Message boards : Theory Application : madgraph5amc: is it possible to raise the number of concurrent threads?
Message board moderation

To post messages, you must log in.

AuthorMessage
Lem Novantotto

Send message
Joined: 24 May 23
Posts: 48
Credit: 4,119,075
RAC: 18,342
Message 51532 - Posted: 13 Feb 2025, 9:36:39 UTC

Theory task, linux native version.

During this stage:
INFO: Starting run 
INFO: Using 2 cores 
INFO: Cleaning previous results 
INFO: Generating events without running the shower. 
INFO: Setting up grids

the app uses 2 threads concurrently. However the task is long.

Is it possible to raise this number, to have - say - 6 or 12 concurrent threads? In other words: is there a place to manually set a value for
nb_core
?
I see this setting in some madgraph5amc configuration files under /var/lib/boinc-client/slots/?*/cernvm/shared/configuration/, however, if the task is already in the slot, I think it's too late...

Thanks.
--
Bye, Lem
ID: 51532 · Report as offensive     Reply Quote
Profile Guy
Avatar

Send message
Joined: 9 Feb 08
Posts: 55
Credit: 1,521,616
RAC: 3,319
Message 51533 - Posted: 13 Feb 2025, 10:49:45 UTC - in response to Message 51532.  

I don't believe Theory tasks are able to use more than one core for the crunching part of a task.
If you look at the LHC@home Applications page, it shows that only CMS and ATLAS jobs are Multi-Threaded - they have "mt"
like
(vbox64_mt_mcore_cms)
(vbox64_mt_mcore_atlas)
(native_mt)
while Theory apps do not. It's the way they're compiled. They can't use more than 1 core/thread.
ID: 51533 · Report as offensive     Reply Quote
Lem Novantotto

Send message
Joined: 24 May 23
Posts: 48
Credit: 4,119,075
RAC: 18,342
Message 51534 - Posted: 13 Feb 2025, 11:12:55 UTC - in response to Message 51533.  

I don't believe Theory tasks are able to use more than one core for the crunching part of a task. [...]
If you look at the LHC@home Applications page, it shows that only CMS and ATLAS jobs are Multi-Threaded - they have "mt"
like
(vbox64_mt_mcore_cms)
(vbox64_mt_mcore_atlas)
(native_mt)
while Theory apps do not. It's the way they're compiled. They can't use more than 1 core/thread.


You're wrong. Usually Theory apps use only one thread. One of these apps (maybe even others?), madgraph5amc, can use more than one thread.

lem@oldie:~$ theorytasks
      1 madgraph5amc
lem@oldie:~$ theorystatus
SLOT 8
===> [runRivet] Wed Feb 12 02:20:32 UTC 2025 [boinc pp zinclusive 13000 -,-,106,- - madgraph5amc 2.6.7.atlas2 nlo2jet 100000 4]

lem@oldie:~$ top -u boinc
[...]
    PID UTENTE   PR   NI    VIRT    RES    SHR S  %CPU  %MEM     TIME+ COMANDO                                                              
2563162 boinc     22   2  208820 105688   7316 R  92,3   0,3  13:18.30 madevent_mintMC                                                      
2563741 boinc     22   2  208812 105860   7536 R  76,9   0,3   7:52.22 madevent_mintMC

--
Bye, Lem
ID: 51534 · Report as offensive     Reply Quote
computezrmle
Volunteer moderator
Volunteer developer
Volunteer tester
Help desk expert
Avatar

Send message
Joined: 15 Jun 08
Posts: 2605
Credit: 262,032,123
RAC: 131,390
Message 51536 - Posted: 13 Feb 2025, 12:49:27 UTC - in response to Message 51532.  

Is it possible to raise this number, to have - say - 6 or 12 concurrent threads? In other words: is there a place to manually set a value for ... madgraph5amc ...

By intention, No.
In the past madgraph tasks allocated all available cores (up to ??) but since this breaks the singlecore BOINC configuration it has been limited to 2.
In practice there's no significant benefit to use more cores but it causes lots of trouble.
ID: 51536 · Report as offensive     Reply Quote

Message boards : Theory Application : madgraph5amc: is it possible to raise the number of concurrent threads?


©2025 CERN