A few numbers for the Gamma-Ray pulsar search (FGRP1):
Validate errors:
- Windows: 0.4% of all results from this app (version)
- Linux: 6.2%
- Mac OS X: 7.8%
Invalid results (on average):
- Windows and Linux: <0.2%
- Mac OS X: 3.7%
As I stated a couple of times: running the FGRP App on the Mac is somewhat more risky than e.g. running it on Windows. We thought about disabling that (OSX) app version alltogether, but I think it's still better to have people decide it on their own. Opting out of FGRP1 via Einstein@Home preferences does work.
In the past weeks and months we had (and still have) all our hands full with work on general BOINC, project infrastructure and the other Apps of this project. I hope that someone will find the time for a closer look at the FGRP search in the next few weeks, but currently I can't promise anything.
When you opt out of FGRP1 with Mac OSX, you stop getting GWS6 LV tasks.
Not in genersl, but it could hsppen. But thst problem lies deep in BOINC istelf and is not easy to fix.
Technically the problem is that what is presented to the user as an "opt-out" is actually implemented as an "opt-in", and thus only affects the applications known at the time when you change your preferences. Applications that are issued afterwards are basically ignored.
I think you opted out of FGRP1 before S6LV1 was issued. If so, changing your preferences on the website first to accept all apps (and save!), then disabling FGRP1 again should fix the problem.
Even after following your suggestions, it is still happening. I will see what happens after I run out of Einstein tasks in a few hours.
Sat 28 Apr 19:01:20 2012 | Einstein@Home | Finished upload of h1_0390.35_S6GC1__2117_S6LV1B_1_0
Sat 28 Apr 22:28:54 2012 | Einstein@Home | Sending scheduler request: To report completed tasks.
Sat 28 Apr 22:28:54 2012 | Einstein@Home | Reporting 5 completed tasks, requesting new tasks for CPU
Sat 28 Apr 22:28:58 2012 | Einstein@Home | Scheduler request completed: got 0 new tasks
Sat 28 Apr 22:28:58 2012 | Einstein@Home | No work sent
Sat 28 Apr 22:28:58 2012 | Einstein@Home | No work is available for Gravitational Wave S6 GC search
Sat 28 Apr 22:28:58 2012 | Einstein@Home | No work is available for Gravitational Wave S6 LineVeto search
Sat 28 Apr 22:28:58 2012 | Einstein@Home | No work available for the applications you have selected. Please check your preferences on the web site.
Sun 29 Apr 03:32:02 2012 | Einstein@Home | Computation for task h1_0390.25_S6GC1__1962_S6LV1B_0 finished
Sun 29 Apr 03:32:02 2012 | Einstein@Home | Starting task h1_0390.30_S6GC1__2118_S6LV1B_0 using einstein_S6LV1 version 112 (SSE2) in slot 2
Sun 29 Apr 03:32:06 2012 | Einstein@Home | Started upload of h1_0390.25_S6GC1__1962_S6LV1B_0_0
Sun 29 Apr 03:32:11 2012 | Einstein@Home | Finished upload of h1_0390.25_S6GC1__1962_S6LV1B_0_0
Sun 29 Apr 05:52:01 2012 | Einstein@Home | Computation for task h1_0390.25_S6GC1__1963_S6LV1B_1 finished
Sun 29 Apr 05:52:01 2012 | Einstein@Home | Starting task h1_0390.35_S6GC1__2118_S6LV1B_1 using einstein_S6LV1 version 112 (SSE2) in slot 1
Sun 29 Apr 05:52:04 2012 | Einstein@Home | Started upload of h1_0390.25_S6GC1__1963_S6LV1B_1_0
Sun 29 Apr 05:52:09 2012 | Einstein@Home | Finished upload of h1_0390.25_S6GC1__1963_S6LV1B_1_0
Is it possible validation errors are related to installed version of BOINC - w/respect to registry - mismatch with actuall executable version? Actual executables = v7.0.25, albeit registry installed version = 6.12.34
Recently I experienced a major malfunction - BSOD c0000021a - w/respect to 12 May 8 Win patch-Tuesday. It was necesary to restore a 30 day old Ghost image of %SystemDrive%. P:1, 2 & 3. %SystemDrive%:\P:1 containes NTLDR & boot.ini, %SystemDrive%:\P:2 contains all installed applications, and %SystemDrive%:\P:3 is used for short-term archive. After restoring all %SystemDrive%:\ partitions individually - and prior to BOINC ever executing - I booted into safe mode and renamed the thereby restored and most current copy of %SystemDrive%:\BOINC_Data. Both the application version found in %SystemDrive%:\%Apps%\BOINC and %SystemDrive%:\BOINC_data are identical.
