=== Martyn- is now known as Martyn | ||
bandwidthcrunch | Hi friends, I have a omap3430 based hardware and i want to compile ubuntu for the armv7 architecture. Any pointers on how to go about the same ? | 07:58 |
---|---|---|
Martyn | You don' thave to | 08:10 |
Martyn | Canonical is working on it, it's just going to take a bit | 08:10 |
Martyn | Are you also working with Cortex-A9 hardware? | 08:11 |
Martyn | or just Cortex-A8? | 08:11 |
bandwidthcrunch | cortex-a8 | 08:11 |
Martyn | Right, 3430 ... du | 08:11 |
Martyn | duh ratehr :) I'm tired, it's 2:11am for me | 08:11 |
bandwidthcrunch | ohh | 08:12 |
bandwidthcrunch | catch some sleep | 08:12 |
Martyn | not quite yet | 08:12 |
Martyn | I'm in the middle of debugging a multicore ARM kernel | 08:12 |
bandwidthcrunch | omap4 ? | 08:12 |
Martyn | I'm not at a place where I can checkpoint .. still doing the instruction trace | 08:12 |
Martyn | Cortex-A9 | 08:12 |
Martyn | but not omap4 SoC | 08:13 |
Martyn | I'm using the realview-eb fastmodel | 08:13 |
bandwidthcrunch | is canonical working on that too ? | 08:13 |
Martyn | (since the only real A9 hardware is the PBX, being released in super-limited quantity, based on an eASIC running at 70Mhz) | 08:13 |
Martyn | no | 08:13 |
Martyn | However, they will be. | 08:14 |
Martyn | ogra and I should both be getting PBX'es around the same time | 08:14 |
Martyn | Better question -- why do you want a v7 compile? | 08:14 |
Martyn | v6->v7 won't result in a huge performance increase | 08:14 |
bandwidthcrunch | optimisations ? Isnt it going to make a difference ? | 08:15 |
Martyn | the difference between the omap 2xxx and 3xxx series isn't much | 08:15 |
Martyn | just speed | 08:15 |
Martyn | if you have a specific application you'd like to compile v7, go for it | 08:15 |
Martyn | cross compile it using CodeSourcery's tools | 08:15 |
Martyn | however, you can't just turn on v7 arch and expect magical optimization and speed | 08:16 |
Martyn | glibc needs to have changes made to take advantage of v7 | 08:16 |
bandwidthcrunch | An application like firefox has a lot of dependecies and i would end up cross compiling a lot more of those | 08:16 |
Martyn | yep | 08:16 |
bandwidthcrunch | i used Codesourcery | 08:16 |
Martyn | What optimization level? | 08:17 |
Martyn | -O2? | 08:17 |
bandwidthcrunch | Just that my folks here want me to get the whole of ubuntu build for armv7 . yeah O2 | 08:17 |
Martyn | so do my folks | 08:18 |
Martyn | we're doing an SoC | 08:18 |
Martyn | but recompiling the entire distro isn't in the works for the moment, simply because we lack the platforms to do native v7 compiles | 08:18 |
Martyn | (and cross compiling is really dicey) | 08:18 |
bandwidthcrunch | i supplied those platforms to canonical 2 months back | 08:19 |
Martyn | Ah! Twas you :) | 08:19 |
bandwidthcrunch | yeah | 08:19 |
Martyn | What SoC are the boards based on? | 08:19 |
Martyn | TI's omap? | 08:20 |
Martyn | or the freescale i.mx51? | 08:20 |
bandwidthcrunch | TI's omap 3503 | 08:20 |
Martyn | Well, hopefully we'll see more v7 compiled code in karmic | 08:21 |
bandwidthcrunch | what would be required to get the whole distro built using codesourcery toolchain ? anything we are missing ? | 08:21 |
Martyn | so much in the air... | 08:22 |
Martyn | I wouldn't want to build it in cross | 08:22 |
Martyn | however, I'm the wrong person to ask about the build engine.. | 08:22 |
Martyn | that's ogra, ncommander, etc... | 08:22 |
* Martyn works on ARM linux porting, and device drivers :) | 08:22 | |
bandwidthcrunch | :) cool got that. I am gonna ping them and see if we can do some magic faster | 08:23 |
Martyn | however, I've been doing some pretty close looking at what v7 buys as far as performance goes using profiling ... | 08:24 |
