Project: Android JB 4.1 with Hardware acceleration

[cleverca22]

yeah, ive now noticed that the framebuffer code was over 6 months old and doing a few things wrong, causing a false error, now that ive updated the framebuffer driver, it 'supports' page flipping at 1280x1024, but the screen is completely blank like it was at 800x600

i'll check irc

[kramlat]

translation: still not working. double buffering is required for JB. result: unusable build. page flipping is required for wine's GDI to do off-screen rendering as well which is why due to the fact that wine's gdiplus does not support it, certain apps, such as synthmaker do not render right. rather strange that a similar problem would plague the android porting project. when you fix that, please contribute to the mono project and the wine project. you will help make a lot of new programs work on x86.

[cleverca22]

after looking at things more, the razdroid 'fork' (it was only 2 commits long) was 'broken' from the very first commit
and checking the original raspberrypi fork of linux, the framebuffer code never had the page flipping problem, ever (it was good from the very first commit)

simply switching over to the raspberrypi branch of linux and enabling the proper android features in menuconfig 'fixes' page flipping 100%

but CM7 has another bug on gralloc, that causes it to just render black if page flipping is supported

[duskman72]

Hi there,

the problem is very simple:

JB needs another HWC (HW Composer) as it was used in ICS, so you/we need to rewrite a bunch of code to get things
working. I've ported JB 4.2.2 CM10.1 to the Galaxy S3 before Samsung was releasing a first leak, there was the the same
problem with wrong versions/method calls from binary firmware drivers.

The correct thing: page flipping isnt needing, but can be used if its in the kernel supported.

As you see, you have to modify things in the kernel (specially framebuffer and KGSL drivers to match the userspace drivers/firmware located in /etc/firmware). The next thing: dont modify the path of the firmware in android - use the physical path as it was on ICS, otherwise the firmware
files cant be found. The next thing: take care of the needed display drivers: as i know, JB supports HDMI a lot better as ICS but in another way,
so you have to modify some surfaceflinger files and the Framebuffer allocator in the Android core framework.

Is there a git repo and rom to flash? I want to have a look at this, may be i can help with some experience of porting android.

[ghans]

http://androidpi.wikia.com/wiki/Android_Pi_Wiki

https://github.com/Razdroid/


Good luck.

ghans

[duskman72]

I'm sorry, but this aren't Android git repos - there are only a few userland files and a kernel source, nothing to play with:

- no Android manifest file
- no Android configuration files (missing the device folder)
- no pre-compiled files

This isn't what i'm searching for, and its nothing to work with. Possibly i have to create a new repo with working files...
this needs some more time start

[ghans]

I don't know what i'm talking about , it seems.
Anything of interest here ?

https://github.com/Mathijsz


ghans

[aaa801]

I'm sorry, but this aren't Android git repos - there are only a few userland files and a kernel source, nothing to play with:

- no Android manifest file
- no Android configuration files (missing the device folder)
- no pre-compiled files

This isn't what i'm searching for, and its nothing to work with. Possibly i have to create a new repo with working files...
this needs some more time start

hi come join us on irc.freenode.net #razdroid
ask mathijs and warg for info

[gryphonB]

My take on the Android-Pi is this: I am looking at creating a cheap, 'thin' PC for educational use. Android Apps have a huge amount of educational resources, it looks like everything you need for a yr 0 - 12 education. I've been searching all day trying to follow up 'Android ports to Pi' and just saw the line at the bottom of the wiki saying 'unusable, since no graphics acceleration'. Others have come up against the same wall. I just found a post saying that Samsung Galaxy Y uses the same Broadcom graphics chip: it has a CyanogenMod unofficial build. Another post says the graphics library links to glibc instead of another library (don't quote me here! It's 500 pages too late for that!)

I have other options, of course. I might be able to build the Davlik VM on top of Linux (or not. Jury is out). I can run a VM to host Android, but this sounds a little ridiculous. I may be able to compile Android-x86 for Pi.

The point really is that Android Apps provide a great resource, meant for a lightweight platform. From a non programmers view, Pi & Android seem made for each other.

[gryphonB]

I have deliberately not read your reply in depth as the first sentence was obviously meant to antagonise. Once again, poor action for an adminisrator. In addition, I have carefully noted the general tone/unecessary shouts provided (thanks) and then researched your very, very many recent posts.

