| === gletelli__ is now known as gletelli | ||
| === rwhitby is now known as rwhitby-n800 | ||
| === rwhitby is now known as rwhitby-n800 | ||
| === rwhitby-n800 is now known as rwhitby | ||
| armin76 | Martyn: pegatron inventec :) | 15:42 |
|---|---|---|
| Martyn | armin76 : Oh hell yea | 15:52 |
| Martyn | armin76 : I have a pegatron netbook on the way to the office. | 15:52 |
| Martyn | armin76 : I have to wait to get my hands on the inventec one though. :( :( | 15:52 |
| Martyn | and on the PXA FPGA based Cortex-A9 from ARM | 15:52 |
| armin76 | gimme one :D | 15:52 |
| Martyn | (which runs at a pokey 70mhz .. but at least is clock-cycle-accurate) | 15:52 |
| Martyn | no. I had to sign >two< NDA's to get the damned pegatron | 15:53 |
| armin76 | yes! | 15:53 |
| Martyn | It's all prerelease hardware too .. so I'm sure it's buggy as all hell (like the Babbage -wasn't-) | 15:53 |
| Martyn | *mrrpf* | 15:53 |
| Martyn | In any case, I've nearly finished adjusting 9.04 for Beagle in a form that might be releaseable | 15:53 |
| Martyn | what I'd like is an armv6 recompile of the the whole tree though | 15:54 |
| Martyn | power consumption is LUDICRIOUSLY high with everything compiled v5 | 15:54 |
| armin76 | use gentoo :) | 15:54 |
| persia | I hear that karmic will be ARMv6 compiled, which may help (but isn't now) | 15:54 |
| Martyn | we'll never convince anyone that 'ubuntu is better than Android' if we can't get a better optimized ARM dist working on the pegatron | 15:55 |
| Martyn | my boss has us doing oprofile() and power work now | 15:55 |
| Martyn | and the news with oprofile() is bad .. it spends far too little time in the idle_pm state | 15:55 |
| ogra | you can run android apps on ubuntu | 15:56 |
| persia | Surely that's not just from the compiler flags. | 15:56 |
| ogra | doesnt that make it classes better already ? :) | 15:56 |
| Martyn | not if your netbook will only run for 2 hours | 15:56 |
| suihkulokki | Martyn: umm.. and you think _that_ will get fixed by fiddlying compiler flags? | 15:56 |
| Martyn | suihkulokki : with proper flags in place? Yes. when you use idle_pm rather than kernel_idle, fully a /third/ of the calls turn into the lower power idle call | 15:57 |
| persia | suihkulokki, I hear that there was a 5% improvement on the Atom processors from flags: it might get an extra 6 minutes :) | 15:57 |
| Martyn | suihkulokki : It's one of the few times that putting in v6 support is a good thing | 15:57 |
| Martyn | persia : This is our beloved ARM .. atom's just don't "get" real power management | 15:58 |
| ogra | you mean 5% will make up 12min then ? | 15:58 |
| persia | ogra, 120 X 0.05 = 6 | 15:58 |
| Martyn | suihkulokki, persia : I'd be glad to show you the oprofile() work and power reductions once the power is complete... | 15:59 |
| persia | Martyn, Ah. I see. | 15:59 |
| Martyn | power study is complete rather | 15:59 |
| ogra | persia, ( 120 X 0.05 )^ARM = 12 | 15:59 |
| ogra | ;) | 15:59 |
| persia | Well, no. If there's that big a difference between idle_pm vs. kernel_idle, then it could be lots more. | 16:00 |
| Martyn | So, this is more like getting 3 hours instead of 2, without doing any hard work... | 16:00 |
| ogra | to be honest i have to see any improvements on power first ... as long as they put the same peripherial stuff around arm as you have on ATOM i really dont see where the saving factor will be | 16:00 |
| Martyn | I think the best thing to do is wait till we finish the study, and then read the paper and see the differences as measured in our lab... | 16:01 |
| Martyn | ogra : The difference is that ARM processors really do power down lots of pieces around the core when you engage the power management features | 16:01 |
| ogra | yeah yeah | 16:02 |
| Martyn | ATOM's dont :) | 16:02 |
| ogra | you have the babbage1 | 16:02 |
| Martyn | or I should say, they don't do it properly. | 16:02 |
| Martyn | Yeah. | 16:02 |
| Martyn | i.mx51 is a dissapointment | 16:02 |
| Martyn | I'm getting better power numbers out of the OMAP3 than I am from the i.mx51 | 16:02 |
| ogra | if the "pieces around" are all connected through unmanageable USB, it doesnt really help | 16:02 |
| persia | Yes, but USB *can* be managed, if it's the right part. | 16:03 |
| ogra | right, and babbage2 improves here | 16:03 |
| persia | Just a matter of good systems integration (mind you, there exists manageable USB for Atom too, although my Atom devices all leave USB powered on all the time (even when sleeping). | 16:04 |
| Martyn | ogra : Have you gotten the babbage2 yet? | 16:04 |
| ogra | yup | 16:04 |
| ogra | sits on my desk | 16:04 |
| Martyn | ogra : I'm skipping babbage2 in favor of the real pegatron hardware ATM | 16:04 |
| Martyn | (i.e. actual netbook) | 16:04 |
| ogra | well, that should be identical | 16:04 |
| Martyn | so I'm told. | 16:05 |
| ogra | i have the guts of the netbook too | 16:05 |
| ogra | but didnt even boot it yet | 16:05 |
| Martyn | ogra : How's the stability of the peripherals? especially ethernet? | 16:05 |
| Martyn | mine doesn't arrive until next Monday | 16:05 |
| ogra | seems ok, i ran the B2 for the last week or so without issues | 16:05 |
| ogra | though we dont have a kernel yet, i'm still using the shipped one | 16:06 |
| ogra | ubuntu kernel will be ready by alpha3 | 16:06 |
| Martyn | excellent | 16:07 |
| Martyn | I will be submitting patches for Cortex-A9 into the mainline kernel in the next couple weeks | 16:07 |
| Martyn | mostly surrounding getting SMP up and going. There are issues in using WFINE in the bootloader to hold off the other cores | 16:07 |
| Martyn | I'd like to move that responsibility totally into the kernel, and not have this bootloader state engine dependency | 16:07 |
| ogra | cool | 16:08 |
| Martyn | okay, catch you all on the flipside... I'm headed to the office | 16:09 |
| Martyn | back | 16:25 |
| Martyn | Okay, I talked to Boss-Man and made noises about getting our hands on a babbage2 | 16:25 |
| Martyn | no point on having me hack on hardware different than is available to Canonical, right? | 16:25 |
| persia | Martyn, Well, if you find that there's some gross incompatibility with other hardware, that becomes interesting, for certain classes of other hardware. | 16:26 |
| persia | Or at least, increases the user base. There's lots of people doing interesting things with Jaunty who have "unsupported" hardware. | 16:26 |
| ogra | well, the pegatron should be identical, i think the B2 was produced in a very limited amount | 16:26 |
| Martyn | limited, but ARM partners should have 'em | 16:28 |
| Martyn | b1 was also quite limited | 16:28 |
| Martyn | enough that if I accidentally break or tear the debug dongle cable, I'm screwed :) | 16:28 |
| cbrake | I get cat: /etc/timezone: No such file or directory when running build-arm-rootfs | 17:56 |
| cbrake | what is the best solution? | 17:56 |
| Martyn | Grrr... getting Babbage to boot from flash image is not fun | 21:39 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!