But this rises another question: should BOINC download CPU tasks on project attach if project preferencies say "don't download CPU tasks" ? BOINC obeys... but only when first tasks already downloaded.

AFAIK BOINC gets the preferences for the venue, to which the machine has been assigned, together with the first scheduler reply, so that behavior seems correct. You should eventually could see it in the log, when it gets an host ID assigned, I'm pretty sure it's not in the moment when it's attached to the project, but when it first asks for work.

---

Ok. I checked original version on brother's smartphone (with ARMv6 VFP). Unfortunatelly I got good results (without NaN's). Do you can tell me what was you crunching workunit (series)? I crunched ps_separation_09_2s workunit. Please check project directory (milkyway) and please do not suspend/resume workunit (too often) when crunched (it impact on the final results). You can simply detach/attach project or reinstall NativeBOINC.

Glad to see PrimeGrid added an app for The Riesel Problem (Sieve). Keep them coming!

Ok. primegrid 5.0 is broken (please lookup on the wiki news: https://github.com/matszpk/native-boinc-for-android/wiki/News), because had bug in the wrapper. Current version (5.1) handles properly unexpected app exits (for example kills by the system).

I have important message: I moved NativeBOINC to my new domain: nativeboinc.org.
New version of the NativeBOINC will be downloading binaries and news from files.nativeboinc.org. Currently I am preparing website for NativeBOINC.
New info at https://github.com/matszpk/native-boinc-for-android/wiki/News and fetched news by NativeBOINC.

Ok. I checked original version on brother's smartphone (with ARMv6 VFP). Unfortunatelly I got good results (without NaN's). Do you can tell me what was you crunching workunit (series)? I crunched ps_separation_09_2s workunit. Please check project directory (milkyway) and please do not suspend/resume workunit (too often) when crunched (it impact on the final results). You can simply detach/attach project or reinstall NativeBOINC.

Sorry for long delay, I was occupied with other things.
Of course can't recall what series were that time so will re-do this experiment again.
Unfortunately, MW is my the only chance to make use of this device for BOINC cause all other projects require too much free inner space and NativeBOINC still can't make use of SD card space.

EDIT: ps_separation_09_2s_sample2_* task was downloaded. Will report how it goes later.
EDIT2: Before re-attaching to MW NativeBOINC manager and core client were updates to latest revisions. Fresh MW app was downloaded at project attaching so I started with fresh MW installation this time.

Well, this task ran better than previous (I did not press power button through whole run, phone screen was always on, dimmed), but not perfect: http://milkyway.cs.rpi.edu/milkyway/workunit.php?wuid=233358927 as one can see, another hosts complete task faster (though it took less than 2 days to get result) and my task got validation error. So, it's unknown is it valid or invalid, server refused to check it.
From device side it finished w/o error and was uploaded so I think it could be validated ok, no computation error.

What strange: NativeBOINC was restarted few times (not only MW task, but very NativeBOINC itself). What caused such restart? Not sure. At least one time I saw kind of SMS message (screen popup) came around time NetiveBOINC was restarted. What caused another restarts - no idea. Of course, NativeBOINC restart means MW task restart too so it was impossible to run w/o restarts at all.

I captured different computation stages of this task as separate SD-card snapshots but cause there was no computation error I'm not sure if these snapshots interesting/useful.
NativeBOINC log doesn't show the cause of restart unfortunately.

Now I run another task from de_separation_09_2s_sample_2_*.
Will do SD-card snapshots while it goes again.
It's url: http://milkyway.cs.rpi.edu/milkyway/workunit.php?wuid=234070478

And this task completed with computation error.... It was restarted more times than prev one.

Also, looks like prev task was invalid.

What is strange: task with computation error now shows 0 seconds of elapsed time.
Also both CPU time and Last checkpoint (CPU time) in task properties are zero. But this task ran quite a few hours before computation error. Why all times are zeroed ? Can it be connected with damaged checkpoint somehow? Maybe checkpointing for this app build has some bug? (also, it shows zero progress too, though it was almost completed before computation error).

EDIT: on ARMv7 CPU in tablet I have no problems with validation:
http://milkyway.cs.rpi.edu/milkyway/results.php?hostid=454110

Maybe ARMv6 CPU can't provide required precision at all ?

