>>>>> varadharajan writes:
varadharajan> Hello all.
varadharajan> I have to do some calculations on sparse matrices using
varadharajan> parallel processors. Which is the best place I can read
varadharajan> about this?
Shared or Distributed memory models? What degree of parallelism? Do you
need data structures that allow parallel random access and dynamic generation
of the structure? What kind of memory latency do you need to worry about? Do
you have any restrictions on how dynamic memory allocation?
Also, what algorithms do you need to implement? Sparse matrix-vector
multiplication is very different (and considerably easier) than sparse
matrix-matrix mult. or addition.
For an introduction to serial sparse matrix algorithms, this is a good online
reference.
Sergio Pissanetzky, sparse matrix technology
http://ceaspub.eas.asu.edu/powerzone/SparseVisual/Default.htm
Several of these algorithms can be parallelized with some modification, using
blocking and other schemes.
I'm sure that I and others would welcome a summary of both your problems and
the references you decide upon.
Ted
--
(In a vain attempt to reduce spam, which will probably find the From: header
in the archives anyway, I've altered my email address below.)
stern'at'cray'dot'com Engineering Applications
Cray Inc. http://www.cray.com
411 First Avenue South, Suite 600 Direct 206-701-2182, Main 206-701-2000
Seattle, WA 98104-2860 FAX 206-701-2500
|