wu limit

[OldChap]

Max 25 per thread there seems no way around it except maybe run two instances of Boinc or run VM's

To run with networking off until the 16th you will need average wu completion time of around 4.8 hours or more... so which projects run that long for you ??

Question: are normal non vina wu's better/faster on 64bit than 32bit os?

[[Ion]]

CEP2 runs for quite a long time...generally 10+ hours
MCM runs for 6+ hours on most of my systems (only exception is the Sandy i5s, since they are fast per-core and don't have HT)--even the Westmere system take 7 hours for them.

[Rob_B]

I can easily abort all Autodock wu's using Boinctasks. But I don't like doing that. And the 25 wu's per thread don't last till morning anyway. (I actually only get 192 wu's max on the 8 thread machines)

I will run only vina wu's or atleast as much as possible during the challenge and go back to MCM and UGM after that.

[Snow Crash]

use cc_config to "simulate " any number of cpu cores using ...

<ncpus>32</ncpus>


then limit the number of concurrent tasks using app_config

<app>
<name>ugm1</name>
<max_concurrent>8</max_concurrent>
</app>


known app names: 'mcm1', 'ugm1', 'cep2', 'fahv', 'faah'

[Rob_B]

Thanks. I will have a look at that.

[OldChap]

I wish I had that info a week ago SC but I know it now

[Rob_B]

It works. But you can only run one project. With ncpu @ 16 on a 4c/8t cpu it will run 16 wu's at once.
Or you have set it back to default when it downloaded enough wu's.

[Snow Crash]

It works. But you can only run one project.

you can set up multiple apps in the config, you would just need to make sure you have enough WUs of each type to feed it properly.

With ncpu @ 16 on a 4c/8t cpu it will run 16 wu's at once.

that is why we make the app_config with max_concurrent, so you can limit the number = to the real number of threads (ie 4C/ 8T) you have
<app>
<name>ugm1</name>
<max_concurrent>4</max_concurrent>
<name>mcm1</name>
<max_concurrent>4</max_concurrent>

</app>

and yes, in this case, where the request is for a 1 time bump in the number of WU to cache you could set it, pull a bunch of work and then set it back. You only need to "advanced --> read config files" to activate changes in either or both cc_config (goes in main boinc data folder) and app_config (goes in wcg project folder).

[Rob_B]

Thanks Snow Crash. I figured that out already.

But the problem I had was that I had plenty of ugm1 wu's that had to finish and I had set the device profile to FAAH only (both autodock and vina)
with 16 cpu's it would run 16 wu's and if limit it to 4 wu's for both in the app_config file it would run only 4 wu's when the ugm1 wu's were all finished.

Now I have 2 crunchers that have plenty of wu's and back to 8 threads. And no network till sunday.

That one cruncher that is allowed network communication is downloading work for 32 cores and only running fahv wu's.
That faah wu's that get downloaded don't start because I set app_config to:
<name>faah</name>
<max_concurrent>-1</max_concurrent>

That way I can abort these when I get home.

[Snow Crash]

nice enhancement on using -1 for faah

[Rob_B]

nice enhancement on using -1 for faah

0 didn't work and I had seen the -1 in cc_config. Tried it and it worked.

Now the crunchers can run overnight without faah wu's starting.

[Otis11]

One note - you don't want to run it though with BOINC thinking you have more cores than you actually do. It forces your CPU into cache swapping which hurts efficiency. Good trick to get more WUs though.

Also, you can always connect and DL more, just set your upload limit to 1kb/s and manually update. You'll get a bunch of WUs before a single one finishes the upload.