Very likely, too many restarts caused incorrect results. Milkyway app doesnt save all work in the checkpoints (only part of precisions). This app uses the Kahan adding, which saves precision. Unfortunatelly it is not saved on the checkpoints.
But I can check again computations on the brother's smartphone to prove this thesis.
Thanks.

Since last post I did more tests. Few tasks were completed on that device working in usual manner. I don't think app was restarted many times for each task but few restarts per taks seem unavoidable (even with "untouched" state I saw some NativeBOINC restarts before).

So far all completed tasks were invalid ones. All those I saw finished with computation error.

Current host web page state:


Task
click for details
Show names Work unit
click for details Sent Time reported
or deadline
explain Status Run time
(sec) CPU time
(sec) Credit Application
304225172 241018442 22 Sep 2012 | 18:21:58 UTC 24 Sep 2012 | 9:12:47 UTC Error while computing 116,170.09 108,015.80 --- MilkyWay@Home
Anonymous platform (CPU)
298594345 236322884 14 Sep 2012 | 17:59:32 UTC 22 Sep 2012 | 18:21:58 UTC Error while computing 125,549.60 108,899.00 --- MilkyWay@Home
Anonymous platform (CPU)


Last result:

<core_client_version>6.12.38</core_client_version>
<![CDATA[
<message>
process exited with code 1 (0x1, -255)
</message>
<stderr_txt>
<search_application> milkyway_separation 1.00 Linux ARM double </search_application>
Unrecognized XML in project preferences: max_gfx_cpu_pct
Skipping: 0
Skipping: /max_gfx_cpu_pct
Unrecognized XML in project preferences: nbody_graphics_poll_period
Skipping: 30
Skipping: /nbody_graphics_poll_period
Unrecognized XML in project preferences: nbody_graphics_float_speed
Skipping: 5
Skipping: /nbody_graphics_float_speed
Unrecognized XML in project preferences: nbody_graphics_textured_point_size
Skipping: 250
Skipping: /nbody_graphics_textured_point_size
Unrecognized XML in project preferences: nbody_graphics_point_point_size
Skipping: 40
Skipping: /nbody_graphics_point_point_size
Error loading Lua script 'astronomy_parameters.txt': [string "number_parameters: 4..."]:1: '<name>' expected near '4'
Error reading astronomy parameters from file 'astronomy_parameters.txt'
Trying old parameters file
<search_application> milkyway_separation 1.00 Linux ARM double </search_application>
Unrecognized XML in project preferences: max_gfx_cpu_pct
Skipping: 0
Skipping: /max_gfx_cpu_pct
Unrecognized XML in project preferences: nbody_graphics_poll_period
Skipping: 30
Skipping: /nbody_graphics_poll_period
Unrecognized XML in project preferences: nbody_graphics_float_speed
Skipping: 5
Skipping: /nbody_graphics_float_speed
Unrecognized XML in project preferences: nbody_graphics_textured_point_size
Skipping: 250
Skipping: /nbody_graphics_textured_point_size
Unrecognized XML in project preferences: nbody_graphics_point_point_size
Skipping: 40
Skipping: /nbody_graphics_point_point_size
Error loading Lua script 'astronomy_parameters.txt': [string "number_parameters: 4..."]:1: '<name>' expected near '4'
Error reading astronomy parameters from file 'astronomy_parameters.txt'
Trying old parameters file
suspend activities
resume activities
suspend activities
resume activities
suspend activities
resume activities
Integral 0 time = 137721.704647 s
Failed to calculate integral 0
Failed to calculate likelihood
<background_integral> NaN </background_integral>
<stream_integral> NaN NaN </stream_integral>
<background_likelihood> NaN </background_likelihood>
<stream_only_likelihood> NaN NaN </stream_only_likelihood>
<search_likelihood> NaN </search_likelihood>
12:38:42 (12370): called boinc_finish

</stderr_txt>
]]>


Prev result:

