Victim cache replacement

• US 8,347,037 B2
• Filed: 10/22/2008
• Issued: 01/01/2013
• Est. Priority Date: 10/22/2008
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Guthrie, Guy L., Jeremiah, Thomas L., McNeil, William L., Patel, Piyush C., Starke, William J., Stuecheli, Jeffrey A.
• Primary Examiner(s): PORTKA, GARY J
• Application Number:
  US12/256,002
• Publication Number:
  US 20100100682A1
• Time in Patent Office: 1,532 Days
• Field of Search: None
• US Class Current: 711/122
• CPC Class Codes: G06F 12/0811 with multilevel cache hierarchies
• CPC Class Codes: G06F 12/0831 using a bus scheme, e.g. with bus monitoring or watching means
• CPC Class Codes: G06F 12/0862 with prefetch
• CPC Class Codes: G06F 12/0897 with two or more cache hierarchy levels with multilevel cache hierarchies G06F12/0811
• CPC Class Codes: G06F 12/128 adapted to multidimensional cache systems, e.g. set-associative, multicache, multiset or multilevel
• CPC Class Codes: G06F 2212/1028 Power efficiency
• CPC Class Codes: Y02D 10/00 Energy efficient computing, e.g. low power processors, power management or thermal management

Abstract

A data processing system includes a processor core having an associated upper level cache and a lower level victim cache. In response to a memory access request of the processor core that specifies a non-modifying access to a target coherency granule, a determination is made whether the memory access request hits or misses in a directory of the lower level victim cache. In response to determining that the memory access request hits in the lower level victim cache in a data-valid coherence state, the lower level victim cache provides the target coherency granule of the memory access request to the upper level cache. The lower level victim cache preserves the target coherency granule in the lower level victim cache in a shared coherence state if the memory access request is of a first type and invalidates the target coherency granule if the memory access request is of a second type.

Claims

1. A method of data processing in a data processing system including a processor core having an associated upper level cache having a lower access latency and a lower level victim cache having a higher access latency, said method comprising:

filling the lower level victim cache with coherency granules evicted from the upper level cache, wherein at least some coherency granules held in the lower level cache are held simultaneously in the upper level cache;

in response to a memory access request of the processor core that missed in the upper level cache and that specifies a non-modifying access to a target coherency granule, determining whether the memory access request hits or misses in a directory of the lower level victim cache;

in response to determining that the memory access request hits in the lower level victim cache in a data-valid coherence state;

the lower level victim cache providing the target coherency granule of the memory access request to the upper level cache;

if the memory access request is of a first type, preserving the target coherency granule in the lower level victim cache and associating the target coherency granule in the lower level victim cache with a shared coherence state indicating the target coherency granule is held in both the upper level cache and the lower level victim cache; and

invalidating the target coherency granule in the lower level victim cache if the memory access request is of a second type.

2. The method of claim 1, wherein:

the first type comprises a demand load request of the processor core generated in response to processing an instruction in the processor core; and

the second type comprises a prefetch request of the processor core.

3. The method of claim 2, wherein the second type comprises a prefetch request of a store prefetch stream.

4. The method of claim 1, wherein the first type comprises an instruction fetch request.

5. The method of claim 1, wherein the first type comprises a translation fetch request requesting information utilized to perform address translation.

6. The method of claim 1, and further comprising;

in response to determining that the memory access request hits in the lower level victim cache in a data-valid coherence state and if the memory access request is of a first type, updating a replacement order of the target coherency granule in the lower level victim cache indicated by a least recently used replacement order field to a position other than most recently used.

7. The method of claim 1, wherein:

the data processing system includes a system memory;

said step of invalidating the target coherency granule if the memory access request is of a second type comprises invalidating the target coherency granule if the memory access request is of a second type and the target coherency granule is modified with respect to a corresponding memory block in the system memory.

