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 . . . 7 · 8 · 9 · 10 · 11 · Next

AuthorMessage
Profile Eric J Korpela
Volunteer moderator
Project administrator
Project developer
Project scientist
Avatar

Send message
Joined: 15 Mar 05
Posts: 1547
Credit: 27,183,456
RAC: 0
United States
Message 46825 - Posted: 9 Sep 2013, 17:23:49 UTC - in response to Message 46820.  
Last modified: 9 Sep 2013, 17:39:36 UTC


My first task just completed and I got about 2 cr/hr. This compares with 6 cr/hr on WCG, 9 cr/hr on Asteroids and 14 cr/hr for Skynet POGS.


Based on your floating point speed ( 159 million ops/sec ) you should be getting 31.9 cr/day or 1.33 cr/hr. Based upon your integer speed (1.490) billion ops/sec you should be getting 300 cr/day or 12.5 cr/hr. Averaged you would get 6.9 credits per hour.

Since SETI@home is mostly floating point operations and is compared against machines that do FLOPS as fast as they do IOPS, it's going to end up close to the floating point benchmarks.

The other projects probably have a separate app for their android jobs so they don't compare to other processors, and grant credit based on the average benchmark, which assumes equal FLOPS and IOPS and that no IOP ever waits for a FLOP to complete.

And then there are the projects that just grant constant credit for their workunits and set the value where ever they want it.

So cross project credits don't mean much any more.
ID: 46825 · Report as offensive
Profile Eric J Korpela
Volunteer moderator
Project administrator
Project developer
Project scientist
Avatar

Send message
Joined: 15 Mar 05
Posts: 1547
Credit: 27,183,456
RAC: 0
United States
Message 46826 - Posted: 9 Sep 2013, 17:29:50 UTC - in response to Message 46823.  

First two sets of tasks logged bugs but completed and validated fine. The third set of tasks crunched for three minutes with no progress then Bugcatcher announced some bugs and then the progress bar started to move.

Bugcatcher reports.


A "no progress" part in the beginning is normal while FFTW does its planning and SETI@home chooses its fastest calculation routines. There are at least two deliberate illegal instruction exceptions (to test exception handling) during that time. I don't know if Bugcatcher catches those exceptions (which are caught and handled by SETI@home) or if it would ignore them.
ID: 46826 · Report as offensive
Alex Storey
Volunteer tester
Avatar

Send message
Joined: 10 Feb 12
Posts: 107
Credit: 305,151
RAC: 0
Greece
Message 46827 - Posted: 9 Sep 2013, 18:30:06 UTC - in response to Message 46768.  

Do the number of errors per task (number of times a task is sent out) also need to be increased? So as not to let good tasks die an unwarranted death?


Cassandra is back! :)

Too many errors (may have bug)
http://setiweb.ssl.berkeley.edu/beta/workunit.php?wuid=5671370
http://setiweb.ssl.berkeley.edu/beta/workunit.php?wuid=5671416
http://setiweb.ssl.berkeley.edu/beta/workunit.php?wuid=5671549
http://setiweb.ssl.berkeley.edu/beta/workunit.php?wuid=5671693
http://setiweb.ssl.berkeley.edu/beta/workunit.php?wuid=5671796

So that's 5 out of 28 tasks I've managed to error out or about 1 in every 5 tasks considering a few of those 28 haven't validated yet. Fortunately 3 of the 5 errors never got done by anybody but 2 are giving some poor guy "Completed, can't validate". Just throwing this info out there in case it's useful.

Also keep in mind the onslaught of The Android Army might make the situation worsen steadily... Front page news on Main makes it to the Notices tab in Boinc which means (for me at least, on Windows 7) I get a balloon message in the taskbar on my desktop. This may turn out to be a very popular Beta Test:)
ID: 46827 · Report as offensive
Alex Storey
Volunteer tester
Avatar

Send message
Joined: 10 Feb 12
Posts: 107
Credit: 305,151
RAC: 0
Greece
Message 46828 - Posted: 9 Sep 2013, 19:40:59 UTC

To drive the point home and to humor you guys a little bit I combed through every single errored task I had, checking the owners of each computer.

