yade-dev team mailing list archive

[Anton Gladky <gladky.anton@xxxxxxxxx>]

> The downside of MPI is that it would make everything (and I mean
> *everything*) upside down, and besides the obvious (passing particles
> around, synchronization etc), there is bunch of other issues to consider,
> such as how to evaluate python expressions which need access to remote nodes
> and similar; the scripting interface would have to change substantially, the
> UI as well (OpenGL would have to get data over the wire) etc etc.

Well, I think it would be easier to start from "very simple" yade-mpi
version. Without UI, OpenGL (only VTK) and python interface.
Just to make it work.

Anton

2011/2/2 Václav Šmilauer <eu@xxxxxxxx>:
>
>> I am really interested in implementation of this feature as I was
>> surprised of ESyS productivity, which uses the same technology.
>> If I have enough time/resources I would try to do it.
>
> IMHO Esys-particles is so efficient because they are much less flexible.
>
> The downside of MPI is that it would make everything (and I mean
> *everything*) upside down, and besides the obvious (passing particles
> around, synchronization etc), there is bunch of other issues to consider,
> such as how to evaluate python expressions which need access to remote nodes
> and similar; the scripting interface would have to change substantially, the
> UI as well (OpenGL would have to get data over the wire) etc etc.
>
> The more I think about it, the more I am sure it is not feasible in 3
> months, that's barely enough to get to understand all the internals...
> perhaps this topic can be left out, being too ambitious.
>
> v.
>
>
>
>
>
>
>
> _______________________________________________
> Mailing list: https://launchpad.net/~yade-dev
> Post to     : yade-dev@xxxxxxxxxxxxxxxxxxx
> Unsubscribe : https://launchpad.net/~yade-dev
> More help   : https://help.launchpad.net/ListHelp
>