Microprocessor circuits, systems, and methods using a combined writeback queue and victim cache

• US 6,038,645 A
• Filed: 08/28/1997
• Issued: 03/14/2000
• Est. Priority Date: 08/28/1996
• Status: Expired due to Term

• Current Assignee: Texas Instruments, Inc.
• Original Assignee: Texas Instruments, Inc.
• Inventors: Nanda, Ashwini K., Shiell, Jonathan H.
• Primary Examiner(s): Chan, Eddie P.
• Assistant Examiner(s): NGUYEN, THAN VINH
• Application Number:
  US08/919,732
• Time in Patent Office: 929 Days
• Field of Search: 711/140, 711/141, 711/142, 711/143, 711/144, 711/146, 711/133, 711/134, 711/136, 711/137, 711/138, 711/118, 711/122
• US Class Current: 711/141
• CPC Class Codes: G06F 12/0804 with main memory updating G06F12/0806 takes precedence

Abstract

A microprocessor (10) comprising a central processor unit core (12) operable to write information during a write cycle and a cache circuit (18) coupled to the central processor unit core and operable to evict information. The microprocessor further includes a combined storage queue (16) coupled to the central processor unit core and to the cache circuit. The combined storage queue includes a set of logical storage blocks (22c) which is operable to store both information written by the central processor unit core and information evicted by the cache circuit. Other circuits, systems, and methods are also disclosed and claimed.

Claims

1. A microprocessor, comprising:

a central processor unit core operable to write information during a write cycle;

a cache circuit coupled to said central processor unit core and operable to evict information;

a combined storage queue coupled to said central processor unit core and said cache circuit, comprising a set of logical storage blocks, wherein each of said set of logical storage blocks is operable to store information selected from either information written by said central processor unit core or information evicted by said cache circuits, said combined storage queue maintaining an indication whether information stored in a logical storage block is information written by said central processor unit core or information evicted by said cache circuit;

selection circuitry for routing information written by said central processor unit core or information evicted by said cache circuit into said combined storage queue on a first-in-first-out basis, said selection circuitry;

responsive to receiving additional information written by said central processor unit core and detecting that each of said set of logical storage blocks stores information, routing said additional information to be stored in said set of logical storage blocks and overwriting from said set of logical storage blocks a least recently stored block of evicted cache information, and

responsive to receiving additional information evicted by said cache circuit and detecting that each of said set of logical storage blocks stores information, routing said additional information to be stored in said set of logical storage blocks and overwriting from said set of logical storage blocks a least recently stored block of evicted cache information.

2. The microprocessor of claim 1 and further comprising:

a memory system higher in hierarchical level than said combined storage queue and for storing information, wherein said set of logical storage blocks storing information evicted by said cache circuit are mapped to locations in said memory system; and

an indicator corresponding to said additional information and for representing that said additional information has been updated in said combined storage queue without updating a corresponding mapped location in said memory system.

3. The microprocessor of claim 2 and further comprising:

circuitry, responsive to said indicator representing that said additional information has been updated in said combined storage queue without updating a corresponding mapped location in said memory system, for updating said corresponding mapped location in said memory system before said selection circuitry overwrites said least recently stored block of evicted cache information.

4. The microprocessor of claim 2 wherein, responsive to receiving additional information written by said central processor unit core and detecting that each of said set of logical storage blocks stores information, said selection circuitry routes said additional information to be stored in said set of logical storage blocks and overwrites from said set of logical storage blocks a least recently stored block of evicted cache information having said indicator representing that said additional information has been updated in said combined storage queue and said corresponding mapped location in said memory system has been updated.

5. The microprocessor of claim 1 and further comprising:

a memory system higher in hierarchical level than said combined storage queue and for storing information, wherein said set of logical storage blocks storing information evicted by said cache circuit are mapped to locations in said memory system; and

an indicator corresponding to said additional information and for representing that said additional information has been updated in said combined storage queue without updating a corresponding mapped location in said memory system.

