librelist archives

« back to archive

Question about p4est for SPH

Question about p4est for SPH

From:
Joseph Smidt
Date:
2015-03-03 @ 02:35
Hey everyone,

     I am writing a new n-body + SPH code and have been investigating
the best way to handle algorithms that use octrees. I came across
p4est today [1] and the fact that it claims to scale to "hundreds of
thousands of processor cores" has me very interested. However a
question:

   The website keeps referring to AMR which often is associated with
grid codes, not Lagrangian codes like SPH. Is there any reason to
think p4est would not suit an nobdy+SPH code that wants to use octree
methods?

   Second, if p4est is for whatever reason not suited for this
problem, are you aware of a better choice?

[1] http://www.p4est.org/



-- 
------------------------------------------------------------------------
Joseph Smidt <josephsmidt@gmail.com>

Theoretical Division
P.O. Box 1663, Mail Stop B283
Los Alamos, NM 87545
Office: 505-665-9752
Fax:    505-667-1931

Re: [p4est] Question about p4est for SPH

From:
Carsten Burstedde
Date:
2015-03-05 @ 03:45
Joseph,

>      I am writing a new n-body + SPH code and have been investigating
> the best way to handle algorithms that use octrees. I came across
> p4est today [1] and the fact that it claims to scale to "hundreds of
> thousands of processor cores" has me very interested. However a
> question:

check out our latest arxiv preprint for updated scaling results.

>    The website keeps referring to AMR which often is associated with
> grid codes, not Lagrangian codes like SPH. Is there any reason to
> think p4est would not suit an nobdy+SPH code that wants to use octree
> methods?

Not that I know of.  p4est manages the parallel octree very efficiently,
and applications are free to interpret the cells any way they like.  It
is perfectly fine to use p4est for a single octree, not using the
forest-of-octrees feature, so no restriction there.

>    Second, if p4est is for whatever reason not suited for this
> problem, are you aware of a better choice?
> 
> [1] http://www.p4est.org/

I'll be at the siam cse conference, feel free to go find me.

Best,

Carsten