Martyn | it's not much | 08:24 |
Martyn | v7 does make a big difference in the power used though. v7 code takes better advantage of the A8's power saving capabilities | 08:26 |
Martyn | Where you get massive performance increases are in any code that takes advantage of vfp/NEON | 08:27 |
Martyn | as well as video decoding | 08:27 |
Martyn | for the same reason .. since you can pass that on to the coprocessor | 08:28 |
bandwidthcrunch | yeah we plan to use neon too | 08:28 |
bandwidthcrunch | my current processor lacks the DSP | 08:28 |
lool | Martyn: I think there could be a significant difference if we'd start using thum or thumb2 in actual binaries, but this sounds like a large project | 08:29 |
Martyn | nod | 08:29 |
lool | *thum | 08:29 |
lool | Rgah | 08:29 |
Martyn | lool : No kidding | 08:29 |
lool | *Thumb | 08:29 |
Martyn | the thumb compression in the A8 is a nice feature | 08:30 |
Martyn | and I'm discovering that pretty significant changes were made to v7 (for the better) on the A9 | 08:30 |
Martyn | Section "Device" | 08:31 |
Martyn | Identifier"Card0" | 08:31 |
Martyn | Driver"fbdev" | 08:31 |
Martyn | Option"fbdev" "/dev/fb0" | 08:31 |
Martyn | VendorName"Unknown" | 08:31 |
Martyn | BoardName"Unknown" | 08:31 |
Martyn | EndSection | 08:31 |
Martyn | Section "Screen" | 08:31 |
Martyn | ... | 08:31 |
Martyn | Device"Card0" | 08:31 |
Martyn | ... | 08:31 |
Martyn | EndSection | 08:31 |
Martyn | Ack! | 08:31 |
Martyn | sorry | 08:31 |
lool | What changed on A9? | 08:32 |
Martyn | thumbEE | 08:34 |
bandwidthcrunch | Ogra , NCommander Any tips on the build process for armv7 ? | 08:34 |
Martyn | L2 is huge | 08:34 |
Martyn | REALLY huge | 08:34 |
Martyn | like "a full meg" | 08:34 |
Martyn | L1's aren't shabby either | 08:34 |
bandwidthcrunch | nice | 08:34 |
Martyn | but the real treat is the four cores | 08:34 |
Martyn | I have a fully multicore jaunty working now | 08:35 |
Martyn | which is quite the achievement :) | 08:35 |
Martyn | "on the shoulders of giants" | 08:35 |
bandwidthcrunch | great work Martyn .. | 08:37 |
lool | Martyn: Wasn't ThubEE on the A8 already?? | 08:37 |
Martyn | lool : It is, but it has been improved in the A9 | 08:38 |
Martyn | instruction execution time is -way- down | 08:38 |
lool | Ah | 08:38 |
Martyn | in theory, there's no reason these chips won't be able to go 1.5Ghz | 08:39 |
lool | Martyn: Are there interesting patches to merge from your work on getting jaunty to work on MP? :) | 08:39 |
Martyn | all the patches minus two are now part of the arm-linux.org tree | 08:39 |
lool | So only kernel? | 08:39 |
Martyn | (course, they are a critical two ... that have to do with using WFE rather than WFI to halt the other cores during bootloading, then releasing them with an SVC call into the secondary_boot process) | 08:40 |
Martyn | lool : Only kernel. | 08:40 |
Martyn | and a BUTTLOAD of drivers | 08:41 |
Martyn | because ARM is so fucking fractured as an architecture | 08:41 |
lool | arm-linux.org => linux-arm I guess | 08:41 |
Martyn | I swear, we need a unified bootloader (UEFI) and then we need a unified boot process | 08:41 |
Martyn | www.linux-arm.org 's git repository, yes | 08:42 |
Martyn | frankly, there are a bunch of merges that have to happen now between that tree, and the mainline linux kernel tree | 08:42 |
Martyn | right now there are four fractures trees that people are building from .. the omap-linux tree, the arm-offical tree, the beagle-flavored tree, and the mainline linux repo | 08:43 |
Martyn | The //problem// is the ARM gatekeeper for the linux kernel | 08:43 |
Martyn | Russel is slowing things down | 08:43 |
lool | Is the OMAP tree going via the ARM tree? | 08:43 |
Martyn | no | 08:44 |
