Arrangements facilitating ordered transactions

• US 20040008677A1
• Filed: 07/12/2002
• Published: 01/15/2004
• Est. Priority Date: 07/12/2002
• Status: Active Grant

• Current Assignee: Intel Corporation
• Original Assignee: Intel Corporation
• Inventors: Cen, Ling
• Granted Patent: US 7,471,675 B2
• Time in Patent Office: Days
• US Class Current: 370/389
• CPC Class Codes: G06F 12/0813 with a network or matrix configuration
• CPC Class Codes: G06F 12/0828 with concurrent directory accessing, i.e. handling multiple concurrent coherency transactions

Abstract

Arrangements facilitating ordered transactions, e.g., ordered writes, in a packet switch system having multiple switch elements.

Claims

1. A method of effecting ordered transactions in a local machine having a plurality of switching elements to route transactions, the method comprising:

for each ordered transaction stream, assigning, out of the plurality of switching elements, a source switching element with the responsibility of tracking progression and ordered completion of transactions within an ordered transaction stream;

the source switching element including operations of;

receiving ordered transactions of the ordered transaction stream from a source node, and enabling tracking of progression of the ordered transactions;

for each ordered transaction of the ordered transaction stream, returning, upon a ready to retire state of the ordered transaction and also completion of all older transactions, a ready for data acknowledgement back to the source node which originated the ordered transaction; and

responsive to the ready for data acknowledgement, receiving source node data for the ordered transaction and forwarding the same to a target node, and verifying completion of the ordered transaction.

2. A method as claimed in claim 1, further comprising:

allowing each ordered transaction of the ordered transaction stream to achieve global visibility within the local machine; and

upon occurrence of a conflicting transaction which conflicts with any respective ordered transaction which has not yet achieved the ready to retire state, invalidating the global visibility of the respective ordered transaction, and permitting processing of the conflicting transaction.

3. A method as claimed in claim 2, further comprising:

for the permitting processing of the conflicting transaction, returning at least one of a snooped status and snooped data to a conflicting node which had requested the conflicting transaction, the snooped data being modified data which was in existence in cache at a time the respective ordered transaction had achieved global visibility.

4. A method as claimed in claim 2, further comprising:

assigning, out of the plurality of switching elements, at least one home switching element which is a target switch for the target node of at least one transaction of the ordered transaction stream; and

,

assigning the responsibility of the invalidating and the permitting processing operations to at least one of the source switching element and the at least one home switching element.

5. A method as claimed in claim 1, where the method is implemented in one of a packet switching system, a fabric switching system and crossbar bus system.

6. A method as claimed in claim 1, where the source switching element autonomously has the responsibility of tracking progression and ordered completion of the transactions within the ordered transaction stream.

7. A method as claimed in claim 1, where the source switching element is a source central tracking buffer element (SCTB) having the responsibility of handling transactions from the source node.

8. A method as claimed in claim 1, where the ordered transaction stream includes snooped writes that are strongly ordered.

9. A machine-readable medium having stored thereon at least one sequence of instructions that, when executed, causes a machine to effect:

a method of ordering transactions in a local machine having a plurality of switching elements operative to route transactions within the local machine, the method comprising;

for each ordered transaction stream, assigning, out of the plurality of switching elements, a source switching element with the responsibility of tracking progression and ordered completion of transactions within an ordered transaction stream;

the source switching element effecting operations including;

receiving ordered transactions of the ordered transaction stream from a source node, and enabling tracking of progression of the ordered transactions;

for each ordered transaction of the ordered transaction stream, returning, upon a ready to retire state of the ordered transaction and also completion of all older transactions, a ready for data acknowledgement back to the source node which originated the ordered transaction; and

responsive to the ready for data acknowledgement, receiving source node data for the ordered transaction and forwarding the same to a target node, and verifying completion of the ordered transaction.

10. A medium as claimed in claim 9, the method further comprising:

allowing each ordered transaction of the ordered transaction stream to achieve global visibility within the local machine; and

upon occurrence of a conflicting transaction which conflicts with any respective ordered transaction which has not yet achieved the ready to retire state, invalidating the global visibility of the respective ordered transaction, and permitting processing of the conflicting transaction.

11. A medium as claimed in claim 10, the method further comprising:

for the permitting processing of the conflicting transaction, returning at least one of a snooped status and snooped data to a conflicting node which had requested the conflicting transaction, the snooped data being modified data which was in existence in cache at a time the respective ordered transaction had achieved global visibility.

12. A medium as claimed in claim 10, the method further comprising:

assigning, out of the plurality of switching elements, at least one home switching element which is a target switch for the target node of at least one transaction of the ordered transaction stream; and

,

assigning the responsibility of the invalidating and the permitting processing operations to at least one of the source switching element and the at least one home switching element.

13. A medium as claimed in claim 9, where the method is implementable in one of a packet switching system, a fabric switching system and crossbar bus system.

