Method for simultaneous overlapped communications between neighboring processors in a multiple

Claim 1

In a method of operating a computing system having a MIMD architecture where a program is concurrently run on a plurality of nodes for reducing total processing time for said program; each node having a processor, a memory, and a communication channel connected to said each node and independently connected to a predetermined number of other nodes; said method including steps of: assigning, via said computing system, groups of adjacent nodes, each said group having communication channels from a central node to four adjacent nodes in up, down, left, and right directions, respectively, each of said nodes being capable of being a central node in a different group; exchanging, via said computing system, messages for each of said nodes; and simultaneously communicating, via said computing system, said messages between adjacent nodes by first sending write messages from each central node sequentially to nodes in the up, down, left and right directions in which subsequent messages are communicated before the transfer of at least one previous message is completed and, then by sequentially receiving read messages from nodes in the down, up, right and left directions in which subsequent messages are received before the transfer of at least one previous message is completed.

Claim 5

A method of operating a digital parallel computing system with a MIMD architecture having input and output devices, where a program is concurrently run on a plurality of nodes, each node having a processor, a memory, and a communication channel connected to said each node and independently connected to a predetermined number of other nodes; said method including the steps of: receiving information from said input device defining a number of nodes to run said program; opening groups of adjacent nodes, each said group having communications channels from a central node to four adjacent nodes in up, down, left, and right directions, respectively, each of said nodes being capable of being a central node in a different group; loading each of said nodes with said program; and executing said program including exchanging messages for each of said nodes by simultaneously communicating said messages between adjacent nodes by first sending write messages from each central node sequentially to nodes in the up, down, left and right directions such than the startup, copy, and transfer time for a subsequent message overlaps the transfer time for a previous message, and that by sequentially receiving read messages from nodes in the down, up, right and left directions such that the startup and transfer time for a subsequent message overlaps the transfer and copy time for a previous message.

Claim 6

A method of operating a digital parallel computing system with a MIMD architecture having input and output devices, where a program is concurrently run on a plurality of nodes, each node having a processor, a memory, and a communication channel connected to said each node and independently connected to a predetermined number of other nodes; said method including the steps of: receiving information from said input device defining a number of nodes to run said program; opening groups of adjacent nodes, each said group having communications channels from a central node to four adjacent nodes in up, down, left, and right directions, respectively, each of said nodes being capable of being a central node in a different group; loading each of said nodes with said program; and executing said program including exchanging messages for each of said nodes by simultaneously communicating said messages between adjacent nodes by first sending write messages from each central node sequentially to nodes in the up, down, left and right directions such than the startup, copy, and transfer time for a subsequent message overlaps the transfer time for a previous message, and that by sequentially receiving read messages from nodes in the down, up, right and left directions such that the startup and transfer time for a subsequent message overlaps the transfer and copy time for a previous message.

Claim 9

A method for interprocessor communication in a N dimension parallel computer system having n dimensional spatial grids with 2n boundaries, where n is less than or equal to N, comprising the steps of: (a) writing messages to each of the selected neighboring processors in paired directions in a specified order of writing first to the neighboring processor in a fore orientation and then writing to the neighboring processor in an aft orientation such that the startup, copy, and transfer time for a subsequent message overlaps the transfer time for a previous message; (b) repeated step (a) until all selected neighboring processors of the n-dimensional subspace have been written to; (c) and, when at a boundary, and in the direction of a boundary where there is only one orientation, writing to only one processor associated with that orientation; (d) subsequently reading messages from each of the selected neighboring processors in paired directions in the same order of directions, but reading first from the neighboring processor in the aft orientation and then reading from the neighboring processor in the fore orientation such that the startup and transfer time for a subsequent message overlaps the transfer and copy time for a previous message; (e) repeating step (d) until all selected neighboring processors of the n-dimensional subspace have been read; (f) and, when at a boundary, and in the direction of a boundary where there is only one orientation, reading form only processor associated with that orientation.

Claim 10

A method for interprocessor communication in a N dimension parallel computer system having n dimensional spatial grids with 2n boundaries, where n is less than or equal to N, comprising the steps of: (a) writing messages to each of the selected neighboring processors in paired directions in a specified order of writing first to the neighboring processor in a fore orientation and then writing to the neighboring processor in an aft orientation such that the startup, copy, and transfer time for a subsequent message overlaps the transfer time for a previous message; (b) repeated step (a) until all selected neighboring processors of the n-dimensional subspace have been written to; (c) and, when at a boundary, and in the direction of a boundary where there is only one orientation, writing to only one processor associated with that orientation; (d) subsequently reading messages from each of the selected neighboring processors in paired directions in the same order of directions, but reading first from the neighboring processor in the aft orientation and then reading from the neighboring processor in the fore orientation such that the startup and transfer time for a subsequent message overlaps the transfer and copy time for a previous message; (e) repeating step (d) until all selected neighboring processors of the n-dimensional subspace have been read; (f) and, when at a boundary, and in the direction of a boundary where there is only one orientation, reading form only processor associated with that orientation.