NHSE Software Catalog

Parallel Matrix Transpose Routines

url

http://sdcd.gsfc.nasa.gov/ESS/exchange/contrib/rood/parallel_transpose.html

abstract

Code performs a global transpose on a matrix P which is domain decomposed
along rows. Each node packs (subroutine 'pack') its columns of P into
N_nodes 2-D buffers and performs a matrix transpose on each one
(subroutine 'internal_transpose'). Each transposed buffer is then sent to its
corresponding node (subroutine 'global_transpose'). Finally, each node
unpacks (subroutine 'unpack') the N_nodes buffers it receives, one from
each node, and loads them into a 2D array. This array contains each node's
columns of the transposed P matrix.

keywords

parallel processing; matrix transpose (GAMS D1b3); numerical routine