No work for cpu
log in

Advanced search

Message boards : CPU : No work for cpu

1 · 2 · Next
Author Message
[VENETO] boboviz
Send message
Joined: 16 Nov 16
Posts: 37
Credit: 1,140,113
RAC: 0
Message 165 - Posted: 20 Jun 2017, 13:07:57 UTC

Now there are 64 wus in cpu queue.
But my pc says "No tasks are available for XaNSoNS BOINC for CPU (multi-threaded)"
In my account preferences cpu is enabled

Vlad
Project administrator
Project developer
Project tester
Project scientist
Help desk expert
Send message
Joined: 26 Oct 16
Posts: 322
Credit: 103,382
RAC: 0
Message 168 - Posted: 20 Jun 2017, 15:15:27 UTC - in response to Message 165.

Now there are 64 wus in cpu queue.
But my pc says "No tasks are available for XaNSoNS BOINC for CPU (multi-threaded)"
In my account preferences cpu is enabled

I don't know what is causing this problem. The scheduler log does not contain any rejections for the host 185.
Here is the plan class description for the CPU app (windows):
<plan_class> <name>mt_windows</name> <min_ncpus>1</min_ncpus> <max_threads>64</max_threads> <nthreads_cmdline>1</nthreads_cmdline> <min_os_version>601760100</min_os_version> <projected_flops_scale>.99</projected_flops_scale> </plan_class>

So, any host that meets these requirements should receive the WUs.
In this projects there are more WUs for the GPU than for the CPU, so there are may be some shortage of CPU tasks.
Also, the status page refreshes once an hour and may provide outdated information.
But, anyway, if you cannot get the CPU tasks, it's better to check the settings. Do you have at least one free CPU core? Do you allowed the client to get the tasks for the xansons_cpu app?

[VENETO] boboviz
Send message
Joined: 16 Nov 16
Posts: 37
Credit: 1,140,113
RAC: 0
Message 170 - Posted: 20 Jun 2017, 17:20:54 UTC - in response to Message 168.

I don't know what is causing this problem. The scheduler log does not contain any rejections for the host 185.


The problem is in host 724
With host 185 i'm crunching cpu wus

Vlad
Project administrator
Project developer
Project tester
Project scientist
Help desk expert
Send message
Joined: 26 Oct 16
Posts: 322
Credit: 103,382
RAC: 0
Message 171 - Posted: 20 Jun 2017, 18:07:49 UTC - in response to Message 170.

The problem is in host 724
With host 185 i'm crunching cpu wus

No rejections for the host 724 in the scheduler log either. Does this host recieve GPU WUs without problem? I'm asking because it has only one GPU WU assigned.

[VENETO] boboviz
Send message
Joined: 16 Nov 16
Posts: 37
Credit: 1,140,113
RAC: 0
Message 181 - Posted: 21 Jun 2017, 6:44:33 UTC - in response to Message 171.

No rejections for the host 724 in the scheduler log either. Does this host recieve GPU WUs without problem? I'm asking because it has only one GPU WU assigned.


I killed the gpu wu, because it's a notebook with an old integrated intel gpu...
I retry to download cpu wus in this nb

mmonnin
Send message
Joined: 28 Nov 16
Posts: 19
Credit: 5,313,490
RAC: 0
Message 183 - Posted: 21 Jun 2017, 9:56:09 UTC - in response to Message 181.

I have only been able to get a steady supply of GPU work in Windows and Linux.

Vlad
Project administrator
Project developer
Project tester
Project scientist
Help desk expert
Send message
Joined: 26 Oct 16
Posts: 322
Credit: 103,382
RAC: 0
Message 184 - Posted: 21 Jun 2017, 10:10:22 UTC - in response to Message 183.

I have only been able to get a steady supply of GPU work in Windows and Linux.

By default the server send 5 tasks for CPU and 15 tasks for GPU at once. Now you have all 5 tasks for CPU marked as in-progress on your Linux and Windows hosts. When the client report to the server that those tasks are ready, you'll get more tasks if any.

