[Gate-users] [EXT] Spam (10.893): Fwd: IS GATE SUPPORTING MULTITHREADING ?

[Mohamed Mohamed]

Dear Bud Wendt,

Thank you for your interresting response, now  it's clear,

Just i want to know the tools to run this?

thank you again
regards.




Le lun. 27 avr. 2020 à 14:58, Wendt III,Richard <rwendt at mdanderson.org> a
écrit :

> Mohamed, et al.-
>
>
>
> It happens that I have some very recent experience that directly addresses
> your question. In order to work from home as we are doing in this age of
> pandemic, I purchased a computer with an 8-core i9-9900K CPU. This has a
> turbo speed of 5 GHz for a single core, and the maximum speed drops off as
> the number of actively working cores increases. It is hyperthreaded, so the
> system monitoring tools show 16 “CPUs”.
>
>
>
> I have a GATE simulation that I have been running on a dual 3.2 GHz Xeon
> system at work that takes about 8 hours a run and is taking 7 hours a run
> on an old AMD 3.6 GHz Phenom Ii computer that I had at home. On this new
> computer, a single job takes less than 3 hours. (Note that my work computer
> has older software whereas the two at home are brand new GATE 9.0
> installations on Ubuntu 19.10.)
>
>
>
> I ran the exact same .mac files in simultaneous jobs by creating a
> separate directory for each job, although I used a random seed (because I
> also wanted to get a sense of the reproducibility of my simulations while I
> was at it). The attached plots show that the average run time per job
> increases slowly with the number of jobs until all 8 physical cores are
> active and then it increases more rapidly. On the basis of work done per
> unit time, the increase is roughly linear with the number of  jobs until
> all of the cores are active and then the rate drops off and actually gets
> worse when all of the 16 threads are running jobs. This point has been made
> in this mailing list in the past that one should run only as many
> simultaneous jobs as there are physical cores. These plots illustrate the
> effect with a current generation CPU.
>
>
>
> For what it is worth, the CPU temperature is around 180F with all 8 cores
> running jobs. The result of my simulations was within 1.5% among all of the
> runs. My simulation uses very little memory, so I did not encounter any
> memory issues and cannot address that aspect of your question.
>
>
>
> Bud Wendt
>
>
>
>
>
> *From:* Gate-users <gate-users-bounces at lists.opengatecollaboration.org> *On
> Behalf Of *Mohamed Mohamed
> *Sent:* Sunday, April 26, 2020 6:12 PM
> *To:* gate-users at lists.opengatecollaboration.org
> *Subject:* [EXT] Spam (10.893):[Gate-users] Fwd: IS GATE SUPPORTING
> MULTITHREADING ?
>
>
>
> *WARNING: *This email originated from outside of MD Anderson. Please
> validate the sender's email address before clicking on links or attachments
> as they may not be safe.
>
>
>
> Hello David,
>
>
>
> I would to hank you warmly for your interresting and clear response, that
> was very helpful and largely sufficient for me,
>
>
>
> Do you mind if i ask you to know about the potential gain from executing
> the same .mac on many cores of my processor (if i double the RAM, then i
> run the .mac on 2 or 3 cores? ), also will i need other tools to do that?
>
>
>
> Finally, how GATE can be multithreaded on GPU and not on CPU? whats the
> difference?
>
>
>
> Thank you again,
>
> Regards.
>
>
>
>
>
> Le ven. 10 avr. 2020 à 14:34, David Boersma <david.boersma at acmit.at> a
> écrit :
>
> Hi Mohamed,
>
>
>
> We (OpenGate collaboration) would love to see Gate multithreaded, but we
> have discussed various approaches to do this and all of them require a lot
> of work (a lot of Gate code is not "thread safe"), so much work that with
> the available manpower it will probably take very long (years?) to complete
> this transition (changing all the code and then validating all of it...),
> so we haven't dared to start this project yet.
>
>
>
> The discussions you have read here in the mailing list recently (and also
> longer ago, actually) are about starting a Gate process many times with the
> same mac file on the same machine, basically once for every physical core
> on the machine. Once they are all finished you combine the results. That's
> a bit resource-wasteful (all processes use RAM and CPU time to construct
> their own copy of the same geometry and the physics tables; with
> multithreading you would do this only once, and then all cores can
> propagate particles in parallel, all using the same one geometry and one
> set of physics tables). But it works, if you install enough RAM.
>
>
>
> Is this sufficient clarification? 😊
>
>
>
> HTH,
>
> David Boersma
> ------------------------------
>
> *Von:* Gate-users <gate-users-bounces at lists.opengatecollaboration.org> im
> Auftrag von Mohamed Mohamed <melkatib1 at gmail.com>
> *Gesendet:* Freitag, 10. April 2020 12:20:35
> *An:* gate-users at lists.opengatecollaboration.org
> *Betreff:* [Gate-users] IS GATE SUPPORTING MULTITHREADING ?
>
>
>
> Hi all,
>
>
>
> I am GATE 8.2 used, when i was trying to install GATE with multithreading
> (after doing same thing for Geant4 installation) i had an errormessage
> indicating that multithreading is not supported, after tha message i
> deactivate multithreadin option and completing installation, BUT recently i
> am hearing some users talkin about capability to run simulation on all
> their threads (real nodes+threads), I WILL APPRECIATE ANY CLARIFICATION
> ABOUT THAT,
>
>
>
> kind regards,
>
> Mohamed
>
> The information contained in this e-mail message may be privileged,
> confidential, and/or protected from disclosure. This e-mail message may
> contain protected health information (PHI); dissemination of PHI should
> comply with applicable federal and state laws. If you are not the intended
> recipient, or an authorized representative of the intended recipient, any
> further review, disclosure, use, dissemination, distribution, or copying of
> this message or any attachment (or the information contained therein) is
> strictly prohibited. If you think that you have received this e-mail
> message in error, please notify the sender by return e-mail and delete all
> references to it and its contents from your systems.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.opengatecollaboration.org/pipermail/gate-users/attachments/20200430/268065e4/attachment.html>