Index  | Recent Threads  | Unanswered Threads  | Who's Active  | Guidelines  | Search
 

Quick Go »
No member browsing this thread
Thread Status: Active
Total posts in this thread: 19
Posts: 19   Pages: 2   [ Previous Page | 1 2 ]
[ Jump to Last Post ]
Post new Thread
Author
Previous Thread This topic has been viewed 5040 times and has 18 replies Next Thread
dskagcommunity
Senior Cruncher
Austria
Joined: May 10, 2011
Post Count: 219
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: Huge difference between Elapsed time and CPU time

I have power savings off since more than a week too.
----------------------------------------
http://www.research.dskag.at
Crunching for my Dog who had "good" Braincancer.


[Aug 5, 2013 7:43:00 PM]   Link   Report threatening or abusive post: please login first  Go to top 
ChrisR1964
Cruncher
Usa
Joined: Dec 5, 2005
Post Count: 17
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: Huge difference between Elapsed time and CPU time

Hello,
thank you for your opinions and ideas where the issue could be.
@lawrence I have 832 MB RAM and I always have over 100 MB free when running the science apps.
For sure there is some bottleneck but I have no idea where it can be I am not very skilled with Android and mobile devices.
I will try to disable Power saving and will share the results after some tests.
Could please somebody try it with me so the results won't be so biased?
Btw. I'm not complaining I just would like to know and maybe improve performance of my Android device for the research.


I've never turned on the power saving mode. I try to stay plugged in since I've installed Boinc on my phone. It seems like I can process more work units by only running one core than when I run more cores. Any info to increase performance would be welcomed.
[Aug 6, 2013 2:22:30 AM]   Link   Report threatening or abusive post: please login first  Go to top 
krab84
Cruncher
Poland
Joined: Nov 25, 2007
Post Count: 3
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: Huge difference between Elapsed time and CPU time

I have got same problem at both of my devices: jiayu g3 and ainol novo 9 spark.
From what i have observed, when android device has screen turned off, cpu goes into some kind of power saving mode, and most of the time they do not crunch. Thats why there is difference between cpu time/elapsed time.

That is independent from:
- cpu governor(im running performance, and setcpu shows me 100% time at top frequency)
- deep sleep(deep sleep info shows me 0% time)

But when my devices have screen turned on, they run beatufuly, and i cant handle the heat :)

Thats the only solution i've found for now.
----------------------------------------
[Edit 1 times, last edit by krab84 at Aug 6, 2013 2:05:46 PM]
[Aug 6, 2013 1:54:34 PM]   Link   Report threatening or abusive post: please login first  Go to top 
dskagcommunity
Senior Cruncher
Austria
Joined: May 10, 2011
Post Count: 219
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: Huge difference between Elapsed time and CPU time

even when you turn off the power saving options its going to sleep?
----------------------------------------
http://www.research.dskag.at
Crunching for my Dog who had "good" Braincancer.


----------------------------------------
[Edit 1 times, last edit by dskagcommunity at Aug 6, 2013 2:45:22 PM]
[Aug 6, 2013 2:43:31 PM]   Link   Report threatening or abusive post: please login first  Go to top 
dskagcommunity
Senior Cruncher
Austria
Joined: May 10, 2011
Post Count: 219
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: Huge difference between Elapsed time and CPU time

As next info. On astroids at home there is the same issue. Single core 40k vs 164k on 4 cores.
----------------------------------------
http://www.research.dskag.at
Crunching for my Dog who had "good" Braincancer.


[Aug 7, 2013 7:41:04 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Randzo
Senior Cruncher
Slovakia
Joined: Jan 10, 2008
Post Count: 339
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: Huge difference between Elapsed time and CPU time

Hello,

so I did a small experiment with power saving.

Efficiency with power saving enabled 17,66%
Efficiency with power saving disabled 19,55%

So it does a small difference, but frankly I would expect more.

So when I run on 4 core simultaneously I am running less than one.

It would be great if Android or ARM (I have no idea whose fault is that) could make it over 90%.

But anyway I will continue to support computation with my Android device.
[Aug 21, 2013 5:35:00 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Former Member
Cruncher
Joined: May 22, 2018
Post Count: 0
Status: Offline
Reply to this Post  Reply with Quote 
Re: Huge difference between Elapsed time and CPU time

Here's my WAG:) I'd think that the easiest way to put BOINC on Android would be to wrap the C++ client in a Java wrapper and communicate through JNI. The problem might be that the client is run through the VM which (kind of) controls processor and memory usage. All tasks that run in the boinc space might need to share whatever resources Java has allocated to that instance of the boinc client.. which would explain why ChrisR1964 returns more result with one core instead of four and would suggest that the per-task cpu usage number doesn't accurately reflect actual usage since "real" usage would roughly be the sum of all tasks:)
[Aug 21, 2013 6:39:09 PM]   Link   Report threatening or abusive post: please login first  Go to top 
krab84
Cruncher
Poland
Joined: Nov 25, 2007
Post Count: 3
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: Huge difference between Elapsed time and CPU time

This problem is software related.

Tried to change build.prop settings like pm.sleep_mode and ro.ril.disable.power.collapse. Not succesful.

BUT!!!
After installing diffrent rom (unofficial cyanogenmod based on android 4.3) on my ainol novo 9 spark, problem disappeared. All cores run 100%, tasks got about the same cpu/elapsed time, even with screen off.

Stil dont know where is the source of problem.
[Oct 7, 2013 5:05:06 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Former Member
Cruncher
Joined: May 22, 2018
Post Count: 0
Status: Offline
Reply to this Post  Reply with Quote 
Re: Huge difference between Elapsed time and CPU time

That's it, krab84!

Android can run quad-cores and ARM cpus are perfectly capable quad-cores but the mobile device makers cripple their devices with a special rom. Probably they are worried about heat and power usage. So the mobile device can run 1 core but additional cores are only intended for handling interrupts or small background processes that might slow down the user while he tries to interact with the device. Different roms probably handle this slightly different ways, but the end result is that most mobile devices should only run BOINC on 1 core.

I feel less frustrated now that I can point to a specific cause - the rom.

Lawrence
[Oct 7, 2013 9:38:18 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Posts: 19   Pages: 2   [ Previous Page | 1 2 ]
[ Jump to Last Post ]
Post new Thread