[VENETO] boboviz
Send message
Joined: 16 Nov 16
Posts: 37
Credit: 1,140,113
RAC: 0
Message 188 - Posted: 21 Jun 2017, 12:23:00 UTC - in response to Message 181.

I killed the gpu wu, because it's a notebook with an old integrated intel gpu...I retry to download cpu wus in this nb


It works, not always (sometimes again this message "No tasks are available for..."), but now i've crunched some cpu wus!

Yavanius
Avatar
Send message
Joined: 18 Nov 16
Posts: 42
Credit: 454,617
RAC: 0
Message 483 - Posted: 15 Sep 2017, 1:14:05 UTC - in response to Message 168.

I haven't gotten any CPU tasks in a whiles myself. I tried pausing all the other projects and all I got out that was the finished tasks cleared out and more GPU tasks came down.

Not really complaining though as I'm getting a steady supply for both Intel and Cuda GPU... just I wanted to crunch some CPU tasks too. :)

~Yav

Yavanius
Avatar
Send message
Joined: 18 Nov 16
Posts: 42
Credit: 454,617
RAC: 0
Message 484 - Posted: 15 Sep 2017, 3:21:48 UTC - in response to Message 483.

On my Dell i5 laptop I turned all 4 cores on and suspended other work and got several CPU WUs.

Wondering if it might have to do something with turning on all cores, I tried that on my ASUS i7 laptop. Fail. I then tried playing with various settings on the client and the project settings also to no luck.

9/14/2017 7:37:26 PM | XANSONS for COD | Sending scheduler request: Requested by user.
9/14/2017 7:37:26 PM | XANSONS for COD | Not requesting tasks: don't need (CPU: ; NVIDIA GPU: ; Intel GPU: job cache full)
9/14/2017 7:37:28 PM | XANSONS for COD | Scheduler request completed


(I had nVidia work off but I can get CUDA work.)

Don't know why suspending everything else on the Dell worked but not on here... O.o

Vlad
Project administrator
Project developer
Project tester
Project scientist
Help desk expert
Send message
Joined: 26 Oct 16
Posts: 322
Credit: 103,382
RAC: 0
Message 485 - Posted: 15 Sep 2017, 16:58:55 UTC - in response to Message 484.

On my Dell i5 laptop I turned all 4 cores on and suspended other work and got several CPU WUs.

Wondering if it might have to do something with turning on all cores, I tried that on my ASUS i7 laptop. Fail. I then tried playing with various settings on the client and the project settings also to no luck.

9/14/2017 7:37:26 PM | XANSONS for COD | Sending scheduler request: Requested by user.
9/14/2017 7:37:26 PM | XANSONS for COD | Not requesting tasks: don't need (CPU: ; NVIDIA GPU: ; Intel GPU: job cache full)
9/14/2017 7:37:28 PM | XANSONS for COD | Scheduler request completed


(I had nVidia work off but I can get CUDA work.)

Don't know why suspending everything else on the Dell worked but not on here... O.o

The project allows to request up to 25 WUs per CPU and up to 50 WUs per GPU at once. However, if I understand how the work fetch works, the client will not request new WUs if it estimates that the assigned WUs (does not matter from this project or from another) may not finish before the deadline. So, do you have any long WUs from other projects currently assigned to this host?

Profile [B@P] Daniel
Send message
Joined: 18 Jun 17
Posts: 25
Credit: 47,963,162
RAC: 0
Message 486 - Posted: 15 Sep 2017, 17:47:56 UTC
Last modified: 15 Sep 2017, 17:51:09 UTC

Check also your NCI apps, BOINC has bug and does not request new WUs for any app if NCI app cannot download new WUs. I had this problem with QCN apps in the past. This problem is probably limited to QCN apps, Monkey apps do not cause this.
____________

Yavanius
Avatar
Send message
Joined: 18 Nov 16
Posts: 42
Credit: 454,617
RAC: 0
Message 487 - Posted: 16 Sep 2017, 4:50:13 UTC - in response to Message 485.
Last modified: 16 Sep 2017, 4:52:50 UTC

However, if I understand how the work fetch works, the client will not request new WUs if it estimates that the assigned WUs (does not matter from this project or from another) may not finish before the deadline. So, do you have any long WUs from other projects currently assigned to this host?


