Method and apparatus for application-specific dynamic cache placement

• US 20080010413A1
• Filed: 07/07/2006
• Published: 01/10/2008
• Est. Priority Date: 07/07/2006
• Status: Abandoned Application

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Sura, Zehra, Kailas, Krishnan Kunjunny, Ravindran, Rajiv Alazhath
• Application Number:
  US11/482,923
• Publication Number:
  US 20080010413A1
• Time in Patent Office: Days
• US Class Current: 711/129
• CPC Class Codes: G06F 12/0842 for multiprocessing or multitasking
• CPC Class Codes: G06F 12/121 using replacement algorithms
• CPC Class Codes: G06F 9/30043 LOAD or STORE instructions; Clear instruction
• CPC Class Codes: G06F 9/3824 Operand accessing

Abstract

One embodiment of the present method and apparatus for application-specific dynamic cache placement includes grouping sets of data in a cache memory system into two or more virtual partitions and processing a load/store instruction in accordance with the virtual partitions, where the load/store instruction specifies at least one of the virtual partitions to which the load/store instruction is assigned.

Claims

1. A method for accessing data in a cache memory system, the method comprising:

grouping sets of data in the cache memory system into two or more virtual partitions; and

processing a load/store instruction in accordance with the two or more virtual partitions, where the load/store instruction specifies at least one of the two or more virtual partitions to which the load/store instruction is assigned.

2. The method of claim 1, wherein the processing comprises:

checking at least a subset of the two or more virtual partitions for a data item referenced by the load/store instruction;

fetching the data item referenced by the load/store instruction from one of the two or more virtual partitions, if the data item is located therein and the load/store instruction specifies a load operation;

storing the data item referenced by the load/store instruction in one of the two or more virtual partitions, if the load/store instruction specifies a store operation; and

loading the data item to the at least one of the two or more virtual partitions to which the load/store instruction is assigned, if the data item is not located in the subset of the two or more virtual partitions.

3. The method of claim 2, wherein the checking comprises:

checking each of the two or more virtual partitions for the data item.

4. The method of claim 2, wherein the checking comprises:

identifying one or more sets in the cache memory system that comprise the at least one of the two or more virtual partitions to which the load/store instruction is assigned; and

checking the one or more sets for the data item.

5. The method of claim 2, wherein the loading and storing comprises:

selecting a virtual partition from among two or more virtual partitions to which the load/store instruction is assigned, in accordance with a cache replacement policy.

6. The method of claim 1, wherein the at least one of the two or more virtual partitions to which the load/store instruction is assigned is specified by an identifier in the load/store instruction.

7. The method of claim 6, wherein the load/store instruction comprises a number of extra bits for specifying the identifier.

8. The method of claim 6, wherein the identifier is implemented as a prefix instruction to the load/store instruction, the prefix instruction specifying one or the two or more virtual partitions that is associated with a next sequential load/store instruction.

9. The method of claim 6, wherein the identifier is specified in accordance with one or more registers implemented as operands of the load/store instruction.

10. The method of claim 1, wherein the grouping comprises:

identifying one or more load/store instructions that are encountered most frequently from among a plurality of load/store instructions;

identifying one or more data items accessed by the one or more load/store instructions;

modeling the one or more load/store instructions as a graph; and

assigning each of the one or more load/store instructions to at least one virtual partition in accordance with the graph.

11. The method of claim 10, wherein the modeling comprises:

dividing each of the one or more load/store instructions into one or more sub-instructions, each of the one or more sub-instructions comprising a node of the graph;

calculating an affinity metric for each pair of load/store instructions within the one or more load/store instructions;

calculating a conflict metric for each pair of load/store instructions; and

assigning weights to edges of the graph in accordance with the affinity metric and the conflict metric, the weights indicating whether each pair of load/store instructions should be assigned to a common virtual partition or to different virtual partitions.

12. A computer readable medium containing an executable program for accessing data in a cache memory system, where the program performs the steps of:

grouping sets of data in the cache memory system into two or more virtual partitions; and

processing a load/store instruction in accordance with the two or more virtual partitions, where the load/store instruction specifies at least one of the two or more virtual partitions to which the load/store instruction is assigned.

13. The computer readable medium of claim 12, wherein the processing comprises:

checking at least a subset of the two or more virtual partitions for a data item referenced by the load/store instruction;

fetching the data item referenced by the load/store instruction from one of the two or more virtual partitions, if the data item is located therein and the load/store instruction specifies a load operation;

storing the data item referenced by the load/store instruction in one of the two or more virtual partitions, if the load/store instruction specifies a store operation; and

loading the data item to the at least one of the two or more virtual partitions to which the load/store instruction is assigned, if the data item is not located in the subset of the two or more virtual partitions.

14. The computer readable medium of claim 13, wherein the checking comprises:

checking each of the two or more virtual partitions for the data item.

15. The computer readable medium of claim 13, wherein the checking comprises:

identifying one or more sets in the cache memory system that comprise the at least one of the two or more virtual partitions to which the load/store instruction is assigned; and

checking the one or more sets for the data item.

16. The computer readable medium of claim 12, wherein the at least one of the two or more virtual partitions to which the load/store instruction is assigned is specified by an identifier in the load/store instruction.

17. The computer readable medium of claim 16, wherein the load/store instruction comprises a number of extra bits for specifying the identifier.

18. The computer readable medium of claim 12, wherein the grouping comprises:

identifying one or more load/store instructions that are encountered most frequently from among a plurality of load/store instructions;

identifying one or more data items accessed by the one or more load/store instructions;

modeling the one or more load/store instructions as a graph; and

assigning each of the one or more load/store instructions to at least one virtual partition in accordance with the graph.

19. The computer readable medium of claim 18, wherein the modeling comprises:

dividing each of the one or more load/store instructions into one or more sub-instructions, each of the one or more sub-instructions comprising a node of the graph;

calculating an affinity metric for each pair of load/store instructions within the one or more load/store instructions;

calculating a conflict metric for each pair of load/store instructions; and

assigning weights to edges of the graph in accordance with the affinity metric and the conflict metric, the weights indicating whether each pair of load/store instructions should be assigned to a common virtual partition or to different virtual partitions.

20. A cache controller for controlling access to a cache memory, the cache controller comprising:

means for grouping sets of data in the cache memory system into two or more virtual partitions; and

means for processing a load/store instruction in accordance with the two or more virtual partitions, where the load/store instruction specifies at least one of the two or more virtual partitions to which the load/store instruction is assigned.