Cache Tag Array Segmentation for Energy Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Highly associative cache memory structures in computer systems require significant energy for parallel tag look-ups, leading to increased power consumption due to the processing power needed for multiple tag examinations.
Innovation Solution
Divide cache elements into groups and perform a serial probing process, exiting the probing process upon a tag hit to reduce unnecessary look-ups and energy usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If parallel tag look-up is used in highly associative cache memory, then cache associativity and hit rate are improved, but energy consumption increases significantly
Solution Approach 1:
The patent divides the cache tag array into multiple banks, allowing parallel look-ups to be performed across banks rather than sequentially within a single bank. This segmentation enables the system to maintain high cache associativity and hit rates while distributing the energy consumption across multiple smaller, more efficient parallel operations.
Solution Approach 2:
The patent implements dynamic control of the tag look-up process by using early exit logic that stops further look-ups within a bank once a match is found. This dynamic approach adapts the actual number of look-ups performed based on the data being searched, reducing average energy consumption while maintaining the capability for full parallel search when needed.
2Speed
If multiple tag locations are examined in parallel, then cache lookup speed is improved, but processing power and energy use increase
Solution Approach 1:
By segmenting the tag array into multiple banks, the patent enables parallel processing across banks while keeping each individual bank's processing requirements manageable. This segmentation allows the system to achieve high lookup speeds through parallelism without requiring excessive processing power in any single processing unit.
Solution Approach 2:
The patent applies partial action by implementing early exit from the tag look-up process within each bank once a match is found. This means that not all tag locations need to be fully examined in every lookup operation, reducing the average processing power required while maintaining the capability for complete parallel examination when necessary for speed.
3Use of energy by moving object
If serial probing is used instead of parallel look-up, then energy consumption is reduced, but cache lookup latency increases
Solution Approach 1:
The patent segments the tag array into multiple banks that can be probed in parallel, fundamentally avoiding the latency penalty of purely serial probing. Each bank can be searched simultaneously, maintaining low lookup latency while the early exit logic within each bank reduces energy consumption compared to examining all tags.
Solution Approach 2:
The patent merges the advantages of parallel processing (low latency) with early exit logic (low energy consumption) by combining multiple parallel bank probes with conditional termination within each bank. This hybrid approach achieves both low latency and low energy consumption simultaneously, resolving the contradiction between these two parameters.
Data Source
AI summary
A method for accessing a cache memory structure includes dividing a multiple cache elements of a cache memory structure into multiple groups. A serial probing process of the multiple groups is performed. Upon a tag hit resulting from the serial probing process, the probing process for remaining groups exits.