Spiral Cache Memory Self-Organization for Low Latency Access

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current cache memory systems face challenges in balancing access time and capacity, with larger caches leading to longer access times and smaller caches resulting in higher access times for frequently accessed values, while also requiring complex routing circuits and LRU logic to manage multiple requests effectively.

Innovation Solution

A spiral cache memory architecture that dynamically organizes its contents to place the most-recently accessed values at the front-most location, using a move-to-front heuristic to minimize access latency and support multiple outstanding requests without the need for complex routing or buffering, leveraging the dimensionality of Euclidean space to reduce worst-case access time.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If the L1 cache size is increased to store more frequently-accessed values, then the number of frequently-accessed values available at short access times is improved, but the access time for typical L1 cache values increases due to physical wiring constraints and signal propagation speed limitations

Engineering Contradiction:
Improvenumber of frequently-accessed valuesVSAvoidaccess time
Core Design Contradiction:
Quantity of substanceVSLoss of time

Solution Approach 1:

The patent implements a dynamic cache organization where cache lines are not statically positioned but are actively moved and repositioned based on access patterns. The spiral arrangement allows cache lines to be dynamically shifted along the spiral path, with recently accessed values moving toward the beginning of the spiral, thereby adapting the cache structure to usage patterns and maintaining fast access times even as cache size increases

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The patent transitions from traditional linear or two-dimensional cache arrays to a spiral organization that effectively utilizes three-dimensional spatial arrangement. This spiral configuration allows for increased cache capacity while maintaining shorter average access paths by wrapping the cache structure in a spiral pattern, reducing the maximum distance any cache line must travel

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

2Loss of time

If the L1 cache size is reduced to decrease access time, then the access time for typical L1 cache values is improved, but the number of frequently-accessed values that can be stored in the L1 cache decreases

Engineering Contradiction:
Improveaccess timeVSAvoidnumber of frequently-accessed values
Core Design Contradiction:
Loss of timeVSQuantity of substance

Solution Approach 1:

The dynamic reorganization mechanism ensures that the most frequently accessed values are always positioned at the beginning of the spiral where access is fastest. This dynamic adaptation allows a smaller cache to effectively store and quickly access the critical few frequently-used values, maximizing the utility of limited cache space without requiring a large static cache structure

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The patent applies different access characteristics to different regions of the cache. The beginning of the spiral provides fast access for frequently-used values, while later portions of the spiral can accommodate less frequently accessed values. This creates a non-uniform access pattern optimized for workload characteristics, allowing small cache sizes to deliver high performance for critical data

Inventive Principle:
Principle #3Local quality

3Productivity

If traditional cache control algorithms with LRU logic and cast-out mechanisms are used to manage cache organization, then the most-frequently-accessed values can be maintained in lower-order caches, but the device complexity increases due to additional access measurement and control structures

Engineering Contradiction:
Improvecache hit rateVSAvoidcontrol structure complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The cache system performs self-organization through the natural dynamics of the spiral structure and access patterns. When a cache line is accessed, it is automatically repositioned along the spiral path based on its access recency, eliminating the need for external control logic to track and manage cache line positions. The structure itself embodies the LRU policy through its geometric organization and access mechanism

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The patent removes the complex LRU control logic, cast-out mechanisms, and access measurement structures from the cache system. Instead of using traditional control algorithms to manage cache organization, the invention extracts these control functions and replaces them with the inherent spatial organization and access dynamics of the spiral structure, significantly reducing device complexity while maintaining cache effectiveness

Inventive Principle:
Principle #2Taking out (Extraction)

4Productivity

If multiple requests are pipelined through traditional cache architectures, then throughput can be improved, but fixed worst-case access latencies and buffering are required to control the flow of pipelined information

Engineering Contradiction:
ImprovethroughputVSAvoidbuffering and flow control complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The spiral cache's dynamic reorganization capability allows it to naturally accommodate pipelined requests without requiring fixed timing structures. As requests are processed, cache lines are dynamically repositioned based on actual access patterns rather than predetermined schedules, allowing the system to handle variable-latency operations efficiently and support multiple outstanding requests without complex flow control

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The cache structure itself manages the flow of pipelined information through its spiral organization and dynamic repositioning mechanism. The geometric structure and access patterns naturally regulate data flow and timing, eliminating the need for external buffering and flow control circuits that would otherwise be required to manage pipelined operations in traditional architectures

Inventive Principle:
Principle #25Self-service

Data Source

PatentEP2353095B1A spiral cache memory and method of operating a spiral cache memory
Publication Date: 2017.06.14 INTERNATIONAL BUSINESS MACHINE CORPORATION
  • EP2353095B1 patent drawingFigure 1A~1C
  • EP2353095B1 patent drawingFigure 2
  • EP2353095B1 patent drawingFigure 3

AI summary

A spiral cache memory provides reduction in access latency for frequently-accessed values by self-organizing to always move a requested value to a front-most central storage element of the spiral. The occupant of the central location is swapped backward, which continues backward through the spiral until an empty location is swapped-to, or the last displaced value is cast out of the last location in the spiral. The elements in the spiral may be cache memories or single elements. The resulting cache memory is self-organizing and for the one- dimensional implementation has a worst-case access time proportional to N, where N is the number of tiles in the spiral. A k-dimensional spiral cache has a worst-case access time proportional to N1/k. Further, a spiral cache system provides a basis for a non-inclusive system of cache memory, which reduces the amount of space and power consumed by a cache memory of a given size.