The above algorithm generalizes in a straight-forward manner
to 
,
where x and y can have any valid vector distribution,
including projected and/or duplicated.
Some care must be taken in creating xdup and ydup.
Notice that xdup must be aligned with the columns of
a (distributed like a row of A , but duplicated),
while ydup must be aligned with the rows
of a (distributed like a column of A , but duplicated).
Since a, x, and y were all
nicely aligned before, this was not an issue.
Creating xdup and ydup is now accomplished through the calls
After this, all required communication and alignment is hidden in the PLA_Copy and PLA_Reduce routines. A code that generalizes even further, implementing the full functionality of the sequentialPLA_Pvector_create_conf_to( a, PLA_PROJ_ONTO_ROW, PLA_ALL_ROWS, &xdup ); PLA_Pvector_create_conf_to( a, PLA_PROJ_ONTO_COL, PLA_ALL_COLS, &ydup );
gemv operation
in given in Figure 
.
PLACE BEGIN HR HERE
PLACE END HR HERE