<core_client_version>6.12.38</core_client_version>
<![CDATA[
<message>
process exited with code 1 (0x1, -255)
</message>
<stderr_txt>
<search_application> milkyway_separation 1.00 Linux ARM double </search_application>
Unrecognized XML in project preferences: max_gfx_cpu_pct
Skipping: 0
Skipping: /max_gfx_cpu_pct
Unrecognized XML in project preferences: nbody_graphics_poll_period
Skipping: 30
Skipping: /nbody_graphics_poll_period
Unrecognized XML in project preferences: nbody_graphics_float_speed
Skipping: 5
Skipping: /nbody_graphics_float_speed
Unrecognized XML in project preferences: nbody_graphics_textured_point_size
Skipping: 250
Skipping: /nbody_graphics_textured_point_size
Unrecognized XML in project preferences: nbody_graphics_point_point_size
Skipping: 40
Skipping: /nbody_graphics_point_point_size
Error loading Lua script 'astronomy_parameters.txt': [string "number_parameters: 4..."]:1: '<name>' expected near '4'
Error reading astronomy parameters from file 'astronomy_parameters.txt'
Trying old parameters file
<search_application> milkyway_separation 1.00 Linux ARM double </search_application>
Unrecognized XML in project preferences: max_gfx_cpu_pct
Skipping: 0
Skipping: /max_gfx_cpu_pct
Unrecognized XML in project preferences: nbody_graphics_poll_period
Skipping: 30
Skipping: /nbody_graphics_poll_period
Unrecognized XML in project preferences: nbody_graphics_float_speed
Skipping: 5
Skipping: /nbody_graphics_float_speed
Unrecognized XML in project preferences: nbody_graphics_textured_point_size
Skipping: 250
Skipping: /nbody_graphics_textured_point_size
Unrecognized XML in project preferences: nbody_graphics_point_point_size
Skipping: 40
Skipping: /nbody_graphics_point_point_size
Error loading Lua script 'astronomy_parameters.txt': [string "number_parameters: 4..."]:1: '<name>' expected near '4'
Error reading astronomy parameters from file 'astronomy_parameters.txt'
Trying old parameters file
<search_application> milkyway_separation 1.00 Linux ARM double </search_application>
Unrecognized XML in project preferences: max_gfx_cpu_pct
Skipping: 0
Skipping: /max_gfx_cpu_pct
Unrecognized XML in project preferences: nbody_graphics_poll_period
Skipping: 30
Skipping: /nbody_graphics_poll_period
Unrecognized XML in project preferences: nbody_graphics_float_speed
Skipping: 5
Skipping: /nbody_graphics_float_speed
Unrecognized XML in project preferences: nbody_graphics_textured_point_size
Skipping: 250
Skipping: /nbody_graphics_textured_point_size
Unrecognized XML in project preferences: nbody_graphics_point_point_size
Skipping: 40
Skipping: /nbody_graphics_point_point_size
Error loading Lua script 'astronomy_parameters.txt': [string "number_parameters: 4..."]:1: '<name>' expected near '4'
Error reading astronomy parameters from file 'astronomy_parameters.txt'
Trying old parameters file
suspend activities
<search_application> milkyway_separation 1.00 Linux ARM double </search_application>
Unrecognized XML in project preferences: max_gfx_cpu_pct
Skipping: 0
Skipping: /max_gfx_cpu_pct
Unrecognized XML in project preferences: nbody_graphics_poll_period
Skipping: 30
Skipping: /nbody_graphics_poll_period
Unrecognized XML in project preferences: nbody_graphics_float_speed
Skipping: 5
Skipping: /nbody_graphics_float_speed
Unrecognized XML in project preferences: nbody_graphics_textured_point_size
Skipping: 250
Skipping: /nbody_graphics_textured_point_size
Unrecognized XML in project preferences: nbody_graphics_point_point_size
Skipping: 40
Skipping: /nbody_graphics_point_point_size
Error loading Lua script 'astronomy_parameters.txt': [string "number_parameters: 4..."]:1: '<name>' expected near '4'
Error reading astronomy parameters from file 'astronomy_parameters.txt'
Trying old parameters file
suspend activities
resume activities
Integral 0 time = 101005.527139 s
Failed to calculate integral 0
Failed to calculate likelihood
<background_integral> NaN </background_integral>
<stream_integral> NaN NaN </stream_integral>
<background_likelihood> NaN </background_likelihood>
<stream_only_likelihood> NaN NaN </stream_only_likelihood>
<search_likelihood> NaN </search_likelihood>
06:28:08 (1095): called boinc_finish

</stderr_txt>
]]>

For last task I did whole NativeBOINC snapshots at ~92%, ~95% task completion and before reporting.
Will extract and upload MW folders.

Looks like checkpointing improving is essential.
From the other side, I'm almost sure there were no restarts between 95% and ultimate computation error. That is, some gradual loss of precision is possible, app not fails immediatelly on resuming from checkpoint, but after some more computations done, at very end of computations.