8. A processing unit for a data processing system, the processing unit comprising:

a processor core; and

a cache hierarchy including an upper level cache coupled to the processor core and a lower level victim cache coupled to the upper level cache, said upper level cache having a lower access latency and the lower level cache having a higher access latency, wherein the lower level victim cache is filled by eviction of coherency granules from the upper level cache and at least some coherency granules held in the lower level cache are held simultaneously in the upper level cache, each of the upper level cache and the lower level victim cache including a respective cache directory and a respective data array, wherein the lower level victim cache, responsive to a memory access request of the processor core that missed in the upper level cache and that specifies a non-modifying access to a target coherency granule, determines whether the memory access request hits or misses in a directory of the lower level victim cache, wherein responsive to determining that the memory access request hits in the lower level victim cache in a data-valid coherence state, the lower level victim cache provides the target coherency granule of the memory access request to the upper level cache and preserves the target coherency granule in the data array of the lower level victim cache if the memory access request is of a first type and invalidates the target coherency granule if the memory access request is of a second type, and wherein the lower level victim cache, if the target coherency granule is preserved in the data array of the lower level victim cache in response to the memory access request, associates the target coherency granule with a shared coherence state indicating the target coherency granule is held in both the upper level cache and the lower level victim cache in the directory of the lower level victim cache.

9. The processing unit of claim 8, wherein:

the first type comprises a demand load request of the processor core generated in response to processing an instruction in the processor core; and

the second type comprises a prefetch request of the processor core.

10. The processing unit of claim 9, wherein the second type comprises a prefetch request of a store prefetch stream.

11. The processing unit of claim 8, wherein the first type comprises an instruction fetch request.

12. The processing unit of claim 8, wherein the first type comprises a translation fetch request requesting information utilized to perform address translation.

13. The processing unit of claim 8, wherein the lower level victim cache, responsive to determining that the memory access request hits in the lower level victim cache in a data-valid coherence state and that the memory access request is of a first type, updates a replacement order of the target coherency granule in the lower level victim cache indicated by a least recently used replacement order field to a position other than most recently used.

14. The processing unit of claim 8, wherein:

the data processing system includes a system memory;

the lower level victim cache invalidates the target coherency granule if the memory access request is of a second type and if the target coherency granule is modified with respect to a corresponding memory block in the system memory.

15. A data processing system, comprising:

at least one system memory; and

a plurality of processing units coupled to the system memory, wherein a processing unit among the plurality of processing units includes;

a processor core; and

a cache hierarchy including an upper level cache coupled to the processor core and a lower level victim cache coupled to the upper level cache, said upper level cache having a lower access latency and the lower level cache having a higher access latency, wherein the lower level victim cache is filled by eviction of coherency granules from the upper level cache and at least some coherency granules held in the lower level cache are held simultaneously in the upper level cache, each of the upper level cache and the lower level victim cache including a respective cache directory and a respective data array, wherein the lower level victim cache, responsive to a memory access request of the processor core that missed in the upper level cache and that specifies a non-modifying access to a target coherency granule, determines whether the memory access request hits or misses in a directory of the lower level victim cache, wherein responsive to determining that the memory access request hits in the lower level victim cache in a data-valid coherence state, the lower level victim cache provides the target coherency granule of the memory access request to the upper level cache and preserves the target coherency granule in the data array of the lower level victim cache if the memory access request is of a first type and invalidates the target coherency granule if the memory access request is of a second type, and wherein the lower level victim cache, if the target coherency granule is preserved in the data array of the lower level victim cache in response to the memory access request, associates the target coherency granule with a shared coherence state indicating the target coherency granule is held in both the upper level cache and the lower level victim cache in the directory of the lower level victim cache.

16. The data processing system of claim 15, wherein:

the first type comprises a demand load request of the processor core generated in response to processing an instruction in the processor core; and

the second type comprises a prefetch request of the processor core.

17. The data processing system of claim 16, wherein the second type comprises a prefetch request of a store prefetch stream.

18. The data processing system of claim 15, wherein the first type comprises an instruction fetch request.

19. The data processing system of claim 15, wherein the first type comprises a translation fetch request requesting information utilized to perform address translation.

20. The data processing system of claim 15, wherein:

the data processing system includes a system memory;

the lower level victim cache invalidates the target coherency granule if the memory access request is of a second type and if the target coherency granule is modified with respect to a corresponding memory block in the system memory.

21. The data processing system of claim 15, wherein the lower level victim cache, responsive to determining that the memory access request hits in the lower level victim cache in a data-valid coherence state and that the memory access request is of a first type, updates a replacement order of the target coherency granule in the lower level victim cache indicated by a least recently used replacement order field to a position other than most recently used.