Project: Android JB 4.1 with Hardware acceleration

[gritz]

lewmur: you don't happen to work for Google by any chance?

It's horses for courses. You wouldn't buy a Ferrari to tow a caravan, or use a chainsaw to trim your nose hair. By all means contribute to the unofficial effort to port Android to the Pi, but personally I don't think that a minimalist desktop format computer with GPIO is Android's best fit.

[lewmur]

lewmur: you don't happen to work for Google by any chance?

It's horses for courses. You wouldn't buy a Ferrari to tow a caravan, or use a chainsaw to trim your nose hair. By all means contribute to the unofficial effort to port Android to the Pi, but personally I don't think that a minimalist desktop format computer with GPIO is Android's best fit.

It isn't Linux's "best fit" either. I can't even run Midori under XFCE4 but have to use Netsurf to browse. And if you check the XDA-Developers forums, you find several *dozen* tablets with about the same CPU-GPU power running Android. This is obviously a case of anti Android, or more probably anti Google, bias. (As your question about my working for Google testifies.)

The fact is, that, at the present time, Android runs better with this of level power because it takes much more advantage of the GPU. That might change if they ever get full hardware accel working with X, but not now. I personally can vouch for the fact that *none* of the Linux ports to the HP TP work anywhere near as well as JB. (Or the earlier ports of Froyo, GB or ICE)

BTW, if you check back a few posts, you'll notice I'm 74 years old. I don't work for anyone. I don't have a dog in that fight.

[gritz]

BTW, if you check back a few posts, you'll notice I'm 74 years old. I don't work for anyone. I don't have a dog in that fight.

I saw that already. I was pulling your leg - it's what we English do.

I tend to agree that Linux isn't best fit either. TBH I've never felt that Linux fitted anything outside of a rack tower and a clump of RJ45 cables, but Linux was the default option because it was free, ubiquitous and allowed the use of a bazillion 3rd party programs and a wide range of external hardware. In a perfect world RISCOS might have complemented the available processing grunt a little better, but unfortunately it doesn't have the large userbase essential to a "community" project, or the level of compatibility yet.

As for CPU / GPU power - it's unfortunately not just a case of comparing bare numbers and saying "it should work". I'm not sure that the BCM2835 was ever designed with Android explicitly in mind - and even if it was then we're talking about a time before Jellybean was even a twinkle in Sergey Brin's eye. Android tends to work well on apparently limited hardware because of tight integration - it's not a "general purpose" O.S. like yer average desktop Linux, which will tend to work (after a fashion) on just about anything (with the notable exception of my old Dell laptop, which has resisted all attempts at Linux installs). Getting 'Droid to run on the Pi in an acceptable manner *might* be just a case of getting the GPU to do the heavy lifting, but it might be a bit more complex than that...

Remember also that the "Android user experience" is an important concept to Google - they're not interested in supporting hardware that's a bit long in the tooth, so expect planned obsolescence, whatever your hardware.

It's also possible that you're expecting a bit too much from this little board.

[jamesh]

lewmur: you don't happen to work for Google by any chance?

It's horses for courses. You wouldn't buy a Ferrari to tow a caravan, or use a chainsaw to trim your nose hair. By all means contribute to the unofficial effort to port Android to the Pi, but personally I don't think that a minimalist desktop format computer with GPIO is Android's best fit.

It isn't Linux's "best fit" either. I can't even run Midori under XFCE4 but have to use Netsurf to browse. And if you check the XDA-Developers forums, you find several *dozen* tablets with about the same CPU-GPU power running Android. This is obviously a case of anti Android, or more probably anti Google, bias. (As your question about my working for Google testifies.)

The fact is, that, at the present time, Android runs better with this of level power because it takes much more advantage of the GPU. That might change if they ever get full hardware accel working with X, but not now. I personally can vouch for the fact that *none* of the Linux ports to the HP TP work anywhere near as well as JB. (Or the earlier ports of Froyo, GB or ICE)

BTW, if you check back a few posts, you'll notice I'm 74 years old. I don't work for anyone. I don't have a dog in that fight.

