I am running pcheev and pcheevx scalapack subroutines on a cluster of computers. i distribute my global matrix to the 8 nodes and call the subroutines to perform the eigne decomposition. this works great but when i used pcelget to collect the local matrices back to the master node. this process slows down the program.
i use pcelset to initialize the local matrices and it works fine
i have narrowed down the problem to being with pcelget-- which tremendously slows down the program time by around ten times that of the lapack subroutine.
does anyone know of any other subroutine that i can use to collect back the local matrices from the different nodes to the master node without using pcelget.
Thanks for the help