Any mismatch to be found is entirely a figment of the registry's imagination.
I'm going to terminate BOINC gracefully, back-up the BOINC_Data folder, uninstall and reinstall BOINC to current v7.0.25, delete the newly created BOINC_data folder and copy back the backed-up data.
Any additional computation or validation errors will compell me to ban any further processing of Gamma Ray Pulsar search v1.0.23 WU's.
After accomplishing the aforementioned 'upgrade' BOINC took off as if there was no difference, i.e., the typical bechnmarking done after version upgrade did NOT occure.
Gamma Ray Pulsar search v1.0.23 is on double secret-triple probation; the slightest hiccup and it gets banned.
New to the project and I have been working on improving my results. I think I have licked most of my validate problems (first night was ugly but few now), but I have questions on these two:
resultid=294264077
resultid=294264040
I did a compare between the first one and the other result for the task 294264078 and did not see a difference in the results other than the GPU statements and dates. I am probably missing something, be I did look more than once.
Did the second one (294264040) get marked invalid because the the task was restarted? I compared it to another result for the task and only difference I detected was the restart info. If that is the case, I will remember to abort tasks on restarts to save the validator and my GPU the effort. I have seen this on a few client restarts.
Your computers are hidden so none of the PC or result infos are available on the public web frontend.
Restarting a job should not have negative effects on validation. If it had, this would be a bug (we actually had some problems like this in rare cases before but this should be fixed now).
The threshold on how similar two results by different hosts must be in order to pass the validation tests are set quite conservatively, to ensure valid science results. We prefer to tolerate some cases where we judge an "acceptable" result invalid to the case were some "bad" result might pass as good (valid) by accident and we might miss something in our search. If the error rate is higher than (say) 3% I would suspect a real problem, e.g. a problem with the hardware (cooling, power supply, memory defects...).
A few numbers for the
)
A few numbers for the Gamma-Ray pulsar search (FGRP1):
Validate errors:
- Windows: 0.4% of all results from this app (version)
- Linux: 6.2%
- Mac OS X: 7.8%
Invalid results (on average):
- Windows and Linux: <0.2%
- Mac OS X: 3.7%
As I stated a couple of times: running the FGRP App on the Mac is somewhat more risky than e.g. running it on Windows. We thought about disabling that (OSX) app version alltogether, but I think it's still better to have people decide it on their own. Opting out of FGRP1 via Einstein@Home preferences does work.
In the past weeks and months we had (and still have) all our hands full with work on general BOINC, project infrastructure and the other Apps of this project. I hope that someone will find the time for a closer look at the FGRP search in the next few weeks, but currently I can't promise anything.
BM
BM
When you opt out of FGRP1
)
When you opt out of FGRP1 with Mac OSX, you stop getting GWS6 LV tasks.
RE: When you opt out of
)
Not in genersl, but it could hsppen. But thst problem lies deep in BOINC istelf and is not easy to fix.
Technically the problem is that what is presented to the user as an "opt-out" is actually implemented as an "opt-in", and thus only affects the applications known at the time when you change your preferences. Applications that are issued afterwards are basically ignored.
I think you opted out of FGRP1 before S6LV1 was issued. If so, changing your preferences on the website first to accept all apps (and save!), then disabling FGRP1 again should fix the problem.
BM
BM
Even after following your
)
Even after following your suggestions, it is still happening. I will see what happens after I run out of Einstein tasks in a few hours.
Sat 28 Apr 19:01:20 2012 | Einstein@Home | Finished upload of h1_0390.35_S6GC1__2117_S6LV1B_1_0
Sat 28 Apr 22:28:54 2012 | Einstein@Home | Sending scheduler request: To report completed tasks.
Sat 28 Apr 22:28:54 2012 | Einstein@Home | Reporting 5 completed tasks, requesting new tasks for CPU
Sat 28 Apr 22:28:58 2012 | Einstein@Home | Scheduler request completed: got 0 new tasks
Sat 28 Apr 22:28:58 2012 | Einstein@Home | No work sent
Sat 28 Apr 22:28:58 2012 | Einstein@Home | No work is available for Gravitational Wave S6 GC search
Sat 28 Apr 22:28:58 2012 | Einstein@Home | No work is available for Gravitational Wave S6 LineVeto search
Sat 28 Apr 22:28:58 2012 | Einstein@Home | No work available for the applications you have selected. Please check your preferences on the web site.
