Message boards : Number crunching : Maximum number of tasks
Message board moderation

To post messages, you must log in.

AuthorMessage
Alessio Susi
Avatar

Send message
Joined: 8 Mar 15
Posts: 12
Credit: 1,299,524
RAC: 49
Message 39094 - Posted: 10 Jun 2019, 7:15:01 UTC

Hi. How can I set a maximum number of tasks for Theory Simulations? I'd like to run almost 3 on my PCs.
ASUS X570 E-Gaming
AMD Ryzen 9 3950X, 16 core / 32 thread 4.4 GHz
AMD Radeon Sapphire RX 480 4GB Nitro+
Nvidia GTX 1080 Ti Gaming X Trio
4x16 GB Corsair Vengeance RGB 3466 MHz

ID: 39094 · Report as offensive     Reply Quote
Toby Broom
Volunteer moderator

Send message
Joined: 27 Sep 08
Posts: 598
Credit: 373,291,597
RAC: 54,690
Message 39096 - Posted: 10 Jun 2019, 9:20:28 UTC

You can adjust settings in LHC@home preferences, under your account at the top right of web page.
ID: 39096 · Report as offensive     Reply Quote
Alessio Susi
Avatar

Send message
Joined: 8 Mar 15
Posts: 12
Credit: 1,299,524
RAC: 49
Message 39099 - Posted: 10 Jun 2019, 10:57:01 UTC
Last modified: 10 Jun 2019, 10:57:24 UTC

I want to change the number of cores only for Theory Simulations, not for the project. Every task uses more than 2GB of RAM and I have other projects to run.
ASUS X570 E-Gaming
AMD Ryzen 9 3950X, 16 core / 32 thread 4.4 GHz
AMD Radeon Sapphire RX 480 4GB Nitro+
Nvidia GTX 1080 Ti Gaming X Trio
4x16 GB Corsair Vengeance RGB 3466 MHz

ID: 39099 · Report as offensive     Reply Quote
Toby Broom
Volunteer moderator

Send message
Joined: 27 Sep 08
Posts: 598
Credit: 373,291,597
RAC: 54,690
Message 39100 - Posted: 11 Jun 2019, 6:09:38 UTC

I think you'll need to use an app_config.xml then.

<app_config>
<app_version>
<app_name>Theory</app_name>
<plan_class>vbox64_mt_mcore</plan_class>
<avg_ncpus>1</avg_ncpus>
</app_version>
</app_config>
ID: 39100 · Report as offensive     Reply Quote
andy_taximan

Send message
Joined: 2 Feb 15
Posts: 2
Credit: 2,849,962
RAC: 0
Message 39453 - Posted: 29 Jul 2019, 15:29:44 UTC

Ive just set up Theory Simulation for 8 tasks x 8cpu (estimated at 17hours) as i have 64 threads/64gig memory , thing is my cpu is only running at 40% with 20% memory usage, would i be better off running 2 tasks x 32 threads or any other combination? I just had a good run on Amicable numbers doing 1 task x 58 threads.
ID: 39453 · Report as offensive     Reply Quote
Toby Broom
Volunteer moderator

Send message
Joined: 27 Sep 08
Posts: 598
Credit: 373,291,597
RAC: 54,690
Message 39455 - Posted: 29 Jul 2019, 16:47:10 UTC

I run single core, however this requires the most RAM.

Run the lowest number of threads that you can that fits in your ram, since you can save plenty of ram by running multicore/thread.
ID: 39455 · Report as offensive     Reply Quote
andy_taximan

Send message
Joined: 2 Feb 15
Posts: 2
Credit: 2,849,962
RAC: 0
Message 39456 - Posted: 29 Jul 2019, 17:09:12 UTC - in response to Message 39455.  

I run single core, however this requires the most RAM.

Run the lowest number of threads that you can that fits in your ram, since you can save plenty of ram by running multicore/thread.


OK by default it wants to do 2 x 32, what do i need to do 1 x 64 ?
ID: 39456 · Report as offensive     Reply Quote
Toby Broom
Volunteer moderator

Send message
Joined: 27 Sep 08
Posts: 598
Credit: 373,291,597
RAC: 54,690
Message 39457 - Posted: 29 Jul 2019, 17:45:35 UTC - in response to Message 39456.  

You can edit your Max # of CPUs for this project to 1 or use the app config that I posted to limit the number of threads to 1.

I can only do about 28 Theory concurrently with 128GB of ram
ID: 39457 · Report as offensive     Reply Quote
Luigi R.
Avatar

Send message
Joined: 7 Feb 14
Posts: 99
Credit: 5,027,000
RAC: 0
Message 39458 - Posted: 29 Jul 2019, 18:00:48 UTC - in response to Message 39456.  