Number of errored taskes: 28
Number of times I was paired with CElliott: Zero

:)
ID: 46828 · Report as offensive
Profile Eric J Korpela
Volunteer moderator
Project administrator
Project developer
Project scientist
Avatar

Send message
Joined: 15 Mar 05
Posts: 1547
Credit: 27,183,456
RAC: 0
United States
Message 46829 - Posted: 9 Sep 2013, 19:45:44 UTC - in response to Message 46828.  
Last modified: 9 Sep 2013, 19:46:29 UTC

Looks like I'll have to fire up the credit granting script for the next few days.
ID: 46829 · Report as offensive
Profile Steve Hawker*
Volunteer tester

Send message
Joined: 31 Oct 12
Posts: 8
Credit: 1,055,466
RAC: 0
United States
Message 46830 - Posted: 9 Sep 2013, 21:22:59 UTC - in response to Message 46825.  


My first task just completed and I got about 2 cr/hr. This compares with 6 cr/hr on WCG, 9 cr/hr on Asteroids and 14 cr/hr for Skynet POGS.


Based on your floating point speed ( 159 million ops/sec ) you should be getting 31.9 cr/day or 1.33 cr/hr. Based upon your integer speed (1.490) billion ops/sec you should be getting 300 cr/day or 12.5 cr/hr. Averaged you would get 6.9 credits per hour.

Since SETI@home is mostly floating point operations and is compared against machines that do FLOPS as fast as they do IOPS, it's going to end up close to the floating point benchmarks.


Really interesting. Thanks for taking the time to respond in such detail, I very much appreciate it and all your efforts here of course.

Maybe this has been answered already but when do you think this beta app will make it to the major league?

Thanks,

Steve
ID: 46830 · Report as offensive
Profile Eric J Korpela
Volunteer moderator
Project administrator
Project developer
Project scientist
Avatar

Send message
Joined: 15 Mar 05
Posts: 1547
Credit: 27,183,456
RAC: 0
United States
Message 46831 - Posted: 9 Sep 2013, 23:18:46 UTC - in response to Message 46830.  

Once we get some of the bugs ironed out we'll release it. Hopefully weeks rather than months.
ID: 46831 · Report as offensive
Profile Dingo
Volunteer tester
Avatar

Send message
Joined: 31 Oct 07
Posts: 5
Credit: 290,257
RAC: 0
Australia
Message 46832 - Posted: 10 Sep 2013, 4:18:08 UTC

I had good results with my Android U2 but I stopped work for a while. I started again a couple of days ago and all but one of the work units have aborted with

Outcome Computation error
Client state Compute error
Exit status 193 (0xc1) EXIT_SIGNAL


http://setiweb.ssl.berkeley.edu/beta/results.php?userid=7648


Proud Founder and member of



Have a look at my WebCam
ID: 46832 · Report as offensive
Profile Raistmer
Volunteer tester
Avatar

Send message
Joined: 18 Aug 05
Posts: 2423
Credit: 15,878,738
RAC: 0
Russia
Message 46835 - Posted: 10 Sep 2013, 8:18:47 UTC - in response to Message 46825.  


So cross project credits don't mean much any more.