I do have one LHC Vbox WU still leftover at this time and had 3 at that time. Although contradictory to the client, BOINC is weird like that, which only figures that time for one core... however, I think by what you say that with multicore that BOINC server may get a flag on the one core regardless that it is multicore and not allow it.

I should finish that LHC WU tonight. I've been whittling down all the work I got when I was trying to spread processing across the projects and ended up making BOINC get more work. I got a 3rd system that I just added Xansons but didn't get any work, BUT, that has a LHC WU 8 hours out.

So tomorrow, I'll see if this theory pans out but I might not get to playing till the evening...

Right now, XANSONS is my project to play with. :) BTW, Any chance for clients for ARM platform, particularly Android?

~Y

frankhagen
Send message
Joined: 5 Jul 17
Posts: 1
Credit: 521,574
RAC: 0
Message 488 - Posted: 16 Sep 2017, 12:26:18 UTC
Last modified: 16 Sep 2017, 12:31:56 UTC

still no cpu-jobs:

Sa 16 Sep 2017 14:16:02 CEST | XANSONS for COD | Requesting new tasks for CPU
Sa 16 Sep 2017 14:16:03 CEST | XANSONS for COD | Scheduler request completed: got 0 new tasks
Sa 16 Sep 2017 14:16:03 CEST | XANSONS for COD | No tasks sent
Sa 16 Sep 2017 14:16:03 CEST | XANSONS for COD | No tasks are available for XaNSoNS BOINC for CPU (multi-threaded)
Sa 16 Sep 2017 14:16:03 CEST | XANSONS for COD | Tasks for NVIDIA GPU are available, but your preferences are set to not accept them
Sa 16 Sep 2017 14:16:03 CEST | XANSONS for COD | Tasks for AMD/ATI GPU are available, but your preferences are set to not accept them
Sa 16 Sep 2017 14:16:03 CEST | XANSONS for COD | Tasks for Intel GPU are available, but your preferences are set to not accept them


host ID: 1130 - you might be able to find it inside your logs..

Vlad
Project administrator
Project developer
Project tester
Project scientist
Help desk expert
Send message
Joined: 26 Oct 16
Posts: 322
Credit: 103,382
RAC: 0
Message 489 - Posted: 16 Sep 2017, 14:01:09 UTC - in response to Message 488.

still no cpu-jobs:

Sa 16 Sep 2017 14:16:02 CEST | XANSONS for COD | Requesting new tasks for CPU
Sa 16 Sep 2017 14:16:03 CEST | XANSONS for COD | Scheduler request completed: got 0 new tasks
Sa 16 Sep 2017 14:16:03 CEST | XANSONS for COD | No tasks sent
Sa 16 Sep 2017 14:16:03 CEST | XANSONS for COD | No tasks are available for XaNSoNS BOINC for CPU (multi-threaded)
Sa 16 Sep 2017 14:16:03 CEST | XANSONS for COD | Tasks for NVIDIA GPU are available, but your preferences are set to not accept them
Sa 16 Sep 2017 14:16:03 CEST | XANSONS for COD | Tasks for AMD/ATI GPU are available, but your preferences are set to not accept them
Sa 16 Sep 2017 14:16:03 CEST | XANSONS for COD | Tasks for Intel GPU are available, but your preferences are set to not accept them


host ID: 1130 - you might be able to find it inside your logs..

Today I had to reduce the number of generated WUs due to server overload. It will be difficult to get new WUs for the next few hours. This issue is not related to the one that Yavanius has.

