Multiprocessor support
Hello, I want to use Gambit for simulation purposes on a multiprocessor machine. I've read in the manual that Gambit does not support multiprocessor machines at the moment. A few months ago there was a discussion on the mailing list about that matter and it sounded as if multiprocessor support was planned. Is there any (preliminary) support? Or can you suggest some kind of workaround? I could imagine to run multiple instances of Gambit that can run on multiple processors and communicate with each other -- perhaps someone has done that already and has a library for that? Best regards Daniel Sadilek
Afficher les réponses par date
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On 5-Dec-06, at 4:00 AM, Daniel Sadilek wrote:
Hello,
I want to use Gambit for simulation purposes on a multiprocessor machine. I've read in the manual that Gambit does not support multiprocessor machines at the moment. A few months ago there was a discussion on the mailing list about that matter and it sounded as if multiprocessor support was planned. Is there any (preliminary) support? Or can you suggest some kind of workaround? I could imagine to run multiple instances of Gambit that can run on multiple processors and communicate with each other -- perhaps someone has done that already and has a library for that?
The design of the multiprocessor support has not advanced. What kind of task are you simulating? Is it one simulation or several small simulations? Marc -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.3 (Darwin) iD8DBQFFdXOF//V9Zc2T/v4RAkMeAJ48+PNIUuHf8H2TMxLp0deUoHLVOwCgoMKG FyfgMf/XSP1XH2T/aWYDJvo= =gR4c -----END PGP SIGNATURE-----
Hi, I want to simulate the behaviour of nodes in a sensor network which communicate with each other. The simulation will be modelled process oriented. I know that parallelising the simulation itself is difficult and perhaps it won't be necessary in terms of performance so perhaps I won't do it at all. But I will use real recorded sensor data as input for the simulation, which need a preprocessing step that is easily parallelizable. So, for the beginning I just want to parallelise the preprocessing but I want the option to parallelise the simulation itself, as well. (I would wish a threading support like that of the Java VM where I just create some threads that are automatically distributed over the available processors.) Best regards Daniel Sadilek On 12/5/06, Marc Feeley <feeley@iro.umontreal.ca> wrote:
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
On 5-Dec-06, at 4:00 AM, Daniel Sadilek wrote:
Hello,
I want to use Gambit for simulation purposes on a multiprocessor machine. I've read in the manual that Gambit does not support multiprocessor machines at the moment. A few months ago there was a discussion on the mailing list about that matter and it sounded as if multiprocessor support was planned. Is there any (preliminary) support? Or can you suggest some kind of workaround? I could imagine to run multiple instances of Gambit that can run on multiple processors and communicate with each other -- perhaps someone has done that already and has a library for that?
The design of the multiprocessor support has not advanced. What kind of task are you simulating? Is it one simulation or several small simulations?
Marc
-----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.3 (Darwin)
iD8DBQFFdXOF//V9Zc2T/v4RAkMeAJ48+PNIUuHf8H2TMxLp0deUoHLVOwCgoMKG FyfgMf/XSP1XH2T/aWYDJvo= =gR4c -----END PGP SIGNATURE-----
Daniel Sadilek wrote:
Hello,
I want to use Gambit for simulation purposes on a multiprocessor machine. I've read in the manual that Gambit does not support multiprocessor machines at the moment. A few months ago there was a discussion on the mailing list about that matter and it sounded as if multiprocessor support was planned. Is there any (preliminary) support? Or can you suggest some kind of workaround? I could imagine to run multiple instances of Gambit that can run on multiple processors and communicate with each other -- perhaps someone has done that already and has a library for that?
Best regards Daniel Sadilek _______________________________________________ Gambit-list mailing list Gambit-list@iro.umontreal.ca https://webmail.iro.umontreal.ca/mailman/listinfo/gambit-list
this sounds like such an overly trivial remark that i'm afraid i might just be uninformed, but here it goes anyway: what about termite? http://toute.ca/ cheers, v.
Hello, I thought about Termite. As far as I know it forbids mutations (set!), but they can be emulated by cells. Does anyone know how they perform compared to simple mutable variables? Best regards Daniel Sadilek On 12/5/06, Vincent Kraeutler <vincent@kraeutler.net> wrote:
Daniel Sadilek wrote:
Hello,
I want to use Gambit for simulation purposes on a multiprocessor machine. I've read in the manual that Gambit does not support multiprocessor machines at the moment. A few months ago there was a discussion on the mailing list about that matter and it sounded as if multiprocessor support was planned. Is there any (preliminary) support? Or can you suggest some kind of workaround? I could imagine to run multiple instances of Gambit that can run on multiple processors and communicate with each other -- perhaps someone has done that already and has a library for that?
Best regards Daniel Sadilek _______________________________________________ Gambit-list mailing list Gambit-list@iro.umontreal.ca https://webmail.iro.umontreal.ca/mailman/listinfo/gambit-list
this sounds like such an overly trivial remark that i'm afraid i might just be uninformed, but here it goes anyway: what about termite? http://toute.ca/
cheers, v. _______________________________________________ Gambit-list mailing list Gambit-list@iro.umontreal.ca https://webmail.iro.umontreal.ca/mailman/listinfo/gambit-list
At 10:00 Uhr +0100 05.12.2006, Daniel Sadilek wrote:
workaround?
(Generally: multiple processes and data passing (open-process, or on unix fork() and pipe() or possibly mmap() (poke me if you want me to release an updated cj-posix library), or (as mentioned) Termite).)
I could imagine to run multiple instances of Gambit that can run on multiple processors and communicate with each other -- perhaps someone has done that already and has a library for that?
I've started writing bindings for PETSc[1] (without a timeline for completion). That library can do multiprocessing (even across a whole cluster). Are you interested in those? Christian. [1] http://www-unix.mcs.anl.gov/petsc/petsc-as/
participants (4)
-
Christian -
Daniel Sadilek -
Marc Feeley -
Vincent Kraeutler