E.g. max 3 tasks for Theory (commented line: max 16 tasks for LHC@home)
<app_config>
  <app>
    <name>Theory</name>
    <max_concurrent>3</project_max_concurrent>
  </app>
  <app_version>
    <app_name>Theory</app_name>
    <plan_class>vbox64_mt_mcore</plan_class>
    <avg_ncpus>1.0</avg_ncpus>
    <cmdline>--nthreads 1</cmdline>
  </app_version>
  <!--<project_max_concurrent>16</project_max_concurrent>-->
</app_config>

For project settings use LHC@home preferences, if you need it one day.

Maybe your job cache on BOINC should be >> 0. Suppose you have a 16-threaded host. I don't know what exactly happens if BOINC downloads 5 Theory + 11 other LHC@home subproject or another project tasks. I think your computer would use only 3+11 threads and 2 would stay idle.
ID: 39458 · Report as offensive     Reply Quote
Alessio Susi
Avatar

Send message
Joined: 8 Mar 15
Posts: 12
Credit: 1,299,524
RAC: 49
Message 40615 - Posted: 23 Nov 2019, 17:07:47 UTC

I want to set maximum number of tasks for both Theory and CMS. How can I do it in a single app_config.xml?
ASUS X570 E-Gaming
AMD Ryzen 9 3950X, 16 core / 32 thread 4.4 GHz
AMD Radeon Sapphire RX 480 4GB Nitro+
Nvidia GTX 1080 Ti Gaming X Trio
4x16 GB Corsair Vengeance RGB 3466 MHz

ID: 40615 · Report as offensive     Reply Quote
computezrmle
Volunteer moderator
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 15 Jun 08
Posts: 1443
Credit: 76,612,229
RAC: 99,971
Message 40616 - Posted: 23 Nov 2019, 17:46:53 UTC - in response to Message 40615.  

I want to set maximum number of tasks for both Theory and CMS. How can I do it in a single app_config.xml?

To download or to run concurrently?

An app_config.xml can only limit the number of tasks that are running concurrently.
See <max_concurrent> and <project_max_concurrent> from Luigi's example.
Simply add a CMS section.

To limit the number of tasks to be downloaded you'll have to set a limit at your web preferences page and/or limit the work buffer of your BOINC client.
ID: 40616 · Report as offensive     Reply Quote
Alessio Susi
Avatar

Send message
Joined: 8 Mar 15
Posts: 12
Credit: 1,299,524
RAC: 49
Message 40619 - Posted: 23 Nov 2019, 18:27:18 UTC - in response to Message 39458.  
Last modified: 23 Nov 2019, 18:28:25 UTC


<app_config>
  <app>
    <name>Theory</name>
    <max_concurrent>3</project_max_concurrent>
  </app>
  <app_version>
    <app_name>Theory</app_name>
    <plan_class>vbox64_mt_mcore</plan_class>
    <avg_ncpus>1.0</avg_ncpus>
    <cmdline>--nthreads 1</cmdline>
  </app_version>
  <!--<project_max_concurrent>16</project_max_concurrent>-->
</app_config>



I wrote this in an xml file but BOINC tells me this:
23/11/2019 19:25:25 | LHC@home | Missing </app> in app_config.xml
and I don't know what I have to write to add CMS settings.
ASUS X570 E-Gaming
AMD Ryzen 9 3950X, 16 core / 32 thread 4.4 GHz
AMD Radeon Sapphire RX 480 4GB Nitro+
Nvidia GTX 1080 Ti Gaming X Trio
4x16 GB Corsair Vengeance RGB 3466 MHz

ID: 40619 · Report as offensive     Reply Quote
computezrmle
Volunteer moderator
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 15 Jun 08
Posts: 1443
Credit: 76,612,229
RAC: 99,971
Message 40620 - Posted: 23 Nov 2019, 19:26:01 UTC - in response to Message 40619.  

This line is wrong:
<max_concurrent>3</project_max_concurrent>

Even in Luigi's example (sorry, didn't see this before)

Strictly follow the template given here:
https://boinc.berkeley.edu/wiki/Client_configuration

Example:
<app_config>
  <app>
    <name>CMS</name>
    <max_concurrent>2</max_concurrent>
  </app>

  <app>
    <name>Theory</name>
    <max_concurrent>3</max_concurrent>
  </app>
  <!--<project_max_concurrent>16</project_max_concurrent>-->
</app_config>


<app_version> sections are no longer required for Theory and CMS since both are singlecore.
A BOINC client restart is required to reset the old settings.
ID: 40620 · Report as offensive     Reply Quote
Aurum
Avatar

Send message
Joined: 12 Jun 18
Posts: 88
Credit: 35,826,709
RAC: 6,682
Message 40625 - Posted: 23 Nov 2019, 23:53:26 UTC - in response to Message 40620.  

A BOINC client restart is required to reset the old settings.
Unless one is using BoincTasks 1.80 and the Extra/Read config files (cc/app_config).
ID: 40625 · Report as offensive     Reply Quote

Message boards : Number crunching : Maximum number of tasks


©2020 CERN