Cache management during asynchronous memory move operations

• US 8,327,101 B2
• Filed: 02/01/2008
• Issued: 12/04/2012
• Est. Priority Date: 02/01/2008
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Arimilli, Ravi K., Blackmore, Robert S., Kim, Chulho, Sinharoy, Balaram, Xue, Hanhong
• Primary Examiner(s): Dudek, Jr., Edward
• Assistant Examiner(s): Verderamo, III, Ralph A
• Application Number:
  US12/024,526
• Publication Number:
  US 20090198897A1
• Time in Patent Office: 1,768 Days
• Field of Search: None
• US Class Current: 711/165
• CPC Class Codes: G06F 12/0815 Cache consistency protocols
• CPC Class Codes: G06F 12/10 Address translation
• CPC Class Codes: G06F 9/30032 Movement instructions, e.g. MOVE, SHIFT, ROTATE, SHUFFLE
• CPC Class Codes: G06F 9/3004 to perform operations on memory
• CPC Class Codes: G06F 9/30043 LOAD or STORE instructions; Clear instruction
• CPC Class Codes: G06F 9/30087 Synchronisation or serialisation instructions
• CPC Class Codes: G06F 9/3017 Runtime instruction translation, e.g. macros
• CPC Class Codes: G06F 9/3877 using a slave processor, e.g. coprocessor peripheral processor G06F13/12; vector processor G06F15/8053

Abstract

A data processing system includes a mechanism for completing an asynchronous memory move (AMM) operation in which the processor receives an AMM ST instruction and processes a processor-level move of data in virtual address space and an asynchronous memory mover then completes a physical move of the data within the real address space (memory). A status/control field of the AMM ST instruction includes an indication of a requested treatment of the lower level cache(s) on completion of the AMM operation. When the status/control field indicates an update to at least one cache should be performed, the asynchronous memory mover automatically forwards a copy of the data from the data move to the lower level cache, and triggers an update of a coherency state for a cache line in which the copy of the data is placed.

Claims

1. A data processing system comprising:

a processor;

a memory coupled to the processor and including a plurality of physical locations with real addresses that are utilized for storing data;

a memory controller that manages memory access operations;

an asynchronous memory mover communicatively coupled to but separate from the processor and the memory controller and which performs a physical move of data during an asynchronous memory move (AMM) operation, which AMM operation is initiated by execution by the processor of an asynchronous memory move (AMM) Store (ST) instruction that causes the processor to perform an effective address move of data as part of a corresponding AMM operation;

at least one lower level cache that buffers data from memory for utilization during processor execution;

processing logic executing on the processor for initiating the AMM operation, the AMM operation comprising;

the processor receiving the AMM ST instruction, wherein the AMM ST instruction comprises a status/control field that includes an indication of a requested treatment of one or more lower level cache(s) on completion of the corresponding AMM operation;

performing an effective address move of data from a first effective address to a second effective address provided by the AMM ST instruction, wherein the effective address move of data comprises the processor allocating the second effective address to the data in place of the first effective address and the effective address move of data triggers the asynchronous memory mover to initiate a physical move of the data within the memory; and

asynchronous memory mover logic executing on the asynchronous memory mover, wherein the asynchronous memory mover logic, in response to being triggered by the processor performing the effective address move of data, performs the functions of;

receiving the real address and other parameters of the AMM ST instruction; and

completing the physical move of the data from a first memory location in the memory having a first real address to a second memory location in the memory having a second real address while the processor continues processing subsequently received instructions, wherein the physical move is coordinated by the asynchronous memory mover along a path that does not include the processor.

2. The data processing system of claim 1, the processing logic further comprising logic for:

checking a status/control field of the AMM ST instruction for an indication of a requested treatment of the at least one lower level cache on completion of the AMM operation;

in response to the status/control field indicating an update to the at least one cache should be performed;

automatically forwarding a copy of the data from the data move to the at least one lower level cache; and

triggering an update of a coherency state for a cache line within the at least one lower level cache in which the copy of the data is placed.

3. The data processing system of claim 2, wherein the asynchronous memory mover contains registers for storing parameters of the AMM ST instruction that enable completion of the AMM operation, and said logic for triggering the update includes logic for retrieving the coherency status from a coherency state register associated with the parameters of the AMM ST instruction, stored within the registers of the asynchronous memory mover.

4. The data processing system of claim 2, wherein forwarding the data further comprises forwarding the data from the source real address and tagging the data with the destination effective address corresponding to the destination real address.

5. The data processing system of claim 2, wherein forwarding the data further comprises:

evaluating a size of the data being moved relative to a size of the at least one lower level cache; and

forwarding only a first subset of the data to the at least one lower level cache when the size of the data exceeds a threshold size which would overrun the lower level cache.

6. The data processing system of claim 5, wherein the at least one lower level cache includes an L2 cache and an L3 cache, and wherein forwarding only a first subset of the data further comprises:

forwarding one or more L2 cache line sized data to the L2 cache from the first portion of the data being moved;

forwarding one or more L3 cache line sized data to the L3 cache from a sequentially following second portion of data being moved; and

forwarding remaining data to the memory;

wherein not all data being moved are placed within the L2 and L3 cache; and

wherein the L2 and L3 caches are not filled with data from the data move operation.

