Each process (including the master process) now reads data. Any partial z-planes read are sent to process along with a copy of the lowest z-plane which is necessary for the solver. The load balancing algorithm which is present in the serial case is not used as this would require complex re-distribution of the data once read.