The ScaLAPACK Users' Guide
http://www.netlib.org/scalapack/slug/node77.html
requires that the LLD_A array descriptor entry satisfies
LLD_A >= max(1, LOC_r(M_A)) .
However, LOC_r(M_A) = 0 is possible: suppose
M_A <= (P_r - 1) * MB_A .
Then, some process rows have no entries of A.
Why must I set LLD_A = 1 on these processes? And, do I have to allocate an empty row?
Thanks,
Nick Knight