6. The microprocessor of claim 1:

further comprising selection circuitry for routing information written by said central processor unit core for storage into said set of logical storage blocks;

wherein each of said set of logical storage blocks is operable to store information equal to or less than a predetermined number of bytes;

wherein, responsive to receiving additional information written by said central processor unit core less than said predetermined number of bytes, said selection circuitry determines whether said combined storage queue is already storing a corresponding block of information which maps to a same block address as said additional information; and

wherein, responsive to said selection circuitry determining that said combined storage queue is already storing said corresponding block of information, said selection circuitry routes said additional information to overwrite selected bytes of said corresponding block of information.

7. The microprocessor of claim 6 and further comprising circuitry for reading said corresponding block of information from said combined storage queue and thereafter writing said corresponding block of information to a memory system higher in hierarchical level than said combined storage queue after said selection circuitry routes said additional information to overwrite selected bytes of said corresponding block of information.

8. A microprocessor, comprising:

a central processor unit core operable to write information during a write cycle;

a cache circuit coupled to said central processor unit core and operable to evict information;

a combined storage queue coupled to said central processor unit core and said cache circuit, comprising a set of logical storage blocks, wherein each of said set of logical storage blocks is operable to store information selected from either information written by said central processor unit core or information evicted by said cache circuit;

wherein said set of logical storage blocks comprises a first set of logical storage blocks operable to store both information written by said central processor unit core and information evicted by said cache circuit; and

wherein said combined storage queue further comprises;

a second set of logical storage blocks, wherein each of said second set of logical storage blocks is operable to store only information written by said central processor unit core; and

a third set of logical storage blocks, wherein each of said third set of logical storage blocks is operable to store only information evicted by said cache circuit.

9. The microprocessor of claim 8 and further comprising selection circuitry for routing information written by said central processor unit core for storage into said second and first sets of logical storage blocks on a first-in-first-out basis such that said information logically routes firstly into said second set of logical storage blocks and secondly from said second set of logical storage blocks to said first set of logical storage blocks.

10. The microprocessor of claim 8 and further comprising selection circuitry for routing information evicted by said cache circuit for storage into said third and first sets of logical storage blocks on a first-in-first-out basis such that said information logically routes firstly into said third set of logical storage blocks and secondly from said third set of logical storage blocks to said first set of logical storage blocks.

11. The microprocessor of claim 8 and further comprising:

selection circuitry for routing information written by said central processor unit core for storage into said second and first sets of logical storage blocks on a first-in-first-out basis such that said information logically routes firstly into said second set of logical storage blocks and secondly from said second set of logical storage blocks to said first set of logical storage blocks; and

selection circuitry for routing information evicted by said cache circuit for storage into said third and first sets of logical storage blocks on a first-in-first-out basis such that said information logically routes firstly into said third set of logical storage blocks and secondly from said third set of logical storage blocks to said first set of logical storage blocks.

12. The microprocessor of claim 11 wherein, responsive to receiving additional information written by said central processor unit core and detecting that each of said first set of logical storage blocks stores information, said selection circuitry routes said additional information to be stored in said second set of logical storage blocks and overwrites from said first set of logical storage blocks a least recently stored block of evicted cache information.

13. The microprocessor of claim 11 wherein, responsive to receiving additional information evicted by said cache circuit and detecting that each of said first set of logical storage blocks stores information, said selection circuitry routes said additional information to be stored in said third set of logical storage blocks and overwrites from said first set of logical storage blocks a least recently stored block of evicted cache information.

14. The microprocessor of claim 11:

wherein, responsive to receiving additional information written by said central processor unit core and detecting that each of said first set of logical storage blocks stores information, said selection circuitry routes said additional information to be stored in said second set of logical storage blocks and overwrites from said first set of logical storage blocks a least recently stored block of evicted cache information; and

wherein, responsive to receiving additional information evicted by said cache circuit and detecting that each of said first set of logical storage blocks stores information, said selection circuitry routes said additional information to be stored in said third set of logical storage blocks and overwrites from said first set of logical storage blocks a least recently stored block of evicted cache information.

