Enabling hardware transactional memory to work more efficiently with readers that can tolerate stale data

• US 9,244,844 B2
• Filed: 03/14/2013
• Issued: 01/26/2016
• Est. Priority Date: 03/14/2013
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: McKenney, Paul E.
• Primary Examiner(s): Rutz, Jared
• Assistant Examiner(s): Baughman, William E
• Application Number:
  US13/827,970
• Publication Number:
  US 20140281267A1
• Time in Patent Office: 1,048 Days
• Field of Search: 711/141
• US Class Current: 1/1
• CPC Class Codes: G06F 12/0815 Cache consistency protocols
• CPC Class Codes: G06F 12/0828 with concurrent directory accessing, i.e. handling multiple concurrent coherency transactions
• CPC Class Codes: G06F 9/467 Transactional memory G06F9/528 takes precedence

Abstract

A technique for enabling hardware transactional memory (HTM) to work more efficiently with readers that can tolerate stale data. In an embodiment, a pre-transaction load request is received from one of the readers, the pre-transaction load request signifying that the reader can tolerate pre-transaction data. A determination is made whether the pre-transaction load request comprises data that has been designated for update by a concurrent HTM transaction. If so, a cache line containing the data is marked as pre-transaction data. The concurrent HTM transaction proceeds without aborting notwithstanding the pre-transaction load request.

Claims

1. A system, comprising:

one or more processors;

a memory coupled to said one or more processors;

control logic implementing a hardware transactional memory (HTM) subsystem using said memory, said control logic being operable to perform operations that prevent a concurrent HTM transaction from aborting due to a concurrent reader, said operations comprising;

receiving a load request from said concurrent reader;

determining whether said load request signifies that said concurrent reader can tolerate pre-transaction data;

determining whether said load request comprises data that has been designated for update by said concurrent HTM transaction;

marking a cache line containing said data as pre-transaction data and returning said data to said concurrent reader if said load request signifies that said concurrent reader can tolerate pre-transaction data and comprises data that has been designated for update by said concurrent HTM transaction; and

if said load request comprises data that has been designated for update by said concurrent HTM transaction, aborting said concurrent HTM transaction in response to said load request if said cache line is not marked as containing pre-transaction data, and proceeding with said concurrent HTM transaction notwithstanding said load request if said cache line is marked as containing pre-transaction data.

2. A system in accordance with claim 1, wherein said operations further include returning said data without marking said cache line if said data has not been designated for update by said concurrent HTM transaction.

3. A system in accordance with claim 1, wherein said data is copied from a transferror cache associated with said concurrent HTM transaction to a transferee cache associated with said concurrent reader, and wherein said marking a cache line comprises marking a cache line containing said data in both said transferror cache and said transferee cache.

4. A system in accordance with claim 3, wherein said transferee cache is on said computer system, representing a local node, and said transferror cache is on a remote computer system, representing a remote node, said local and remote nodes being in communication with each other via a node interconnect.

5. A system in accordance with claim 4, wherein said method further includes exchanging request and response messages between said local and remote nodes as part of copying said data from said transferror cache to said transferee cache, said request and response messages signifying that said concurrent reader can tolerate pre-transaction data.

6. A system in accordance with claim 1, wherein said load request signifying that said concurrent reader can tolerate pre-transaction data comprises a pre-transaction load instruction that is either explicitly specified by said concurrent reader or emitted by a compiler during source code compilation of a non-transaction-specific load instruction.

7. A system in accordance with claim 1, wherein said marking a cache line is performed following a wait period in which said concurrent HTM transaction fails to complete.

8. A computer program product, comprising:

one or more machine-useable storage media;

control logic provided by said one or more media to implement a hardware transactional memory (HTM) subsystem, said control logic being operable to perform operations that prevent a concurrent HTM transaction from aborting due to a concurrent reader, said operations comprising;

receiving a load request from said concurrent reader;

determining whether said load request signifies that said concurrent reader can tolerate pre-transaction data;

determining whether said pre-transaction load request comprises data that has been designated for update by said concurrent HTM transaction;

marking a cache line containing said data as pre-transaction data and returning said data to said concurrent reader if said load request signifies that said concurrent reader can tolerate pre-transaction data and comprises data that has been designated for update by said concurrent HTM transaction; and

if said load request comprises data that has been designated for update by said concurrent HTM transaction, aborting said concurrent HTM transaction in response to said load request if said cache line is not marked as containing pre-transaction data, and proceeding with said concurrent HTM transaction notwithstanding said load request if said cache line is marked as containing pre-transaction data.

9. A computer program product in accordance with claim 8, wherein said operations further include returning said data without marking said cache line if said data has not been designated for update by said concurrent HTM transaction.

10. A computer program product in accordance with claim 8, wherein said data is copied from a transferror cache associated with said concurrent HTM transaction to a transferee cache associated with said concurrent reader, and wherein said marking a cache line comprises marking a cache line containing said data in both said transferror cache and said transferee cache.

11. A computer program product in accordance with claim 10, wherein said transferee cache is on a first computer system, representing a local node, and said transferror cache is on a second computer system, representing a remote node, said local and remote nodes being in communication with each other via a node interconnect.

12. A computer program product in accordance with claim 11, wherein said method further includes exchanging request and response messages between said local and remote nodes as part of copying said data from said transferror cache to said transferee cache, said request and response messages signifying that said concurrent reader can tolerate pre-transaction data.

13. A computer program product in accordance with claim 8, wherein said load request signifying that said concurrent reader can tolerate pre-transaction data comprises a pre-transaction load instruction that is either explicitly specified by said concurrent reader or emitted by a compiler during source code compilation of a non-transaction-specific load instruction.

14. A computer program product in accordance with claim 8, wherein said marking a cache line is performed following a wait period in which said concurrent HTM transaction fails to complete.