System for augmenting two dimensional data sets in a two dimensional parallel computer system

A method of data transfer between processors connected in an array while maintaining the overall organization of the data, each processor storing two dimensional subsets of a data set distributed over the processors, and data subset storage areas being arranged in an array with a center subset storage area and storage areas for side and corner subsets, the method comprising the steps of:

• (a) transferring contents of the center subset storage area to one of the neighboring processors;
• (b) receiving contents of a neighboring center subset storage area and storing the contents of neighboring center subset storage area in the side storage area;
• (c) transferring contents of the side storage area to the one of the neighboring processors;
• and
• (d) receiving contents of a neighboring side subset storage area and storing the contents of the neighboring side subset storage area in the corner subset storage area.

A system, comprising:

• an array of connected processing systems, said processing systems together storing a two dimensional data set and each of said processing systems storing a two-dimensional subset of the data set, each processing system being connected to a plurality of neighboring processing systems each of said processing systems comprising:
• a main processor including a memory storing the two dimensional subset and for performing an array data operation;
• and
• data transfer processors coupled to the memory and each neighboring processing system to which the processing system is connected, and comprising:
• means for transferring a local data subset to neighboring processing systems;
• means for receiving and storing first foreign data subsets from the neighboring processing systems in a position related to the position of the neighboring processor from which the first foreign data subset was received;
• means for transferring the first foreign data subset to the neighboring processing systems;
• and
• means for receiving and storing second foreign data subsets from the neighboring processing systems in a position related to the position of the neighboring processor from which the second foreign data subset was received.