Please use stock apps in SETI@home Beta.

Message boards : News : Please use stock apps in SETI@home Beta.
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3 · 4 · 5 · Next

AuthorMessage
Profile Raistmer
Volunteer tester
Avatar

Send message
Joined: 18 Aug 05
Posts: 2423
Credit: 15,878,738
RAC: 0
Russia
Message 43295 - Posted: 13 Jul 2012, 11:03:14 UTC

Increased cache to max, got another ati_opencl_100 task.
Initial estimation is: 742h
742/24 = 30 days.
deadline is 7.08.2012. So, less than 30 days until deadline.

This rises 2 questions and 2 bugreports:

1) why so big initial estimation? it's a first bug.
2) why server sent me this task at all? it's second bug, server should refuse to send work to host if it thinks it will take more time than remained to deadline.

In my case server thinks that task will take 30 days, less than 30 days until deadline but task was sent !
[perhaps with so screwed estimations no one should recive ATi work at all :D, or server just ignores its own estimation... maybe it's even good and not a bug but nice feature instead ;) ]
News about SETI opt app releases: https://twitter.com/Raistmer
ID: 43295 · Report as offensive
Josef W. Segur
Volunteer tester

Send message
Joined: 14 Oct 05
Posts: 1137
Credit: 1,848,733
RAC: 0
United States
Message 43300 - Posted: 13 Jul 2012, 14:34:42 UTC - in response to Message 43295.  

Increased cache to max, got another ati_opencl_100 task.
Initial estimation is: 742h
742/24 = 30 days.
deadline is 7.08.2012. So, less than 30 days until deadline.

This rises 2 questions and 2 bugreports:

1) why so big initial estimation? it's a first bug.
2) why server sent me this task at all? it's second bug, server should refuse to send work to host if it thinks it will take more time than remained to deadline.

In my case server thinks that task will take 30 days, less than 30 days until deadline but task was sent !
[perhaps with so screwed estimations no one should recive ATi work at all :D, or server just ignores its own estimation... maybe it's even good and not a bug but nice feature instead ;) ]

The server always sends at least one task if the client is requesting work (and there's a suitable application). That's fundamental to BOINC's design, BOINC has no idea how a new host will perform for instance. The server only looks at the estimated run time after assigning a task, subtracts it from the request to see if it should send another.

You don't yet have enough "completed" for the servers to use the APR. With the standard BOINC server code, the flops sent in the app_version would be based on an average of all hosts which do have enough completed. However, I think that's the area Eric is trying to improve so there may be more factors now. Please take a look in client_state.xml and say what the flops are for those apps, also what your DCF is since that is also still affecting estimates.

The bad estimates only affect work requesting really, the BOINC client doesn't discard a task just because the estimate is longer than deadline. You could edit DCF to a very small number which gives a reasonable estimate and get normal work requests that way, but I suggest setting a short cache if you do. That's to protect against the transition in server logic when the host gets 10 or 11 "completed".
                                                                  Joe
ID: 43300 · 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: 27,079,749
RAC: 4,356
United States
Message 43301 - Posted: 13 Jul 2012, 14:40:18 UTC - in response to Message 43300.  

I've at least found some clues as to the bad work requests and credit grants. It appears the plan_class_spec mechanism isn't consistent about whether it is using CPU speed or GPU speed estimates for those things. So work estimates that should be based on GPU Flops*GPU FLOPS Scale, are being based on CPU Flops*GPU FLOPS Scale.

The code is spaghetti, so I haven't figured out how to fix it yet. I probably won't get time to look at it until tomorrow (Saturday) morning.
ID: 43301 · Report as offensive
BetelgeuseFive
Volunteer tester

Send message
Joined: 3 Jun 12
Posts: 64
Credit: 2,513,247
RAC: 504
Netherlands
Message 43326 - Posted: 15 Jul 2012, 12:31:12 UTC

I don't know what has changed, but estimates for new tasks are pretty accurate now. Credits are all over the place however:

Peanuts: http://setiweb.ssl.berkeley.edu/beta/workunit.php?wuid=4049598
Jackpot: http://setiweb.ssl.berkeley.edu/beta/workunit.php?wuid=4062790

Tom
ID: 43326 · 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: 27,079,749
RAC: 4,356
United States
Message 43327 - Posted: 15 Jul 2012, 20:46:18 UTC - in response to Message 43326.  