Martyn | it's got it's own little world | 08:44 |
Martyn | and it's become quite the fork at this point, especially the framebuffer code | 08:44 |
lool | Right, it seems they have a pile of not-easily upstreamable patches | 08:44 |
Martyn | And it's resulting in this kind of mess: http://osdir.com/ml/linux-kernel/2009-06/msg05879.html | 08:45 |
Martyn | I mean, you don't see the x86 side of the kernel doing this kind of insanity | 08:45 |
Martyn | nor the MIPS guys, really | 08:45 |
Martyn | but ARM? We've got the worst part of the Linux source tree .. our branches look like a teenager's room after an all-week party kegger | 08:46 |
lool | Having per subarch or even per board trees is not really nice indeed | 08:48 |
Martyn | sub arch is acceptable | 08:49 |
Martyn | <board> is ... crazy | 08:49 |
Martyn | it's partially ARM's fault | 08:50 |
Martyn | and partially the phoneset manufacturer's fault (odd bootloaders, TrustZone, etc) | 08:50 |
Martyn | but mostly the Linux Maintainer's fault | 08:50 |
lool | bandwidthcrunch: Hey, who are you? :) | 10:11 |
bandwidthcrunch | Hi lool | 10:12 |
bandwidthcrunch | Manav Gautama | 10:12 |
ogra | bandwidthcrunch, i doubt using codesourcery to cross-build *the whole distro* is easily possible without a year of building infrastructure :) | 10:12 |
ogra | we build the whole system natively | 10:13 |
lool | bandwidthcrunch: I'm working on a tool to rebuild the distro with a different toolchain | 10:13 |
bandwidthcrunch | Willing to chuck codesourcery | 10:13 |
lool | It wouldn't be crossbuilt though, but natively built within qemu | 10:13 |
Le1 | Is it a buildd or obs like system? | 10:14 |
ogra | launchpad buildds | 10:14 |
bandwidthcrunch | That should also be ok. I was interested in having armv7 using qemu or building them natively | 10:14 |
bandwidthcrunch | have some omap3 board | 10:14 |
bandwidthcrunch | also shipped one to canonical | 10:15 |
lool | bandwidthcrunch: Who did you ship it to? | 10:15 |
bandwidthcrunch | give me a minute | 10:15 |
lool | Le1: obs? | 10:15 |
ogra | lool, suse | 10:15 |
lool | Oh | 10:15 |
Le1 | opensuse build service | 10:16 |
lool | Ubuntu uses Launchpad-driven buildds; my system will use EC2 running Debian-style buildd tools | 10:16 |
ogra | Le1, https://launchpad.net/builders | 10:16 |
ogra | (scroll down to armel, they are integrated and used natively as any other arch buildd) | 10:17 |
lool | That's how we do it in for Ubuntu itself | 10:18 |
Le1 | i see. | 10:20 |
Le1 | I've thought it's a local buildd. :P | 10:20 |
bandwidthcrunch | lool , shipped to to kress | 10:22 |
lool | bandwidthcrunch: If this is pubic information, could you share details about what board this is? | 10:24 |
lool | *public | 10:24 |
bandwidthcrunch | its not yet public but its been a while i heard anything. So am trying to setup a build out here and see if we can do things faster | 10:26 |
lool | bandwidthcrunch: I think you should talk to Mike Kress and check directly with him | 10:26 |
bandwidthcrunch | I just got his address and seeing if i can get any more information | 10:27 |
lool | *I* personally neverheard of your project/board, so if it's not public we can't really discuss anything; we're a larger company now, so it's probably other people are working on it | 10:27 |
=== Le1 is now known as eggonlea | ||
bandwidthcrunch | but at the same side i would also like to start builds here at my end . I understand lool. Am just checking up with anyone else is working on a armv7 port | 10:28 |
ogra | the whole 9.10 release will be armv6+ | 10:28 |
lool | bandwidthcrunch: The main Ubuntu armel port is going to move to ARMv6 + VFP when our buildds are moved to this | 10:28 |
bandwidthcrunch | mojo ? | 10:28 |