No No No! I am NOT anti Android or anti Google (and neither are the other posters above) - not in the slightest. Android will run on a Raspi - and it will run fairly well when using the GPU for the graphics. I'm not arguing that Android should not be ported - I think it should; what I'm arguing is that it is NOT suitable for teaching programming, and therefore should not be the highest priority item on the Foundations list. And even if it were the highest item on the Foundations list, they are still limited by the resources that Broadcom can throw at the problem, as the Foundation employs NO engineers themselves....

Please don't mistake this argument as being anti-Android.

[magno23]

i think that what all of you want to say is , android could be ported to the rpi but since it isn't a good developers tool it's not in the priority list of things to do right now.

[poglad]

I don't actually see /any/ benefit in using Android over Linux, even if it was ready to go now, let alone putting aside other stuff to make it work.

Your opinions may vary, of course.

It will probably be of more educational value than Plan9, possibly RISC OS too, and they are available for the Pi. Mind you, if they had required significant input from Broadcom then they probably would still be on the wish list like Android. I think the view of the Pi as a 'Linux box', repeated daily on the official front page, is unfortunate and limits the imagination somewhat. For Android, we'll have to wait until usable firmware emerges from the more pressing work Broadcom is doing, but once that happens it would be silly not to pursue it as an option.

[jamesh]

I don't actually see /any/ benefit in using Android over Linux, even if it was ready to go now, let alone putting aside other stuff to make it work.

Your opinions may vary, of course.

It will probably be of more educational value than Plan9, possibly RISC OS too, and they are available for the Pi. Mind you, if they had required significant input from Broadcom then they probably would still be on the wish list like Android. I think the view of the Pi as a 'Linux box', repeated daily on the official front page, is unfortunate and limits the imagination somewhat. For Android, we'll have to wait until usable firmware emerges from the more pressing work Broadcom is doing, but once that happens it would be silly not to pursue it as an option.

Broadcom didn't have any input on Plan9, but one of the major RISCOS people works for Broadcom, so that was a help there. However, I do think RISCOS has a certain amount of possibility in the teaching market.
Linux was the first OS on the Raspi, has a huge amount of software available for it, and is relatively easy to learn the basics. It's also a useful (read sellable) skill to have as you advance in to the jobs market. So it's a good strap line.

[lewmur]

Remember also that the "Android user experience" is an important concept to Google - they're not interested in supporting hardware that's a bit long in the tooth, so expect planned obsolescence, whatever your hardware.

It's also possible that you're expecting a bit too much from this little board.

It may be an important concept at Google but that doesn't mean anything to 99% of the devices running Android. No one looks to Google for support except the rare device that was actually designed by Google. Anymore than I would look to the Foundation for Pi support. Android support, like Pi support, comes from the community.

And, just to make things clear, I wouldn't recommend, by any means, that JB be the version of Android devs should be looking to port first. Sure, everyone wants "the latest and greatest". But 90% of "The Android Experience" can be had with Android 2.3. And that version works well with an 800mhz CPU and 400mhz GPU even with only 256mb of RAM.

From my own experience, early on I had a zt180 "china made" tablet. It had the spec I quoted above and came with Android 2.0. But the community quickly developed ROMs for 2.3 and ported Debian in a dual boot config. Of course, like the Pi, the Debian port didn't have hardware accel. But it ran OK. But everyone agreed that Android was much, much better.

BTW, anyone who wants verification of this can check the "Flytouch III" forums at Slashdroid from a year or so back.

[tufty]

I don't actually see /any/ benefit in using Android over Linux, even if it was ready to go now, let alone putting aside other stuff to make it work.

It will probably be of more educational value than Plan9

*coff* you what?

No, really. You what?

