AP V7

Message boards : Number crunching : AP V7
Message board moderation

To post messages, you must log in.

Previous · 1 . . . 4 · 5 · 6 · 7 · 8 · 9 · 10 . . . 20 · Next

AuthorMessage
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14649
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1585293 - Posted: 11 Oct 2014, 15:45:02 UTC - in response to Message 1585292.  

the whole point of excluding these Wu's is to get a stable APR.

Speed estimate, from consistent runtimes.
ID: 1585293 · Report as offensive
Profile Cliff Harding
Volunteer tester
Avatar

Send message
Joined: 18 Aug 99
Posts: 1432
Credit: 110,967,840
RAC: 67
United States
Message 1585295 - Posted: 11 Oct 2014, 15:54:19 UTC - in response to Message 1585289.  

I'm now have 24 AP v7 GPU validated tasks listed. But the 21:42:15 estimated run time still exists for the remaining 103 AP tasks, 4 of which are running all running high priority. This is well beyond the supposed level of 11 tasks required for balancing the estimated run time.

No, you only have 13 non-Outlier tasks completed, not 24, the 'Consecutive valid tasks' value includes all the outlier tasks, tasks that end early don't count.

http://setiathome.berkeley.edu/host_app_versions.php?hostid=5501972

Since you're only just past your 11 validated tasks, try another request for GPU work, it should get updated when you receive some.
(It should have already done so, since you're received two Wu's since you posted that)

Claggy


You lost me with the 13 non-Outlier completed. When looking at my tasks list for this host, I see now 25 validated tasks http://setiathome.berkeley.edu/results.php?hostid=5501972&offset=0&show_names=0&state=4&appid=20. How do I distinguish between non & Outlier tasks? What determines an early task? The only ones that I see that have a short run time are 100% blanked. The scheduler still hasn't updated my manager, as I have not only requested an update manually, but also have stopped and started BOINC with the same results.


If you looks at the Application Details for your host you will see the number of tasks completed.



As I only installed the v7 apps late yesterday afternoon and reset the project after suspending all non running tasks and waited until all running tasks had completed at:

10-Oct-2014 21:29:08 [SETI@home] Resetting project
10-Oct-2014 21:29:28 [SETI@home] work fetch resumed by user
10-Oct-2014 21:29:33 [SETI@home] Master file download succeeded
10-Oct-2014 21:29:38 [SETI@home] Sending scheduler request: To fetch work.
10-Oct-2014 21:29:38 [SETI@home] Requesting new tasks for CPU and NVIDIA GPU
10-Oct-2014 21:29:41 [SETI@home] Scheduler request completed: got 20 new tasks

Where is the disconnect between the application details page and the host task list?


I don't buy computers, I build them!!
ID: 1585295 · Report as offensive
Profile HAL9000
Volunteer tester
Avatar

Send message
Joined: 11 Sep 99
Posts: 6534
Credit: 196,805,888
RAC: 57
United States
Message 1585300 - Posted: 11 Oct 2014, 16:00:24 UTC - in response to Message 1585295.  
Last modified: 11 Oct 2014, 16:04:44 UTC

I'm now have 24 AP v7 GPU validated tasks listed. But the 21:42:15 estimated run time still exists for the remaining 103 AP tasks, 4 of which are running all running high priority. This is well beyond the supposed level of 11 tasks required for balancing the estimated run time.

No, you only have 13 non-Outlier tasks completed, not 24, the 'Consecutive valid tasks' value includes all the outlier tasks, tasks that end early don't count.

http://setiathome.berkeley.edu/host_app_versions.php?hostid=5501972

Since you're only just past your 11 validated tasks, try another request for GPU work, it should get updated when you receive some.
(It should have already done so, since you're received two Wu's since you posted that)

Claggy


You lost me with the 13 non-Outlier completed. When looking at my tasks list for this host, I see now 25 validated tasks http://setiathome.berkeley.edu/results.php?hostid=5501972&offset=0&show_names=0&state=4&appid=20. How do I distinguish between non & Outlier tasks? What determines an early task? The only ones that I see that have a short run time are 100% blanked. The scheduler still hasn't updated my manager, as I have not only requested an update manually, but also have stopped and started BOINC with the same results.


