Cluster Memory Broadcast Control for Multi-PE Shared Access

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Efficiently controlling access to a common shared memory by a cluster of multiple processing units or processing elements in complex data processing systems is challenging, particularly in terms of time and energy efficiency, especially when multiple processing elements need to access data simultaneously without causing data collisions.

Innovation Solution

A Cluster Controller is introduced to manage memory access by a cluster of processing elements, initiating data broadcasting only after receiving requests from all participating elements, and processing elements are configured to interwork with this controller to enable memory access, allowing simultaneous data transfers without synchronization.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If multiple processing elements access common memory simultaneously using sequential control, then data collision risk is reduced, but time efficiency and productivity deteriorate

Engineering Contradiction:

Improvedata collision avoidanceVSAvoiddata transfer efficiency

Core Design Contradiction:

ReliabilityVSProductivity

Solution Approach 1:

A cluster controller is introduced as an intermediary between processing elements and common memory. The controller receives broadcast requests from multiple processing elements, manages the broadcasting process, and coordinates data transfers. This mediator enables simultaneous access by multiple processing elements while maintaining data integrity and avoiding collisions through centralized control logic.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The cluster controller waits to initiate broadcasting only after receiving requests from all participating processing elements. This preliminary action ensures that all necessary data is prepared and ready before the broadcast begins, allowing multiple processing elements to proceed simultaneously without waiting for sequential authorization, thereby improving productivity while maintaining reliability.

Inventive Principle:

Principle #10Preliminary action

2Reliability

If processing elements are synchronized for memory access, then data collision risk is reduced, but time efficiency deteriorates due to waiting

Engineering Contradiction:

Improvedata access safetyVSAvoidsynchronization waiting time

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

Processing elements independently send broadcast requests to the cluster controller without waiting for synchronization signals from other processing elements. Each processing element operates autonomously, initiating its data transfer when ready, while the cluster controller manages the actual broadcasting. This eliminates synchronization waiting time while maintaining data access safety through the controller's coordination.

Inventive Principle:

Principle #25Self-service

3Reliability

If sequential data transfer control is used, then data collision risk is reduced, but energy efficiency deteriorates due to repeated access control operations

Engineering Contradiction:

Improvedata transfer integrityVSAvoidenergy consumption per data transfer

Core Design Contradiction:

ReliabilityVSUse of energy by moving object

Solution Approach 1:

Multiple data transfer operations are merged into a single broadcast operation managed by the cluster controller. Instead of each processing element performing separate access control sequences, the controller consolidates these operations into one coordinated broadcast that serves all processing elements simultaneously. This reduces the total number of control operations and associated energy consumption while maintaining data transfer integrity.

Inventive Principle:

Principle #5Merging (Combining)

Data Source

PatentUS20260010492A1Controlling access to a memory shared by a cluster of multiple processing elements

Publication Date: 2026.01.08 TELESIS INNOVATION AB

AI summary

There is provided a Cluster Controller, CC, configured to control memory access by a cluster of multiple processing units or processing elements, PEs, to a common Cluster Memory, CM, shared by the processing elements within the cluster. The Cluster Controller is configured to receive requests, referred to as cluster broadcast requests, from at least a subset of the multiple processing elements within the cluster for broadcasting of data from the common cluster memory to said at least a subset of the multiple processing elements. The Cluster Controller is configured to initiate broadcasting of data in response to the received cluster broadcast requests only after broadcast requests have been received from all processing elements of said at least a subset of said multiple processing elements that are participating in the broadcast.