I note you use the word "we" quite a lot.

If you really are a RPi foundation member I recommend you review your recent posts hereabouts, take a deep breath and reconsider the tone of your replies as a reflection of the perception that others may have of the Foundation.

For KARMA see http://en.wikipedia.org/wiki/Invocation

For progress, see elsewhere, as I have freely supplied after much effort. I'm simply trying to help others hereabouts who may be puzzled at the bizzarre situation created by 4/5month old dead-end topics that result in hours of wasted effort/circular links/reinvention of what appears right now to be a square wheel. If that's my fault - hang me!

You are quite simply, with your continued digging at a moderator posts, asking for a ban. Is that what you want? We try and keep things civil here, but there are limits to my patience. I re-read my posts and have been civil, albeit slightly sarcastic. Since you, by your own admission have NOT read my all posts, I think it's you who need to consider what they are writing - it's definitely bad form to criticise other poeple post without reading them. Everything I have written, has been, to the best of my knowledge, completely accurate. And tbh, my knowledge, due to working at Broadcom with Eben and knowing him and Liz personally, is quite accurate (not necessarily perfect, as I do not know everything)

Get over it. The guy just asked what was happening. I'd like to know too. Everyone thinks that Android is just Angry Birds, but there are 1/2 a million apps on Google, and leveraging that kind of resource is what non programming babes-in-the-woods like me would love to do. The whole idea of a great concept like the RPi is that it goes in unexpected directions, like this one.

We all get the picture that there is a great deal of work to do and not everyone can do it. I keep on getting tantalising glimpses that it is possible: the video library is now open, yet we're still unable to actually utilise the RPi's supposedly stellar graphics capabilities. I can't yet find an in depth spec sheet. I think this is very much a 'build it & they will come' scenario.

[jamesh]

I have deliberately not read your reply in depth as the first sentence was obviously meant to antagonise. Once again, poor action for an adminisrator. In addition, I have carefully noted the general tone/unecessary shouts provided (thanks) and then researched your very, very many recent posts.

I note you use the word "we" quite a lot.

If you really are a RPi foundation member I recommend you review your recent posts hereabouts, take a deep breath and reconsider the tone of your replies as a reflection of the perception that others may have of the Foundation.

For KARMA see http://en.wikipedia.org/wiki/Invocation

For progress, see elsewhere, as I have freely supplied after much effort. I'm simply trying to help others hereabouts who may be puzzled at the bizzarre situation created by 4/5month old dead-end topics that result in hours of wasted effort/circular links/reinvention of what appears right now to be a square wheel. If that's my fault - hang me!

You are quite simply, with your continued digging at a moderator posts, asking for a ban. Is that what you want? We try and keep things civil here, but there are limits to my patience. I re-read my posts and have been civil, albeit slightly sarcastic. Since you, by your own admission have NOT read my all posts, I think it's you who need to consider what they are writing - it's definitely bad form to criticise other poeple post without reading them. Everything I have written, has been, to the best of my knowledge, completely accurate. And tbh, my knowledge, due to working at Broadcom with Eben and knowing him and Liz personally, is quite accurate (not necessarily perfect, as I do not know everything)

Get over it. The guy just asked what was happening. I'd like to know too. Everyone thinks that Android is just Angry Birds, but there are 1/2 a million apps on Google, and leveraging that kind of resource is what non programming babes-in-the-woods like me would love to do. The whole idea of a great concept like the RPi is that it goes in unexpected directions, like this one.

We all get the picture that there is a great deal of work to do and not everyone can do it. I keep on getting tantalising glimpses that it is possible: the video library is now open, yet we're still unable to actually utilise the RPi's supposedly stellar graphics capabilities. I can't yet find an in depth spec sheet. I think this is very much a 'build it & they will come' scenario.

Can I suggest that digging up a post from nearly a year ago, then telling a moderator to 'get over it' might not be the best approach for a second post on here.

As for Android, there has been no work done on it since the last time I said there had been no work done it, for exactly the same reasons. Close of 2M Raspi's sold, none running Android. Not exactly a commercial disaster due to lack of it.

The Raspi's graphics capabilities are all available for use using standard libraries (and always have been - search for OpenGLES, OpenVG, EGL and OpenMAX), or you can use code already available (camera apps, OMXplayer, XBMC etc). You don't need Android to use them. In fact you will get better performance using them from Linux than going through the Android eco-system.