Just as in the past. They never meant anything. So, flopcounting for SETI not bad option. CreditNew is futile.
ID: 46835 · Report as offensive
Profile Raistmer
Volunteer tester
Avatar

Send message
Joined: 18 Aug 05
Posts: 2423
Credit: 15,878,738
RAC: 0
Russia
Message 46836 - Posted: 10 Sep 2013, 8:38:04 UTC
Last modified: 10 Sep 2013, 8:40:24 UTC

And more on topic.
Now I have 3 "identical" G-tabs under testing. One of them still experience regular freezes of 7.08 apps. I need to restart BOINC few times daily.
But custom app never freezes. Also, never noticed 7.07 freeze so far. After restart elapsed time greatly reduced (that means no checkpointing during freeze). What I don't quite understand - why BOINC (being so "careful" about windows GPU apps and killing them too often w/o real need) ignores this app freezes. I would say it's some kind of design flaw in it. Looks like "no heartbeat" exit totally in app process own responsibility (timer thread inside process). So, when process hard locked for some reason thit measure doesn't work.

EDIT: and no such freezes on other device. Third one recived 7.09 instead, no freezes so far.
ID: 46836 · Report as offensive
Profile Eric J Korpela
Volunteer moderator
Project administrator
Project developer
Project scientist
Avatar

Send message
Joined: 15 Mar 05
Posts: 1547
Credit: 27,183,456
RAC: 0
United States
Message 46838 - Posted: 10 Sep 2013, 20:28:40 UTC - in response to Message 46836.  
Last modified: 10 Sep 2013, 20:37:24 UTC

Next time you get a freeze, could you suspend the workunit, wait a few minutes, and then unsuspend to see if it starts running. That might let me know if it's a workunit dependent freeze or dependent on some random (or non-random) variable (i.e. cycle counter/timer value, free memory, phase of the moon).
ID: 46838 · Report as offensive
Profile Raistmer
Volunteer tester
Avatar

Send message
Joined: 18 Aug 05
Posts: 2423
Credit: 15,878,738
RAC: 0
Russia
Message 46839 - Posted: 10 Sep 2013, 21:23:04 UTC - in response to Message 46838.  

Next time you get a freeze, could you suspend the workunit, wait a few minutes, and then unsuspend to see if it starts running. That might let me know if it's a workunit dependent freeze or dependent on some random (or non-random) variable (i.e. cycle counter/timer value, free memory, phase of the moon).

Please be more specific.
Currently I restart NativeBOINC clien via its GUI (shutdown client, start client). Then task continues from last checkpoint. Suspend resume single task will not restart app in current config cause it's "leave app in memory".
Do you want that I disable leaving task in memory?
ID: 46839 · Report as offensive
Profile Raistmer
Volunteer tester
Avatar

Send message
Joined: 18 Aug 05
Posts: 2423
Credit: 15,878,738
RAC: 0
Russia
Message 46840 - Posted: 10 Sep 2013, 21:37:36 UTC
Last modified: 10 Sep 2013, 21:39:09 UTC

Well, checking now I found device in "semi-freeze" state (usually both tasks are freezed). Now 7.08 continues to make progress, 7.07 frozen. I suspended 7.07 task, BOINC switched to custom 7.00 app from SETi main. Resuming 7.07 task didn't change anything, now it waits for SETI main task finish.

After SETI main task suspend 7.07 continues to run... but looks like it was restarted from last checkpoint (though "leave app in memory" enabled (!). Elasped time dropped back to ~1h from >2h that was before initial suspend.

Now both tasks (7.08, 7.07) make progress. Cause i see this situation only on single device quite probably it's hardware-related issue... but sad thing is BOINC's reaction - it doesn't restart app. App can be frozen many hours. I.e. only "max elapsed time exceeded" or operator attention will save device from wasting cycles . And "max elapsed time" for such device could be weeks... leaving "babysitting" as only real option.
ID: 46840 · Report as offensive
Profile Raistmer
Volunteer tester
Avatar

Send message
Joined: 18 Aug 05
Posts: 2423
Credit: 15,878,738
RAC: 0
Russia
Message 46841 - Posted: 10 Sep 2013, 21:44:02 UTC - in response to Message 46838.  
Last modified: 10 Sep 2013, 21:45:20 UTC

That might let me know if it's a workunit dependent freeze or dependent on ... phase of the moon.