[poglad]

I don't actually see /any/ benefit in using Android over Linux, even if it was ready to go now, let alone putting aside other stuff to make it work.

It will probably be of more educational value than Plan9

*coff* you what?

No, really. You what?

Since you can't see any possible benefit of having Android available, there's not much point me answering that is there.

[wayner]

Some of us would like to use an RPi as a Netflix client. Assuming that the RPi does run Linux with decent video performance doesn't that mean that Android should facilitate this, since Netflix doesn't (really) run on Linux?

Sure that is not really an educational use case, but many RPis are being sold for other purposes, and I am sure that Netflix has educational videos available as well.

[tufty]

there's not much point me answering that is there.

Of course there is. All you have to do is explain how an OS front end designed for passive media consumption and, as was rather astutely mentioned earlier, "data mining", is somehow more educational than an experimental OS designed to run in a completely distributed manner, something that pushes the underlying concepts of Linux (and, thus, Android) to their limits.

Y'know, that sort of thing.

[gritz]

On a more philosophical note I think that we're just seeing some "growing pains" here.

Those x86 boxes that so many of us are familiar with may be big, ugly and power hungry, but it's a stable platform with good software, hardware and legacy support and if it's relatively modern box it will have ample processing power for any "normal" application. Low power ARM hardware doesn't have the luxury of general-purpose processing power to spare, so the hardware / software has to have more synergy and you have to choose your hardware to explicitly fit your application, or be prepared for some pain.

Perhaps there is some merit to Android for e.g. idiot-proof media consumption, but given the obsolescence of Android devices it would be a cinch to pick up a cheap plug-and-play 'Droid device on ebay or similar and save one's self the bother of having to pester people to write e.g. a GPU driver for free, in their spare time. As a bonus you may even end up with something newer (and less vulnerable) than 2.3 Gingerbread to boot. And you won't be running a touchscreen OS on a desktop. If Android is your favoured OS then not being able to run it is a bummer, but hey - I'm still quite content with Windows XP. Time marches on though and pragmatic choices have to be made.

I would always advise someone (unless they are a talented and experienced hacker) to buy something solely on the strength of what it can do now. Forget about promised features that may or may not materialise, or possibilities suggested by reading between the lines of a specsheet and extrapolating. Been there, done that, felt the disappointment. .

[poglad]

Y'know, that sort of thing.

1. They can learn how the world's leading mobile operating system works under the hood, with a hackable installation on hardware that uses a SoC found in real mobile devices.
2. They can access a growing library of educational software available via Google Play.

Bear in mind that the individual may not already have an Android phone, and may not be able to afford one.

I'm only suggesting it be made available if the work that's being done by Broadcom anyway becomes available to us. I honestly can't think of a downside.

[jamesh]

Some of us would like to use an RPi as a Netflix client. Assuming that the RPi does run Linux with decent video performance doesn't that mean that Android should facilitate this, since Netflix doesn't (really) run on Linux?

Sure that is not really an educational use case, but many RPis are being sold for other purposes, and I am sure that Netflix has educational videos available as well.

THANK YOU! Someone has actually stated a perfectly valid reason for an Android port. Rather than "We want Android".

Also, Poglad's case of educational software from Google play is also a good point.

It's only taken 2 pages of comments to extract those!

With regards to the port, it's likely when/if its ready to be a Jelly Bean port (that where all Broadcom's efforts are), fully using the GPU for video acceleration, camera system (yes, that should work with the foundation camera board). All this code is already in place, and mostly working. What needs to be done is bolting it altogether in a Raspberry Pi build, taking in to account the Raspi hardware. Now that's been got going once, but with some issues, so hopefully, as soon as time pressures reduce on actual work stuff, it's shouldn't be a huge job to get it all going.

What I don't know is if there are any licence issues with a release of Android. All the legal cruft can be a bit of a minefield, but Eben's good at that stuff.