[gryphonB]

As an aside, I'd be interested in what people actually want an Android port for! May seem like an odd question, but it's useless as a dev platform, it's not portable enough to be a mobile device, so it must purely be for access to thinks like the app store. So it's of relevance purely from a consumer of information point of view. WHIch is fine,. But exactly NOT what the Foundation designed the Raspi for!

As I said, I'd be interested in knowing what people actually want Android for.

Easy answer: access to huge amounts of very cheap educational software. You wouldn't believe that so many people have done so much. Must be all that free time teachers have. Everything from preschool to high school. Chemistry, history, languages, unameit. Dictionary, encyclopedia, thesaurus. An entire education on a phone/RPi. And then there's all the other stuff. For $1. I'd guess that's exactly what the Foundation designed for. Think of it as something like the $100 laptop.

I keep on getting hints that it's possible. CyanogenMod have Android (unofficial, ie buggy) running on the Samsung Galaxy Y, which uses a Broadcom video chip. The Pandaboard, an ARM, is a target for the Linaro Android team. I can get the Android Source, I also may be able to run the Dalvik VM on Linux, which seems to just leave the problem of the JNI. Someone managed to compile an Android source for RPi but could not link the Broadcom EGL libs. It seems possible now, or is that just my ignorance showing?

[jamesh]

As an aside, I'd be interested in what people actually want an Android port for! May seem like an odd question, but it's useless as a dev platform, it's not portable enough to be a mobile device, so it must purely be for access to thinks like the app store. So it's of relevance purely from a consumer of information point of view. WHIch is fine,. But exactly NOT what the Foundation designed the Raspi for!

As I said, I'd be interested in knowing what people actually want Android for.

Easy answer: access to huge amounts of very cheap educational software. You wouldn't believe that so many people have done so much. Must be all that free time teachers have. Everything from preschool to high school. Chemistry, history, languages, unameit. Dictionary, encyclopedia, thesaurus. An entire education on a phone/RPi. And then there's all the other stuff. For $1. I'd guess that's exactly what the Foundation designed for. Think of it as something like the $100 laptop.

I keep on getting hints that it's possible. CyanogenMod have Android (unofficial, ie buggy) running on the Samsung Galaxy Y, which uses a Broadcom video chip. The Pandaboard, an ARM, is a target for the Linaro Android team. I can get the Android Source, I also may be able to run the Dalvik VM on Linux, which seems to just leave the problem of the JNI. Someone managed to compile an Android source for RPi but could not link the Broadcom EGL libs. It seems possible now, or is that just my ignorance showing?

It's quite a complicated task, otherwise, well, it would already be done by now. If it was quick and easy (or even not quite so quick and not quite so easy), it would be a no brainer to spend some time on it.

I'm sure there will be lots of educational stuff for all sorts of subject matter, but I'm not convinced on the quality of educational materials available for Android for the core purpose of the Raspi - teaching about the low level device, coding etc. Android abstracts too much of that sort of stuff away.

[redhawk]

I'm sure Android on the Pi would be nice but even if drivers did exist the performance wouldn't be great and software limited to ARMv6 code whereas ARMv7 with it's super fast integer/single precision float would enable Adobe Flash and a few other ARMv7 compatible programs.
Unless the Pi foundation upgrade their SoC with ARMv7 capability (which I don't doubt will happen in the future) then the current Android project is pretty much a pointless venture.

Just out of curiosity is there an ARMv7 SoC by Broadcom which is drop in compatible with the Pi board layout??

Richard S.

[jamesh]

I'm sure Android on the Pi would be nice but even if drivers did exist the performance wouldn't be great and software limited to ARMv6 code whereas ARMv7 with it's super fast integer/single precision float would enable Adobe Flash and a few other ARMv7 compatible programs.
Unless the Pi foundation upgrade their SoC with ARMv7 capability (which I don't doubt will happen in the future) then the current Android project is pretty much a pointless venture.

Just out of curiosity is there an ARMv7 SoC by Broadcom which is drop in compatible with the Pi board layout??

Richard S.

No. Any new SoC would require a complete board redesign.

[gryphonB]

You win. Completely & utterly sorry I asked.