If you looks at the Application Details for your host you will see the number of tasks completed.

[img ]http://www.hal6000.com/seti/images/completed_tasks.png[/img]


As I only installed the v7 apps late yesterday afternoon and reset the project after suspending all non running tasks and waited until all running tasks had completed at:

10-Oct-2014 21:29:08 [SETI@home] Resetting project
10-Oct-2014 21:29:28 [SETI@home] work fetch resumed by user
10-Oct-2014 21:29:33 [SETI@home] Master file download succeeded
10-Oct-2014 21:29:38 [SETI@home] Sending scheduler request: To fetch work.
10-Oct-2014 21:29:38 [SETI@home] Requesting new tasks for CPU and NVIDIA GPU
10-Oct-2014 21:29:41 [SETI@home] Scheduler request completed: got 20 new tasks

Where is the disconnect between the application details page and the host task list?

Only tasks that are applied to APR are listed as completed on the Application Detail page. The task list is just that. The list of tasks and their various status.
Not every task you do will be counted towards the APR & you don't want every task to be. Otherwise you would end up with highly fluctuating estimates.
SETI@home classic workunits: 93,865 CPU time: 863,447 hours
Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[
ID: 1585300 · Report as offensive
Profile Cliff Harding
Volunteer tester
Avatar

Send message
Joined: 18 Aug 99
Posts: 1432
Credit: 110,967,840
RAC: 67
United States
Message 1585304 - Posted: 11 Oct 2014, 16:14:05 UTC - in response to Message 1585278.  

I'm now have 24 AP v7 GPU validated tasks listed. But the 21:42:15 estimated run time still exists for the remaining 103 AP tasks, 4 of which are running all running high priority. This is well beyond the supposed level of 11 tasks required for balancing the estimated run time. With the remaining 99 tasks all having the expected deadline/report date of 11/04, they also will be running in a HP state. Questions - When does the scheduler update the estimated run times for a specific app on a specific host? Is anyone else experiencing this situation?

The estimated time will be for new tasks you download. Tasks currently in your queue retain their estimated run time.


So, I can expect all remaining initial tasks to run as High Priority? There should be some way other than resetting the project to get the manager to reset times already in the queue.


I don't buy computers, I build them!!
ID: 1585304 · Report as offensive
Profile HAL9000
Volunteer tester
Avatar

Send message
Joined: 11 Sep 99
Posts: 6534
Credit: 196,805,888
RAC: 57
United States
Message 1585309 - Posted: 11 Oct 2014, 16:18:20 UTC - in response to Message 1585304.  

I'm now have 24 AP v7 GPU validated tasks listed. But the 21:42:15 estimated run time still exists for the remaining 103 AP tasks, 4 of which are running all running high priority. This is well beyond the supposed level of 11 tasks required for balancing the estimated run time. With the remaining 99 tasks all having the expected deadline/report date of 11/04, they also will be running in a HP state. Questions - When does the scheduler update the estimated run times for a specific app on a specific host? Is anyone else experiencing this situation?

The estimated time will be for new tasks you download. Tasks currently in your queue retain their estimated run time.


So, I can expect all remaining initial tasks to run as High Priority? There should be some way other than resetting the project to get the manager to reset times already in the queue.

There is if you remember how to do <flops> in your app_info.xml
SETI@home classic workunits: 93,865 CPU time: 863,447 hours
Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[
ID: 1585309 · Report as offensive
Profile Cliff Harding
Volunteer tester
Avatar

Send message
Joined: 18 Aug 99
Posts: 1432
Credit: 110,967,840
RAC: 67
United States
Message 1585313 - Posted: 11 Oct 2014, 16:21:56 UTC - in response to Message 1585309.  

I'm now have 24 AP v7 GPU validated tasks listed. But the 21:42:15 estimated run time still exists for the remaining 103 AP tasks, 4 of which are running all running high priority. This is well beyond the supposed level of 11 tasks required for balancing the estimated run time. With the remaining 99 tasks all having the expected deadline/report date of 11/04, they also will be running in a HP state. Questions - When does the scheduler update the estimated run times for a specific app on a specific host? Is anyone else experiencing this situation?

