Load/store unit employing last-in-buffer indication for rapid load-hit-store

• US 6,393,536 B1
• Filed: 05/18/1999
• Issued: 05/21/2002
• Est. Priority Date: 05/18/1999
• Status: Active Grant

• Current Assignee: Advanced Micro Devices, Inc.
• Original Assignee: Advanced Micro Devices, Inc.
• Inventors: Roberts, James Scott, Hughes, William Alexander
• Primary Examiner(s): Kim, Matthew
• Assistant Examiner(s): Vital, Pierre M.
• Application Number:
  US09/314,180
• Time in Patent Office: 1,099 Days
• Field of Search: 711/118, 711/144, 711/156, 711/159, 712/218, 712/225
• US Class Current: 711/159
• CPC Class Codes: G06F 9/3834 Maintaining memory consistency
• CPC Class Codes: G06F 9/3838 Dependency mechanisms, e.g. register scoreboarding

Abstract

A load/store unit includes a buffer configured to retain store memory operations which have probed the data cache. Each entry in the buffer includes a last-in-buffer (LIB) indication which identifies whether or not the store in that entry is the youngest store in the buffer to update the memory locations specified by the corresponding store address. Load addresses are compared to the store addresses, and the comparison result is qualified with the corresponding LIB indication such that only the youngest store is identified as a hit. At most one load hit store is detected. In one embodiment, the buffer also stores loads which have probed the data cache. To associate the load with the youngest store which is older than the load, the buffer records the store instruction tag of a store which is hit by the load during the initial probe (according to the LIB indications during the initial probe). During reprobes, the LIB indications are ignored and instead the store instruction tags are compared to the recorded store instruction tag. As stores are inserted into the buffer, the store address is compared to the addresses in the buffer. If a hit is detected, the LIB indication for the hit store is set to a state indicating that the hit store in not the last in buffer to update the corresponding store address. The LIB indication for the newly inserted store is set to a last-in-buffer state.

Claims

1. A load/store unit comprising:

a buffer including a plurality of entries, each of said plurality of entries configured to store a data address and a last-in-buffer (LIB) indication, wherein said LIB indication, in a first state, is indicative that a corresponding store memory operation is a youngest store memory operation within said buffer to update a memory location identified by said data address; and

control logic coupled to said buffer and to receive a first data address probing a data cache, wherein said control logic is configured to identify a first entry of said plurality of entries for which;

(i) said data address stored in said first entry matches said first data address, and (ii) said LIB indication stored in said first entry is in said first state.

2. The load/store unit as recited in claim 1 wherein said control logic, in response to store data being valid within said first entry and said first address corresponding to a first load memory operation, is configured to cause said store data from said first entry to be forwarded as a result of said first load memory operation.

3. The load/store unit as recited in claim 2 wherein said first entry is further configured to store a store instruction tag identifying a first store memory operation assigned to said first entry, and wherein said control logic is configured to select said store instruction tag from said first entry if said store data within said first entry is invalid.

4. The load/store unit as recited in claim 3 wherein said control logic is configured to store said store instruction tag into a second entry of said plurality of entries, said second entry assigned to said first load memory operation.

5. The load/store unit as recited in claim 4 wherein said control logic is configured to select said first load memory operation to reprobe said data cache from said second entry.

6. The load/store unit as recited in claim 5 wherein said control logic, during said reprobing, is configured to compare said store instruction tag from said second entry to a plurality of instruction tags stored in said buffer.

7. The load/store unit as recited in claim 6 wherein said control logic, in response to a match between said store instruction tag from said second entry and a store instruction tag in said first entry, is configured to cause said store data from said first entry to be forwarded even if said LIB indication in said first entry is in a second state indicative that said first store memory operation is not youngest within said buffer to update said memory location identified by said data address, provided said store data is valid.

8. The load/store unit as recited in claim 1 wherein, in response to said first data address corresponding to a second store memory operation, said control logic is configured to set said LIB indication to a second state indicative that said first store memory operation is not youngest within said buffer to update said memory location identified by said data address.

9. The load/store unit as recited in claim 8 wherein said LIB indication comprises an LIB bit and a clear LIB bit.

10. The load/store unit as recited in claim 9 wherein said first state comprises said LIB bit and said clear LIB bit having opposite binary values.

11. The load/store unit as recited in claim 9 wherein said second state comprises said LIB bit and said clear LIB bit having a same binary value.

