Cache metadata identifiers for isolation and sharing

• US 20080040551A1
• Filed: 08/06/2007
• Published: 02/14/2008
• Est. Priority Date: 12/07/2005
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Harris, Timothy, Smith, Burton, Gray, Jan, Larus, James
• Granted Patent: US 8,225,297 B2
• Time in Patent Office: Days
• US Class Current: 711/130
• CPC Class Codes: G06F 12/084 with a shared cache

Abstract

Various technologies and techniques are disclosed for providing software accessible metadata on a cache of a central processing unit. A multiprocessor has at least one central processing unit. The central processing unit has a cache with cache lines that are augmented by cache metadata. The cache metadata includes software-controlled metadata identifiers that allow multiple logical processors to share the cache metadata. The metadata identifiers and cache metadata can then be used to accelerate various operations. For example, parallel computations can be accelerated using cache metadata and metadata identifiers. As another example, nested computations can be accelerated using metadata identifiers and cache metadata. As yet another example, transactional memory applications that include parallelism within transactions or that include nested transactions can be also accelerated using cache metadata and metadata identifiers.

Claims

1. A system comprising:

a multiprocessor having at least one central processing unit, the central processing unit having a cache consisting of a plurality of cache lines augmented by a plurality of cache metadata, the cache metadata including software-controlled metadata identifiers that allow a plurality of logical processors to share the cache metadata.

2. The system of claim 1, wherein the metadata identifiers are used to achieve isolation between logically distinct accesses to the cache metadata.

3. The system of claim 1, wherein the cache metadata can be shared among multiple logical processors by assigning each of the multiple logical processors a same metadata identifier.

4. The system of claim 1, wherein each metadata identifier is associated with a separate logical processor to achieve software thread isolation.

5. The system of claim 1, wherein the metadata identifiers are operable to accelerate parallelism within transactions by allowing a plurality of software threads within a same transaction to share a same one of the metadata identifiers and a corresponding metadata state.

6. The system of claim 5, wherein the plurality of software threads are isolated from other software threads that are not part of the same transaction.

7. The system of claim 1, wherein a plurality of nested transactions are accelerated by assigning a unique one of the metadata identifiers to each separate nesting level of the nested transactions.

8. The system of claim 7, wherein each separate nesting level uses the cache metadata to eliminate redundant transactional memory operations.

9. The system of claim 1, wherein a plurality of software subsystems each use a distinct sub-range of the metadata identifiers to accelerate one or more computations by metadata instructions.

10. The system of claim 1, wherein a plurality of software subsystems each use a distinct sub-range of the metadata identifiers to access a corresponding portion of the cache metadata.

11. The system of claim 1, wherein a plurality of software subsystems can share the cache metadata during a same interval of time.

12. The system of claim 1, wherein the cache metadata are present at a plurality of levels in the cache.

13. A method for accelerating a parallel computation using cache metadata and metadata identifiers, comprising the steps of:

obtaining a metadata identifier;

when parallelizing a computation, assigning the metadata identifier to a plurality of parallel threads in the computation; and

performing at least some metadata accesses of the cache metadata from the plurality of parallel threads using the metadata identifier.

14. The method of claim 13, wherein the cache metadata are used to avoid redundant operations within the computation.

15. The method of claim 13, wherein the computation is a transaction.

16. A computer-readable medium having computer-executable instructions for causing a computer to perform the steps recited in claim 13.

17. A method for accelerating nested computations using multiple metadata identifiers comprising the steps of:

when entering a new nest level of a computation, obtaining a new metadata identifier for the computation;

performing accesses to the cache metadata using the new metadata identifier;

upon exiting the new nest level of the computation and resuming at a prior nest level, obtaining a different metadata identifier; and

performing accesses to the cache metadata using the different metadata identifier.

18. The method of claim 17, wherein the different metadata identifier is a prior metadata identifier, and wherein upon entering the new nest level, the prior metadata identifier is saved, and wherein upon resuming at the prior nest level, the prior metadata identifier is retrieved and used for performing accesses to the cache metadata.

19. The method of claim 17, wherein the computation is a nested transaction.

20. A computer-readable medium having computer-executable instructions for causing a computer to perform the steps recited in claim 17.