15. The microprocessor of claim 8:

further comprising selection circuitry for routing information written by said central processor unit core for storage into said second and first sets of logical storage blocks;

wherein each of said second and first sets of logical storage blocks is operable to store information equal to or less than a predetermined number of bytes;

wherein, responsive to receiving additional information written by said central processor unit core less than said predetermined number of bytes, said selection circuitry determines whether said combined storage queue is already storing a corresponding block of information which maps to a same block address as said additional information; and

wherein, responsive to said selection circuitry determining that said combined storage queue is already storing said corresponding block of information, said selection circuitry routes said additional information to overwrite selected bytes of said corresponding block of information.

16. The microprocessor of claim 15 and further comprising circuitry for reading said corresponding block of information from said combined storage queue and thereafter writing said corresponding block of information to a memory system higher in hierarchical level than said combined storage queue after said selection circuitry routes said additional information to overwrite selected bytes of said corresponding block of information.

17. The microprocessor of claim 8 wherein each of said second set of logical storage blocks and said third set of logical storage blocks equals a same integer number of logical storage blocks.

18. The microprocessor of claim 8 wherein each of said second set of logical storage blocks and said third set of logical storage blocks equals four logical storage blocks.

19. The microprocessor of claim 8 wherein said first set of logical storage blocks equals twelve logical storage blocks.

20. The microprocessor of claim 8:

wherein each of said second set of logical storage blocks and said third set of logical storage blocks equals four logical storage blocks; and

wherein said first set of logical storage blocks equals twelve logical storage blocks.

21. A microprocessor, comprising:

a central processor unit core operable to write information during a write cycle;

a cache circuit coupled to said central processor unit core and operable to evict information;

a combined storage queue coupled to said central processor unit core and said cache circuit, comprising a set of logical storage blocks, wherein each of said set of logical storage blocks is operable to store information selected from either information written by said central processor unit core or information evicted by said cache circuit;

wherein said set of logical storage blocks comprises a first set of logical storage blocks operable to store both information written by said central processor unit core and information evicted by said cache circuit; and

wherein said set of logical storage blocks further comprises a second set of logical storage blocks, wherein each of said second set of logical storage blocks is operable to store only information written by said central processor unit core.

22. The microprocessor of claim 21 and further comprising selection circuitry for routing information written by said central processor unit core for storage into said second and first sets of logical storage blocks on a first-in-first-out basis such that said information logically routes firstly into said second set of logical storage blocks and secondly from said second set of logical storage blocks to said first set of logical storage blocks.

23. The microprocessor of claim 22 wherein, responsive to receiving additional information written by said central processor unit core and detecting that each of said first set of logical storage blocks stores information, said selection circuitry routes said additional information to be stored in said second set of logical storage blocks and overwrites from said first set of logical storage blocks a least recently stored block of evicted cache information.

24. The microprocessor of claim 21:

further comprising selection circuitry for routing information written by said central processor unit core for storage into said second and first sets of logical storage blocks;

wherein each of said second and first sets of logical storage blocks is operable to store information equal to or less than a predetermined number of bytes;

wherein, responsive to receiving additional information written by said central processor unit core less than said predetermined number of bytes, said selection circuitry determines whether said combined storage queue is already storing a corresponding block of information which maps to a same block address as said additional information; and

wherein, responsive to said selection circuitry determining that said combined storage queue is already storing said corresponding block of information, said selection circuitry routes said additional information to overwrite selected bytes of said corresponding block of information.

25. The microprocessor of claim 24 and further comprising circuitry for reading said corresponding block of information from said combined storage queue and thereafter writing said corresponding block of information to a memory system higher in hierarchical level than said combined storage queue after said selection circuitry routes said additional information to overwrite selected bytes of said corresponding block of information.

