Hi Jason,
not sure I understand your doubt.
Once a simulation has finished, you need to sum up the results to calculate the average and uncertainty of your estimator.
(in FLUKA this is the only way available to compute the statistical uncertainty, that’s why you need to run more jobs).
The computed uncertainty would diminish as a function of the square root of the number of histories.
If the uncertainty is < 10%, the result is typically considered as reliable, otherwise you need to run more histories (of course by changing the initial random seed).
I also recommend you to have a look at these slides (from page 29 in particular)
https://indico.cern.ch/event/694979/contributions/2927088/
Best, luigi
On 21 Aug 2019, at 09:47, TYRRELL, JASON <J.TYRRELL3_at_nuigalway.ie<mailto:J.TYRRELL3_at_nuigalway.ie>> wrote:
Apologises, 60 million.
Get Outlook for Android<
https://aka.ms/ghei36>
________________________________
From: TYRRELL, JASON
Sent: Wednesday, August 21, 2019 8:40:29 AM
To: Luigi Salvatore Esposito <luigi.salvatore.esposito_at_cern.ch<mailto:luigi.salvatore.esposito_at_cern.ch>>
Cc: fluka-discuss_at_fluka.org<mailto:fluka-discuss_at_fluka.org> <fluka-discuss_at_fluka.org<mailto:fluka-discuss_at_fluka.org>>
Subject: Re: [fluka-discuss]: Utilizing four cores
Hi Luigi,
That's great thanks, I'll have a look through that pdf.
To clarify, I am running four jobs( one in each core) at the moment, with 15 million histories for each, all running in parallel. Therefore, when I combine each job the average all four will reduce my statistical uncertainty and produce much more accurate results.
In my second email, I thought that if you combine 4 jobs of 15 million histories each, these add to produce results equating to 50 million histories. But this is probably incorrect?
Thanks again!
Jason.
Get Outlook for Android<
https://aka.ms/ghei36>
________________________________
From: Luigi Salvatore Esposito <luigi.salvatore.esposito_at_cern.ch<mailto:luigi.salvatore.esposito_at_cern.ch>>
Sent: Wednesday, August 21, 2019 8:06:16 AM
To: TYRRELL, JASON <J.TYRRELL3_at_nuigalway.ie<mailto:J.TYRRELL3_at_nuigalway.ie>>
Cc: fluka-discuss_at_fluka.org<mailto:fluka-discuss_at_fluka.org> <fluka-discuss_at_fluka.org<mailto:fluka-discuss_at_fluka.org>>
Subject: Re: [fluka-discuss]: Utilizing four cores
Hi Jason,
the statement in your first email is correct.
Not sure I completely understood your second email, but the combination of the results of different jobs is usually fast (few seconds to few minutes depending on the size of the scoring files).
As mentioned before, you need to combine the results from several jobs in order to calculate the averages and statistical uncertainties.
Please give also a look at these slides
https://indico.cern.ch/event/694979/contributions/2927109/
Hope this clarify your questions.
Best luigi
On 20 Aug 2019, at 19:58, TYRRELL, JASON <J.TYRRELL3_at_nuigalway.ie<mailto:J.TYRRELL3_at_nuigalway.ie>> wrote:
In addition to the previous email.
Therefore, when I run 1 million histories I am actually getting 4 million histories, in a shorter time after I combine them?
Jason
Get Outlook for Android<
https://aka.ms/ghei36>
________________________________
From: TYRRELL, JASON
Sent: Tuesday, August 20, 2019 6:55:32 PM
To: Luigi Salvatore Esposito <luigi.salvatore.esposito_at_cern.ch<mailto:luigi.salvatore.esposito_at_cern.ch>>
Cc: fluka-discuss_at_fluka.org<mailto:fluka-discuss_at_fluka.org> <fluka-discuss_at_fluka.org<mailto:fluka-discuss_at_fluka.org>>
Subject: Re: [fluka-discuss]: Utilizing four cores
Ok great.
Can you correct me if I'm wrong please.
I have set my number of histories to 1 million and set spawn equal to 4. Therefore, each core is running 1 million histories, with an ETA of 1 hour and 30 mins. Each job is assigned a different random number seed with each job saved separately. Then, when the run is complete, I use the data window to combine each job into one?
