Welcome to MilkyWay@home

no "multicore" tasks


Advanced search

Questions and Answers : Preferences : no "multicore" tasks
Message board moderation

To post messages, you must log in.

AuthorMessage
ursmii

Send message
Joined: 8 Aug 19
Posts: 1
Credit: 34,376,634
RAC: 16,251
30 million credit badge2 year member badge
Message 69544 - Posted: 17 Feb 2020, 19:40:47 UTC

hi
running a 8 and now a 16 core CPU it's ugly for other projects when there are 8 or 16 CPU tasks have been build.
so how can I suppress downloading such tasks. I would rather have all normal 1 core tasks to give other projects to run in parallel

this should be possible having a fair core distribution policy!
ID: 69544 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profilemikey
Avatar

Send message
Joined: 8 May 09
Posts: 2544
Credit: 462,666,679
RAC: 64
300 million credit badge12 year member badgeextraordinary contributions badge
Message 69545 - Posted: 18 Feb 2020, 18:22:29 UTC - in response to Message 69544.  

hi
running a 8 and now a 16 core CPU it's ugly for other projects when there are 8 or 16 CPU tasks have been build.
so how can I suppress downloading such tasks. I would rather have all normal 1 core tasks to give other projects to run in parallel

this should be possible having a fair core distribution policy!


This part "to give other projects to run in parallel" won't happen 99% of the time anyway as that's not how Boinc works. Normally you run project A on all your cpu cores then after a bit project B will be running on alll of your cpu cores. There is an algorythm to do it that's very complicated but it ends doing what I said. You will NEVER be able to schedule one cpu core to run project A and 7 cpu cores run project B for ever, Boinc is NOT programmed to do that.

What you can do is change the resource share of your projects so project A gets less time than project B so for instance you could set project A to a resource share of 25 and project B to a resource share of 75 and project B would have 3 times as much RAC as project A does. It can take awhile for it to even out but it will do it in time. You can do this over as many projects as you have,just pickanumber as the max of either 100 or 1000 and then work out the RAC percentge you want each project to have. The one problem to this is unequal length workunits at the different projects which end up having widely varying credit rewards, that can make the evening out to your percentage numbers even longer.
ID: 69545 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
jdzukley

Send message
Joined: 26 May 11
Posts: 30
Credit: 33,139,843
RAC: 17
30 million credit badge10 year member badge
Message 70655 - Posted: 9 Mar 2021, 13:16:05 UTC
Last modified: 9 Mar 2021, 13:36:58 UTC

I have the reverse problem, I have not been able to get any N-Body simulation jobs in the last 24+ hours.

In preferences I have CPU set to yes, and N-Body Simulation set to yes.

The server status page says there are 1000 jobs for N-Body in que.

In options, the store at least que has 2 days of work so there is plenty of que space available for work.

The event log shows the following messages:
3/9/2021 8:11:59 AM | Milkyway@Home | Sending scheduler request: To fetch work.
3/9/2021 8:11:59 AM | Milkyway@Home | Requesting new tasks for CPU and NVIDIA GPU
3/9/2021 8:12:00 AM | Milkyway@Home | Scheduler request completed: got 0 new tasks
3/9/2021 8:12:00 AM | Milkyway@Home | No tasks sent
3/9/2021 8:12:00 AM | Milkyway@Home | No tasks are available for Milkyway@home N-Body Simulation
3/9/2021 8:12:00 AM | Milkyway@Home | Tasks for AMD/ATI GPU are available, but your preferences are set to not accept them
3/9/2021 8:12:00 AM | Milkyway@Home | Project requested delay of 91 seconds

I would appreciate any help.


NEVER MIND - MOSTLY. While writing this post, I finally go a load of CPU N-Body jobs. My CPU que totally emptied out during the night, and I was without CPU tasks.

Bottom line, I currently have plenty of work. I do play with my preferences as I would like to run only CPU multi core N-Body work. I down load as many jobs as the system will allow, Then I update preferences to get GPU work. I rather not run separation work on my CPU when I run loads of them on my GPU's.

Are there total job que limits that kick in? For example, it appears that the max GPU separation jobs to download is 600. Could that limit my CPU jobs, even if my CPU que is totally empty?

It would be nice if the preference (like with Einstein@home) that do not download CPU jobs where GPU tasks are available. That is what I am overall attempting to manipulate...
ID: 70655 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profilemikey
Avatar

Send message
Joined: 8 May 09
Posts: 2544
Credit: 462,666,679
RAC: 64
300 million credit badge12 year member badgeextraordinary contributions badge
Message 70656 - Posted: 10 Mar 2021, 0:48:42 UTC - in response to Message 70655.  

I have the reverse problem, I have not been able to get any N-Body simulation jobs in the last 24+ hours.

In preferences I have CPU set to yes, and N-Body Simulation set to yes.

The server status page says there are 1000 jobs for N-Body in que.

In options, the store at least que has 2 days of work so there is plenty of que space available for work.