7. The data processing system of claim 1, wherein said processing logic further comprises logic for:

determining when the destination effective address of the AMM ST instruction is an effective address that is represented within the lower level cache; and

in response to the destination effective address being represented within the lower level cache, performing at least one of;

triggering an automatic update of the data within cache lines addressed with the corresponding effective address of the lower level cache; and

triggering an update of the cache state for the cache lines to an invalid (I) state, to force the cache to load coherent data from the memory when the data associated with the effective address is requested by the processor.

8. The data processing system of claim 1, wherein the status/control field includes one or more specific cache states for assigning to the data by the asynchronous memory mover.

9. The data processing system of claim 1, wherein the status/control field provides at least one of:

an indication to store the data to the at least one lower level cache;

an indication to not store the data to the at least one lower level cache;

a tag of the first effective address of the data with one of an Exclusive (E), Shared (S), or Invalid state within the cache; and

a tag of the second effective address of the data with one of a Modified, E, or S state within the cache.

10. The data processing system of claim 1, wherein the asynchronous memory mover provides the data to the one or more lower level cache(s) via cache injection.

11. In a data processing system having a processor, an asynchronous memory mover, a memory subsystem including a memory, a memory controller, and at least one lower level cache, a method comprising:

completing, via the processor, a move of data in virtual address space, in response to receiving an asynchronous memory move (AMM) store (ST) instruction, wherein the move includes a processor-level allocation of a second effective address to the data in place of a first effective address, to represent an effective address move of the data in virtual address space, wherein execution, by the processor, of the effective address move of data in virtual address space triggers the asynchronous memory mover to initiate a physical move of the data within the memory;

in response to the trigger caused by the processor execution of the effective address move of data in virtual address space, completing, via the asynchronous memory mover, an asynchronous memory move (AMM) operation generated by the AMM ST instruction, wherein the AMM operation completes a physical move of the data from a first memory location in the memory having a first real address to a second memory location in the second memory having a second real address, while the processor continues processing subsequently received instructions, wherein the physical move is coordinated by the asynchronous memory mover along a path that does not include the processor; and

performing an update to the at least one lower level cache based on the completion of the physical move and an included indication, within a status/control field of the AMM ST instruction, of a requested treatment of one or more lower level cache(s) on completion of the AMM operation;

wherein the asynchronous memory mover is communicatively coupled to but separate from the processor and the memory controller; and

wherein the asynchronous memory mover performs the physical move of data responsive to receipt of real addresses and other parameters associated with the AMM ST instruction.

12. The method of claim 11, wherein the performing an update further comprises:

checking a status/control field of the AMM ST instruction for an indication of a requested treatment of the at least one lower level cache on completion of the AMM operation; and

in response to the status/control field indicating an update to the at least one cache should be performed;

automatically forwarding a copy of the data from the physical move to the at least one lower level cache; and

triggering an update of a coherency state for a cache line within the at least one lower level cache in which the copy of the data is placed.

13. The method of claim 12, wherein the asynchronous memory mover contains registers for storing parameters of the AMM ST instruction that enable completion of the AMM operation, and said triggering the update further comprises retrieving the coherency status from a coherency state register associated with the parameters of the AMM ST instruction, stored within the registers of the asynchronous memory mover.

14. The method of claim 11, wherein the performing an update comprises forwarding the data from a source real address and tagging the data with a destination effective address corresponding to the destination real address.

15. The method of claim 11, wherein the performing an update comprises:

evaluating a size of the data being moved relative to a size of the at least one lower level cache; and

forwarding only a first subset of the data to the at least one lower level cache when the size of the data exceeds a threshold size which would overrun the lower level cache.

16. The method of claim 15, wherein the at least one lower level cache includes an L2 cache and an L3 cache, wherein said forwarding only a first subset of the data comprises:

forwarding one or more L2 cache line sized data to the L2 cache from the first portion of the data being moved;

forwarding one or more L3 cache line sized data to the L3 cache from a sequentially following second portion of data being moved; and

forwarding remaining data to the memory;

wherein not all data being moved are placed within the L2 and L3 cache and wherein the L2 and L3 caches are not filled with data from the data move operation.

17. The method of claim 11, further comprising:

determining when the destination effective address of the AMM ST instruction is an effective address that is represented within the lower level cache; and

responding to the effective addresses being the same by performing one of;

triggering an automatic update of the data within the lower level cache within cache lines addressed with the corresponding effective address; and

triggering an update of the cache state for the cache lines to an invalid (I) state, to force the cache to load coherent data from the memory when the data associated with the effective address is requested by the processor.

18. The method of claim 11, wherein the status/control field includes one or more specific cache states for assigning to the data by the asynchronous memory mover.

19. The method of claim 11, wherein the status/control field provides at least one of:

an indication to store the data to the at least one lower level cache;

an indication to not store the data to the at least one lower level cache;

a tag of the first effective address of the data with one of an Exclusive (E), Shared (S), or Invalid state within the cache; and

a tag of the second effective address of the data with one of a Modified, E, or S state within the cache.

20. The method of claim 11, wherein the asynchronous memory mover provides the data to the one or more lower level cache(s) via cache injection.