The £5,000 Challenge

[WizardStan]

I'm wondering if that much cash would get the kernel updated to 2.6.37

I would throw so much money at the person who did that!

is there an hibernate/sleep mode on pandora? if yes, which quantity of battery consumption does it use?


how long it can stay in this mode before the battery needs to be charged again?

There's no actual hibernate yet, just a low power mode which turns down the CPU and shuts off a bunch of the hardware. A full battery will last for about 30 hours in this low power mode. With proper tweaks, we should theoretically be able to get something that lasts days on a full battery, which is what I think any sort of large money challenge should go to.

 

[tsh]

I would throw so much money at the person who did that!

It would be interesting to see an educated guess of the amount of work involved. Probably a fair bit of manual merging of patches, but there might not be a lot that is rocket science.

 

[Dead1nside]

Maybe we should be holding coding competitions, including much needed software optimisation and upgrades, to within the community, so that it encourages future contributions rather than one-offs. Maybe that also gives incentive to long standing contributors to the OS, and the software ecosystem too, to take part, and feel like they're getting something other than just our eternal praise and appreciation.

 

[urjaman]

I think that for £5000 one could get a kernel developer (you'd get me, thats for sure) to work full-time on the pandora for atleast 2 months. One could get a huge amount of work done compared to the current speed

 

[myownlittlworld]

Maybe we should be holding coding competitions, including much needed software optimisation and upgrades, to within the community, so that it encourages future contributions rather than one-offs. Maybe that also gives incentive to long standing contributors to the OS, and the software ecosystem too, to take part, and feel like they're getting something other than just our eternal praise and appreciation.

That's a good point. We could certainly setup coding sprints, like the PyPy project, which could even be used toward answering any development bounties that people post.


To hijack my own thread, does anybody feel like organizing a coding sprint? You'd need to organize dates, locations, lodging, and sprint targets (what goal you're actually hacking toward). We might even be able to use the development fund to help people with transportation or lodging costs. Interestingly, the OPT is already doing sprints, they're just the (unending) weekly assembly build-athons. Sprints are also a great place for people new to software development to get into the game.


As I mentioned in the original post, I'm really not worried about the Pandora's cold boot times myself, I just thought I'd post the link for anybody who might be interested. Furthermore, both the dev fund (don't have a link for that), development bounties, and the Pandora donations page are probably better places to throw your money around.


----

> I'm wondering if that much cash would get the kernel updated to 2.6.37


I would throw so much money at the person who did that!


> is there an hibernate/sleep mode on pandora? if yes, which quantity of battery consumption does it use?


> how long it can stay in this mode before the battery needs to be charged again?


There's no actual hibernate yet, just a low power mode which turns down the CPU and shuts off a bunch of the hardware. A full battery will last for about 30 hours in this low power mode. With proper tweaks, we should theoretically be able to get something that lasts days on a full battery, which is what I think any sort of large money challenge should go to.

That sounds like a pair of bounties. Done.

 

[Norfen]

One could get a huge amount of work done compared to the current speed

Which is why ED decided to stop with the wideo a day and turn his attention elsewhere. I still think the community should get on with the kernel, and with more people with Pandoras to do so, I think we'll see some good progress.

 

[doublebeta]

I made a post somewhere else, but ...erm... I don't know where the hell it's gone! Was there a forum rollback? It's not in my 'posts made' thing either.


Basically...


Their demos (at least the one where it says so) are using NOR flash!


What does this mean?


NOR Flash, unlike the NAND flash used in the Pandora, is VERY VERY fast to read, but *really* slow to write to. You really would not want to use it even for just saving casual, sub 100k docs. And i'd imagine using THunar would be very unpleasant (it caches thumbnails to NAND, unless this issue has been addressed). Plus, there's lots of batch1 out already, and as ED has said before, we can't change the hardware now or it'll piss them off (maybe not in the same words )


GBA Flash carts use NOR memory because the code is run straight from the cart due to there being little RAM (LOL I automatically typed 'crap'). NAND memory is too slow and games can't possibly run from it, without wasting a lot of cycles in wait states.


What advantage will be gained by having them 'have a look at'(i.e. fuck with!) with our kernel/OS?


We'll gain maybe a second off boot time, if (they actually do) anything. But we'll no doubt drop functionality with that!


And there's &5000 gone - for probably little more than a second?


Anyone who's actually going to try and optimise the monster that is the Linux kernel (graceful (er, comparatively - and I could do no better!) at run time, lesser so at compile time), deserves my best D:/ :blink:/


From somewhere in the BOFH articles ( © Simon Travaglia),


Something like this:


Drop the clock to 1Mhz, slowly wind it back up each week as 'upgrades'.


