DSP Processor MMU Segmentation for Parallel External Memory Fetching

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing digital signal processors face inefficiencies in accessing external memory due to serial operations in memory management units, leading to increased maintenance costs and reduced system efficiency when multiple processors are integrated, and there is a risk of memory stomping.

Innovation Solution

Implementing separate memory management units for program and data ports in digital signal processors, allowing parallel fetching of instructions and data, and incorporating translation lookaside buffers (TLBs) to manage access addresses, which simplifies hardware design and supports parallel operations.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If a unified MMU is used for both program and data ports, then device complexity is reduced, but system efficiency deteriorates due to serial operations

Engineering Contradiction:
ImproveMMU structure complexityVSAvoidinstruction and data fetching efficiency
Core Design Contradiction:
Device complexityVSProductivity

Solution Approach 1:

The patent divides the unified MMU into separate program port MMU and data port MMU modules. Each MMU independently handles address translation for its respective port, enabling parallel operations. This segmentation resolves the contradiction by sacrificing some structural simplicity to achieve significant efficiency improvement through parallel processing.

Inventive Principle:
Principle #1Segmentation

2Device complexity

If multiple digital signal processors share off-chip memory space without individual allocation, then device complexity is reduced, but reliability deteriorates due to memory stomping risks

Engineering Contradiction:
Improvememory allocation scheme complexityVSAvoidmemory access safety
Core Design Contradiction:
Device complexityVSReliability

Solution Approach 1:

The patent implements separate off-chip memory space allocation for each digital signal processor through individual MMU modules. Each processor gets dedicated memory regions with isolated address translation, eliminating memory stomping risks. This segmentation approach maintains low complexity while significantly improving reliability.

Inventive Principle:
Principle #1Segmentation

3Device complexity

If program and data fetching are performed serially through a single MMU, then device complexity is reduced, but productivity deteriorates

Engineering Contradiction:
ImproveMMU module structureVSAvoidfetching operation speed
Core Design Contradiction:
Device complexityVSProductivity

Solution Approach 1:

The patent creates separate MMU modules for program and data ports, allowing simultaneous address translation and fetching operations. The program MMU handles instruction fetches while the data MMU handles data access, doubling the effective throughput. This segmentation directly addresses the productivity issue while maintaining manageable complexity.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent combines multiple MMU modules under a unified memory management architecture, allowing parallel operations while maintaining coordinated control. The merged structure enables both program and data fetching to proceed simultaneously through different MMU paths, resolving the serial operation bottleneck.

Inventive Principle:
Principle #5Merging (Combining)

Data Source

PatentEP3842913B1DSP processor and system, and access method for external storage space
Publication Date: 2025.10.15 ZTE CORP
  • EP3842913B1 patent drawingFigure 1~2
  • EP3842913B1 patent drawingFigure 3
  • EP3842913B1 patent drawingFigure 4

AI summary

A DSP processor, a system, and a method for accessing external memory space are disclosed. The DSP processor may include: a DSP core; and a program port and a data port which are connected to the DSP core and configured to access an external memory, where the program port and the data port are respectively configured to communicate with a memory management unit configured for management of an access address. Since the program port and the data port are respectively configured to communicate with a memory management unit for management of an access address, an operation of fetching instructions and an operation of fetching data can be carried out in parallel, thereby improving the efficiency of a DSP system.