Posts by Raubhautz*
log in
1) Message boards : Number crunching : Uploading WUs (Message 17233)
Posted 1629 days ago by Profile Raubhautz*
Hi,

I have one that is performing the same way (failed ul). Other wu have come and gone since, but this one stays.

Raubhautz190

109 Collatz Conjecture 07/31/2013 20:29:00 Started upload of solo_collatz_2380303395902813088104_824633720832_0_0
110 Collatz Conjecture 07/31/2013 20:29:02 [error] Error reported by file upload server: can't open file /home/boincadm/projects/collatz/upload/d9/solo_collatz_2380303395902813088104_824633720832_0_0: Permission denied
111 Collatz Conjecture 07/31/2013 20:29:02 Temporarily failed upload of solo_collatz_2380303395902813088104_824633720832_0_0: transient upload error
112 Collatz Conjecture 07/31/2013 20:29:02 Backing off 05:13:48 on upload of solo_collatz_2380303395902813088104_824633720832_0_0


Thank you in advance for any assistance.

Phil
2) Message boards : Number crunching : Not Requesting Tasks.... I want tasks! (Message 16999)
Posted 1662 days ago by Profile Raubhautz*
If that doesn't work, let me know and I'll figure something out.

Eric at Seti Beta is still trying to get anonymous platform to work for Intel GPUs without much luck, he has done a few fixes:

Astropulse 6.05 for OpenCL on Intel GPUs released.

Changeset 4e338e94 in boinc-v2

Changeset 03e64f72 in boinc-v2

So I think some debugging is still in order.

Claggy


Thanks Claggy, I'll be keeping an eye out for any successes on the intel anon. platform. I appreciate the links.

Phil
3) Message boards : Number crunching : Not Requesting Tasks.... I want tasks! (Message 16998)
Posted 1662 days ago by Profile Raubhautz*
It looks like the WUs are all for one computer and that that computer has an app_info file. Maybe it thinks you have WUs for apps which you no longer have selected in your prefs? (e.g. all the ones listed as in progress are mini and collatz and the only Intel GPU app is for solo). MY suggestion would be to first try to reset the project. If that doesn't work, disconnect and reconnect. If that doesn't work, let me know and I'll figure something out.

Or... enable mini and collatz and let the WUs get reset. Set the project to no new work. Abort the tasks. Change the prefs to solo only. Then allow new tasks. That __should__ do the trick.


Okay, reset did not do much, replaced some of the png files, and same issue; however, the detach/reattach worked by purging everything in the project directory (including the directory itself), then set solo only in preferences, and upon reattach, I started getting work! :) So, I left the config alone (no anonymous) and only added an app_config.xml to work more than one wu at a time.

Thank you for the resolution - it is much appreciated.

Phil
4) Message boards : Number crunching : Not Requesting Tasks.... I want tasks! (Message 16985)
Posted 1663 days ago by Profile Raubhautz*
Okay, this might help; Upon further investigation, I found the client_state.xml file contains a bunch of result entries... How did I 'lose' the wu, without the appropriate entries in the client_state file.

Better yet, if this is the cause of my problem, how best to fix it?

Thank you.
5) Message boards : Number crunching : Not Requesting Tasks.... I want tasks! (Message 16984)
Posted 1663 days ago by Profile Raubhautz*
Hi,

I am trying to get my Intel HD4000 card to take more work for Collatz. It was working fine about 3 weeks ago for both the Intel and NVIDIA work. Intel performed work on Solo project, NVIDIA did Collatz and mini's.