Thanks again
Jason.
Get Outlook for Android<
https://aka.ms/ghei36>
________________________________
From: Luigi Salvatore Esposito <luigi.salvatore.esposito_at_cern.ch<mailto:luigi.salvatore.esposito_at_cern.ch>>
Sent: Tuesday, August 20, 2019 6:14:55 PM
To: TYRRELL, JASON <J.TYRRELL3_at_nuigalway.ie<mailto:J.TYRRELL3_at_nuigalway.ie>>
Cc: fluka-discuss_at_fluka.org<mailto:fluka-discuss_at_fluka.org> <fluka-discuss_at_fluka.org<mailto:fluka-discuss_at_fluka.org>>
Subject: Re: [fluka-discuss]: Utilizing four cores
Yes!
You need to do combine all the jobs to compute the statistical uncertainty of you estimators.
Best, luigi
On 20 Aug 2019, at 19:02, TYRRELL, JASON <J.TYRRELL3_at_nuigalway.ie<mailto:J.TYRRELL3_at_nuigalway.ie>> wrote:
That's great thank Luigi. Will I also have to combine the 4 jobs at the end?
Regards,
Jason.
Get Outlook for Android<
https://aka.ms/ghei36>
________________________________
From: Luigi Salvatore Esposito <luigi.salvatore.esposito_at_cern.ch<mailto:luigi.salvatore.esposito_at_cern.ch>>
Sent: Tuesday, August 20, 2019 5:04:09 PM
To: TYRRELL, JASON <J.TYRRELL3_at_nuigalway.ie<mailto:J.TYRRELL3_at_nuigalway.ie>>
Cc: fluka-discuss_at_fluka.org<mailto:fluka-discuss_at_fluka.org> <fluka-discuss_at_fluka.org<mailto:fluka-discuss_at_fluka.org>>
Subject: Re: [fluka-discuss]: Utilizing four cores
Dear Jason,
in the spawn column (next to the input you want to run) you can digit the number of desired spawn jobs.
When you start the run, the input number of jobs will run in parallel (in your case at most 4).
If you click on the Help icon in the Run view, the Flair manual will be opened and you can see a picture showing an example.
Hope this could help.
Cheers luigi
On 20 Aug 2019, at 17:54, TYRRELL, JASON <J.TYRRELL3_at_nuigalway.ie<mailto:J.TYRRELL3_at_nuigalway.ie>> wrote:
Hi Luigi,
I am using flair. I saw the spawn option in a thread but I didn't understand. How exactly would I do that? I am coming to the end of my project and it is quite a problem.
Thanks a lot,
Jason.
Get Outlook for Android<
https://aka.ms/ghei36>
________________________________
From: Luigi Salvatore Esposito <luigi.salvatore.esposito_at_cern.ch<mailto:luigi.salvatore.esposito_at_cern.ch>>
Sent: Tuesday, August 20, 2019 4:51:17 PM
To: TYRRELL, JASON <J.TYRRELL3_at_nuigalway.ie<mailto:J.TYRRELL3_at_nuigalway.ie>>
Cc: fluka-discuss_at_fluka.org<mailto:fluka-discuss_at_fluka.org> <fluka-discuss_at_fluka.org<mailto:fluka-discuss_at_fluka.org>>
Subject: Re: [fluka-discuss]: Utilizing four cores
Dear Jason,
you can run more jobs in parallel.
For example, if you use Flair you can do it from the Run view and choose the number of spawned jobs.
Best luigi
On 20 Aug 2019, at 16:32, TYRRELL, JASON <J.TYRRELL3_at_nuigalway.ie<mailto:J.TYRRELL3_at_nuigalway.ie>> wrote:
Hello Everyone,
I am running FLUKA on a PC with four cores however, FLUKA is only using two of them. Can someone explain how I can utilize all four cores?
Thanks a lot,
Jason.
__________________________________________________________________________
You can manage unsubscription from this mailing list at
https://www.fluka.org/fluka.php?id=acc_info
Received on Wed Aug 21 2019 - 16:12:35 CEST