It appears that garbage will continue for a while at least. At least the jackpot results seem more frequent.
ID: 43327 · Report as offensive
Matthias Lehmkuhl
Volunteer tester

Send message
Joined: 15 Jul 05
Posts: 176
Credit: 1,674,830
RAC: 0
Germany
Message 43332 - Posted: 16 Jul 2012, 8:42:50 UTC

Got 50 AP GPU results after reseting the project with 45 sec. estimated runtime.

the manual download of the ap_grafics exe for the client does not work.
seti beta likes to overwrite the manually downloaded file with the corrupt one again.
so I need to reset the project to get a new start.
Matthias

ID: 43332 · Report as offensive
Matthias Lehmkuhl
Volunteer tester

Send message
Joined: 15 Jul 05
Posts: 176
Credit: 1,674,830
RAC: 0
Germany
Message 43333 - Posted: 16 Jul 2012, 11:04:08 UTC - in response to Message 43332.  

Got 50 AP GPU results after reseting the project with 45 sec. estimated runtime.

the manual download of the ap_grafics exe for the client does not work.
seti beta likes to overwrite the manually downloaded file with the corrupt one again.
so I need to reset the project to get a new start.


First results have downloaded and calculated,
http://setiweb.ssl.berkeley.edu/beta/result.php?resultid=11029728
http://setiweb.ssl.berkeley.edu/beta/result.php?resultid=11029730
and some more results
all my results got this error

<core_client_version>6.12.34</core_client_version>
<![CDATA[
<message>
Maximum elapsed time exceeded
</message>
<stderr_txt>
Running on device number: 0
DATA_CHUNK_UNROLL at default:2
DATA_CHUNK_UNROLL at default:2
Matthias

ID: 43333 · Report as offensive
Profile Raistmer
Volunteer tester
Avatar

Send message
Joined: 18 Aug 05
Posts: 2423
Credit: 15,878,738
RAC: 0
Russia
Message 43334 - Posted: 16 Jul 2012, 11:35:34 UTC
Last modified: 16 Jul 2012, 11:36:06 UTC

<message>
Maximum elapsed time exceeded
</message>


just very same eror I feared because of incorrect estimations...
ID: 43334 · Report as offensive
kevin6912
Volunteer tester

Send message
Joined: 11 Aug 08
Posts: 4
Credit: 3,402,282
RAC: 0
United States
Message 43335 - Posted: 16 Jul 2012, 13:35:29 UTC

To bad each these "Max elapsed time exceeded" errors are dropping the max tasks per day value for the application.
ID: 43335 · Report as offensive
Profile CElliott
Volunteer tester

Send message
Joined: 16 Aug 05
Posts: 79
Credit: 71,936,490
RAC: 0
United States
Message 43339 - Posted: 16 Jul 2012, 16:23:24 UTC - in response to Message 43181.  

Pls look at http://setiweb.ssl.berkeley.edu/beta/workunit.php?wuid=4051356. The server assigned 700 credits to a WU that errored out and 53.99 credits to those who compled the WU successfully.
ID: 43339 · Report as offensive
Matthias Lehmkuhl
Volunteer tester

Send message
Joined: 15 Jul 05
Posts: 176
Credit: 1,674,830
RAC: 0
Germany
Message 43342 - Posted: 16 Jul 2012, 21:19:09 UTC

got also 700 credits for my "Maximum elapsed time exceeded" results
sample:
http://setiweb.ssl.berkeley.edu/beta/workunit.php?wuid=4064191
Matthias

ID: 43342 · 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: 27,079,749
RAC: 4,356
United States
Message 43343 - Posted: 16 Jul 2012, 21:34:39 UTC - in response to Message 43342.  

Must be a problem in my script. It should be giving 700 credits to any result that got less than 300 credits.

The scheduler and validator have been a mess today due to an extra "," in a SQL statement. That bug was my fault. Hopefully we're back on track again.

