Distributed allocation of system hardware resources for multiprocessor systems

• US 7,124,410 B2
• Filed: 01/09/2002
• Issued: 10/17/2006
• Est. Priority Date: 01/09/2002
• Status: Expired due to Fees

• Current Assignee: Forum Health & Personal Care LLC, International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Gilbert, Bruce M., Lloyd, Stacey G., Berg, Thomas B.
• Primary Examiner(s): BULLOCK JR, LEWIS ALEXANDER
• Application Number:
  US10/045,925
• Publication Number:
  US 20030131043A1
• Time in Patent Office: 1,742 Days
• Field of Search: 718100-104, 710107-113, 710/306, 710309-312, 710/317
• US Class Current: 718/104
• CPC Class Codes: G06F 9/544 Buffers; Shared memory; Pipes

Abstract

A method is provided for allocating system resources across multiple nodes of a system communicating through a hardware device. The method provides for allocation of transaction units or identifiers in an allocating component for use in a multiple target component which may be in a distinct target node within the multiple node system. Based on the operations or requests that a target node receives from multiple external request source nodes, each requiring the use of target transaction unit objects such as transaction identification bits, the method provides inclusion of such information in the initial request to a target node which allows any data transmission between the source node and the target node, or the target node and the source node to be accomplished without any further intervention by the allocating component. Such component may be a local memory control agent or device.

Claims

1. A method for handling resources of a multiprocessor system, comprising the steps of:

a requestor group of one or more processors of the system allocating a resource of a target group of one or more processors of the system, the target being interconnected to the requestor, the resource being associated with a request of the requestor, allocating the resource comprising;

the requestor sending a request to a central crossbar interconnecting the requestor and the target and separate from the requestor and the target;

the central crossbar allocating the resource at the target group;

the requestor sending a result of the request to the target; and

the target directly receiving the result by using the resource without intermediate handling of the resource by the target.

2. The method of claim 1, further comprising the central crossbar assigning an identifier to the request.

3. The method of claim 1, further including the step of the target deallocating the resource.

4. The method of claim 1, wherein the resource controls access to a portion of a shared system memory of the target.

5. The method of claim 1, wherein the request is associated with a transaction between the requestor and the target.

6. The method of claim 5, wherein the resource is a transaction identifier of the transaction.

7. The method of claim 1, wherein the target comprises a first group and a second group of the processors, the first and second groups being interconnected.

8. A computing system, comprising:

a requestor group of one or more processors;

a resource allocator responsive to requests of the requestor; and

a target group of one or more processors, the target being interconnected to the requestor and responsive to a resource provided by the resource allocator and to a result of the request from the requestor;

a central crossbar interconnecting the requestor and the target and separate from the requestor and the target, the central crossbar comprising the resource allocator and requesting the resource at the target group;

wherein the target directly receives the result responsive to the resource without intermediate handling of the resource by the target.

9. The system of claim 8, wherein the resource allocator comprises one or more pipelines for handling requests.

10. The system of claim 8, wherein the central crossbar comprises a first crossbar for data interconnection and a second crossbar for control interconnection.

11. The system of claim 10, wherein the second crossbar is a tag and address crossbar.

12. The system of claim 8, wherein the resource controls allocation of a portion of shared system memory of the target.

13. The system of claim 8, wherein the resource is a transaction identifier.

14. The system of claim 8, wherein the target group of processors comprises interconnected first and second groups of processors.

15. In a multiprocessor computer system comprising a plurality of processing nodes;

a shared, distributed system memory; and

a central hardware device comprising a communications pathway connecting said processing nodes;

wherein each one of said processing nodes includes at least one processor; and

a portion of said shared system memory coupled to said processor and said communication pathway;

said communications pathway comprised of communications ports each dedicated to communicating with one of said processing nodes wherein said communications pathway is separate from and interconnects said processing nodes and is further comprised of a tag and address crossbar to communicate tag and address information, and a data crossbar means for communicating data within the system;

a method for handling tagging and addressing within the system comprising the steps of;

a first node communicating to said tag and address crossbar, a request for a transaction in the system;

allocating a transaction identifier in the tag and address crossbar for use by the identified transaction;

conveying said transaction identifier to an originating node;

attaching said transaction identifier to said transaction; and

communicating said transaction from the first node to the originating node which receives the transaction directly through said data crossbar means.

16. The method of claim 15, wherein the step of conveying said transaction identifier to the originating node is accomplished by conveying said identifier from said tag and address crossbar to said first node.

17. In a multiprocessor computer system capable of being partitioned into one or more independently functioning processing systems comprising:

a plurality of processing nodes, each node capable of operating independently;

a shared, distributed system memory;

and a communications pathway which interconnects said plurality of processing nodes;

wherein each one of said processing nodes includes at least one processor and a portion of said shared system memory coupled to said processor and said communication pathway, and said communication pathway is comprised of a central crossbar including tag and address means for communicating the identification of data transactions being processed through the system connected to said plurality of processing nodes, said tag and address means including means for storing information related to the identification of data including the target location for said data; and

means for defining which memory location in one or more of said nodes, said data should be sent;

the central crossbar interconnecting the nodes and is separate from the nodes;

a method for handling a request from a requestor node of the system to a target node of the system comprising the steps of;

the requestor node sending a request to the central crossbar;

the central crossbar allocating a resource to receive results of the request at the target node;

the central crossbar assigning an identifier to the request;

the requestor sending the results to the target node; and

the target node receiving the results directly into the resource without intermediate buffering of the results at the target node.

18. The method of claim 17, wherein said target node is comprised of more than one additional node.