System, apparatus and method for performing look-ahead lookup on predictive information in a cache memory

A data return cache for managing predictive accesses to a memory comprising:

• a short term cache memory configured to store predictions having an age less than a threshold;
• a long term cache memory configured to store predictions having an age greater than or equal to said threshold, said long term cache memory having more memory capacity than said short term cache;
• and
• an interface configured to detect in parallel whether multiple predictions are stored in either said short term cache memory or said long term cache memory, or both,
• wherein said interface uses at least two representations of each of said multiple predictions when examining said short term cache and said long term cache memories.

A method for managing predictions to access a memory comprising:

• generating sequential and nonsequential predictions;
• examining a subset of said predictions against multiple cache memories in parallel, said multiple cache memories including a short term cache and a long term cache;
• and
• disregarding a prediction of said subset if said prediction is stored in any of said multiple cache memories.

A computing system configured to manage accesses to a memory comprising:

• a central processing unit for executing program instructions and processing program data;
• a system memory containing said program instructions and said processing program data;
• and
• a memory processor coupled to said central processing unit and said system memory and configured to control predictive accesses to said system memory, said memory processor including
• a nonsequential predictor configured to generate nonsequential predictions,
• a short term cache memory, each entry having an age less than a threshold,
• a long term cache memory receiving data only from said short term cache, each entry of said long term cache having an age equal to or greater than said threshold,
• a cache filter configured to compare a subset of said nonsequential predictions in parallel against the contents of both said short term and long term cache memories, and
• a cache memory manager configured to communicate a prediction identifier to said nonsequential predictor when said central processing unit requests predictive information associated with said prediction identifier.