Sun 29 Apr 03:32:02 2012 | Einstein@Home | Computation for task h1_0390.25_S6GC1__1962_S6LV1B_0 finished
Sun 29 Apr 03:32:02 2012 | Einstein@Home | Starting task h1_0390.30_S6GC1__2118_S6LV1B_0 using einstein_S6LV1 version 112 (SSE2) in slot 2
Sun 29 Apr 03:32:06 2012 | Einstein@Home | Started upload of h1_0390.25_S6GC1__1962_S6LV1B_0_0
Sun 29 Apr 03:32:11 2012 | Einstein@Home | Finished upload of h1_0390.25_S6GC1__1962_S6LV1B_0_0
Sun 29 Apr 05:52:01 2012 | Einstein@Home | Computation for task h1_0390.25_S6GC1__1963_S6LV1B_1 finished
Sun 29 Apr 05:52:01 2012 | Einstein@Home | Starting task h1_0390.35_S6GC1__2118_S6LV1B_1 using einstein_S6LV1 version 112 (SSE2) in slot 1
Sun 29 Apr 05:52:04 2012 | Einstein@Home | Started upload of h1_0390.25_S6GC1__1963_S6LV1B_1_0
Sun 29 Apr 05:52:09 2012 | Einstein@Home | Finished upload of h1_0390.25_S6GC1__1963_S6LV1B_1_0
Gots two after mondo
)
Gots two after mondo crunching time:
http://einsteinathome.org/task/290601834
http://einsteinathome.org/task/290890304
Qwxion:
Is it possible validation errors are related to installed version of BOINC - w/respect to registry - mismatch with actuall executable version? Actual executables = v7.0.25, albeit registry installed version = 6.12.34
Recently I experienced a major malfunction - BSOD c0000021a - w/respect to 12 May 8 Win patch-Tuesday. It was necesary to restore a 30 day old Ghost image of %SystemDrive%. P:1, 2 & 3. %SystemDrive%:\P:1 containes NTLDR & boot.ini, %SystemDrive%:\P:2 contains all installed applications, and %SystemDrive%:\P:3 is used for short-term archive. After restoring all %SystemDrive%:\ partitions individually - and prior to BOINC ever executing - I booted into safe mode and renamed the thereby restored and most current copy of %SystemDrive%:\BOINC_Data. Both the application version found in %SystemDrive%:\%Apps%\BOINC and %SystemDrive%:\BOINC_data are identical.
Any mismatch to be found is entirely a figment of the registry's imagination.
I'm going to terminate BOINC gracefully, back-up the BOINC_Data folder, uninstall and reinstall BOINC to current v7.0.25, delete the newly created BOINC_data folder and copy back the backed-up data.
Any additional computation or validation errors will compell me to ban any further processing of Gamma Ray Pulsar search v1.0.23 WU's.
After accomplishing the
)
After accomplishing the aforementioned 'upgrade' BOINC took off as if there was no difference, i.e., the typical bechnmarking done after version upgrade did NOT occure.
Gamma Ray Pulsar search v1.0.23 is on double secret-triple probation; the slightest hiccup and it gets banned.
I've detached from this
)
I've detached from this project.
I receive on all my works
)
I receive on all my works units this error message. How can I do for resolve this problem. I Use a NVIDIA GTX 560Ti and I use Windows 7 64 bits.
Message error:
7.0.28
couldn't start Can't link app version file: -108
The link of my results list: http://einsteinathome.org/account/tasks
Thx for your help!
New to the project and I have
)
New to the project and I have been working on improving my results. I think I have licked most of my validate problems (first night was ugly but few now), but I have questions on these two:
resultid=294264077
resultid=294264040
I did a compare between the first one and the other result for the task 294264078 and did not see a difference in the results other than the GPU statements and dates. I am probably missing something, be I did look more than once.
Did the second one (294264040) get marked invalid because the the task was restarted? I compared it to another result for the task and only difference I detected was the restart info. If that is the case, I will remember to abort tasks on restarts to save the validator and my GPU the effort. I have seen this on a few client restarts.
Thanks
Your computers are hidden so
)
Your computers are hidden so none of the PC or result infos are available on the public web frontend.
Restarting a job should not have negative effects on validation. If it had, this would be a bug (we actually had some problems like this in rare cases before but this should be fixed now).
The threshold on how similar two results by different hosts must be in order to pass the validation tests are set quite conservatively, to ensure valid science results. We prefer to tolerate some cases where we judge an "acceptable" result invalid to the case were some "bad" result might pass as good (valid) by accident and we might miss something in our search. If the error rate is higher than (say) 3% I would suspect a real problem, e.g. a problem with the hardware (cooling, power supply, memory defects...).
CU
HB