send_offset = (rank * block_size) + (i * block_size * num_procs);
recv_offset = (my_z_base * block_size) + (i * block_size);
MPI_Sendrecv(tmp_buff1 + send_offset, send_count, send_type, rank, tag,
send_offset = (rank * block_size) + (i * block_size * num_procs);
recv_offset = (my_z_base * block_size) + (i * block_size);
MPI_Sendrecv(tmp_buff1 + send_offset, send_count, send_type, rank, tag,