Back to results
US 7,406,568 B2Granted
Buffer allocation for split data messages
AssigneeIntel Corporation
InventorBenjamin Tsien
Life of this patent
Estimating the term…
Abstract
A technique to store a plurality of addresses and data to address and data buffers, respectively, in an ordered manner. More particularly, one embodiment of the invention stores a plurality of addresses to a plurality of address buffer entries and a plurality of data to a plurality of data buffer entries according to a true least-recently-used (LRU) allocation algorithm.
Claims30
An apparatus comprising:
- an address buffer to store a first address in a first address buffer entry;
- a data buffer coupled to the address buffer to store a first data in a first data buffer entry;
- allocation logic to perform a true least-recently-used (LRU) buffer entry allocation algorithm to allocate either or both of the first address buffer entry and the first data buffer entry within the first address buffer and the first data buffer, respectively,
- wherein each of the address buffer and data are to be logically or physically partitioned into a plurality of partitions, wherein each partition of the address buffer is store address information generated by a select one of a plurality of cache agents and each partition of the data buffer is to store data generated by a select one of the plurality of cache agents.
1 more element
A system comprising:
- a first processor including address allocation logic to allocate address entries among a first group of address buffer entries in an address buffer according to a first true least-recently-used (LRU) allocation algorithm independently of data entries to be allocated among a first group of data buffer entries in a data buffer according to a second LRU allocation algorithm;
- a target recipient device to receive messages generated by the first processor,
- wherein each of the address buffer and data are to be logically or physically partitioned into a plurality of partitions, wherein each partition of the address buffer is to store address information generated by a select one of a plurality of agents and each partition of the data buffer is to store data generated by a select one of the plurality of agents.
A method comprising:
- generating a plurality of addresses and data;
- allocating entries within an address buffer according to a true least-recently-used (LRU) allocation algorithm to cause a least-recently de-allocated address buffer entry to be allocated;
- allocating entries within a data buffer according to the true LRU allocation algorithm to cause a least-recently de-allocation data buffer entry to be allocated;
- merging at least one of the plurality of data with at least one of the plurality of addresses into at least one of a plurality of messages,
- wherein each of the address buffer and data are to be logically or physically partitioned into a plurality of partitions, wherein each partition of the address buffer is to store address information generated by a select one of a plurality of cache agents and each partition of the data buffer is to store generated by a select one of the plurality of cache agents.
2 more elements
A processor comprising:
- an address allocation logic to allocate least-recently de-allocated address buffer entries in an address buffer in order to store a plurality of addresses;
- a data allocation logic to allocate least-recently de-allocated data buffer entries in a data buffer in order to store a plurality of data;
- an arbitration logic to de-allocate the least-recently used address buffer entries and the least-recently used data buffer entries and update least-recently-used state information in the address allocation logic and the data allocation logic;
- merge logic to combine the plurality of addresses and the plurality of data into a plurality of messages;
- a plurality of cache agents, at least one of which is to generate the plurality of addresses and the plurality of data to be stored in the address buffer entries and the data buffer entries, respectively,
- wherein each of the address buffer and data are to be logically or physically partitioned into a plurality of partitions, wherein each partition of the address buffer is to store address information generated by a select one of the plurality of cache agents and each partition of the data buffer is to store data generated by a select one of the plurality of cache agents.
3 more elements
Dependencies are inferred from the claim text.