12. A processor comprising:

a data cache configured to store data; and

a load/store unit coupled to said data cache, said load store unit including;

a buffer including a plurality of entries, each of said plurality of entries configured to store a data address and a last-in-buffer (LIB) indication, wherein said LIB indication, in a first state, is indicative that a corresponding store memory operation is a youngest store memory operation within said buffer to update a memory location identified by said data address; and

control logic coupled to said buffer and to receive a first data address probing said data cache, wherein said control logic is configured to identify a first entry of said plurality of entries for which;

(i) said data address stored in said first entry matches said first data address, and (ii) said LIB indication stored in said first entry is in said first state.

13. The processor as recited in claim 12 wherein said control logic, in response to store data being valid within said first entry and said first address corresponding to a first load memory operation, is configured to cause said data cache to forward said store data from said first entry as a result of said first load memory operation.

14. The processor as recited in claim 13 wherein said first entry is further configured to store a store instruction tag identifying a first store memory operation assigned to said first entry, and wherein said control logic is configured to select said store instruction tag from said first entry if said store data within said first entry is invalid.

15. The processor as recited in claim 14 wherein said control logic is configured to store said store instruction tag into a second entry of aid plurality of entries, said second entry assigned to said first load memory operation.

16. The processor as recited in claim 15 wherein said control logic is configured to select said first load memory operation to reprobe said data cache from said second entry.

17. The processor as recited in claim 16 wherein said control logic, during said reprobing, is configured to compare said store instruction tag from said second entry to a plurality of instruction tags stored in said buffer.

18. The processor as recited in claim 17 wherein said control logic, in response to a match between said store instruction tag from said second entry and a store instruction tag in said first entry, is configured to cause said data cache to forward said store data from said first entry even if said LIB indication in said first entry is in a second state indicative that said first store memory operation is not youngest within said buffer to update said memory location identified by said data address, provided said store data is valid.

19. The processor as recited in claim 12 wherein, in response to said first data address corresponding to a second store memory operation, said control logic is configured to set said LIB indication to a second state indicative that said first store memory operation is not youngest within said buffer to update said memory location identified by said data address.

20. The processor as recited in claim 19 wherein said LIB indication comprises an LIB bit and a clear LIB bit.

21. The processor as recited in claim 20 wherein said first state comprises said LIB bit and said clear LIB bit having opposite binary values.

22. The processor as recited in claim 20 wherein said second state comprises said LIB bit and said clear LIB bit having a same binary value.

23. A computer system comprising:

a processor including;

a data cache configured to store data; and

a load/store unit coupled to said data cache, said load store unit including;

a buffer including a plurality of entries, each of said plurality of entries configured to store a data address and a last-in-buffer (LIB) indication, wherein said LIB indication, in a first state, is indicative that a corresponding store memory operation is a youngest store memory operation within said buffer to update a memory location identified by said data address; and

control logic coupled to said buffer and to receive a first data address probing said data cache, wherein said control logic is configured to identify a first entry of said plurality of entries for which;

(i) said data address stored in said first entry matches said first data address, and (ii) said LIB indication stored in said first entry is in said first state; and

an I/O device coupled to said processor, said I/O device configured to communicate between said computer system and another computer system to which said I/O device is coupled.

24. The computer system as recited in claim 23 wherein said I/O device comprises a modem.

25. The computer system as recited in claim 24 further comprising a second processor coupled in parallel with said processor.

26. A method comprising:

probing a data cache with a first data address corresponding to a first memory operation; and

identifying a first entry within a buffer of memory operations, said first entry storing a second data address of a second memory operation and a last-in-buffer (LIB) indication, said identifying including;

determining that said second address matches said first address; and

determining that said LIB indication is in a first state indicative that said second memory operation comprises a store memory operation which is youngest in said buffer to update a memory location identified by said second address.

27. The method as recited in claim 26 further comprising, if said first memory operation comprises a load memory operation:

determining if store data within said first entry is valid; and

forwarding said store data as a result of said load memory operation.

28. The method as recited in claim 27 further comprising, if said store data is invalid:

selecting a store instruction tag from said first entry; and

storing said instruction tag in a second entry within said buffer, said second entry allocated to said load memory operation.

29. The method as recited in claim 28 further comprising:

selecting said load from said second entry to reprobe said data cache; and

comparing said store instruction tag in said second entry to a plurality of instruction tags in said buffer.

30. The method as recited in claim 29 further comprising, if said store instruction tag matches said store instruction tag in said first entry, forwarding said store data from said first entry as a result of said load memory operation even if said LIB indication in said first entry is in a second state indicative that said store memory operation is not youngest within said buffer to update said memory location identified by said first data address.

31. The method as recited in claim 26 further comprising, if said first memory operation comprises a second store memory operation, setting said LIB indication in said first entry to a second state indicative that said store memory operation is not youngest within said buffer to update said memory location identified by said first data address.

32. The method as recited in claim 31 further comprising setting a second LIB indication in a second entry allocated to said second store memory operation to said first state.