Cache Line Replacement via Segmented Access Tracking

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current cache management systems incur significant system overheads and low replacement efficiency due to the need to traverse all cache lines to select a replacement line, which affects overall system performance.

Innovation Solution

A cache management method where a cache controller allocates a tag area to each cache line, tracks access frequency and time, and divides access frequency counts into segments to pre-select a set of cache lines for replacement, reducing the need for extensive traversal and improving replacement efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If the controller traverses all original cache lines to select a to-be-replaced cache line based on access counters, then the cache replacement decision is made with statistical accuracy, but the system overhead increases and replacement efficiency decreases

Engineering Contradiction:
Improvecache replacement decision accuracyVSAvoidcache line replacement efficiency
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The patent segments the cache lines into multiple groups based on their access patterns and characteristics. Each group is managed independently with its own replacement policy, avoiding the need to traverse all cache lines globally. This segmentation reduces the search space for replacement candidates while maintaining accurate replacement decisions within each segment.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary classification of cache lines into groups based on access frequency and patterns before the actual replacement decision is needed. By pre-organizing cache lines into manageable groups with distinct access characteristics, the system prepares replacement candidates in advance, reducing the computational overhead during actual replacement operations.

Inventive Principle:
Principle #10Preliminary action

2Quantity of substance

If the cache stores more cache lines to increase caching capacity, then the caching amount increases, but the time to traverse and select replacement lines increases

Engineering Contradiction:
Improvecaching amountVSAvoidreplacement selection time
Core Design Contradiction:
Quantity of substanceVSLoss of time

Solution Approach 1:

By dividing the large cache into multiple smaller groups, the patent enables the system to store more cache lines overall while keeping the traversal scope for each group limited. The segmentation allows parallel or independent management of multiple groups, reducing the time penalty associated with increased cache capacity.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent applies different replacement strategies to different groups, using more aggressive replacement policies for groups with lower access frequency. This partial action approach allows the system to efficiently manage large cache capacity by focusing intensive replacement operations only on specific groups that need it, rather than uniformly processing all cache lines.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP3388935B1Cache management method, cache controller and computer system
Publication Date: 2021.09.29 HUAWEI TECH CO LTD
  • EP3388935B1 patent drawingFigure 1
  • EP3388935B1 patent drawingFigure 2
  • EP3388935B1 patent drawingFigure 3

AI summary

A cache management method, a cache controller, and a computer system are provided. In the method, the cache controller obtains an operation instruction; when a destination address in the operation instruction hits no cache line cache line in a cache of the computer system, and the cache includes no idle cache line, the cache controller selects a to-be-replaced cache line from a replacement set, where the replacement set includes at least two cache lines; and the cache controller eliminates the to-be-replaced cache line from the cache, and stores, in the cache, a cache line obtained from the destination address. According to the cache management method, system overheads of cache line replacement can be reduced, and cache line replacement efficiency can be improved.