"With our advanced analysis techniques and experience in Linux tuning, boot times can be reduced in some cases to less than a second." - sounds like BS to me. Hey, I don't mind being proved wrong, but this sounds pretty BOFH.


And look at their flowchart with timings thing - check out 'filesystem mount'. Fishy.


I don't mean this post as a kind of "RTFM, STFW" post, or as any stab in any way, but, really consider if this place is legit. Prove me wrong, I wanna see fast boot times. But I have reasons to believe it's bullshit and we'll get nowhere . Maybe we should review any unneeded kernel stuff ourselves? Or, port faster, smaller OSes, like maybe Haiku - i'd love to work on Haiku but lack the incentive, maybe there are others. Pandora might be the key.


We could look at things like, is the issue in CPU doing all its init stuff, or disk(most likely). If it's disk, maybe we can (um, shit I don't know much about arm) run some stuff in Thumb? I don't know whether we can have kernel mode in thumb while userspace runs x without too much overhead.


Look out for any testimonials. I did some quick google searches and found none.


Good Luck, it's a good idea. Don't give up because I said this, just maybe avoid this lot .


Edit: Btw >1 second boot times are undesirable

 

[mvickers03]

They need to sort out the actual suspend feature before we worry about boot times. My battery keeps going flat because it turns itself back on, seemingly at random (but only when it's in my bag or my pocket). The number of times I've taken it out of my bag after four or five hours only to find I'm at 10% battery is getting boring now.


D.

Dunny, I have seen this issue come up before it's because the battery is jolting inside, if you connect the battery while Pandora is off, it switches on. It can easily be fixed by putting a peice of folded up tissue under the battery cover

 

[Wootson]

No matter how long you are going to have to wait, after a while the new situation will also feel like an era of pain.


For example:


Just like the trains they are making faster. Today they go from Amsterdam to Paris in two hours, and next year in only one hour. Thing is, it's still an hour. AN HOUR I TELL YOU! It's better to invest half the costs into making the two hour ride more pleasurable, than to invest twice as much into making the trip faster. Because people will always complain about travel times, no matter what.


The very same thing goes for the booting time of the Pandora. Now it's 30 seconds. And with the investment of €5000+ it will be halved. Thing is, it will still be 15 seconds.


15 SECONDS I TELL YOU!


I would rather pay someone €5000+ to create a random-intro-generator with funky graphics and sounds.

 

[WizardStan]

I made a post somewhere else, but ...erm... I don't know where the hell it's gone! Was there a forum rollback? It's not in my 'posts made' thing either.

You posted on the gp32x forum.

Dunny, I have seen this issue come up before it's because the battery is jolting inside, if you connect the battery while Pandora is off, it switches on. It can easily be fixed by putting a peice of folded up tissue under the battery cover

That wouldn't explain why it keeps starting BattleJewels.

 

[doublebeta]

You posted on the gp32x forum.

Yeah, just realised. Oops. Ah well, I could do with a silly little mistake. hah .


I have a tendency to do dumb stuff (and get corrected very nicely) here now don't I? Thankyou

 

[moxie]

That wouldn't explain why it keeps starting BattleJewels.

But that one is quite easy, isn't it? He has the icon for Battlejewels on the desktop (or on the first screen of minimenu, whatever he uses) and either one of the buttons stick up long enough to touch the touchscreen at the right place in a bump, or the bump is enough to jolt the right nub into a doublecklick position (i.e. straight up) for a moment.


...would be my guess, anyway.

 

[WizardStan]

Too many moving parts in that scenario for me to make it first consideration. You're suggesting that the battery is being jostled, causing it to reset, and the 'b' and 'enter' buttons are being pressed sometime after it was jostled without jostling it again.


More likely it's something bumping the power switch. The 'b' button can still be bumped, but it's independent of when the power switch is being bumped.


He could easily test the reset theory though: just type uptime in a terminal. If it is the battery coming loose, uptime will be minutes.

 

[Cloudef]

"With our advanced analysis techniques and experience in Linux tuning, boot times can be reduced in some cases to less than a second." - sounds like BS to me. Hey, I don't mind being proved wrong, but this sounds pretty BOFH.

There are already people booting Linux under 5 sec, in example some arch users with their custom init scripts or quick-init scripts from AUR.


Here in example 2 second boot to Awesome WM.


And here 1 second cold boot, the technical slides they made out looks legit, but who knows.