LAPACK/ScaLAPACK Development

Hello

In the current version of my code I build the LHS matrix 'A' on different processors of our cluster
since it's too big to fit on a single machine. (A is a dense matrix)
The way I build the matrix on different processors is compatible with "two-dimensional block cyclic decomposition"
and I have all the information about which blocks of matrix are on which machines.
So I was wondering if I can use ScaLAPACK to solve A*X=B equation and at the end just gather the solution onto master machine.

any input and/or hint is appreciated.

thanks in advance

I am a new user of ScaLapack and have a similar question. I have a matrix that has to be computed over distributed nodes (each entry is obtained by calling some subprogram). But I cannot gather the entries back to form a single matrix ('cuz it's too large). How can I then call, say, PDSYEVR routine to compute its eigenvalues ? In an eigen-solver routine in Scalapck, the matrix variable "A" is the actual whole matrix or what ?

I really appreciate anyone that can give a hint.