Cache Array Power Reduction via Subset Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The high power consumption in processor caches due to multi-set associativity, where multiple cache sets are accessed in parallel, leading to unnecessary power usage as only one set typically contains the desired data, becomes more significant with increasing processing cores.
Innovation Solution
A method and system where cache sets are grouped into subsets, allowing only the relevant subset to be accessed based on a set selection signal, thereby reducing active switching power by gating non-identified cache sets and preventing unnecessary access.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If multiple cache sets are accessed in parallel to reduce access latency and improve performance, then the processor performance is improved, but the power consumption increases significantly
Solution Approach 1:
The cache array is segmented into multiple subsets, where each subset contains a portion of the total cache sets. Instead of accessing all cache sets in parallel, the system segments the access operation to only activate and read from the relevant subset identified by the set selection signal, thereby reducing power consumption while maintaining performance
Solution Approach 2:
Different subsets of cache sets are activated based on the specific access requirements. The system applies local quality by making each subset accessible only when needed, rather than uniformly accessing all cache sets, thus optimizing power consumption according to actual access patterns
2Reliability
If all cache sets are accessed simultaneously to ensure data is found, then the probability of finding desired data is maximized, but unnecessary power is wasted on cache sets that do not contain the desired instruction or data
Solution Approach 1:
The set selection signal is generated in advance to identify which subset of cache sets contains the desired data before the actual read operation begins. This preliminary identification action allows the system to pre-activate only the necessary subset, avoiding the need to access all cache sets and thereby preventing energy waste while ensuring the data can be found
Data Source
AI summary
Embodiments of the disclosure include a cache array having a plurality of cache sets grouped into a plurality of subsets. The cache array also includes a read line configured to receive a read signal for the cache array and a set selection line configured to receive a set selection signal. The set selection signal indicates that the read signal corresponds to one of the plurality subsets of the cache array. The read line and the set selection line are operatively coupled to the plurality of cache sets and based on the set selection signal the subset that corresponds to the set selection signal is switched.


