It appears so far, that the first project on Android 10 (Universe@home) is running on my Essential Phone (with Android 10).
It runs a snapdragon 835 processor, with 4GB of ram.
The snapdragon 835 runs a big little configuration, with 4 fast cores running at 2,45Ghz, and 4 slower ones running at 1,9Ghz.
I've tried running 4 tasks, which get loaded in the little little cores. 6 tasks, 2 apparently occasionally blipping the big cores to 2.4 GHz. 8 tasks keep the CPU fully occupied, according to CPUZ.
The power draw on this system is low, it remains charged on a 5 watt (slow) charger, and battery temperatures remain below 39C, much lower than gaming.

It took nearly 6 hours to get 85% of a task done, and I'm hoping it will finish the task.

Most other projects still give a compute error.

It'd be nice to hear from others running Android 10.

Ok, the first sets of tasks have been successfully uploaded, and credits awarded.

I know the SETI guys were working on getting it to work on Android, but recently got a notification they're stalling certain projects, due to a backlog of work?

There was one more project addressing the Android 10 issue, I know of, and they were close to releasing it. Last update was a few weeks ago.

Seems like I was mistaken. Running 8 threads takes more than twice as long, as running 4.

---

Good news!
Rosetta now has WUs available for Android 10 (since today).
They're from the ARM COVID19 batch.
Mine took a while to download, but it seems to crunch them fine.
That's 2 projects fully working under Android 10!

Hi，
While my PC is crunching F@H I set up BOINC on my tablet and phone, hoping for some COVID WUs. Luckily recently Rosetta has sent out some WUs for Android. Prior to that I was crunching WCG and Universe@Home.

I'm trying to work out the best config for # of processors, and number of active tasks.

Both devices have 8 cores in a big.LITTLE config. The Tablet has 3GB and phone has 6GB RAM.

When I set BOINC to run on all 8 cores, things get hot very quickly. But I'm curious if one active task can run on multiple cores.

Eg, if I run one task at a time, will the work get split across say 2 cores or more? Or will the task be confined to one core only? Is it one task to one core?

What's the most efficient way to run Android tasks?

When I set BOINC to run on all 8 cores, things get hot very quickly. But I'm curious if one active task can run on multiple cores.

Actually what happens is that by setting BOINC to use 8 cores, you're using 4 and double load them. Android itself does not allow you to use the fastest cores, as it wants to keep them for its own operations. So on big.little configurations it's better to set the amount to half the cores you have, that way a) things won't heat up so much, and b) you're running one task per core, albeit on the slower cores.
It also helps to just put a fan on the device(s). I have three Android devices doing the calculations, they lie in the path of two 140 mm fans that spin at 5V to keep the devices reasonably cool.

---

Actually what happens is that by setting BOINC to use 8 cores, you're using 4 and double load them.

Ok so I guess that leads me into my next question:
If you set BOINC to use 4 cores, but only run two projects at a time, do the two project get split over all four "available" cores? Would that increase the speed that a project can get done because it has more processing power behind it? Or is BOINC set to run one task per core only?

I have had success running Android 10 with:

World Community Grid - Mapping Cancer Markers
Rosetta@home - Rosetta for Portable Devices

Einstein@Home - Binary Radio Pulsar works on Android 9, but in the logs the app is blocked by a OS version check from running on Android 10. Seems like a simple server side change

If you look at the Einstein application list, you'll see that there are four different application versions available for Android on ARM - (NEON, VFP, ASIMDPIE, NEONPIE). That's not as bad as SETI@Home, where there are 12 versions, but still...

It may be as simple as incrementing the version check on one of the last two, but I rather doubt it.

Just FYI for everyone out there. I was trying to run 8 cores and had all kinds of weird symptoms/problems. It would suspend saying I needed to plug in to a charger when I already was. It would also say battery was too low and will resume when charge reached 90% and then show the current charge above 90. Also was getting computation errors. When I changed settings back to use only 4 cores all these issues went away.

Just FYI for everyone out there. I was trying to run 8 cores and had all kinds of weird symptoms/problems. It would suspend saying I needed to plug in to a charger when I already was. It would also say battery was too low and will resume when charge reached 90% and then show the current charge above 90. Also was getting computation errors. When I changed settings back to use only 4 cores all these issues went away.