EDIT:
BTW, if you need to run some debug/verbose build for checking why error occurs I will glad to run such build on my device.

EDIT2:
Link with mentioned MW state snapshots: http://dl.dropbox.com/u/60381958/NativeBOINC_MW_app_debug.rar

ok. I prepared debug version: available at http://files.nativeboinc.org/debug/milkyway-debug-distrib.zip.
Anybody, who want to send me debug results (for debugging this app) can use this via NativeBOINC (update from SDCard) or as standalone application (running from terminal emulator or adb shell). This version creates 'debug_data.txt', which contains intermediate results (desired by me for debugging).

to Raistmer: I consider, that you problems (with milkyway) caused by hardware or
very strange OS incompatibility (I really dont know why). You can help me to investigate their reasons.

PS.: I am sorry for delayed answer. I am working on the SDCard installation feature (it needed testing, research :) )
Thanks.

ok. I prepared debug version: available at http://files.nativeboinc.org/debug/milkyway-debug-distrib.zip.
Anybody, who want to send me debug results (for debugging this app) can use this via NativeBOINC (update from SDCard) or as standalone application (running from terminal emulator or adb shell). This version creates 'debug_data.txt', which contains intermediate results (desired by me for debugging).

to Raistmer: I consider, that you problems (with milkyway) caused by hardware or
very strange OS incompatibility (I really dont know why). You can help me to investigate their reasons.

PS.: I am sorry for delayed answer. I am working on the SDCard installation feature (it needed testing, research :) )
Thanks.

Sure will do. I was distracted with other activity too, now will continue with NativeBOINC testing.
BTW, what preferable way to report issues?
I used GitHub issues before, now there is Trac (so I already created Trac ticket for some issue linked with new SD card ability). Trac ticket or GitHub Issue: what is more convenient to you ?

Tried to install debug binary into NativeBOINC via update binary from SD. Cant, it says no new binaries found

Ok. rename zip file to 'MilkyWay@Home.zip' or 'Milkyway@Home.zip'. One of the given names should be work. NativeBOINC searches binaries by project name or distribution project name.

Hi.
Used 100% OK on a generic 7" tablet and an ODROID-X.
Big, big thanks.

Ok. rename zip file to 'MilkyWay@Home.zip' or 'Milkyway@Home.zip'. One of the given names should be work. NativeBOINC searches binaries by project name or distribution project name.

Thanks, second one go OK.

What I should do to get useful debug info? Should I do snapshots on some % done or just suspend network and sent final state to you?

Hi,
NativeBOINC downloaded, installed and running great on my Ice Cream Motorolla Razr. Pauses when on battery, nice status screen and options, big thumbs up here awesome work.

R.

Simply, suspend workunit, dump boinc directory and get debug_data.txt on the slot directory of a workunit. I dont need debug for a fully crunched workunit.

Ok, here is the file: https://dl.dropbox.com/u/60381958/debug_data.txt.7z

I think you find some interesting there:

streamData:
0: sum=-9.378893700808814431e+60,corr=3.143412892622864462e-10
1: sum=-3.269667431103558338e+43,corr=-8.283224747184779746e-01
2: sum=4.564673771342388591e+02,corr=-1.648059257626835993e-14
BgProb: sum=1.499540830335913631e-04,corr4.068722784552303161e-21
step: nu=406,mu=0
streamData:
0: sum=-9.378893700808814431e+60,corr=-4.166763153152042154e-02
1: sum=-3.269667431103558338e+43,corr=-3.346770785495461809e+00
2: sum=-1.897804144654852500e+15,corr=-4.992987497661696694e-02
BgProb: sum=NaN,corrNaN
step: nu=407,mu=0
streamData:
0: sum=5.477781213074433216e+63,corr=-4.788975681913954402e-20
1: sum=-3.269667431103558338e+43,corr=9.789898484307859405e+01
2: sum=-1.897804144654973250e+15,corr=-6.339234217429283447e-02
BgProb: sum=NaN,corrNaN


Looks like loss of precision happened quite early in processing...

matszpk, do you have plans on getting nativeBOINC to work on the x86 Android version? I currently am testing 4.1.1 in a Virtualbox. Your app will download and start installing, but it gets to the point of unpacking native client and then never finishes. I have tried installing it the normal way and with the older device option as well with the same result. I'm sure as Intel continues to work on x86 handheld devices, this will become a necessity.