The estimated time will be for new tasks you download. Tasks currently in your queue retain their estimated run time.


So, I can expect all remaining initial tasks to run as High Priority? There should be some way other than resetting the project to get the manager to reset times already in the queue.

There is if you remember how to do <flops> in your app_info.xml


I've never done <flops>, will be greatly appreciative if you enlighten me.


I don't buy computers, I build them!!
ID: 1585313 · Report as offensive
Cherokee150

Send message
Joined: 11 Nov 99
Posts: 192
Credit: 58,513,758
RAC: 74
United States
Message 1585318 - Posted: 11 Oct 2014, 16:34:19 UTC

Is there some written documentation on all the options available in the app_info file and their effects on the application? That would be most helpful to me and, I suspect, to many of us. :-)
ID: 1585318 · Report as offensive
Profile HAL9000
Volunteer tester
Avatar

Send message
Joined: 11 Sep 99
Posts: 6534
Credit: 196,805,888
RAC: 57
United States
Message 1585319 - Posted: 11 Oct 2014, 16:37:33 UTC - in response to Message 1585318.  
Last modified: 11 Oct 2014, 16:41:17 UTC

Is there some written documentation on all the options available in the app_info file and their effects on the application? That would be most helpful to me and, I suspect, to many of us. :-)

http://boinc.berkeley.edu/wiki/Anonymous_platform

I forget the method for how to calculate flops. I suspend processing and then adjust it until I find the correct value.

Beware modifying you app_info.xml incorrectly can cause BOINC to dump all of the work in your queue.
SETI@home classic workunits: 93,865 CPU time: 863,447 hours
Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[
ID: 1585319 · Report as offensive
Cherokee150

Send message
Joined: 11 Nov 99
Posts: 192
Credit: 58,513,758
RAC: 74
United States
Message 1585320 - Posted: 11 Oct 2014, 16:40:15 UTC - in response to Message 1585319.  

Thank you! That is most helpful. :-)
ID: 1585320 · Report as offensive
Josef W. Segur
Volunteer developer
Volunteer tester

Send message
Joined: 30 Oct 99
Posts: 4504
Credit: 1,414,761
RAC: 0
United States
Message 1585407 - Posted: 11 Oct 2014, 20:02:39 UTC - in response to Message 1585286.  
Last modified: 11 Oct 2014, 20:03:28 UTC

The AP Validator basically just checks two values to decide whether a result should be flagged as a runtime outlier.

1. Blanking higher than 10%. That catches the most common cases where runtime has been significantly affected. For now that's overkill for the OpenCL GPU apps, but is needed for CPU processing.

2. Exit before 95% progress. That catches the 100% blanking cases which exit at effectively 0% progress as well as most 30/30 overflows.

As Claggy pointed out earlier, there's a third case where processing is somewhat faster; when the limit of 30 repetitive pulses has been found the application stops looking for them. That optimization has been in the Lunatics AP v6 app versions for years and is now in all the AP v7 app versions. The Validator does not mark them as runtime outliers, so they do affect the averages. Such tasks happen often enough that it's probably best to include them in the averages, and the effect is small enough that it doesn't cause any problem.
                                                                   Joe
ID: 1585407 · Report as offensive
Claggy
Volunteer tester

Send message
Joined: 5 Jul 99
Posts: 4654
Credit: 47,537,079
RAC: 4
United Kingdom
Message 1585411 - Posted: 11 Oct 2014, 20:17:01 UTC - in response to Message 1585407.  
Last modified: 11 Oct 2014, 20:18:46 UTC

Thanks for your more concise answer, it's a bit better than my vague answer, that wasn't totally accurate.

