Store instruction ordering for multi-core processor
View Patent ↗A method and apparatus for minimizing stalls in a pipelined processor is provided. Instructions in an out-of-order instruction scheduler are executed in order without stalling the pipeline by sending store data to external memory through an ordering queue.
1. A processor comprising:
a write buffer having a plurality of write buffer entries for storing data to be stored in external memory; and
a system interface, the system interface comprising:
a first logic circuit that sends data from write buffer entries to external memory; and
a second logic circuit which upon detecting execution of a store ordering instruction ensures that earlier stores complete before subsequent stores complete, the first and second logic circuits further configured to prevent overflow of an ordering queue and to eliminate stalling of any store ordering instructions at the system interface.
2. The processor of claim 1 , wherein the second logic circuit comprises ordering queue logic that ensures that the write buffer addresses are sent by monitoring a commit indicator returned from external memory for each write buffer address.
3. The processor of claim 1 , wherein upon detecting active write buffer entries in the write buffer, the system interface allocates an ordering queue entry and identifies the active write buffer entries in the ordering queue entry.
4. The processor of claim 1 , wherein upon detecting no active write buffer entries in the write buffer and at least one ordering queue entry in the ordering queue, the system interface sets a store ordering instruction execution indicator in a last ordering queue entry.
5. The processor of claim 1 , wherein upon detecting no active write buffer entries in the write buffer and no ordering queue entries in the ordering queue, the system interface allocates an ordering queue entry and sets a store ordering instruction execution indicator in the allocated ordering queue entry.
6. The processor of claim 1 wherein the write buffer has N write buffer entries and the ordering queue has N+1 ordering queue entries.
7. The processor of claim 6 , wherein N=16.
8. The processor of claim 1 , wherein the write buffer entry includes a buffer for storing data to be stored in external memory and a mask having a bit corresponding to each byte in the buffer.
9. The processor of claim 1 , wherein an ordering queue entry includes a store ordering instruction execution indicator and an ordering mask having a bit corresponding to each write buffer entry.
10. A method for ordering store data in a processor comprising:
storing data to be stored in external memory in a write buffer entry in a write buffer;
sending data stored in the write buffer to the external memory; and
upon detecting execution of a store ordering instruction, ensuring that earlier stores complete before subsequent stores complete, preventing overflow of an ordering queue and eliminating stalling of any store ordering instructions at a system interface of the processor.
11. The method of claim 10 , wherein ensuring comprises monitoring a commit indicator returned from external memory for each write buffer address.
12. The method of claim 10 further comprising:
upon detecting active write buffer entries in the write buffer, allocating an ordering queue entry in the ordering queue and identifying the active write buffer entries in the ordering queue entry.
13. The method of claim 10 further comprising:
upon detecting no active write buffer entries in the write buffer and at least one ordering queue entry in an ordering queue, setting a store ordering instruction execution indicator in a last ordering queue entry.
14. The method of claim 10 further comprising:
upon detecting no active write buffer entries in the write buffer and no ordering queue entries in the ordering queue, allocating an ordering queue entry and setting a store ordering instruction execution indicator in the allocated ordering queue entry.
15. The method of claim 10 , wherein the write buffer entry includes a buffer for storing data to be stored in external memory and a mask having a bit corresponding to each byte in the buffer.
16. The method of claim 10 , wherein an ordering queue entry in the ordering queue includes a store ordering instruction execution indicator and an ordering mask having a bit corresponding to each write buffer entry.
17. The method of claim 10 , wherein the write buffer has N write buffer entries and the ordering queue has N+1 ordering queue entries.
18. The method of claim 17 , wherein N=16.
19. A network services processor comprising:
cache memory; and
at least one processor core, the processor core comprising:
a write buffer having a plurality of write buffer entries for storing data to be stored in the cache memory; and
a system interface, the system interface comprising:
a first logic circuit sending data from write buffer entries to the cache memory; and
a second logic circuit which upon detecting execution of a store ordering instruction ensures that earlier stores complete before subsequent stores complete, the first and second logic circuits further configured to prevent overflow of an ordering queue and eliminating stalling of any store ordering instructions at the system interface.