ogra | not sure thats sufficient for you, but if it is you wont need to rebuild the whole of it | 10:28 |
lool | bandwidthcrunch: What with mojo? | 10:29 |
bandwidthcrunch | isnt that armv6 +VFP ? | 10:29 |
bandwidthcrunch | i already have a rootfs based off it and it looks pretty decent. | 10:29 |
ogra | they stopped a year ago, didnt they ? | 10:29 |
lool | bandwidthcrunch: mojo provides various optimized *arm* archives already (Note: not "armel"); if that works for you, that's ok; note that you wont be compatible with the Debian and Ubuntu armel ports | 10:31 |
ogra | for rolling a 9.04 rootfs (which is armv5 though) just see the topic :) | 10:31 |
bandwidthcrunch | Actually i have both jaunty and mojo working for me . | 10:31 |
bandwidthcrunch | Just that i wanted to have armv7 port and source control of the whole rootfs and be able to build the same natively or in qemu | 10:33 |
lool | ogra: Did you rename the project holding the rootfs builder? :-) | 10:34 |
ogra | lool's work in karmic should help with that | 10:34 |
ogra | lool, not yet | 10:34 |
lool | ogra: Did you pick a name? | 10:35 |
ogra | no, then i would have asked for renaming already | 10:35 |
bandwidthcrunch | these launchpad builders armel builds happen on canonical servers ? any way i can throw a build there or setup the same infrastructure at my end ? | 10:38 |
lool | bandwidthcrunch: The official Ubuntu buildds are administered/hosted by Canonical; they are using Debianish tools + Launchpad | 10:40 |
ogra | with lool's tool you will be able to | 10:40 |
bandwidthcrunch | where can i locate the tools ? | 10:41 |
lool | bandwidthcrunch: So you can't just throw your buildd; to setup your own buildd, you should check the Debian tools, but that's not enough to rebuild the distro | 10:41 |
lool | bandwidthcrunch: Start reading up from sbuild | 10:41 |
lool | (The package) | 10:41 |
lool | It should help you setup dchroot and the like | 10:41 |
ogra | lool, how about "rootstock" for the project ? | 10:41 |
ogra | not very original, i know and less funny than your suggestions ... but its a name | 10:42 |
lool | ogra: It's a decent name | 10:43 |
ogra | right, i'll take that one then | 10:43 |
* ogra waits for an answer in #launchpad ... | 10:45 | |
lool | You probably want to file it as a question, but well | 10:46 |
ogra | i wanted to ask first | 10:49 |
ogra | https://answers.launchpad.net/ubuntu-arm-roofs-builder/+question/74975 | 10:49 |
ogra | *sigh* why do i get all that cyrillic spam ... i cant even read it | 10:51 |
bandwidthcrunch | thanks lool, ogra , will look up the tools you have mentioned and see if i can cook up something tho it sounds to be quite a task | 10:58 |
rek_ | heloooo | 17:42 |
rek_ | where can i download ubuntu arm and ubnt arm packages? | 17:43 |
ogra | http://ports.ubuntu.com/ has all armel packages | 17:44 |
rek_ | arm ? | 17:45 |
rek_ | can i install them with smart q5 ? | 17:45 |
ogra | if you have a kernel and bootloader 9.04 should work ... see the topic | 17:47 |
rek_ | what? | 17:47 |
ogra | "Build a rootfs from scratch: https://wiki.ubuntu.com/ARM/RootfsFromScratch" | 17:47 |
rek_ | armel ihow can i port ubuntu into my device? | 17:48 |
ogra | you need a kernel and bootloader for your device, with the script from the wikipage you can build an ubuntu rootfs for it | 17:48 |
rek_ | is not simle | 17:49 |
rek_ | simple* | 17:49 |
rek_ | r u able to do dis ? | 17:49 |
ogra | well, i wrote the script ... | 17:50 |
ogra | so if i would have such a device, a kernel and a bootloader i surely could do it | 17:50 |
rek_ | really?? | 17:50 |
rek_ | amzing....baffling you're the man... | 17:50 |
Martyn | rek_ I'ts not hard, really. | 17:51 |
rek_ | i should read | 17:51 |
