TLB Replacement Policy for Large Page Persistence

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Translation lookaside buffers (TLBs) face inefficiencies in managing page sizes, leading to increased latency and reduced performance due to the possibility of smaller pages evicting larger page translations, which are more persistent and cover a larger portion of system memory.

Innovation Solution

The method involves updating the replacement policy data to indicate that TLB entries for larger pages are not the least-recently-used, thereby preventing smaller pages from evicting them, ensuring that translations for larger pages remain persistent and service a larger portion of system memory, reducing misses and latency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If the TLB uses standard LRU replacement policy, then smaller pages can be replaced when cache space is full, but larger page translations are evicted prematurely reducing cache effectiveness

Engineering Contradiction:
Improvecache coverageVSAvoidtranslation latency
Core Design Contradiction:
Adaptability or versatilityVSLoss of time

Solution Approach 1:

The patent applies local quality by differentiating replacement behavior based on page size. Larger pages receive preferential treatment in the replacement policy, ensuring they are not evicted by smaller pages. The LRU data is updated differently depending on whether the accessed page is large or small, creating localized quality differences in cache management that prioritize large page translations.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The patent implements dynamics by making the replacement policy adaptive rather than static. The LRU data is dynamically updated based on page size characteristics - when a large page is accessed, the LRU data is updated to prevent its eviction, while small page accesses follow standard LRU behavior. This dynamic adjustment allows the cache to adapt to the importance of different page sizes.

Inventive Principle:
Principle #15Dynamics

2Adaptability or versatility

If the TLB prioritizes larger page translations, then cache coverage increases, but replacement policy complexity increases

Engineering Contradiction:
Improvecache coverageVSAvoidreplacement policy complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent applies segmentation by dividing the replacement policy into distinct handling paths for large and small pages. The system segments cache management based on page size characteristics, with separate update rules for LRU data depending on whether the accessed page is large or small. This segmentation simplifies the overall complexity by creating clear, rule-based distinctions rather than requiring complex unified logic.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP3497577B1Updating least-recently-used data for greater persistence of higher-generality cache entries
Publication Date: 2023.05.31 ADVANCED MICRO DEVICES INC
  • EP3497577B1 patent drawingFigure 1
  • EP3497577B1 patent drawingFigure 2
  • EP3497577B1 patent drawingFigure 3

AI summary

Techniques for improving translation lookaside buffer (TLB) operation are disclosed. A particular entry of the TLB is to be updated with data associated with a large page size. The TLB updates replacement policy data for that TLB entry for that large page size to indicate that the TLB entry is not the least-recently-used. To prevent smaller pages from evicting the TLB entry for the large page size, the TLB also updates replacement policy data for that TLB entry for the smaller page size to indicate that the TLB entry is not the least-recently-used.