Method and apparatus for optimizing line writes in cache coherent systems
AssigneeIntel Corporation
InventorsSujat Jamil, Hang T. Nguyen, Samantha J. Edirisooriya, David E. Miner, R. Frank O'Bleness, Steven J. Tu
Life of this patent
Estimating the term…
Abstract
A method and apparatus for optimizing line writes in cache coherent systems. A new cache line may be allocated without loading data to fill the new cache line when a store buffer coalesces enough stores to fill the cache line. Data may be loaded to fill the line if an insufficient number of stores are coalesced to fill the entire cache line. The cache line may be allocated by initiating a read and invalidate request and asserting a back-off signal to cancel the read if there is an indication that the coalesced stores will fill the cache line.
Claims22
A method comprising:
- allocating a new cache line corresponding to a memory location without loading data associated with the memory location to fill the new cache line.
A method comprising:
- allocating a cache line;
- loading data to fill the cache line responsive to an indication that a plurality of buffered stores does not fill the cache line;
- and
- not loading data to fill the cache line responsive an indication that the plurality of buffered stores does fill the cache line.
A method comprising:
- initiating a read and invalidate request for a cache line;
- and
- asserting a back-off signal responsive to an indication that a coalesced plurality of stores fills the cache line.
An apparatus comprising:
- a store buffer to hold a coalesced plurality of stores;
- a flag associated with the store buffer to indicate whether the coalesced plurality of stores fills a cache line;
- and
- control logic coupled with the store buffer to allocate the cache line, the control logic loading data to fill the cache line responsive to said flag indicating that the coalesced plurality of stores does not fill the cache line, and the control logic not loading data to fill the cache line responsive to the flag indicating that the coalesced plurality of stores does fill the cache line.
A data processing system comprising:
- an addressable memory to store display frame data;
- a first agent including:
- a first buffer to hold a plurality of stores;
- and
- a first cache control logic coupled with the first buffer to allocate a cache line, the first cache control logic signaling to load data to fill the cache line responsive to one or more buffered stores of the plurality of stores not filling the cache line, and the first cache control logic signaling not to load data to fill the cache line responsive to a coalesced set of stores of the plurality of stores filling the cache line;
- and
- memory control logic coupled with the addressable memory and responsive to the cache control logic signaling to load data, to transfer display frame data to fill the cache line, and responsive to the cache control logic signaling not to load data, not to transfer display frame data to fill the cache line.
Dependencies are inferred from the claim text.