|
|
Cart3D's Flow
Solvers
Tell
me a little about flowCart?
What is the difference between "flowCart" and "mpi_flowCart"
Can I still get a copy of Tiger?
Tell me a little about FlowCart?
flowCart
is the current solver being released with Cart3D. It is a scalable,
multilevel, linearly-exact upwind solver and uses
domain-decomposition to achieve very
good scalability. It is among the most scalable, accurate and
robust codes in the industry (see some news items
from 2005 & 2006). On most modern desktop machines it can
converge well over 1 million cells-per-hour-per processor, and it does
very well on multi-core CPUs. flowCart is
very tightly integrated into Cart3D and all of
our automation tools are built around it. Since it is a
multilevel
code, it converges very quickly and
includes
our latest work on low-dissipation approaches, solid wall boundaries,
mesh interfaces and limiters. Both the parallelization and multigrid
are completely transparent to
the
user and are turned on by simple command line arguments to encourage
their
use.
What is the difference
between "flowCart" and "mpi_flowCart"
Its basically a choice between shared-memory or distributed
memory. Cart3D's default solver module is
called "flowCart" which is a shared-memory build and
can be run in parallel within a single comput node. "mpi_flowCart"
is a distributed memory build. Both builds use the same
domain-decomposition strategy and have very similar scalability on most
systems. So...which build should you use? We recommend all users start
with flowCart, and then, after warming up, start experimenting with
mpi_flowCart if they happen to be in a distributed computing
environement. With multi-core CPUs, shared memory is more relevant than
ever, but if you're in a cluster-computing environment, distributed
memory is still of interest for large problems. So, if you have a
cluster of 8 boxes, each of which has dual quad-core cpu's (8
processing threads per cluster node), you can run flowCart on up to 8 threads in each
box. If you need to run more threads, or if you need more memory than
you have in a single box you will need to run mpi_flowCart to distribute your job
over multiple cluster nodes. flowCart has
been run in shared memory on over 500 processors, and mpi_flowCart has
been run on well over 4000 processors.
(top)
Can I still get a copy of Tiger?
Tiger14
was the last distributed version of Tiger. It was included with Cart3D v1.2, but has not been
included in the distribution since. If you want a copy
contact me.
Tiger is a
uni-processor code that runs well on scalar CPUs. It
uses central-difference,
finite-volume
solver with blended second and fourth order dissipation. Click here
to see the v1.2 release
docs.
(top)
last update June 2004 M. Aftosmis
|
|