Data decomposition and implementation

[vaira_st]

Hi i am new to forum as well as to LAPACK.
I have a small doubt. if i have an array of A and B (square or rectangular) and i use xGEMM to multiply them on shared memory multicore machine, how will the data be decomposed and what type of algorithm is used and does this differ in ScaLAPACK? I was searching in LAPACK working notes but couldn't find a proper explanations.

If its very obvious please ignore my innocence.
Cheers!