[portets]

One reason for me is that I'm currently very low on money and would love to have an Android development platform. Maybe I could even make some spare change if I make a decent app.

[gritz]

One reason for me is that I'm currently very low on money and would love to have an Android development platform. Maybe I could even make some spare change if I make a decent app.

You don't need Android to develop an Android application. The dev. tools are cross platform. A resourceful and pragmatic person could snag a capable enough (x86) computer from the free ads of their local paper for peanuts, or round up something for free from a relative / friend who has upgraded.

No point in making things harder than they need to be!

http://developer.android.com/guide/comp ... index.html

[lewmur]

It's only taken 2 pages of comments to extract those!

With regards to the port, it's likely when/if its ready to be a Jelly Bean port (that where all Broadcom's efforts are), fully using the GPU for video acceleration, camera system (yes, that should work with the foundation camera board). .

James;
It's nice to hear that at least something is in the pipeline. But JB is going to leave out all of the early Pi adopters because, unlike GB, it won't run with only 256mb of RAM.

Then again, that will be an excuse for them to "upgrade" to the 512mb version.
Lew

[aaa801]

It's only taken 2 pages of comments to extract those!

With regards to the port, it's likely when/if its ready to be a Jelly Bean port (that where all Broadcom's efforts are), fully using the GPU for video acceleration, camera system (yes, that should work with the foundation camera board). .

James;
It's nice to hear that at least something is in the pipeline. But JB is going to leave out all of the early Pi adopters because, unlike GB, it won't run with only 256mb of RAM.

Then again, that will be an excuse for them to "upgrade" to the 512mb version.
Lew

It will run fine on 256mb of ram.

[lewmur]

James;
It's nice to hear that at least something is in the pipeline. But JB is going to leave out all of the early Pi adopters because, unlike GB, it won't run with only 256mb of RAM.

Then again, that will be an excuse for them to "upgrade" to the 512mb version.
Lew

It will run fine on 256mb of ram.

That depend on your definition of "fine". I'm just quoting the devs at XDA. They claim the performance with 256mb is "flaky". Lots of FC's.

[jamesh]

It's only taken 2 pages of comments to extract those!

With regards to the port, it's likely when/if its ready to be a Jelly Bean port (that where all Broadcom's efforts are), fully using the GPU for video acceleration, camera system (yes, that should work with the foundation camera board). .

James;
It's nice to hear that at least something is in the pipeline. But JB is going to leave out all of the early Pi adopters because, unlike GB, it won't run with only 256mb of RAM.

Then again, that will be an excuse for them to "upgrade" to the 512mb version.
Lew

For goodness sake, some people are NEVER satisfied. It will probably run OK on 256MB.

[lewmur]

For goodness sake, some people are NEVER satisfied...

I would hope not! The only people who should be "satisfied" with the status quo, are those lying in their graves!

[wkulesza]

My team would have another idea for an educational project.
We would like to use Raspi (together with GPS receiver and 3g modem, both on usb) as a cheap tracking device. These devices are mostly expensive, but there are open source clients on android to save&send current position from gps.
Android on Raspberry would allow to install such client and on another raspberry we would plan to install a server software to receive the readings from clients and put them on a map in order to allow tracking.
Would this be a valuable new reason for existance of Android for Raspi? My programming skills currently are not sufficient to write a client from scratch on Raspbian and thus, I would like to start with the client that's ready and go from there to develop 100% own solution.

Keep up the good work !

[jamesh]

TBH, that's doesn't sound like a big job in Python or even C, so would be perfectly fine running on Raspbian. GPS dongles certainly work - I had them going myself. I *think* others have had 3G working.

[jamesh]

For goodness sake, some people are NEVER satisfied...

I would hope not! The only people who should be "satisfied" with the status quo, are those lying in their graves!

I never said anything about the status quo...at the moment we don't even have Android (of any description), now is NOT the time to complain about it perhaps not working on the 256 device! Run before walking etc.