SETI@home v7 7.06 released for ARMv7 Android

Message boards : News : SETI@home v7 7.06 released for ARMv7 Android
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3 · 4 · 5 . . . 11 · Next

AuthorMessage
Profile Raistmer
Volunteer tester
Avatar

Send message
Joined: 18 Aug 05
Posts: 2423
Credit: 15,878,738
RAC: 0
Russia
Message 46579 - Posted: 3 Aug 2013, 19:51:06 UTC - in response to Message 46577.  
Last modified: 3 Aug 2013, 19:53:32 UTC

That 0.9% is a sign that it's restarting all the time. I plan to have a version that can run in less memory out next week.

I'll talk to the Einstein guys to see what they are doing. One way around it would be to do memory allocations through a memory mapped file on the internal flash drive, which would allow memory to be swapped out. I'm not doing that because it's a good way to reduce the lifetime of a phone, flash just can't handle that many write cycles for long.

Perhaps better to get in touch with Mateusz who has already working SETI v7 app for Android.

EDIT: here is his contact data: http://nativeboinc.org/site/uncat/aboutme
ID: 46579 · Report as offensive
Profile Eric J Korpela
Volunteer moderator
Project administrator
Project developer
Project scientist
Avatar

Send message
Joined: 15 Mar 05
Posts: 1546
Credit: 26,470,697
RAC: 886
United States
Message 46580 - Posted: 4 Aug 2013, 2:51:49 UTC - in response to Message 46579.  

I already grabbed his optimized NEON and VFP routines and they are in 7.05 and 7.06. I don't think we (UC) can use the FFTS library due to license issues (it has an attribution during runtime clause that might be tricky). That's not a legal opinion nor should it be considered legal advice or binding on anyone but UCB. If FFTW is causing the memory issues we can go back to using the ooura FFT, but it will be slower than the Neon FFTW routines.
ID: 46580 · Report as offensive
Profile Mr. Kevvy
Volunteer tester
Avatar

Send message
Joined: 21 Apr 13
Posts: 23
Credit: 2,253,909
RAC: 0
Canada
Message 46581 - Posted: 4 Aug 2013, 12:27:54 UTC
Last modified: 4 Aug 2013, 12:33:02 UTC

Yay new beta!

Having a surplus of 'Droids these days I've attached three: A Google Nexus 7, Samsung Galaxy Tab 3 and Kyocera Rise phone (unfortunately I only rooted the phone so far.) Using the official BOINC app. from Play.

Edit: Make that four devices... forgot my cheap D2-912 tablet. Well, it's forgettable. :^)
ID: 46581 · Report as offensive
Profile Raistmer
Volunteer tester
Avatar

Send message
Joined: 18 Aug 05
Posts: 2423
Credit: 15,878,738
RAC: 0
Russia
Message 46582 - Posted: 5 Aug 2013, 9:13:18 UTC
Last modified: 5 Aug 2013, 9:28:11 UTC

Have first positive results for Android stock app:

http://setiweb.ssl.berkeley.edu/beta/show_host_detail.php?hostid=64310 So, Asus Transformer Pad can do it 4 at once (4 cores CPU) w/o issues.
Unfortunately I should return this device to owner soon and failed to convince her to install BOINC on permanent basis there :)

Another device, Viewsonic G-Tab:
http://setiweb.ssl.berkeley.edu/beta/show_host_detail.php?hostid=64312

again, no issues so far. On this device I hope to get offline speed comparisons between Metteusz's and stock apps later.

And one another G-tab, almost identical to prev one, has some issue:

http://setiweb.ssl.berkeley.edu/beta/results.php?hostid=64311

Elapsed time considerably bigger than CPU one. I had same issue on this device with Mateusz's app too before. He recommended to manually enable secondary CPU core (device rooted so it's possible to do). Now I checked both cores "online" but big discrepance between elapsed and CPU still. I think it's device own issue rather than app one, but any suggestions how to fix it are welcomed.

EDIT: another observation for Transformer device: big difference in completion time (~170ks vs ~100ks) when device works with all cores busy vs some cores free.
Both tasks are VLARs so I think should have similar work amount to do.
Perhaps, embedded frequency control for cores? Depending on device temp/power consumption? If so there is some room for hacking: if this embedded control could be disabled (provided good external cooling) device could perform considerably faster on full load.
ID: 46582 · Report as offensive
Profile Eric J Korpela
Volunteer moderator
Project administrator
Project developer
Project scientist
Avatar

Send message
Joined: 15 Mar 05
Posts: 1546
Credit: 26,470,697
RAC: 886
United States
Message 46584 - Posted: 5 Aug 2013, 15:25:34 UTC - in response to Message 46582.  

I was also wondering if this was due to a power saving mode or thermal control in some of these devices. I see this on a lot of android devices. It could even be how the BOINC thermal control setting manifests itself.

I don't see a huge problem with it. It'll boost granted credit if one of these devices validates against another, but I don't think we should mess with user or OS settings to fix it.

