuntu specific [17:25] dmart: works fine on fedora and debian [17:26] ok then [17:26] it definately removes all device nodes if you klick the little eject thingie in nautilus [17:27] have a nice rest of day [17:27] hrw: later === hrw is now known as hrw|gone [17:27] ogra: i'll have to dig into that, possible the syntax has changed [17:27] * ogra goes back to TV to watch snooker champoinship while the beagle server install runs ... watching d-i work is tiring [17:30] ogra: fun fun, thanks for the info [17:30] dmart: you too [17:58] ndec, hmm, smells like there are additional changes in the omapbf package needed [17:58] *omapfb [17:58] your log definately shows that the loading works correctly [18:45] laughs, flash-kernel works a little too great... but the omap kernel works good on my Bx beagle.. [18:58] ok, server install on beagle works fine as well ! [18:58] \o/ [18:59] two happy images already [18:59] sounds great ogra.. ;) [18:59] rcn-ee, what dont you like about flash-kernel ? [19:00] laughs, it's overriding my external kernel.. ;) I'm trying to add a /etc/flash-kernel.conf override shortly after debian-install creates the partition.. ;) [19:00] /etc/flash-kernel.conf isnt used [19:01] But i discovered the omap kernel works great on a Bx, otg/dvi etc.. [19:01] edit /etc/kernel-img.conf or flash-kernel itself [19:01] yeah, that one.. sorry was looking at both... [19:01] or even uninstall flash-kernel [19:02] since you would also have to edit update-initramfs otherwise [19:02] i have it echo "you are using the community kernel, remove this file to use ubuntu's" then it calles exit, which works... [19:02] safest is to just apt-get remove it [19:02] Yeah, that would be the safest.. Except i decided to hack into the NetInstall... ;) [19:03] ah, heh [19:03] ps: i put your wiki link here: http://elinux.org/BeagleBoardUbuntu#NetInstall so you shoudl get more users, i didn't know what else to put other then the http link... [19:04] awesome, thanks ! [19:04] i'll also blog about it once i have all images tested [19:04] though netbook is still building and its my turn with cooking today [19:05] that's going to be a fun week, I'll be in esc so i'm going to miss all the thursday/friday release fun.. [19:05] so i'll call it a day now ... [19:05] * ogra has to cook a trout ... never done that ... will be intresting [19:07] Hey, ogra (or anyone else) .. who made a server armel image? [19:07] where is it hiding? [19:08] (I'm sitting in an office with Rob and Dustin) [19:10] I don't know if he uploaded it yet, he used the NetInstall methiod.. [19:13] http://cdimage.ubuntu.com/ubuntu-server/ports/daily/current/ [19:16] thanks for the link [21:35] I'm new to ARM development. I have a proprietary library (camera driver) that "has EABI version 0" and I am trying to compile it with an "EABI version 4" toolchain. Or at least I think that's what my console is telling me. Can somebody point me in the right direction on what it means and how to resolve this? [21:39] I think its more likely its an OABI binary [21:40] entirely possible - can you give me the short version of what that means? [21:40] how I could determine that? [21:43] I guess what I'm trying to determine is which of the following is true: [21:43] * I can "wrap" the proprietary library somehow and link it against other code. [21:43] * I could install an older version of linux/compilers/etc. and make everything the same version as the library. [21:43] * I'm up a creek unless I can get a different version of the library. [21:50] chrismurf: up the creek is the safest estimation, you can make an OABI chroot, but that depends on ubuntu kernel fully working with OABI [21:51] XorA, what if I went back several versions of ubuntu/debian/whatever? I'm happy to do that if it would fix the problem. [21:51] we had similar issues with GPS chip in OpenMoko and it was always pain [21:51] or does Cortex-A8 mandate the EABI? [21:51] chrismurf: pre armel version of debian should work [21:51] okay [21:51] and I'll just see decreased floating point performance and what have you [21:51] chrismurf: the main issue is no-one is doing it so support rapidly rots [21:51] duly noted [21:52] the evils of proprietary libraries [21:52] sadly sometimes we don't have choices in these matters [21:52] could try sending ubuntu dudes to chat with the vendor [21:52] we just shot the vendor an email [21:52] they've been decent folks in the past, maybe we'll get lucky [21:53] suspect they just haven't had call for a recent ARM version [21:53] Ubuntu is the current buzzword for PHB so just mentioning their stuff doesnt work might be enough [21:54] duly noted ;-) [22:14] re [22:15] After meeting with Dustin and Rob, we've got a good idea now for what blueprints Smooth-stone will be working on for UDS-M === JaMae is now known as JaMa [22:26] anyone run ubuntu on dockstar ? it is an arm chip [22:26] http://ahsoftware.de/dockstar/ [22:32] it says you can install ubuntu .. but not in much details [22:32] http://ahsoftware.de/dockstar/ [22:41] this page says debian does not support seagate freeagent dockstar [22:41] http://www.cyrius.com/debian/kirkwood/sheevaplug/plugs.html [22:41] is that mean ubuntu arm won't compile in here ? [22:53] axisys: Ubuntu Lucid ( and M ) will not work on the sheeva based device [22:54] because sheeva is arm version 5 code, and lucid/M are compiled thumb2 which requires arm version 7 code [22:54] you can, however, use debian if you like\ [23:05] Dang. [23:48] how about karmic ? will sheeva based devices work with karmic ? [23:49] nope karmic is armv6 === XorA is now known as XorA|gone [23:50] looks like jaunty will work [23:50] armv5t [23:54] the sheeva is shipped with Jaunty isn't it [23:55] Kinda. It's shipped with a custom derivative of jaunty. [23:59] tormod: oh ok [23:59] persia: thanks