PLD Parallelism Adjustment for Bus Bandwidth Bottleneck

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

In information processing systems with FPGAs, a bottleneck on the bus bandwidth can occur due to the dynamic reconfiguration and operation of multiple logic circuits, leading to decreased parallelism and increased execution times, as the existing methods do not efficiently manage the degree of parallelism across different logic circuits.

Innovation Solution

An information processing apparatus that includes a processor and a programmable logic circuit device (PLD) with a reconfiguration region, where the processor compares execution times and adjusts the degree of parallelism by decreasing it for one logic circuit and increasing it for another, within the limits of the bus bandwidth, to optimize resource utilization and prevent bottlenecks.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If multiple logic circuits are dynamically reconfigured and operated in parallel in the FPGA, then the processing capability and functionality are improved, but a bottleneck is generated on the bus band between the FPGA and memory

Engineering Contradiction:
Improveprocessing capabilityVSAvoidbus band bottleneck
Core Design Contradiction:
ProductivityVSObject-generated harmful factors

Solution Approach 1:

The patent implements dynamic adjustment of the degree of parallelism for logic circuits based on real-time bus band utilization monitoring. When the bus band utilization reaches a threshold, the system dynamically reduces the parallelism degree of selected logic circuits to prevent bottlenecks, thereby maintaining system productivity while avoiding harmful bus band saturation

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The system changes the parameter of parallelism degree (an operational parameter) of logic circuits in response to bus band conditions. By adjusting this parameter dynamically, the system optimizes the balance between processing capability and bus band utilization, preventing bottlenecks while maintaining high productivity

Inventive Principle:
Principle #35Parameter changes

2Loss of time

If the degree of parallelism of logic circuits is increased to maintain processing speed, then the execution time is reduced, but the bus band utilization increases and may reach the upper limit

Engineering Contradiction:
Improveexecution timeVSAvoidbus band saturation
Core Design Contradiction:
Loss of timeVSObject-generated harmful factors

Solution Approach 1:

The patent implements a feedback mechanism where the system monitors bus band utilization and executes time continuously. Based on this feedback, the control unit adjusts the parallelism degree of logic circuits to maintain optimal execution time while preventing bus band saturation, creating a closed-loop control system

Inventive Principle:
Principle #23Feedback

3Reliability

If the bus band utilization is reduced to prevent bottlenecks, then the system stability is improved, but the degree of parallelism and processing speed decrease

Engineering Contradiction:
Improvesystem stabilityVSAvoidprocessing speed
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The system performs periodic monitoring of bus band utilization and executes time, and adjusts the parallelism degree in periodic cycles. This periodic action allows the system to maintain stability by preventing bottlenecks while recovering processing speed when bus band conditions improve

Inventive Principle:
Principle #19Periodic action

Data Source

PatentUS10534621B2Information processing apparatus, PLD management program and PLD management method
Publication Date: 2020.01.14 FUJITSU LTD
  • US10534621B2 patent drawing
  • US10534621B2 patent drawing
  • US10534621B2 patent drawing

AI summary

An information processing apparatus has a processor and a programmable logic circuit device (PLD) that includes a reconfiguration region to configure a logic circuit requested by a configuration request from the processor. The processor compares a first execution time of a plurality of the logic circuits for a case when a degree of parallelism adjustment is performed by decreasing a degree of parallelism of a first logic circuit and increasing a degree of parallelism of a second logic circuit and a second execution time of the plurality of logic circuits for a case when the degree of parallelism adjustment is not performed, and requests the degree of parallelism adjustment to the PLD when the first execution time is shorter than the second execution time, and does not request the degree of parallelism adjustment to the PLD when the first execution time is not shorter than the second execution time.