I'll probably get the new Android version out on Wednesday.
ID: 46584 · Report as offensive
Profile Raistmer
Volunteer tester
Avatar

Send message
Joined: 18 Aug 05
Posts: 2423
Credit: 15,878,738
RAC: 0
Russia
Message 46585 - Posted: 5 Aug 2013, 15:56:27 UTC - in response to Message 46584.  
Last modified: 5 Aug 2013, 15:56:39 UTC

I was also wondering if this was due to a power saving mode or thermal control in some of these devices. I see this on a lot of android devices. It could even be how the BOINC thermal control setting manifests itself.

Power saving mode is separate story on its own... working differently on different devices. For this Asus device I saw full stop of processing when "power" button pressed. And found no other way to turn screen off (w/o stopping computations) so processing was possible only with active screen.
From other side, G-tab and Samsung Galaxy mini can continue computations (and so far looks like even w/o freq drop) when power button pressed to switch off screen...

I don't think we should mess with user or OS settings to fix it.

Sure. Any such hacking should go from device owner/user himself, not from any BOINC's project. But to find what could be done in this direction would be nice.
ID: 46585 · Report as offensive
Claggy
Volunteer tester

Send message
Joined: 29 May 06
Posts: 1037
Credit: 8,438,782
RAC: 26
United Kingdom
Message 46586 - Posted: 5 Aug 2013, 17:30:50 UTC - in response to Message 46585.  

I was also wondering if this was due to a power saving mode or thermal control in some of these devices. I see this on a lot of android devices. It could even be how the BOINC thermal control setting manifests itself.

Power saving mode is separate story on its own... working differently on different devices. For this Asus device I saw full stop of processing when "power" button pressed. And found no other way to turn screen off (w/o stopping computations) so processing was possible only with active screen.
From other side, G-tab and Samsung Galaxy mini can continue computations (and so far looks like even w/o freq drop) when power button pressed to switch off screen...

I don't think we should mess with user or OS settings to fix it.

Sure. Any such hacking should go from device owner/user himself, not from any BOINC's project. But to find what could be done in this direction would be nice.

You might want to try the Official Boinc on a couple of your devices, it doesn't link computation to the screen being active,

Claggy
ID: 46586 · Report as offensive
Profile Raistmer
Volunteer tester
Avatar

Send message
Joined: 18 Aug 05
Posts: 2423
Credit: 15,878,738
RAC: 0
Russia
Message 46587 - Posted: 5 Aug 2013, 19:17:01 UTC - in response to Message 46586.  

NativeBOINC doesn't link to screen also. It's device feature, it put CPU to sleep perhaps when screen switched off. Very doubtful that "official" BOINC will behave differently on that device. And I returned it to owner already so no more experiments with Asus Transformer.
ID: 46587 · Report as offensive
Profile Eric J Korpela
Volunteer moderator
Project administrator
Project developer
Project scientist
Avatar

Send message
Joined: 15 Mar 05
Posts: 1546
Credit: 26,470,697
RAC: 886
United States
Message 46592 - Posted: 7 Aug 2013, 15:30:56 UTC - in response to Message 46587.  

I think that's going to be an issue with a lot of Android tablets. Phones have to maintain some background tasks in order to handle phone calls rapidly. A tablet manufacturer, OTOH, can decide to do a full suspend when the screen blanks.
ID: 46592 · Report as offensive
Profile Eric J Korpela
Volunteer moderator
Project administrator
Project developer
Project scientist
Avatar

Send message
Joined: 15 Mar 05
Posts: 1546
Credit: 26,470,697
RAC: 886
United States
Message 46593 - Posted: 7 Aug 2013, 23:54:23 UTC - in response to Message 46592.  

I've released 7.07, which on small memory machines (<384MB) will use less than 58MB at most. It still has problems with the function timing on my 256MB machine, so I'll probably need further mods to only do function timing if memory is above 384MB, but to assume that the neon routines will be fastest on a machine with neon.
ID: 46593 · Report as offensive
Claggy
Volunteer tester

Send message
Joined: 29 May 06
Posts: 1037
Credit: 8,438,782
RAC: 26
United Kingdom
Message 46594 - Posted: 8 Aug 2013, 13:05:23 UTC - in response to Message 46593.  

I've released 7.07, which on small memory machines (<384MB) will use less than 58MB at most. It still has problems with the function timing on my 256MB machine, so I'll probably need further mods to only do function timing if memory is above 384MB, but to assume that the neon routines will be fastest on a machine with neon.

Are there any tasks available? as Both the Scheduler and the Server Status page says there aren't any available.

Maybe it's time you put a limit of in progress tasks here, it'll stop CElliott from being So plain Greedy and hoovering almost half of the inprogress Wu's:

In progress tasks for computer 58103: 4336

In progress tasks for computer 58772: 3929

In progress tasks for computer 52900: 6166

In progress tasks for computer 9196: 2456

Total: 16,889 out of 37,896 Results in progress.

Claggy
ID: 46594 · Report as offensive
Richard Haselgrove
Volunteer tester

Send message
Joined: 3 Jan 07
Posts: 1444
Credit: 3,264,298
RAC: 0
United Kingdom
Message 46596 - Posted: 8 Aug 2013, 13:43:25 UTC - in response to Message 46594.  