Now, all I get from requests is "...Not Requesting Tasks..." My Intel SDK is working (still), and the NVIDIA pulls work from all other projects. Pertinent info from my log is as follows:

Raubhautz190

13 06/27/2013 16:36:48 CUDA: NVIDIA GPU 0: Quadro K2000M (driver version 320.00, CUDA version 5.50, compute capability 3.0, 2048MB, 1982MB available, 572 GFLOPS peak)
14 06/27/2013 16:36:48 OpenCL: NVIDIA GPU 0: Quadro K2000M (driver version 320.00, device version OpenCL 1.1 CUDA, 2048MB, 1982MB available, 572 GFLOPS peak)
15 06/27/2013 16:36:48 OpenCL: Intel GPU 0: Intel(R) HD Graphics 4000 (driver version 9.17.10.2843, device version OpenCL 1.1, 1624MB, 1624MB available, 51 GFLOPS peak)
18 Collatz Conjecture 06/27/2013 16:36:48 Found app_info.xml; using anonymous platform
22 Collatz Conjecture 06/27/2013 16:36:48 Found app_config.xml
23 Collatz Conjecture 06/27/2013 16:36:48 app solo_collatz not found in app_config.xml
33 06/27/2013 16:36:48 Config: report completed tasks immediately
34 06/27/2013 16:36:48 Config: use all coprocessors
35 06/27/2013 16:36:48 Config: fetch on update
43 Collatz Conjecture 06/27/2013 16:36:48 URL http://boinc.thesonntags.com/collatz/; Computer ID 128097; resource share 2501
122 Collatz Conjecture 06/27/2013 16:38:10 update requested by user
123 Collatz Conjecture 06/27/2013 16:38:11 Sending scheduler request: Requested by user.
124 Collatz Conjecture 06/27/2013 16:38:11 Not requesting tasks
125 Collatz Conjecture 06/27/2013 16:38:13 Scheduler request completed


Thank you in advance for any assistance provided.



**** Addendum ****

Looking at my account on Collatz shows that I have 113 various tasks assigned to this computer, issued around the 1st of June; I do not have these tasks - I have no wu files in my project folder.
6) Message boards : News : Intel GPUs Supported (Message 16793)
Posted 1679 days ago by Profile Raubhautz*
v4.08 has been released for solo_collatz. It __should__ run OK when in standalone mode and also should checkpoint properly (I hope). If it runs OK, let me know and I will deploy it for collatz and mini_collatz as well.


Well, it is running fine on my intel i7-3920XM/4000 HD Lenovo laptop which uses Optimus, so I have the laptop display running the computer, the NVIDIA is dedicated to BOINC, and now I am running solo_collatz on the intel and all is well. Took 5h10m and completed without problems. The cpu utilization is between 65 and 75%, a bit high... I even put the laptop on sleep and hibernate to see if the check-pointing worked, it does. :D So, bring on the mini's!


PS - I did a little searching, but did not find anything: Is it possible to run one project's gpu wu on 2 different gpus in the same machine? For now I don't mind just leaving the intel on collatz, but... Thanks!
7) Message boards : News : Intel GPUs Supported (Message 16792)
Posted 1679 days ago by Profile Raubhautz*
Hello.

Intel 2600K GPU HD3000 can run on solo_collatz ?




My understanding of the new Open CL is that it is restricted to the Ivy CPU w/4000 HD. Out of curiosity, I'll try installing the driver and get it to work on my Sandy i7-2960XM/3000 HD - but I am not hopeful. I'll post results later.
8) Message boards : News : Intel GPUs Supported (Message 16791)
Posted 1679 days ago by Profile Raubhautz*
[...]

<exclude_gpu>
<url>http://boinc.thesonntags.com/collatz/</url>
<device_num>0</device_num>
<type>intel_gpu</type>
<app>mini_collatz</app>
</exclude_gpu>
<exclude_gpu>
<url>http://boinc.thesonntags.com/collatz/</url>
<device_num>0</device_num>
<type>intel_gpu</type>
<app>solo_collatz</app>
</exclude_gpu>
<exclude_gpu>
<url>http://boinc.thesonntags.com/collatz/</url>
<device_num>0</device_num>
<type>intel_gpu</type>
<app>collatz</app>
</exclude_gpu> [...]



