Interconnect fabric for a data processing system

• US 7,944,932 B2
• Filed: 04/01/2008
• Issued: 05/17/2011
• Est. Priority Date: 02/10/2005
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Clark, Leo J., Starke, William J., Fields, James S. Jr., Guthrie, Guy L.
• Primary Examiner(s): Kumar; Pankaj
• Assistant Examiner(s): Foud; Hicham B
• Application Number:
  US12/060,751
• Publication Number:
  US 20080181244A1
• Time in Patent Office: 1,141 Days
• Field of Search: None
• US Class Current: 370/400
• CPC Class Codes: G06F 15/16 Combinations of two or more digital computers each having at least an arithmetic unit, a program unit and a register, e.g. for a simultaneous processing of several programs coordinating program control therefor G06F9/52; in regulating and control system G05B

Abstract

A data processing system includes a plurality of processing units each having a respective point-to-point communication link with each of multiple others of the plurality of processing units but fewer than all of the plurality of processing units. Each of the plurality of processing units includes interconnect logic, coupled to each point-to-point communication link of that processing unit, that broadcasts operations received from one of the multiple others of the plurality of processing units to one or more of the plurality of processing units.

Claims

1. A data processing system, comprising:

at least three processing nodes each including at least four processing units, wherein;

each of the processing units is connected by a respective point-to-point communication link with each other processing unit in its own processing node ;

at least one processing unit in each of the processing nodes is connected by a point-to-point communication link to a processing unit in a different processing node;

no more than one processing unit in each processing node is connected by a point-to-point communication link to a processing unit in a given other processing node;

wherein each of said processing units includes interconnect logic, coupled to each point-to-point communication link of that processing unit, that broadcasts operations received from others of said processing units to one or more of said processing units.

2. The data processing system of claim 1, wherein:

each point-to-point communication link coupling processing units within said plurality of processing nodes comprises a first tier link;

each point-to-point communication link coupling processing units in different processing nodes comprises a second tier link;

processing units coupled by said first tier links all follow a first uniform connectivity rule; and

processing units coupled by said second tier links all follow a second uniform connectivity rule.

3. The data processing system of claim 1, wherein each point-to-point communication link comprises a uni-directional communication link.

4. The data processing system of claim 3, wherein each said point-to-point communication link comprises one of a pair of point-to-point communication links having opposing directions of information flow.

5. The data processing system of claim 1, wherein each of said processing nodes comprises a multi-chip module (MCM).

6. The data processing system of claim 1, wherein each processing node of the data processing system is directly connected to each other processing node by a respective one of multiple point-to-point communication links between pairs of processing units.

7. A processing unit for a multi-processor data processing system, wherein said multi-processor data processing system includes at least three processing nodes, said processing unit comprising:

at least one processor core that executes instructions;

at least one master that initiates operations communicated to other processing units;

at least one snooper that receives operations from the other processing units; and

interconnect logic having;

connections for a plurality of first tier point-to-point communication links that links the processing unit to each other of at least four local processing units in a processing node to which the processing unit belongs; and

a connection for a second tier point-to-point communication link that links the processing unit to a remote processing unit in a different processing node than the processing unit;

wherein said interconnect logic broadcasts operations received from local processing units via the first tier point-to-point communication links to said remote processing unit via the second tier point-to-point communication link and broadcasts operations received from the remote processing unit via the second tier point-to-point communication link to each other of the local processing units via the first tier point-to-point communication links;

wherein the processing unit is a sole processing unit in its processing node coupled to any processing unit in the different processing node.

8. The processing unit of claim 7, wherein each point-to-point communication link comprises a uni-directional communication link.

9. The processing unit of claim 8, wherein each said point-to-point communication link comprises one of a pair of point-to-point communication links having opposing directions of information flow.

10. A multi-chip module, comprising:

a plurality of processing units including the processing unit of claim 7; and

a package containing said plurality of processing units.

11. A method in a data processing system including at least three processing node each including at least four processing units, said method comprising:

connecting each of the processing units in the data processing system by a respective point-to-point communication link with each other processing unit in its processing node;

connecting at least one processing unit in each of the processing nodes by a point-to-point communication link to a processing unit in a different processing node;

connecting no more than one processing unit in each processing node by a point-to-point communication link to a processing unit in a given other processing node; and

configuring each of said plurality of processing units to broadcast operations received from others of said processing units to one or more of said processing units.

12. The method of claim 11, wherein:

each point-to-point communication link coupling processing units within said plurality of processing nodes comprises a first tier link;

each point-to-point communication link coupling processing units in different processing nodes comprises a second tier link; and

said providing comprises providing a data processing system in which processing units coupled by said first tier links all follow a first uniform connectivity rule and processing units coupled by said second tier links all follow a second uniform connectivity rule.

13. The method of claim 11, and further comprising:

reconfiguring the data processing system from a first configuration to a second configuration during operation of the data processing system without interrupting communication of operations between processing units in the second configuration.

14. The method of claim 13, wherein said reconfiguring comprises coupling one or more additional processing units to the data processing system.

15. The method of claim 11, and further comprising directly connecting each processing node to each other processing node in the data processing system by a respective one of multiple point-to-point communication links between pairs of processing units.