Martyn | You have to pay attention to detail, but the building of a rootFS is straightforward | 17:51 |
Martyn | ogra : BTW .. looks like I might have to replicate the build system here in house in the next couple months. | 17:52 |
Martyn | ogra : The interest in a v7 build is increasing, and I've been put on UEFI duty | 17:52 |
rek_ | however my device i'm gonna buy has alread installed a linux distro... i think it's arm ubuntu... they said so... so i'm worried about how could i install new programs... how could i run a new os ecc | 17:52 |
rek_ | it's smart q5 | 17:53 |
* ogra points Martyn to https://blueprints.launchpad.net/ubuntu/+spec/mobile-karmic-arm-cloud-builds | 17:53 | |
rek_ | so ogra is a genius ? | 17:53 |
ogra | Martyn, though i doubt v7 will give you much more than v6+vfp and thats what karmic will default to anyway | 17:53 |
Martyn | ogra : I need NEON support | 17:54 |
rek_ | the processor is arm11 ... | 17:54 |
Martyn | ogra : amongst other things .. however for me it's a Mandate From Above(tm) | 17:54 |
ogra | isnt that a matter of kernel support and drivers ? | 17:54 |
rek_ | is this good? | 17:54 |
ogra | first of all at least ... | 17:54 |
Martyn | Also application support, and apps have to be compiled with support for the NEON extensions | 17:55 |
rek_ | what? | 17:55 |
ogra | ah, right ... | 17:55 |
ogra | Martyn, so you will try NEON on the babbage ?? | 17:55 |
Martyn | ogra : So it's a big, hairy ball of a mess | 17:55 |
rek_ | ogra is it simple to add spplications? | 17:55 |
Martyn | ogra : no, I'm going to use NEON on our chip once it tapes out | 17:56 |
Martyn | and on the PBX | 17:56 |
Martyn | which just arrived in a nice big box. The PBX is -huge- man, and s l o w | 17:56 |
ogra | rek_, if there is already ubuntu on the device that should be straightforward, yes | 17:56 |
Martyn | but dual core, implemented on an eASIC | 17:56 |
rek_ | but | 17:56 |
Martyn | no buts :) | 17:56 |
rek_ | but.... it's arm i think | 17:56 |
Martyn | so? | 17:56 |
rek_ | which packages do i need? | 17:56 |
Martyn | the pools are all compiled | 17:56 |
Martyn | just install the rootfs, in it's entirely | 17:56 |
rek_ | no butts lol | 17:57 |
Martyn | you'll have a base rootFS (no graphics) then you'll have to configure and install xorg-core | 17:57 |
Martyn | etc. | 17:57 |
rek_ | i don't have to...may be if i change os | 17:57 |
Martyn | OR .. you can run the version of Intrepid that the smart Q5 comes with, and /not worry about it/ | 17:57 |
ogra | Martyn, the cloud spec should make it easy to rebuild parts or the whole archive | 17:57 |
rek_ | ninfact... | 17:57 |
Martyn | in fact, after poking around and doing a bit of research, it seems that the Smart Q5 may even have Jaunty o it | 17:57 |
Martyn | Jaunty on it | 17:58 |
rek_ | do u know smart q5 ? | 17:58 |
Martyn | ogra : I agree. Looks good. | 17:58 |
Martyn | ogra : What's the timeframe for working on the tool? | 17:58 |
rek_ | is the touch good? | 17:58 |
ogra | Martyn, ask lool, he implements it | 17:58 |
Martyn | rek_ : There aren't a lot of ARM platforms that I haven't played with at this point. | 17:58 |
rek_ | what? | 17:58 |
rek_ | the processor is arm11 samsung 667mhz what coul i install? | 17:59 |
Martyn | ogra : Will do. | 17:59 |
Martyn | rek_ : Anything you want | 17:59 |
rek_ | lol :-) what? | 17:59 |
rek_ | i think only some distro for pda | 17:59 |
Martyn | anything you want. If it's in the armel software pools, you can aptitude install <xxx> whatever you want | 17:59 |
Martyn | the thing runs ubuntu perfectly. | 18:00 |
rek_ | what's armel ? | 18:00 |
Martyn | rek_ : *groan* Sir, you need to do a lot more reading please. | 18:00 |
rek_ | i know but could u tell me? | 18:00 |
rek_ | arm is armel? | 18:01 |