The above works great, thanks. This reminds me to read my manuals before I start playing... Hah, like that will ever happen!!! ;)


It isn't efficient, but since BOINC doesn't use a DTD, I don't think you can repeat the app element within the exclude_gpu element. In other words, I'm not sure the following would work, but you could give it a try instead:

<exclude_gpu>
<url>http://boinc.thesonntags.com/collatz/</url>
<device_num>0</device_num>
<type>intel_gpu</type>
<app>collatz</app>
<app>mini_collatz</app>
<app>solo_collatz</app>
</exclude_gpu>

Also, it isn't documented well, but I believe the device number starts at 0 for the first device. Note that the device number is the device number for that type of GPU. With one Intel and one NVIDIA, they would each be device 0 but the type would be different.


You postulated correctly, the multiple app elements do not work; like a variable in programming code, its value is that of the last entry.

Regarding the device number, you are correct, your first gpu of a given type will always be 0, next is 1, etc. So if you have a NVIDIA and a 4000 HD, you will have 2 gpu's with device 0, just of different type.
9) Message boards : Windows : Team being dropped from my account!? (Message 16682)
Posted 1687 days ago by Profile Raubhautz*
Thanks for the response slicker.
10) Message boards : Windows : Team being dropped from my account!? (Message 16669)
Posted 1691 days ago by Profile Raubhautz*
Guten Tag,

I noticed this evening that my Collatz project did not have a team listing in BOINC manager.


  1. Seeing this I attempted to update the project from the client machine, no change to the team entry.
  2. I performed an update with BOINCStats from the client machine, no change to the team entry.
  3. I then logged into BOINCStats and verified the teams that I have assigned to my projects, and found that all my projects had 'Sicituradastra.' properly assigned to them, including Collatz Conjuncture.
  4. I then logged into Collatz' site using my account to verify my team status and found that I was now team-less in Collatz!?



Now, I can easily add the team back, I am curious as to how and when this 'quitting a team' happened. I am respectfully requesting that someone look into your logs and report for me the date & time (please let me know if it is my local or UTC time); as well as the computer name that was logged on when this occurred.

Thank you in advance for your assist in this matter.

Raubhautz* (Phil)

11) Message boards : News : ByeBye Wingman (Message 16413)
Posted 1720 days ago by Profile Raubhautz*
I am concerned about this change. Milkyway changed their apps a while back and rendered two of my machines useless on their project. It is beginning to look like volunteered computers are irrelevant. I hope the same thing does not happen here.



Not sure about MW, but I can no longer use Collatz on my machines... Which is too bad. It was running the 23's fine, but now, just errors. I had to move to POEM to keep my gpu's busy.


Umm that might not work, Poem is regulating the amount of gpu units per day to the number they can actually review, meaning it is not nearly enough to keep everyones caches full and people often run out of them.


Working for me so far! 96 wu/day is enough per machine. PG is a good backup. Thanks for the concern, however.

I don't see any improvements here, so business as usual.
12) Message boards : News : ByeBye Wingman (Message 16368)
Posted 1726 days ago by Profile Raubhautz*
I am concerned about this change. Milkyway changed their apps a while back and rendered two of my machines useless on their project. It is beginning to look like volunteered computers are irrelevant. I hope the same thing does not happen here.



Not sure about MW, but I can no longer use Collatz on my machines... Which is too bad. It was running the 23's fine, but now, just errors. I had to move to POEM to keep my gpu's busy.
13) Message boards : News : ByeBye Wingman (Message 16314)
Posted 1730 days ago by Profile Raubhautz*
A new application has been added named "solo_collatz". This application has a quorum of 1 so credit will be granted as soon as the result is validated. Rather than wait for a second computer to return the same result, the solo application uses the CPU to double check the GPU results eliminating the need for a wingman. Multiple instances of the solo CUDA and OpenCL applications can be run on a GPU at a time via the use of an app_config.xml file if GPU utilization is lower than with previous Collatz applications.

If you wish to run both GPU and CPU applications, it is suggested that you uncheck both "collatz" and "mini_collatz", check "solo_collatz" and also check "If no work for selected applications is available, accept work from other applications?" That will enable you to get mini_collatz work for you CPU and solo_collatz work for your GPU.