Claggy
ID: 1585411 · Report as offensive
Profile Jord
Volunteer tester
Avatar

Send message
Joined: 9 Jun 99
Posts: 15184
Credit: 4,362,181
RAC: 3
Netherlands
Message 1585415 - Posted: 11 Oct 2014, 20:48:43 UTC
Last modified: 11 Oct 2014, 20:51:18 UTC

I tried leaving some mention of AP v6 in my app_info.xml, but apparently that won't work. Or according to EL:

11-Oct-2014 22:39:26 [SETI@home] Found app_info.xml; using anonymous platform
11-Oct-2014 22:39:26 [SETI@home] State file error: bad application name astropulse_v6
11-Oct-2014 22:39:26 [SETI@home] State file error: bad application name astropulse_v6
11-Oct-2014 22:39:26 [SETI@home] State file error: bad application name astropulse_v6
11-Oct-2014 22:39:26 [SETI@home] State file error: bad application name astropulse_v6

Oh well, removed all trace of it then. For those 17 APs I got the last 3 years, it's probably not worth it to leave AP v6 in anyway. :)

P.S: I did find that the Lunatics installer had put 8 mentions of the optimized AP v6 606 in my app_info.xml, a bit much. Only one mention of 604 and 608.
ID: 1585415 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14649
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1585438 - Posted: 11 Oct 2014, 21:19:18 UTC - in response to Message 1585415.  

P.S: I did find that the Lunatics installer had put 8 mentions of the optimized AP v6 606 in my app_info.xml, a bit much. Only one mention of 604 and 608.

Yes, I do that, and you'll find that v7 is much the same. The idea is to cover every version listed on the applications page - you'll find a similar list of v6 versions there too.

You'll only be using one of them, but I don't know in advance which one...
ID: 1585438 · Report as offensive
Profile Jord
Volunteer tester
Avatar

Send message
Joined: 9 Jun 99
Posts: 15184
Credit: 4,362,181
RAC: 3
Netherlands
Message 1585552 - Posted: 12 Oct 2014, 0:15:42 UTC - in response to Message 1585438.  

I know what the idea is, but the thing here was that for 606 there were 8 entries, all exactly the same. Then one for 604, one for 608. Just one 606 was all it needed here as well.

Well, moot now, as I changed it to one entry for v7 and 704. So your installer can go add crud again. ;-)
ID: 1585552 · Report as offensive
Josef W. Segur
Volunteer developer
Volunteer tester

Send message
Joined: 30 Oct 99
Posts: 4504
Credit: 1,414,761
RAC: 0
United States
Message 1585607 - Posted: 12 Oct 2014, 2:25:05 UTC - in response to Message 1585552.  

An app_version is distinguished by four values; app_name, version_num, platform, and plan_class. When the new Lunatics 64 bit installer comes out, the Astropulse v7 OpenCL_NV app version will need eight entries. All will be for astropulse_v7 and version_num 705, but combinations of platform and plan_class are needed:

platform           plan_class
----------------   -----------------
windows_x86_64     opencl_nvidia_100
windows_x86_64     cuda_opencl_100
windows_x86_64     opencl_nvidia_cc1
windows_x86_64     cuda_opencl_cc1
windows_intelx86   opencl_nvidia_100
windows_intelx86   cuda_opencl_100
windows_intelx86   opencl_nvidia_cc1
windows_intelx86   cuda_opencl_cc1


It's not crud, without covering the full set of possibilities the BOINC client may delete work on hand as the installation completes. However, we do not attempt to cover the situation where a user has chosen a personal plan_class name.
                                                                  Joe
ID: 1585607 · Report as offensive
Profile cov_route
Avatar

Send message
Joined: 13 Sep 12
Posts: 342
Credit: 10,270,618
RAC: 0
Canada
Message 1585612 - Posted: 12 Oct 2014, 3:36:45 UTC

I obtained ap7_win_x64_sse3_cpu_r2692_vc10_pgoed.7z from Mikes site. It includes the debug symbols AP7_win_x64_SSE3_CPU.pdb. I didn't deploy that file, just the exe.

I suppose that's right?
ID: 1585612 · Report as offensive
Profile Raistmer
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 16 Jun 01
Posts: 6325
Credit: 106,370,077
RAC: 121
Russia
Message 1585629 - Posted: 12 Oct 2014, 5:08:46 UTC - in response to Message 1585612.  