Second more viable ;) All tasks that experienced such freeze were ultimately finished OK. Some of them after few restarts. So, resume from last checkpoint always allowed to continue progress. Looks like all those tasks were VLARs though. But last freeze broke this "tradition" too, not only it was 7.07 instead of 7.08, task wasn't VLAR.
ID: 46841 · Report as offensive
Profile Eric J Korpela
Volunteer moderator
Project administrator
Project developer
Project scientist
Avatar

Send message
Joined: 15 Mar 05
Posts: 1547
Credit: 27,183,456
RAC: 0
United States
Message 46843 - Posted: 10 Sep 2013, 22:34:20 UTC - in response to Message 46841.  

I've seen such frozen states on my low memory droid. They sit there taking very little memory and using no CPU. Because of the memory weirdness on android, I wondered if it was putting processes into a "wait until I have more free memory" state when they try to allocate too much. Of course I have no evidence to support that, though. Apart from never having seen it happen on the high memory machines I have.
ID: 46843 · Report as offensive
Profile Raistmer
Volunteer tester
Avatar

Send message
Joined: 18 Aug 05
Posts: 2423
Credit: 15,878,738
RAC: 0
Russia
Message 46850 - Posted: 11 Sep 2013, 6:32:41 UTC - in response to Message 46843.  

Is there some good system state (CPU load, available memory, temps and so on) logging tool for Android?
I could install it on all 3 devices to see why only one experience freezes.

Regarding big difference between elapsed and CPU times on one G-tab (no such difference on another one):
In Cyanogen mod settings under "erformance->CPU settings default CPU lowest freq is 216MHz and CPU highest freq is 1000MHz. On first Gtab highest freq is changed to 912MHz (perhaps, to provide more stable operation at times it crunched MW tasks). It's the single meaningful difference (besides different wallpapers on desktop :) ) I found so far. Strange, usually OS correctly converts CPU cycles into passed seconds, but here looks like OS thinks that CPU still makes 1e9 cycles per second while it makes less. That could explain difference between elapsed and CPU times (CPU time would be underestimated in this case).
ID: 46850 · Report as offensive
Jeffrey Dahn
Volunteer tester

Send message
Joined: 9 Sep 13
Posts: 2
Credit: 815
RAC: 0
Message 46878 - Posted: 11 Sep 2013, 15:59:26 UTC

On my HP slate 7 tablet, the status shows "computing", but at the same time it shows SETI suspended at 63.1% after 25:45 of running. What's happening?
ID: 46878 · Report as offensive
Jeffrey Dahn
Volunteer tester

Send message
Joined: 9 Sep 13
Posts: 2
Credit: 815
RAC: 0
Message 46879 - Posted: 11 Sep 2013, 16:03:16 UTC - in response to Message 46878.  

Never mind, I got it back up and running by manually suspending the work unit and then re-opening it.
ID: 46879 · Report as offensive
EdwardPF
Volunteer tester

Send message
Joined: 8 Sep 05
Posts: 82
Credit: 545,522
RAC: 0
United States
Message 46882 - Posted: 11 Sep 2013, 16:17:08 UTC - in response to Message 46879.  
Last modified: 11 Sep 2013, 16:18:47 UTC

flops / cpu sec = 284,310,872.53
Min's / CS = 21.33

4 down and verified working on v7.10 (all have failed so far)

Ed F
ID: 46882 · Report as offensive
Snug-Beta
Volunteer tester

Send message
Joined: 5 Sep 13
Posts: 1
Credit: 37,346
RAC: 0
Message 46904 - Posted: 12 Sep 2013, 20:54:41 UTC

Been using the beta on my Kindle Fire HD for a couple of weeks now, running on battery. Messaging with settings, running the battery low, suspending etc.

All seems good no problems and still humming away.

Good job guys, hopes it helps get through the back log.
ID: 46904 · Report as offensive
Previous · 1 . . . 7 · 8 · 9 · 10 · 11 · Next

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


 
©2020 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.