Memory Manager Buffer Allocation Across Multiple DDR Interfaces

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current memory system-on-chip (SoC) designs with multiple DDR memory interfaces face challenges in efficient bandwidth and capacity utilization due to manual buffer allocation, leading to potential waste and increased complexity in on-chip interconnects, especially when usage patterns are variable.

Innovation Solution

A memory management device with a controller that automatically divides buffers equally across multiple memories, providing a virtual unified memory system and simplifying buffer allocation, thereby optimizing bandwidth and capacity usage without requiring manual software design intervention.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If manual buffer allocation is used in two DDR memory interface systems, then bandwidth capacity can be provided, but device complexity and design time increase

Engineering Contradiction:
Improvebandwidth capacityVSAvoiddesign complexity
Core Design Contradiction:
Quantity of substanceVSDevice complexity

Solution Approach 1:

The memory management unit automatically performs buffer allocation and distribution across multiple DDR memory interfaces without requiring manual software designer intervention. The system self-manages buffer allocation based on actual usage patterns, eliminating the need for manual configuration while maintaining optimal bandwidth utilization across all memory interfaces.

Inventive Principle:
Principle #25Self-service

2Quantity of substance

If manual buffer allocation is performed, then bandwidth can be managed, but adaptability to variable usage patterns deteriorates

Engineering Contradiction:
Improvebandwidth managementVSAvoidflexibility to variable usage
Core Design Contradiction:
Quantity of substanceVSAdaptability or versatility

Solution Approach 1:

The memory management unit dynamically adjusts buffer allocation based on actual runtime usage patterns rather than static manual configuration. The system automatically redistributes buffers between DDR memory interfaces according to real-time demand, providing adaptability to variable usage patterns while maintaining efficient bandwidth management throughout operation.

Inventive Principle:
Principle #15Dynamics

3Quantity of substance

If manual buffer allocation is used, then bandwidth capacity can be provided, but loss of time in design and programming increases

Engineering Contradiction:
Improvebandwidth capacityVSAvoiddesign time
Core Design Contradiction:
Quantity of substanceVSLoss of time

Solution Approach 1:

The memory management unit performs automatic buffer allocation and distribution without requiring manual software configuration. This self-service approach eliminates time-consuming manual allocation tasks during system programming, allowing developers to focus on higher-level application logic while the memory subsystem automatically optimizes its own resource distribution.

Inventive Principle:
Principle #25Self-service

4Quantity of substance

If two DDR memory interfaces are used, then bandwidth requirement can be met, but interconnect complexity increases

Engineering Contradiction:
ImprovebandwidthVSAvoidinterconnect complexity
Core Design Contradiction:
Quantity of substanceVSDevice complexity

Solution Approach 1:

The memory management unit acts as an intermediary between the processor and multiple DDR memory interfaces. It centralizes the complexity of buffer allocation and distribution in a dedicated management component, simplifying the overall interconnect architecture by abstracting the multi-memory access patterns into a unified management layer that handles routing and allocation automatically.

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentUS9164903B2Memory manager
Publication Date: 2015.10.20 STMICROELECTRONICS (RES & DEV) LTD
  • US9164903B2 patent drawing
  • US9164903B2 patent drawing
  • US9164903B2 patent drawing

AI summary

A memory management device including a plurality of outputs, each output configured to interface to respective one of a plurality of memories; and a controller configured to cause each buffer allocated to the memories to be divided up substantially equally between each of the plurality of memories.