From boinc alpha newsletter:

Dear alpha testers:

We have released a new version of BOINC (7.1.10) for testing.

This release includes, for the first time, a version for Android devices (phones, tables, ultrabooks).
If you have such a device, please help us test BOINC on it.

We are distributing the Android version through the Google Play Store instead of the BOINC web site.
To get the test version,
you must first join a Google Group called "BOINC Android Testing".
Instructions for doing this are here:
http://boinc.berkeley.edu/trac/wiki/AlphaInstructions#Installingtestsoftware

Currently only two listed projects (OProject and Asteroids@home) offer Android apps.
You can also manually attach to the Einstein@home test project, ​ http://albert.phys.uwm.edu/ This project provides images and icons that are shown in the Android GUI; the others do not.

I hope other projects will release android app....rosetta??? :-)

---

I hope other projects will release android app....rosetta??? :-)

+1

Also +1.

Naturally, the output of android devices isn't that much, but it can only go upward.

From what I've read, Android 4.2 should make things easier - RenderScript seems to provide a unified language that can make use of whatever is capable of computing(i.e. CPU, GPU, DSP?), sort of like OpenCL, I gather.
(Of course, I have only what limited knowledge I gained from searching the internet, no practical experience with that stuff...)

Anyway, my phone is currently burning power for Albert@Home - while worthwhile, I'd prefer Rosetta@Home (Or FAH, for that matter, but that seems very unlikely.)

Certainly an interesting development, but wouldn't this cause the phone to drain its battery a lot quicker? As it is, my droid basically lasts about a day before I have to recharge it.

I would guess that the chances of the whole Rosetta program being ported to ARM are slim in the near-future because of the time required to port it, the problems and effort required to maintain two code-bases, and the comparatively low performance of ARM devices.

Here's a comparison of the MS Surface RT against some relatively quick ARM CPUs. That's an i5 at 800MHz (I think), so you could increase the gap by 5x for a 4GHz desktop i3 or 10x for a desktop i7, and that's not doing FPU work.

My Nexus 4 has 4 cores and 2GB RAM so 512MB per core - might be enough per core (after removing some for the OS and other running things). I'm not sure how Android deals with swap-space or whether it just closes apps that aren't being used? Either way, that would probably be very slow or a problem.

In response to Kenneth's post, yes - this would drain the battery pretty quickly. It would only be practical to run a DC app when charging, so presumably overnight, and then leave it off through the majority of the day when the phone is likely to be in use. Tablets might get more run-time. My phone would likely need to sit on a heatsink too as it'd hit its thermal envelope pretty quickly I think (and a glass back isn't great for disippating heat!)...

There are probably lots of smaller projects that it will be more practical to port, hopefully freeing up some x86 power for Rosetta.

---

Certainly an interesting development, but wouldn't this cause the phone to drain its battery a lot quicker?

Definitely.
The amount of battery/heat depends on the number of cores used, naturally.

However, if you only compute while charging, it should be easy to find a point where the charger can supply enough to not actually drain the battery or heat it up too much.

I don't have absolute numbers with me right now, but the last time I checked, giving Boinc 2 of the 4 cores allowed the charger to actually recharge the phone and compute at the same time.
According to the power meter thingy, it takes about 5 watt doing so.

With 2 cores, the phone is handwarm, I'd say. If I can trust the battery sensor it heats to about 30°C (varying, it can get to 40 or more, the default cutoff of boinc is 40, I think).

It certainly gets a lot hotter when playing Ingress :-)

Of course, on my phone right now there is only Albert (and Einstein) plus WUProp doing any work - I have no idea how thorough they really can make use of the cores (I know that FAH doesn't use the whole GPU on my PC, for example).
Nor do I know if they can make use of the GPU in the Tegra 3 chip (which would add more computation power but also battery drain and heat).

That's why my impression is that Renderscript could change a lot - it would reduce the required amount of specialised code.


Right now, I think it would be interesting to know how many Android devices might be available for computation at all.
It may take thousands of cores to replace a single PC+GPU, but they likely are far more numerous too.

If regular computations are too much for Android, are there other tasks that Rosetta needs to do, that could be outsourced?
Continuus/limitless internet access doesn't seem to be so uncommon anymore.

There are probably lots of smaller projects that it will be more practical to port, hopefully freeing up some x86 power for Rosetta.

Mmmm, Correlizer??

---

1st advantage is PC is nowadays for the mass almost a dying culture vs android devices are invading through all user friendliness factors not to mention those cpu's are constantly expanding 8 core and 15 core android cpu's are already near their launch.

2nd for heating problems of android devices I am sure there are hundreds of brains already working on heatsink solutions since the thin shape of devices provide some advantage on quicker transfering of heat.

---

WCG have launched a FightAIDS@Home version for Android:

http://www.worldcommunitygrid.org/about_us/viewNewsArticle.do?articleId=318&utm_source=email&utm_medium=email&utm_campaign=android_launch_20130718

---

WCG have launched a FightAIDS@Home version for Android:

http://www.worldcommunitygrid.org/about_us/viewNewsArticle.do?articleId=318&utm_source=email&utm_medium=email&utm_campaign=android_launch_20130718

Joined and already returning some results, while tweaking some settings.

Pretty hungry on battery, but it can be set to run while charging only, when connected by wifi only and with core temperature below 40deg and it seems to run without affecting normal use. Fine for the moment.

---

Now the octa core (1.9GHz Exynos Octa CPU) devices are coming in millions how would a 4 core and 8 core android cpu crunching power scale as PC crunching scale?



- Looks like crunching ready!!

---

Depends if the architecture can use the cores at the same time. My Quad in the Galaxy S3 can only use one CPU with full speed. Using all 4 cores increasing the crunchtime by 4. So crunching with only 1 core dont eat up the battery to much too, so it doesnt matter when your only crunch with 1 core on THIS device ^^ Its only usefull when running multiple projects at the same time, but this can be done too with serial computing on one core ^^

---

DSKAG Austria Research Team: http://www.research.dskag.at

Depends if the architecture can use the cores at the same time. My Quad in the Galaxy S3 can only use one CPU with full speed. Using all 4 cores increasing the crunchtime by 4. So crunching with only 1 core dont eat up the battery to much too, so it doesnt matter when your only crunch with 1 core on THIS device ^^ Its only usefull when running multiple projects at the same time, but this can be done too with serial computing on one core ^^

I presume you're referring to thermal-throttling/TDP-throttling/turbo-boost which means it can't run all cores at full speed at the same time, but it certainly doesn't mean it increases the crunching time x4. You'll get far more productivity with more cores, although there are other issues like cache use and memory bandwidth. I expect 2 cores is a decent compromise.

---

Now is out 7.2.17 version (beta) with a lot of improvements for android...

---