But seriously. Take a look at what is available in Google Apps. It's insane. 'A thousand monkeys typing for a thousand years' insane. Just because it doesn't fit the original remit of the Foundation doesn't mean it's not a really good idea, hugely expanding the reach of RPi & the number of people exposed to it. Especially since Broadcom already have the drivers/libs from other work.

[jamesh]

You win. Completely & utterly sorry I asked.

But seriously. Take a look at what is available in Google Apps. It's insane. 'A thousand monkeys typing for a thousand years' insane. Just because it doesn't fit the original remit of the Foundation doesn't mean it's not a really good idea, hugely expanding the reach of RPi & the number of people exposed to it. Especially since Broadcom already have the drivers/libs from other work.

There's not need to get all snotty about it, just because it's not what you want to hear.

Yes, there is LOADS of stuff in Google Play/Apps or whatever they call it this week. But very little of it is applicable to the Raspi core purpose (some of it was actually written by monkeys, who are not renowned for high quality SW) - which is what any available money is spent on. At the moment, the money is better spent on things that have a more direct bearing on the charity's aims.

As I work at Broadcom, sometimes on Android stuff, I can say that the amount of work required to get all the libraries etc working correctly for the Raspi is not insignificant. That would cost quite a bit of money - it's work that would be done by Broadcom, and they want paying.

[redhawk]

As an unregistered device the Pi would be blocked from accessing the Google play store anyway, this is common problem with some cheap and nasty Chinese tablets or when accessing Google play app from an Android emulator.
For what it's worth I think Raspbian, RiscOS and XBMC are doing a pretty good job for coding, hacking, education and entertainment.
As for Android I really don't see any benefit apart from "it works - isn't that amazing??" or people desperate for Angry Birds and Skype.

Richard S.

[jamesh]

As an unregistered device the Pi would be blocked from accessing the Google play store anyway, this is common problem with some cheap and nasty Chinese tablets or when accessing Google play app from an Android emulator.
For what it's worth I think Raspbian, RiscOS and XBMC are doing a pretty good job for coding, hacking, education and entertainment.
As for Android I really don't see any benefit apart from "it works - isn't that amazing??" or people desperate for Angry Birds and Skype.

Richard S.

I think you could probably sideload apps should you want to, bypassing the store completely. And I'm more of a Bad Piggies fan.....

[redhawk]

Sideloading would be possible (if you know how) but you still need to find a reliable source of APK files to install from and to be honest I wouldn't trust anything unless it came from the Play Store or from my Android phones /data/app folder.
Android APK files are easy to repack with additional code which means any 3rd party download site could be a potential source of apps with malware.

Richard S.

[aaa801]

Nothings stopping the end user from installing gapps,
apart from googles terms

[adsp]

jamesh wrote (a few posts back):

As an aside, I'd be interested in what people actually want an Android port for! [...] of relevance purely from a consumer of information point of view. WHIch is fine,. But exactly NOT what the Foundation designed the Raspi for!

Here is one answer. I couldn't care less for Android / Google / Broadcom /PlayStore / consumer (I agree with you here...).
I'm following this thread just because I would like a decent browser on RPi.
Can I take this conclusion that a decent browser is NOT very high on Foundation list?

[jamesh]

jamesh wrote (a few posts back):

As an aside, I'd be interested in what people actually want an Android port for! [...] of relevance purely from a consumer of information point of view. WHIch is fine,. But exactly NOT what the Foundation designed the Raspi for!

Here is one answer. I couldn't care less for Android / Google / Broadcom /PlayStore / consumer (I agree with you here...).
I'm following this thread just because I would like a decent browser on RPi.
Can I take this conclusion that a decent browser is NOT very high on Foundation list?

No, you cannot make that conclusion. The only advantage in a browser running on Android is some HW acceleration. So, if you get the HW acceleration working for Linux rather than just Android, you get the same effect. So that's where the effort is going - Wayland/Weston acceleration will accelerate the whole desktop for Linux, including the browser.

[tufty]

The only advantage in a browser running on Android is some HW acceleration

And even that only on a platform where HW acceleration is available.

[redhawk]

Android may offer a more compatible browser with JavaScript support but it's not going to satisfy the "I want YouTube" crowd nor can it run Adobe Flash Player.
Speed wise it won't be great either and I know this for a fact since I have an ARMv6 based (Qualcomm MSM7227) Android phone to form a reliable opinion with.
Android on the Pi would provide no real advantage over Raspbian or RiscOS except novelty value.

Richard S.