Eric
ID: 43343 · Report as offensive
Matthias Lehmkuhl
Volunteer tester

Send message
Joined: 15 Jul 05
Posts: 176
Credit: 1,674,830
RAC: 0
Germany
Message 43345 - Posted: 17 Jul 2012, 8:18:40 UTC

146 SETI@home Beta Test 17.07.2012 10:13:52 Requesting new tasks for NVIDIA GPU
147 SETI@home Beta Test 17.07.2012 10:13:55 Scheduler request completed: got 0 new tasks
148 SETI@home Beta Test 17.07.2012 10:13:55 No tasks sent
149 SETI@home Beta Test 17.07.2012 10:13:55 No tasks are available for AstroPulse v6
150 SETI@home Beta Test 17.07.2012 10:13:55 This computer has finished a daily quota of 1 tasks
http://setiweb.ssl.berkeley.edu/beta/results.php?hostid=53666
Matthias

ID: 43345 · Report as offensive
Urs Echternacht
Volunteer tester
Avatar

Send message
Joined: 18 Jan 06
Posts: 1038
Credit: 18,734,730
RAC: 0
Germany
Message 43347 - Posted: 17 Jul 2012, 13:04:12 UTC - in response to Message 43343.  

Must be a problem in my script. It should be giving 700 credits to any result that got less than 300 credits.

The scheduler and validator have been a mess today due to an extra "," in a SQL statement. That bug was my fault. Hopefully we're back on track again.

Eric

Guess it would be also a good idea to "limit" the credit-"jackpots" if i look at this one :

http://setiweb.ssl.berkeley.edu/beta/workunit.php?wuid=4064599


_\|/_
U r s
ID: 43347 · Report as offensive
Grumpy Swede
Volunteer tester
Avatar

Send message
Joined: 10 Mar 12
Posts: 1666
Credit: 13,132,357
RAC: 12,879
Sweden
Message 43349 - Posted: 17 Jul 2012, 13:40:41 UTC - in response to Message 43345.  
Last modified: 17 Jul 2012, 13:41:40 UTC

146 SETI@home Beta Test 17.07.2012 10:13:52 Requesting new tasks for NVIDIA GPU
147 SETI@home Beta Test 17.07.2012 10:13:55 Scheduler request completed: got 0 new tasks
148 SETI@home Beta Test 17.07.2012 10:13:55 No tasks sent
149 SETI@home Beta Test 17.07.2012 10:13:55 No tasks are available for AstroPulse v6
150 SETI@home Beta Test 17.07.2012 10:13:55 This computer has finished a daily quota of 1 tasks
http://setiweb.ssl.berkeley.edu/beta/results.php?hostid=53666


You have errored out, and/or aborted 96 tasks, that's why the quota is down to just one task. Finish some work, and the quota will go up again.
ID: 43349 · Report as offensive
Matthias Lehmkuhl
Volunteer tester

Send message
Joined: 15 Jul 05
Posts: 176
Credit: 1,674,830
RAC: 0
Germany
Message 43350 - Posted: 17 Jul 2012, 14:05:04 UTC - in response to Message 43349.  

150 SETI@home Beta Test 17.07.2012 10:13:55 This computer has finished a daily quota of 1 tasks
http://setiweb.ssl.berkeley.edu/beta/results.php?hostid=53666


You have errored out, and/or aborted 96 tasks, that's why the quota is down to just one task. Finish some work, and the quota will go up again.


Sten-Arne, yes you are right. It depends on the problem of the estimated runtime that was to short for GPU AP results.


OK, got one GPU result for this computer. So it seem to be not that problem as I thought.
And this GPU result has a "normal" estimated runtime of round 12 hours.
Matthias

ID: 43350 · Report as offensive
Matthias Lehmkuhl
Volunteer tester

Send message
Joined: 15 Jul 05
Posts: 176
Credit: 1,674,830
RAC: 0
Germany
Message 43354 - Posted: 18 Jul 2012, 12:06:42 UTC - in response to Message 43347.  


Guess it would be also a good idea to "limit" the credit-"jackpots" if i look at this one :

http://setiweb.ssl.berkeley.edu/beta/workunit.php?wuid=4064599


Found also one of these WU
http://setiweb.ssl.berkeley.edu/beta/workunit.php?wuid=4051027
Matthias

ID: 43354 · Report as offensive
Urs Echternacht
Volunteer tester
Avatar

Send message
Joined: 18 Jan 06
Posts: 1038
Credit: 18,734,730
RAC: 0
Germany
Message 43355 - Posted: 18 Jul 2012, 14:50:09 UTC - in response to Message 43354.  