14. A medium as claimed in claim 9, where the source switching element autonomously has the responsibility of tracking progression and ordered completion of the transactions within the ordered transaction stream.

15. A medium as claimed in claim 9, where the source switching element is a source central tracking buffer element (SCTB) having the responsibility of handling transactions from the source node.

16. A medium as claimed in claim 9, where the ordered transaction stream includes snooped writes that are strongly ordered.

17. A switching system to effect ordered transactions in a local machine having a plurality of switching elements operative to route transactions within the local machine, the switching system comprising:

a source switching element having, for an ordered transaction stream delivered to the source switching element, a responsibility of tracking progression and ordered completion of transactions within the ordered transaction stream;

the source switching element operable to;

receive ordered transactions of the ordered transaction stream from a source node, and track progression of the ordered transactions;

return, for each ordered transaction of the ordered transaction stream, upon a ready to retire state of the ordered transaction and also completion of all older transactions, a ready for data acknowledgement back to the source node which originated the ordered transaction; and

receive, responsive to the ready for data acknowledgement, source node data for the ordered transaction and forward the same to a target node, and verify completion of the ordered transaction.

18. A switching system as claimed in claim 17, further comprising the source switching element operable to:

allow each ordered transaction of the ordered transaction stream to achieve global visibility within the local machine; and

upon occurrence of a conflicting transaction which conflicts with any respective ordered transaction which has not yet achieved the ready to retire state, invalidate the global visibility of the respective ordered transaction, and permit processing of the conflicting transaction.

19. A switching system as claimed in claim 18, further comprising the source switching element operable to:

return at least one of a snooped status and snooped data to a conflicting node which had requested the conflicting transaction, the snooped data being modified data in existence in cache at a time the respective ordered transaction had achieved global visibility.

20. A switching system as claimed in claim 18, further comprising:

at least one home switching element which is a target switch for the target node of at least one transaction of the ordered transaction stream;

the at least one home switching element having an ability to invalidate the global visibility of the respective ordered transaction, and permit processing of the conflicting transaction.

21. A switching system as claimed in claim 17, where the switching system is one of a packet switching system, a fabric switching system and crossbar bus system.

22. A switching system as claimed in claim 17, where the source switching element autonomously has the responsibility of tracking progression and ordered completion of the transactions within the ordered transaction stream.

23. A switching system as claimed in claim 17, where the source switching element is a source central tracking buffer element (SCTB) having the responsibility of handling transactions from the source node.

24. A switching system as claimed in claim 17, where the ordered transaction stream includes snooped writes that are strongly ordered.

25. A local machine comprising:

a source node;

at least one target node;

a switching system having a plurality of switching elements operative to route transactions within the local machine, and a source switching element having, for an ordered transaction stream delivered to the source switching element, a responsibility of tracking progression and ordered completion of transactions within the ordered transaction stream;

the source switching element operable to;

receive ordered transactions of the ordered transaction stream from the source node, and track progression of the ordered transactions;

return, for each ordered transaction of the ordered transaction stream, upon a ready to retire state of the ordered transaction and also completion of all older transactions, a ready for data acknowledgement back to the source node which originated the ordered transaction; and

receive, responsive to the ready for data acknowledgement, source node data for the ordered transaction and forward the same to a target node of the at least one target node, and verify completion of the ordered transaction.

26. A local machine as claimed in claim 25, further comprising the source switching element operable to:

allow each ordered transaction of the ordered transaction stream to achieve global visibility within the local machine; and

upon occurrence of a conflicting transaction which conflicts with any respective ordered transaction which has not yet achieved the ready to retire state, invalidate the global visibility of the respective ordered transaction, and permit processing of the conflicting transaction.

27. A local machine as claimed in claim 26, further comprising the source switching element operable to:

return at least one of a snooped status and snooped data to a conflicting node which had requested the conflicting transaction, the snooped data being modified data in existence in cache at a time the respective ordered transaction had achieved global visibility.

28. A local machine as claimed in claim 26, further comprising:

at least one home switching element which is a target switch for the target node of at least one transaction of the ordered transaction stream;

the at least one home switching element having an ability to invalidate the global visibility of the respective ordered transaction, and permit processing of the conflicting transaction.

29. A local machine as claimed in claim 25, where the switching system is one of a packet switching system, a fabric switching system and crossbar bus system.

30. A local machine as claimed in claim 25, where the source switching element autonomously has the responsibility of tracking progression and ordered completion of the transactions within the ordered transaction stream.

31. A local machine as claimed in claim 25, where the source switching element is a source central tracking buffer element (SCTB) having the responsibility of handling transactions from the source node.

32. A local machine as claimed in claim 25, where the ordered transaction stream includes snooped writes that are strongly ordered.

33. A local machine as claimed in claim 25, where the local machine is one of a personal computer, a server, a mainframe computer and an electronic apparatus.