I did as this post suggests, and all I get are 4.2 mini's and collatz' wu. I have config for 5.0 and get none; message claims I don't have config for app_info (which I then guesstimated), and it still fails. I even tried to reset the project to 'attract' the solo files, but no dice.

So, 2 questions:

1. Why am I not getting cuda5.0 wu?
2. Where is this config for solo wu?

Thank you in advance.
14) Message boards : News : Collatz 4.xx CUDA and CPU Applications Released for Windows x86_64 (Message 16198)
Posted 1745 days ago by Profile Raubhautz*
Two new v4.xx applications were released today, one for 64-bit Windows and the other for CUDA 5.0 for 64-bit Windows. These now contain internal validation which will reduce the "inconclusive" results and, once all platforms and plan classes have been upgraded, reduce the quorum to 1. Versions for Linux, OS/X, and 32-bit Windows will be released as testing on each is completed.



Errr, I have never had an inconclusive result on Collatz.... until using the v4.02 drivers, and it seems that half my regular collatz get inconclusive (with 2 confirmed invalids).

Ideas? Thoughts? I do see a speed increase on both collatz and mini (mini's do not get any invalids).

Thank you in advance.


My thoughts are that the ptx code generated fail for certain hardware/compute combinations. For whatever reason, the reduction code which uses shared memory and inline functions seems to fail on some hardware but not all hardware and even that depends upon driver version. Since the reductions really don't affect CPU usage very much at all, the easiest solution is to use reduce_on_cpu=1 in the collatz.config file.


Cool, thanks for the information... I am changing the config file now; will see what future wu bring! To be continued...


Alright, the addition of the 'reduce_on_cpu=1' line worked and while it was there no units went invalid, but then I booted my machine for an update (non-boinc related sw) and I noted that almost everything Collatz went invalid again; and after checking, the config file was reset! I did note that Zombie67 had similar issues, but I did not note any fix. Is there any? Is it restarting the system or just restarting boinc that initiates a 'new' config file?

Thanks, Phil
15) Message boards : News : Collatz 4.xx CUDA and CPU Applications Released for Windows x86_64 (Message 16182)
Posted 1747 days ago by Profile Raubhautz*
Two new v4.xx applications were released today, one for 64-bit Windows and the other for CUDA 5.0 for 64-bit Windows. These now contain internal validation which will reduce the "inconclusive" results and, once all platforms and plan classes have been upgraded, reduce the quorum to 1. Versions for Linux, OS/X, and 32-bit Windows will be released as testing on each is completed.



Errr, I have never had an inconclusive result on Collatz.... until using the v4.02 drivers, and it seems that half my regular collatz get inconclusive (with 2 confirmed invalids).

Ideas? Thoughts? I do see a speed increase on both collatz and mini (mini's do not get any invalids).

Thank you in advance.


My thoughts are that the ptx code generated fail for certain hardware/compute combinations. For whatever reason, the reduction code which uses shared memory and inline functions seems to fail on some hardware but not all hardware and even that depends upon driver version. Since the reductions really don't affect CPU usage very much at all, the easiest solution is to use reduce_on_cpu=1 in the collatz.config file.


Cool, thanks for the information... I am changing the config file now; will see what future wu bring! To be continued...
16) Message boards : News : Collatz 4.xx CUDA and CPU Applications Released for Windows x86_64 (Message 16162)
Posted 1751 days ago by Profile Raubhautz*
Two new v4.xx applications were released today, one for 64-bit Windows and the other for CUDA 5.0 for 64-bit Windows. These now contain internal validation which will reduce the "inconclusive" results and, once all platforms and plan classes have been upgraded, reduce the quorum to 1. Versions for Linux, OS/X, and 32-bit Windows will be released as testing on each is completed.



Errr, I have never had an inconclusive result on Collatz.... until using the v4.02 drivers, and it seems that half my regular collatz get inconclusive (with 2 confirmed invalids).

Ideas? Thoughts? I do see a speed increase on both collatz and mini (mini's do not get any invalids).

Thank you in advance.




Main page · Your account · Message boards


Copyright © 2018 Jon Sonntag; All rights reserved.