Cache Associativity via Iterative Set Lookup
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Direct mapping in storage structures like caches can lead to data eviction issues, particularly in transactional environments where associativity is limited, resulting in increased circuitry requirements and power consumption due to higher lookup pressures.
Innovation Solution
Implementing a set-associative storage structure with lookup circuitry that iterates through candidate sets using an access policy to select an appropriate storage location, effectively increasing associativity beyond physical ways by utilizing pseudo-ways, and managing transactional data through read and write indicators.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If the associativity of the storage structure is directly increased to provide more alternative storage locations, then the ability to store transactional data is improved, but the amount of circuitry required and power consumption increase
Solution Approach 1:
The patent segments the storage structure into multiple candidate sets that can be iterated through sequentially. Instead of providing all alternative storage locations simultaneously with full parallel lookup capability, the system divides the search space into segments (candidate sets) and processes them in sequence, reducing the circuitry required for parallel comparisons while maintaining the ability to access multiple storage locations.
Solution Approach 2:
The patent introduces dynamic iteration through candidate sets, where the lookup process adapts by sequentially examining different candidate sets based on transactional indicators. This dynamic approach allows the system to effectively increase associativity by exploring multiple storage locations over time rather than requiring all locations to be simultaneously accessible through static parallel circuitry.
2Adaptability or versatility
If the associativity of the storage structure is directly increased to provide more alternative storage locations, then the ability to store transactional data is improved, but the number of parallel lookups increases causing timing pressures
Solution Approach 1:
The patent segments the lookup process into iterative steps through different candidate sets. By dividing the search into manageable segments that can be processed sequentially, the system avoids the timing pressures of simultaneous parallel lookups across all possible storage locations, while still achieving high effective associativity through the iterative search process.
Solution Approach 2:
The patent uses read and write indicators to preliminarily identify candidate sets that are likely to contain the required data. This preliminary action filters the search space before the actual lookup occurs, reducing the number of candidate sets that need to be examined and thereby reducing timing pressure while maintaining high associativity for transactional data.
3Reliability
If set-associative storage structure is used to resolve data eviction issues, then data retention capability is improved, but the device complexity increases
Solution Approach 1:
The patent segments the associativity function into multiple candidate sets with iterative access. This segmentation approach achieves the data retention capability of high-associativity structures without requiring the complex simultaneous comparison circuitry of traditional set-associative caches, thereby reducing device complexity while maintaining reliability.
Solution Approach 2:
The patent introduces read and write indicators as intermediary elements that mediate between the storage locations and the lookup process. These indicators guide the iterative search through candidate sets, enabling the system to achieve high effective associativity and data retention capability without the complexity of full set-associative structures, as the intermediaries manage the search process efficiently.
Data Source
AI summary
There is provided an apparatus that includes storage circuitry. The storage circuitry is made up from a plurality of sets, each of the sets having at least one storage location. Receiving circuitry receives an access request that includes an input address. Lookup circuitry obtains a plurality of candidate sets that correspond with an index part of the input address. The lookup circuitry determines a selected storage location from the candidate sets using an access policy. The access policy causes the lookup circuitry to iterate through the candidate sets to attempt to locate an appropriate storage location. The appropriate storage location is accessed in response to the appropriate storage location being found.