Multi-Way Read Logic for CAM Tag Arrays

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

High-performance CPU architectures face challenges in achieving efficient tag searches in content addressable memories (CAMs) due to the adverse impact of tag compression on power, performance, and area, particularly in high bus width applications where higher order tag bits rarely change, leading to timing penalties and inefficiencies in tag array operations.

Innovation Solution

The proposed solution involves a method and architecture for multi-way reads in highly associative CAM arrays, utilizing a tag array with multi-way read logic and conditional read logic to efficiently read and compare subsets of stored tag bits, reducing the need for duplicated peripheral circuits and dynamic energy consumption by employing a static read port and match line receiver circuit.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Area of stationary object

If tag compression is used in CAM arrays, then area is reduced, but timing performance deteriorates due to lookup penalties

Engineering Contradiction:
Improvetag array areaVSAvoidtag search timing
Core Design Contradiction:
Area of stationary objectVSSpeed

Solution Approach 1:

The patent divides the tag array into multiple segments or ways, allowing parallel read operations on different segments simultaneously. This segmentation enables the system to achieve both area efficiency through compression and timing performance through parallel processing, resolving the contradiction between compact storage and fast access.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent implements dynamic control of read operations, where the system can selectively activate different read paths and control logic based on operational requirements. This dynamic approach allows the tag array to adapt between compressed and uncompressed read modes, optimizing both area utilization and timing performance for different search scenarios.

Inventive Principle:
Principle #15Dynamics

2Productivity

If multi-way reads are implemented, then search parallelism is improved, but area increases due to duplicated peripheral circuits

Engineering Contradiction:
Improvesearch parallelismVSAvoidperipheral circuit area
Core Design Contradiction:
ProductivityVSArea of stationary object

Solution Approach 1:

The patent designs universal read ports and control logic that can serve multiple ways simultaneously. Instead of duplicating peripheral circuits for each way, the same read port infrastructure is shared across multiple ways through time-multiplexed or space-multiplexed access, enabling multi-way reads without proportionally increasing peripheral circuit area.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The patent merges the read control logic for multiple ways into a unified control structure. By combining the control signals and read ports that serve different ways, the system achieves multi-way parallel search capability while minimizing the total area of peripheral circuits through consolidation rather than duplication.

Inventive Principle:
Principle #5Merging (Combining)

3Area of stationary object

If higher order tag bits are compressed, then area is saved, but power consumption increases due to additional lookup operations

Engineering Contradiction:
Improvetag storage areaVSAvoidpower consumption
Core Design Contradiction:
Area of stationary objectVSUse of energy by stationary object

Solution Approach 1:

The patent performs preliminary organization of compressed tag bits during the write operation, pre-positioning them in a format that facilitates efficient parallel comparison. This preliminary action eliminates the need for additional dynamic lookup operations during reads, reducing power consumption while maintaining area efficiency through compression.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent creates static copies of compressed higher order tag bits that are held in a buffer or latch structure. These copied bits are then compared in parallel without requiring repeated lookup operations, thereby reducing dynamic power consumption while maintaining the area benefits of compression.

Inventive Principle:
Principle #26Copying

Data Source

PatentEP3519973B1Area efficient architecture for multi way read on highly associative content addressable memory (CAM) arrays
Publication Date: 2021.12.01 QUALCOMM INC
  • EP3519973B1 patent drawingFigure 1
  • EP3519973B1 patent drawingFigure 2
  • EP3519973B1 patent drawingFigure 3

AI summary

Aspects disclosed herein relate to techniques and an efficient architecture for enabling multi-way reads on highly associative content addressable memory (CAM) arrays. For example, a method for performing a tag search of a tag array can include reading a first subset of stored tag bits from multiple entries of the tag array, and comparing a second subset of stored tag bits from a one or more entries of the tag array against a search-tag to produce one or more possible way hit signals.