Vlad
Project administrator
Project developer
Project tester
Project scientist
Help desk expert
Send message
Joined: 26 Oct 16
Posts: 322
Credit: 103,382
RAC: 0
Message 490 - Posted: 16 Sep 2017, 17:31:22 UTC - in response to Message 487.

BTW, Any chance for clients for ARM platform, particularly Android?

No, sorry. Massive calculations for this project will be completed in about 3 weeks, there is no need for Android devices. By the way, on the BOINC:FAST'17 conference there was a report on the upcoming BOINC project dedicated to mobile devices. Here is their presentation: http://boincfast.ru/files/7415/0520/3622/PresentacionBoinc.pdf.

Yavanius
Avatar
Send message
Joined: 18 Nov 16
Posts: 42
Credit: 454,617
RAC: 0
Message 491 - Posted: 16 Sep 2017, 18:20:18 UTC - in response to Message 485.

So, do you have any long WUs from other projects currently assigned to this host?


Hi Vlad,

I confirmed the theory. If there is any one core working on a long WU (I would take that to be greater than the amount of work you have set to be requested) then the scheduler won't serve multi-core WUs.

I was able to get some multicores on my Asus i7 while I was asleep before they ran out. The older HP Pavilion unfortunately did not finish up in time to get work.

One way around this might be user definable number of cores to use. So if one had 8 cores, they could set only use 4 cores max. ATLAS, part of LHC@home, I believe allows you to do this.

Don't know how long the project will be active, but if you are going to keep going it might be worth investigating.

Thanks for the troubleshooting tip and hope that info is helpful. I haven't done a lot of multicore at other projects (the few that have them) so I can't say if this is a common occurrence.

~Y

Yavanius
Avatar
Send message
Joined: 18 Nov 16
Posts: 42
Credit: 454,617
RAC: 0
Message 492 - Posted: 16 Sep 2017, 18:34:21 UTC - in response to Message 490.

So what's next after the next 3 weeks?

That's quite an interesting study. Didn't quite follow some of the more technical stuff but I get the idea and the numbers were very interesting.

ARM7 architecture just keeps getting more and more powerful and we have 64 bit processors out there now.

'Desktop' CPUs probably will continue to dominate the hard number crunching, but it looks like ARM can provide quite the nice alternative on some tasks, especially those that aren't time critical...

Thanks for the link.

~Y

Vlad
Project administrator
Project developer
Project tester
Project scientist
Help desk expert
Send message
Joined: 26 Oct 16
Posts: 322
Credit: 103,382
RAC: 0
Message 494 - Posted: 16 Sep 2017, 19:08:48 UTC - in response to Message 491.

One way around this might be user definable number of cores to use. So if one had 8 cores, they could set only use 4 cores max. ATLAS, part of LHC@home, I believe allows you to do this.

It's possible to do this via app_config.xml. For example, in this app_config the number of max CPU cores is set to 2. You can put this file to the project directory, just remove the second app_version section for the GPU app. If you want to change the number of cores from 2 to 4, replace:
<cmdline>--nthreads 2</cmdline> <avg_ncpus>2</avg_ncpus>
with
<cmdline>--nthreads 4</cmdline> <avg_ncpus>4</avg_ncpus>

Vlad
Project administrator
Project developer
Project tester
Project scientist
Help desk expert
Send message
Joined: 26 Oct 16
Posts: 322
Credit: 103,382
RAC: 0
Message 495 - Posted: 16 Sep 2017, 19:20:45 UTC - in response to Message 492.

So what's next after the next 3 weeks?

The project will start processsing the entries added to COD in real time. However, there are only about 1500 new entries per month. Only a few hosts are needed to process them.

1 · 2 · Next

Message boards : CPU : No work for cpu


Main page · Your account · Message boards


© 2021 Vladislav Neverov (NRC 'Kurchatov institute'), Nikolay Khrapov (Institute for Information Transmission Problems of RAS)