With average turnaround times of

20.51 days
15.17 days
41.21 days
20.08 days

respectively. That doesn't help anybody.
ID: 46596 · Report as offensive
Profile Raistmer
Volunteer tester
Avatar

Send message
Joined: 18 Aug 05
Posts: 2423
Credit: 15,878,738
RAC: 0
Russia
Message 46598 - Posted: 8 Aug 2013, 17:55:59 UTC - in response to Message 46596.  

+1
Even my phone has smaller turnaround times :P
ID: 46598 · Report as offensive
Profile Eric J Korpela
Volunteer moderator
Project administrator
Project developer
Project scientist
Avatar

Send message
Joined: 15 Mar 05
Posts: 1546
Credit: 26,470,697
RAC: 886
United States
Message 46599 - Posted: 8 Aug 2013, 18:03:34 UTC

I figured out the problem with the splitter, so there should be more work out there.


ID: 46599 · Report as offensive
TRuEQ & TuVaLu
Volunteer tester
Avatar

Send message
Joined: 28 Jan 11
Posts: 619
Credit: 2,580,051
RAC: 0
Sweden
Message 46601 - Posted: 9 Aug 2013, 9:46:51 UTC

Still kinda strange.
My phone is suspended and visible on web page.
Wife's phone I click update and the webpage updates the time last conected.
But the task that is in progress does not show....
Only the tasks from my phone.

Some id of the 2 sony ericsson phones might be needed.

I will detach and re attach my phone now.
Second phone will complete it's task first.
ID: 46601 · Report as offensive
TRuEQ & TuVaLu
Volunteer tester
Avatar

Send message
Joined: 28 Jan 11
Posts: 619
Credit: 2,580,051
RAC: 0
Sweden
Message 46602 - Posted: 9 Aug 2013, 9:57:18 UTC

More strange...

I re-attached phone 2 again....Got me a task.
Phone 1 got it's task abandoned.
ID: 46602 · Report as offensive
Claggy
Volunteer tester

Send message
Joined: 29 May 06
Posts: 1037
Credit: 8,438,782
RAC: 26
United Kingdom
Message 46604 - Posted: 9 Aug 2013, 12:48:07 UTC - in response to Message 46602.  
Last modified: 9 Aug 2013, 12:48:34 UTC

More strange...

I re-attached phone 2 again....Got me a task.
Phone 1 got it's task abandoned.

Both hosts are using the same host ID, Try Boinc 7.2.9:

client (Android): allow GUI to pass MAC address to client via set_host_info(). If domain name is "localhost" and MAC address is nonempty, use MAC address instead of domain name when showing host on project web page


Claggy
ID: 46604 · Report as offensive
Profile Raistmer
Volunteer tester
Avatar

Send message
Joined: 18 Aug 05
Posts: 2423
Credit: 15,878,738
RAC: 0
Russia
Message 46605 - Posted: 9 Aug 2013, 14:03:17 UTC
Last modified: 9 Aug 2013, 14:13:10 UTC

Installed SETI beta on my Samsung Galaxy mini phone.
It downloaded 7.07 app. Unfortunately, it failed almost immediately:

<core_client_version>7.0.36</core_client_version>
<![CDATA[
<message>
process got signal 4
</message>
<stderr_txt>

</stderr_txt>
]]>


Host: http://setiweb.ssl.berkeley.edu/beta/show_host_detail.php?hostid=64393

Need to note that this device works OK with SETI v7 tasks on SETI main running Mateusz's custom app.

EDIT: memory freeing and phone rebooting didn't help.
ID: 46605 · Report as offensive
TRuEQ & TuVaLu
Volunteer tester
Avatar

Send message
Joined: 28 Jan 11
Posts: 619
Credit: 2,580,051
RAC: 0
Sweden
Message 46606 - Posted: 9 Aug 2013, 15:56:31 UTC - in response to Message 46604.  
Last modified: 9 Aug 2013, 16:04:29 UTC

More strange...

I re-attached phone 2 again....Got me a task.
Phone 1 got it's task abandoned.

Both hosts are using the same host ID, Try Boinc 7.2.9:

client (Android): allow GUI to pass MAC address to client via set_host_info(). If domain name is "localhost" and MAC address is nonempty, use MAC address instead of domain name when showing host on project web page


Claggy


I will. As soon it is available on Google play apps page.

Thank you!!
ID: 46606 · Report as offensive
Profile Eric J Korpela
Volunteer moderator
Project administrator
Project developer
Project scientist
Avatar

Send message
Joined: 15 Mar 05
Posts: 1546
Credit: 26,470,697
RAC: 886
United States
Message 46609 - Posted: 9 Aug 2013, 18:48:55 UTC - in response to Message 46605.  

Signal 4: Illegal instruction.

Looking at your phone, it's an ARMv6 processor. The stock BOINC requires ARMv7, so I had compiled for that. I'll see if I can make a ARMv6 compile. Shouldn't be difficult and shouldn't change performance significantly.
ID: 46609 · Report as offensive
Previous · 1 · 2 · 3 · 4 · 5 . . . 11 · Next

Message boards : News : SETI@home v7 7.06 released for ARMv7 Android


 
©2018 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.