Found also one of these WU
http://setiweb.ssl.berkeley.edu/beta/workunit.php?wuid=4051027

That is how it started for me a few days ago. But i'm more worried about the currently still rising credits :
2000, 2100 ... 7700 ... 20000 ... 23000 ... 30000+ /wu, there seems to be no limit !
Here is a list of my results :
http://setiweb.ssl.berkeley.edu/beta/workunit.php?wuid=4062488
http://setiweb.ssl.berkeley.edu/beta/workunit.php?wuid=4062492
http://setiweb.ssl.berkeley.edu/beta/workunit.php?wuid=4062501
http://setiweb.ssl.berkeley.edu/beta/workunit.php?wuid=4062614
http://setiweb.ssl.berkeley.edu/beta/workunit.php?wuid=4063093
http://setiweb.ssl.berkeley.edu/beta/workunit.php?wuid=4064608
http://setiweb.ssl.berkeley.edu/beta/workunit.php?wuid=4064599
http://setiweb.ssl.berkeley.edu/beta/workunit.php?wuid=4064774
http://setiweb.ssl.berkeley.edu/beta/workunit.php?wuid=4064198
http://setiweb.ssl.berkeley.edu/beta/workunit.php?wuid=4064611

Does not seem fair to me to get that much credit for some results while other results are credited normally.

Here is another issue i'm watching (colored).

SETI@home v7 6.97 windows_intelx86
Number of tasks completed 74
Max tasks per day 111
Number of tasks today 19
Consecutive valid tasks 80
Average processing rate 13.97113031241
mittlere Taskdauer 23.78 days
AstroPulse v6 6.01 windows_intelx86
Number of tasks completed 13
Max tasks per day 43
Number of tasks today 9
Consecutive valid tasks 12
Average processing rate 35.949628117514
mittlere Taskdauer 7.27 days
AstroPulse v6 6.04 windows_intelx86 (opencl_ati_100)
Number of tasks completed 10
Max tasks per day 7038
Number of tasks today 8
Consecutive valid tasks 20
Average processing rate 386.55599890193
mittlere Taskdauer 0.41 days
AstroPulse v6 6.04 windows_intelx86 (ati_opencl_100)
Number of tasks completed 31
Max tasks per day 87
Number of tasks today 0
Consecutive valid tasks 54
Average processing rate 446.74331174927
mittlere Taskdauer 0.43 days

_\|/_
U r s
ID: 43355 · 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: 27,079,749
RAC: 4,356
United States
Message 43356 - Posted: 18 Jul 2012, 16:11:12 UTC - in response to Message 43355.  

Yeah, I'll need to build something that will fix these as well. Some of the jackpot workunits are getting credit at rate seen in other projects (which shall remain nameless) that seem to think GPUs actually perform work at their theoretical rates.
ID: 43356 · Report as offensive
Matthias Lehmkuhl
Volunteer tester

Send message
Joined: 15 Jul 05
Posts: 176
Credit: 1,674,830
RAC: 0
Germany
Message 43359 - Posted: 18 Jul 2012, 20:33:09 UTC - in response to Message 43350.  
Last modified: 18 Jul 2012, 20:42:21 UTC

And this GPU result has a "normal" estimated runtime of round 12 hours.


Now I'm getting again short time results (estimated runtime 00:01:41)
Got 17 results with this short GPU runtime.
Sample:
http://setiweb.ssl.berkeley.edu/beta/result.php?resultid=11033798

Edit:
here is one result with "Maximum elapsed time exceeded" on AstroPulse v6 v6.04 (ati_opencl_100) of the WU where I got one result.
http://setiweb.ssl.berkeley.edu/beta/result.php?resultid=11030228

Edit2:
One point more, Max tasks per day = 2 is set for this AstroPulse v6 6.04 windows_intelx86 (cuda_opencl_100)
So how can I get 17 results of this App?
Matthias

ID: 43359 · Report as offensive
Previous · 1 · 2 · 3 · 4 · 5 · Next

Message boards : News : Please use stock apps in SETI@home Beta.


 
©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.