26. A microprocessor having an architected state, said microprocessor comprising:

a central processor unit core operable to write information during a write cycle, wherein said information is actual writeback information to affect said architected state of said microprocessor;

a cache circuit coupled to said central processor unit core and operable to evict information; and

a combined storage queue coupled to said central processor unit core and said cache circuit, comprising;

a first set of logical storage blocks, wherein each of said set of logical storage blocks is operable to store information selected from either said information written by said central processor unit core or information evicted by said cache circuit; and

a second set of logical storage blocks, wherein each of said second set of logical storage blocks is operable to store only said information written by said central processor unit core;

selection circuitry for routing said information written by said central processor unit core for storage into said first and second sets of logical storage blocks;

wherein each of said first and second sets of logical storage blocks is operable to store information equal to or less than a predetermined number of bytes;

wherein, responsive to receiving additional information written by said central processor unit core less than said predetermined number of bytes, said selection circuitry determines whether said combined storage queue is already storing a corresponding block of information which maps to a same block address as said additional information; and

wherein, responsive to said selection circuitry determining that said combined storage queue is already storing said corresponding block of information, said selection circuitry routes said additional information to overwrite selected bytes of said corresponding block of information.

27. The microprocessor of claim 26 and further comprising selection circuitry for routing information written by said central processor unit core for storage into said second and first sets of logical storage blocks on a first-in-first-out basis such that said information logically routes firstly into said second set of logical storage blocks and secondly from said second set of logical storage blocks to said first set of logical storage blocks.

28. The microprocessor of claim 27 wherein, responsive to receiving additional information written by said central processor unit core and detecting that each of said first set of logical storage blocks stores information, said selection circuitry routes said additional information to be stored in said second set of logical storage blocks and overwrites from said first set of logical storage blocks a least recently stored block of evicted cache information.

29. The microprocessor of claim 26 and further comprising circuitry for reading said corresponding block of information from said combined storage queue and thereafter writing said corresponding block of information to a memory system higher in hierarchical level than said combined storage queue after said selection circuitry routes said additional information to overwrite selected bytes of said corresponding block of information.

30. A microprocessor, comprising:

a central processor unit core operable to write information during a write cycle, said information including actual information in response to execution of an instruction intended to take effect upon an architected state of the microprocessor and speculative information in response to a speculatively executed instruction;

a cache circuit coupled to said central processor unit core and operable to evict information;

a combined storage queue coupled to said central processor unit core and said cache circuit, comprising a set of logical storage blocks, wherein each of said set of logical storage blocks is operable to store information selected from either actual information written by said central processor unit core, speculative information written by said central processor unit core or information evicted by said cache circuit;

a memory system higher in hierarchical level than said combined storage queue and for storing information, wherein said set of logical storage blocks storing information evicted by said cache circuit are mapped to locations in said memory system; and

wherein said combined storage queue maintains an indication whether information written by said central processor unit core is actual information or speculative information; and

selection circuitry for routing information written by said central processor unit core or information evicted by said cache circuit into said combined storage queue on a first-in-first-out basis, responsive to receiving additional information written by said central processor unit core or information evicted by said cache circuit and upon determining that information written by said central processor unit core stored in a particular logical block is to be evicted said selection circuitry;

evicting said information and writing said information to said higher level memory system if said information is actual information, and

evicting said information without writing to said higher level memory system if said information is speculative information.

31. The microprocessor of claim 30, wherein:

responsive to receiving additional information written by said central processor unit core said selection circuitry

determines whether said combined storage queue is already storing a corresponding block of information which maps to a same block address as said additional information,

routes said additional information to overwrite said corresponding logical storage block if said corresponding information is actual information, and

routes said additional information to another logical storage block if said corresponding information is speculative information.

32. The microprocessor of claim 30 wherein said circuitry for indicating that said information written by said central processor unit core is speculative information written by said central processor unit is operable to further indicate that said information written by said central processor unit core is converted from speculative information to actual information in response to completion of an instruction corresponding to said speculative information.