Parallel algorithm for tridiagonal matrix
- From: arpiruk@xxxxxxxxx
- Date: Wed, 22 Aug 2007 11:59:24 -0700
Hello everyone,
I have been searching for parallel algorithm for compact scheme, i.e.
Lele1993 (J.Comp.Phy).
This means I have to solve Ax=Bf. The matrix A I intended to use is a
tridiagonal matrix.
I believe it will always be positive definite on ordinary smooth grid.
(still don't know how to prove it)
My data will be distributed in domain decomposition fashion. Topology
is 2D grid.
Due to the coupling nature of compact scheme, each grid cannot solve
the solution on its own and need some kind of communication with other
grid.
I did some research and believe that fine-grained parallelism such as
Recursive doubling (Stone1975 TOMS) and Cyclic reduction(Hockney1965
J.ACM ) are not suited for current supercomputers which are now very
coarse grained ( A cluster with multi-core node ).
Family of pipelining algorithm also requires too much communications.
It seems that the reduced parallel diagonal dominant of SUN (http://
mack.ittc.ku.edu/sun95application.html) are the most efficient for
this problem.
The complexity is 5n/p+4J (p=#of proc, J = truncated bandwidth) which
almost matches the best serial code ( which is 5n-3) .
Another approach is transposed algorithm
(as I was suggested from http://www.cfd-online.com/Forum/main.cgi?read=54361)
where we collect the strip of data on each grid line
and send to certain processor in that line. Then the data in each
strip line can be solve on one processor. After solving it, the data
can be distributed back to original processor.
Are you aware of any algorithm that is more efficient than this two
algorithms ?
Are there any particular problem that I should be aware of concerning
these two method?
My target problem will be a direct numerical simulation. The problem
may take a size of NxNyNz=1000^3. Time step can take up to 10E6. This
problem should be well scalable up to 512 processors.
Any suggestions are appreciated.
Regards,
Arpiruk
(also posted in math group)
.
- Follow-Ups:
- Re: Parallel algorithm for tridiagonal matrix
- From: William R. Frensley
- Re: Parallel algorithm for tridiagonal matrix
- Prev by Date: Re: conjugate gradient stuck at inflection point
- Next by Date: Re: conjugate gradient stuck at inflection point
- Previous by thread: Asymptotic of Bessel function of complex argument
- Next by thread: Re: Parallel algorithm for tridiagonal matrix
- Index(es):
Relevant Pages
|