Martyn | rek_ : You have a rather large gap in your knowledge, and this is mostly a developer-related channel as opposed to a user-install channel. | 18:01 |
rek_ | user-install chan lol :-) | 18:01 |
ogra | there are different flavours of ARM CPU ... google for the difference between big endian and little endian :) | 18:01 |
rek_ | i'm sorry... | 18:01 |
Martyn | not that I want to put you off, please understand, but I don't want to spend a great deal of time educating :) | 18:01 |
ogra | armel means the packages are built for ARM with little endian | 18:01 |
ogra | (arm-endian-little = armel) | 18:02 |
rek_ | i like to being teached if it's correct | 18:02 |
rek_ | little endian? | 18:03 |
ogra | http://en.wikipedia.org/wiki/Endianness | 18:03 |
Martyn | rek_ : wikipedia is your friend | 18:03 |
rek_ | i'd like to buy also the axim x30 ... but i want linux there thx fr da link | 18:04 |
rek_ | but i know the port is not completed | 18:05 |
* ogra calls it a day (late here) | 18:06 | |
rek_ | what do you call? | 18:06 |
ogra | rek_, it means "gooybye, i'm off for today" | 18:07 |
rek_ | really? | 18:07 |
rek_ | means u ar tired ? | 18:07 |
ogra | yes, i'm working since 11h | 18:08 |
rek_ | working on ? | 18:08 |
ogra | ubuntu ... | 18:08 |
rek_ | can i find a iso image of arm ubuntu? | 18:08 |
* ogra is really gone now, before the shops close | 18:08 | |
rek_ | really are u a developer? | 18:08 |
rek_ | ah what's the time ? | 18:08 |
Martyn | rek : he's in Germany | 18:09 |
Martyn | i'm in the US (texas) | 18:09 |
rek_ | 19.09 here | 18:09 |
Martyn | people are spread around | 18:09 |
Martyn | 12:09 here | 18:09 |
Martyn | rek : I too am going offline for a bit, for lunch | 18:09 |
Martyn | and yes, ogra's an ubuntu developer | 18:09 |
rek_ | it's my fault maybe | 18:09 |
Martyn | so are lool, amitk, ncommander, armin76 .. etc.. | 18:09 |
rek_ | congratulation ogra | 18:09 |
Martyn | I'm a linux kernel developer, and work in the x86 and ARM tree | 18:10 |
rek_ | i'd like to ask a lot of things but.... | 18:10 |
rek_ | good | 18:10 |
rek_ | how old are u? | 18:10 |
Martyn | not associated with any particular distro, but more Debian/Ubuntu of late because the company I'm with has connections to Canonical | 18:10 |
Martyn | Now that's a bit of a personal question, don't you think? | 18:10 |
Martyn | i'm in my late 30's | 18:11 |
rek_ | canonical? yes i think so... | 18:11 |
Martyn | And it's not a question relevant to ubuntu development. | 18:11 |
rek_ | i'm 19-20 | 18:11 |
Martyn | Very good. Well, please pardon me .. but I do have to go as well. Lunch awaits. | 18:11 |
rek_ | yes because... it's smthg difficult so you cannot be there at an early age ain't it ? | 18:11 |
rek_ | ok have a nice dinne | 18:12 |
rek_ | r | 18:12 |
lool | Martyn: bon appétit | 18:12 |
lool | Martyn: Re: the rebuild tool: you might have seen some efforts in building a good qemu setup; that's part of this project | 18:13 |
rek_ | buon appetito | 18:13 |
Martyn | lool : awesome | 18:13 |
lool | I'm almost done with that part, I'd like to rebase on a .30 and use highmem | 18:13 |
armin76 | rofl | 18:17 |
armin76 | Martyn: i'm not, i'm a gentoo one | 18:17 |
rek_ | what sould i do to install zydas zd1211 on damn small ? | 18:18 |
NCommander | Anyone got any idea what this does or why it might be necessary? http://paste.ubuntu.com/201603/ | 19:12 |
NCommander | (from thunderbird 3.0 source, no comment explaining it) | 19:12 |
Martyn | re | 19:28 |
rek_ | ciao | 19:29 |
rek_ | hello | 19:30 |
Martyn | re | 19:34 |
Martyn | Back from lunch | 19:34 |
Martyn | armin76 : Pardon me :) I even _knew_ you were into gentoo... | 19:34 |
Martyn | Slipped my mind :) | 19:34 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!