The event log shows the following messages:
3/9/2021 8:11:59 AM | Milkyway@Home | Sending scheduler request: To fetch work.
3/9/2021 8:11:59 AM | Milkyway@Home | Requesting new tasks for CPU and NVIDIA GPU
3/9/2021 8:12:00 AM | Milkyway@Home | Scheduler request completed: got 0 new tasks
3/9/2021 8:12:00 AM | Milkyway@Home | No tasks sent
3/9/2021 8:12:00 AM | Milkyway@Home | No tasks are available for Milkyway@home N-Body Simulation
3/9/2021 8:12:00 AM | Milkyway@Home | Tasks for AMD/ATI GPU are available, but your preferences are set to not accept them
3/9/2021 8:12:00 AM | Milkyway@Home | Project requested delay of 91 seconds

I would appreciate any help.


NEVER MIND - MOSTLY. While writing this post, I finally go a load of CPU N-Body jobs. My CPU que totally emptied out during the night, and I was without CPU tasks.

Bottom line, I currently have plenty of work. I do play with my preferences as I would like to run only CPU multi core N-Body work. I down load as many jobs as the system will allow, Then I update preferences to get GPU work. I rather not run separation work on my CPU when I run loads of them on my GPU's.

Are there total job que limits that kick in? For example, it appears that the max GPU separation jobs to download is 600. Could that limit my CPU jobs, even if my CPU que is totally empty?

It would be nice if the preference (like with Einstein@home) that do not download CPU jobs where GPU tasks are available. That is what I am overall attempting to manipulate...


Yes there are download limits here as sometimes they have limited numbers of tasks and a 16 or 32 core Ryzien pc could come in and get 99% of them with a 10 day cache. There is also a 10 minute time-out inbetween requests before you get the gpu tasks, a past admin did that and the current ones can't find the setting to undo it, but gpu's also have a max limit. One thing most of us do is run a 2nd gpu project that responds quickly, like Collatz, as a zero resource share to run tasks inbetween batches of MW gpu tasks. That way when MW says nope too soon Collatz says here crunch this, then when MW refills the cache Collatz just sits their and waits until you are out of MW units again.
ID: 70656 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Michael Matthias

Send message
Joined: 9 Jul 20
Posts: 1
Credit: 26,042,967
RAC: 37,065
20 million credit badge1 year member badge
Message 70718 - Posted: 8 Apr 2021, 19:15:30 UTC

I have dozens of tasks for MilkyWay@home, almost 200. Is this normal? Seems it would be an inefficient way of doing things.
ID: 70718 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profilemikey
Avatar

Send message
Joined: 8 May 09
Posts: 2544
Credit: 462,666,679
RAC: 64
300 million credit badge12 year member badgeextraordinary contributions badge
Message 70720 - Posted: 9 Apr 2021, 10:50:37 UTC - in response to Message 70718.  

I have dozens of tasks for MilkyWay@home, almost 200. Is this normal? Seems it would be an inefficient way of doing things.


It's their project and they like multicore tasks as opposed to running one task per cpu core, you can always limit it with an app_config file
ID: 70720 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Aurum
Avatar

Send message
Joined: 11 Jul 17
Posts: 11
Credit: 1,324,823,618
RAC: 210
1 billion credit badge4 year member badge
Message 70754 - Posted: 26 Apr 2021, 17:12:10 UTC - in response to Message 70720.  

I have dozens of tasks for MilkyWay@home, almost 200. Is this normal? Seems it would be an inefficient way of doing things.

It's their project and they like multicore tasks as opposed to running one task per cpu core, you can always limit it with an app_config file
This works for me:
<app_config>
<app>
    <name>milkyway</name>
    <gpu_versions>
        <cpu_usage>1.0</cpu_usage>
        <gpu_usage>0.5</gpu_usage>
    </gpu_versions>
    <max_concurrent>4</max_concurrent>
</app>
<app_version>
    <app_name>milkyway_nbody</app_name>
    <plan_class>mt</plan_class>
    <avg_ncpus>1</avg_ncpus>
    <cmdline>--nthreads 1</cmdline>
    <max_concurrent>4</max_concurrent>
</app_version>
</app_config>

ID: 70754 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profilemikey
Avatar

Send message
Joined: 8 May 09
Posts: 2544
Credit: 462,666,679
RAC: 64
300 million credit badge12 year member badgeextraordinary contributions badge
Message 70755 - Posted: 26 Apr 2021, 23:23:41 UTC - in response to Message 70754.  
Last modified: 26 Apr 2021, 23:26:41 UTC

I have dozens of tasks for MilkyWay@home, almost 200. Is this normal? Seems it would be an inefficient way of doing things.

It's their project and they like multicore tasks as opposed to running one task per cpu core, you can always limit it with an app_config file
This works for me:
<app_config>
<app>
    <name>milkyway</name>
    <gpu_versions>
        <cpu_usage>1.0</cpu_usage>
        <gpu_usage>0.5</gpu_usage>
    </gpu_versions>
    <max_concurrent>4</max_concurrent>
</app>
<app_version>
    <app_name>milkyway_nbody</app_name>
    <plan_class>mt</plan_class>
    <avg_ncpus>1</avg_ncpus>
    <cmdline>--nthreads 1</cmdline>
    <max_concurrent>4</max_concurrent>
</app_version>
</app_config>


Thanks I made a copy and saved it so I can also use a single cpu core at a time!!
ID: 70755 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote

Questions and Answers : Preferences : no "multicore" tasks

©2021 Astroinformatics Group