I obtained ap7_win_x64_sse3_cpu_r2692_vc10_pgoed.7z from Mikes site. It includes the debug symbols AP7_win_x64_SSE3_CPU.pdb. I didn't deploy that file, just the exe.

I suppose that's right?

right... until some crash occured and no info will be given about where.
ID: 1585629 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14649
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1585689 - Posted: 12 Oct 2014, 8:16:14 UTC - in response to Message 1585607.  

An app_version is distinguished by four values; app_name, version_num, platform, and plan_class. When the new Lunatics 64 bit installer comes out, the Astropulse v7 OpenCL_NV app version will need eight entries. All will be for astropulse_v7 and version_num 705, but combinations of platform and plan_class are needed:

platform           plan_class
----------------   -----------------
windows_x86_64     opencl_nvidia_100
windows_x86_64     cuda_opencl_100
windows_x86_64     opencl_nvidia_cc1
windows_x86_64     cuda_opencl_cc1
windows_intelx86   opencl_nvidia_100
windows_intelx86   cuda_opencl_100
windows_intelx86   opencl_nvidia_cc1
windows_intelx86   cuda_opencl_cc1


It's not crud, without covering the full set of possibilities the BOINC client may delete work on hand as the installation completes. However, we do not attempt to cover the situation where a user has chosen a personal plan_class name.
                                                                  Joe

And unfortunately the pre-release packs which Raistmer prepared, and which are now hosted by Mike, Arkayn, and probably team sites around the world, used version 701. So that would make 16 combinations.....

I'm sorry, I'm not going there. The installer is designed to last for many months, perhaps a year, before the next major round of updates. New volunteers discover the SETI project every day, sign up, and download one or more stock applications automatically. As they come to know and love the community here, they may wish to deepen their involvement at some later stage. I will always try to look ahead and make the upgrade path for those new users in the future as smooth as possible. But - and I really am sorry about this, but I have to draw the line somewhere - I'm not going to bend over backwards to provide transitional paths for experienced users who simply want to jump the gun.
ID: 1585689 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14649
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1585700 - Posted: 12 Oct 2014, 8:55:11 UTC - in response to Message 1585691.  

And unfortunately the pre-release packs which Raistmer prepared, and which are now hosted by Mike, Arkayn, and probably team sites around the world, used version 701. So that would make 16 combinations.....

I'm sorry, I'm not going there. The installer is designed to last for many months, perhaps a year, before the next major round of updates. New volunteers discover the SETI project every day, sign up, and download one or more stock applications automatically. As they come to know and love the community here, they may wish to deepen their involvement at some later stage. I will always try to look ahead and make the upgrade path for those new users in the future as smooth as possible. But - and I really am sorry about this, but I have to draw the line somewhere - I'm not going to bend over backwards to provide transitional paths for experienced users who simply want to jump the gun.

Well, I do not know about most experienced users, or if I am to be counted amongs them, but I always just rip out from the installer the apps I want, and manually install the needed files, and edit the app_info file. I haven't used the installer as an installer for years.

I downloaded the v7 files from Mike's site, and immediately noted that the aistub used <version_num>701</version_num>, and I changed that faster than you can say "hi", to 704.

Checking your computer, and the applications page, 704 is correct for your ATI card. Congratulations, you are hereby accredited an 'advanced user'. For NVidia cards, Joe's value of 705 is correct.
ID: 1585700 · Report as offensive
OTS
Volunteer tester

Send message
Joined: 6 Jan 08
Posts: 369
Credit: 20,533,537
RAC: 0
United States
Message 1585782 - Posted: 12 Oct 2014, 14:21:27 UTC

I see that there is a 32-bit SSE3 version of the AP7 app for Windows (AP7 32-bit SSE3 CPU r2691), but there is none for Linux. Is that because there is no real advantage to using SSE3 instructions with Linux or is it just not posted yet?
ID: 1585782 · Report as offensive
Previous · 1 . . . 4 · 5 · 6 · 7 · 8 · 9 · 10 . . . 20 · Next

Message boards : Number crunching : AP V7


 
©2024 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.