That supports what was said earlier, that Android reserves 4 cores for the OS and leaves 4 cores for apps. By running 8 cores on BOINC, yours just running two jobs per core on the 4 available for apps, and therefore maxing out your processor and slowing down the overall processing. 4 cores is definitely the way to go.

Ok so I guess that leads me into my next question:
If you set BOINC to use 4 cores, but only run two projects at a time, do the two project get split over all four "available" cores?

Eventually yes, as long as you don't go micromanage how BOINC should do things. Then you'll see that some cores will run other tasks.

Would that increase the speed that a project can get done because it has more processing power behind it?

It will only speed up from the cores not being double loaded. There's still no multi-threading applications for Android, as far as I know. So...

Or is BOINC set to run one task per core only?

Not BOINC, but the science applications.

Btw, that was 3 questions. ;-)
And sorry I answered so late, hadn't seen your questions.

---

Good news. I reported the issue to Einstein and they fixed the version check. It is now compatible with Android 10

And sorry I answered so late, hadn't seen your questions.

Thanks. Since dropping down to 4 cores, my two devices have been behaving much better.

4 threads unfortunately is the max big-little core configurations can handle.
There is a possibility that 5 threads might actually work, by allowing 1 thread to be processed in between the spaces when the others are going through their CPU cycles.
In my measurments, both 4 and 5 threads came out identical in performance, but heated up the ARM CPU by a lot.
More than 5 threads, and there's a reduction in speed (eg: running 8 threads, per thread run at less than 0.5x the speed of when running 4 threads on a big-little config).

Android treats the big cores like Intel would it's turbo boost.
It'll only load them when loading the apps or programs. Then moves everything to the little cores for continuous data processing.

There may be a possibility, by switching the OS to Linux, to get the full performance out of all the cores, but this would not be the thread for this.

Many people report even with the smaller cores crunching data continuously on cellphones, battery decay (blow up in size), pushing the screen out of the device.
This due to the close proximity of the CPU to the battery.
And Li batteries don't handle heat very well.

So if you do have a device like that, it'll be best to open the device, and physically remove the battery from it's location.
And run on the AC adapter only.

For cooling you can easily plug a 120mm fan (12V) into a USB (5V) charger (with a small USB to 3/4 pin fan adapter), and have it keep 3 or 4 cellphones cool.
I personally have an Essential phone, which has a built in magnet. So long I place the phone on a magnetic surface, temperatures remain below 40C when crunching, as the metal surface extracts heat better than air does.

I have only worked with Android AMLogic TV boxes, which have quad cores, and have no experience with Big-Little configurations under Linux.
However it would be worth to try, If they can run all 8 cores, at $50 per box, they would be a serious contender against the AMD Threadrippers.
Not only more cores, but also much lower power consumption.

They would also run more projects on Linux than on Android 10, or even 9 or 8.
I believe that Android projects would still run in Linux.

PS : I can confirm that Einstein seems to be running (so far)...
Waiting for it to complete and getting a score.

[...]do the two project get split over all four "available" cores? [...]Or is BOINC set to run one task per core only?

Wordy Reply:

Under big.LITTLE, BOINC "sees" all cores, but is set to use the LITTLE cores only, regardless of how many tasks you have downloaded - in this case, it wouldn't matter if they're all from one project or 8 different projects. So, when you set the '# of cores' slider, it starts with 1 task per core until all LITTLE cores are running (so if you set '3', it will only run 3 tasks on 3 cores, even if you have 18 tasks for example).
When you start going above 50% of your cores (#'s 5-8, in this case), each LITTLE core will start switching between 2 task, and each task will receive 50% of the CPU cycles... but each task's "Wall clock" run time keeps running as if it were getting 100% (so the net "wall clock" time is 2x, for the same amount of net 'cpu time's).

---

https://boinc.berkeley.edu/dl/

Confirmed Einstein does NOT work on Android 10.
In all this time, I haven't had a single point added since I've added it in my last post.
Which means currently only Rosetta and Universe work.

At this moment Einstein has different problems - an outdated certificate used for communications. This makes it that some BOINC versions, especially 7.4.53 (the present stable) can not report finished tasks.

Another project that is not successful at the higher Android numbers is Moo! Wrapper. WU's error out without exception on 17.16.3

Bill F

---

In October 1969 I took an oath to support and defend the Constitution of the United States against all enemies, foreign and domestic;
There was no expiration date.