Accelerator Offload Device CPU Core Segmentation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional methods for acquiring operation results from accelerators in offloading techniques suffer from increased processing time due to interrupt overhead in the interrupt method and wasteful CPU resource usage in the polling method, leading to reduced CPU utilization efficiency.

Innovation Solution

An accelerator offload device is designed with separate CPU cores for request processing and response processing, allowing for efficient offloading of specific application program processing to an accelerator, thereby reducing interrupt overhead and improving CPU utilization.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Speed

If the interrupt method is used to acquire operation results from accelerators, then the processing speed is improved, but the processing time increases due to interrupt overhead

Engineering Contradiction:
Improveprocessing speedVSAvoidprocessing time
Core Design Contradiction:
SpeedVSLoss of time

Solution Approach 1:

The patent divides the CPU into multiple cores, with dedicated accelerator management cores separate from application processing cores. This segmentation allows interrupt handling to occur on management cores without blocking application processing on other cores, thereby reducing the impact of interrupt overhead on overall processing time while maintaining fast response to accelerator completion events.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces a ring buffer as an intermediary data structure between the accelerator and CPU. The ring buffer acts as a buffer zone that decouples the accelerator's operation completion from the CPU's interrupt handling, allowing the CPU to process interrupts at its own pace without directly blocking the accelerator pipeline, thus reducing processing time overhead.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Loss of time

If the polling method is used to acquire operation results from accelerators, then the processing time is reduced, but CPU resource usage increases wastefully

Engineering Contradiction:
Improveprocessing timeVSAvoidCPU resource usage
Core Design Contradiction:
Loss of timeVSUse of energy by moving object

Solution Approach 1:

Instead of continuous polling, the system uses periodic interrupt-driven notification where the accelerator controller notifies the CPU only when operation results are ready. This periodic action occurs only when necessary (upon completion), eliminating the wasteful continuous CPU resource consumption of polling while maintaining fast response times comparable to polling.

Inventive Principle:
Principle #19Periodic action

Solution Approach 2:

The accelerator controller autonomously manages the notification mechanism, generating interrupts only when operation results are ready. This self-service approach eliminates the need for continuous CPU intervention to check status, reducing CPU resource usage while maintaining efficient processing time through event-driven notification.

Inventive Principle:
Principle #25Self-service

3Productivity

If separate CPU cores are used for request and response processing, then CPU utilization efficiency is improved, but device complexity increases

Engineering Contradiction:
ImproveCPU utilization efficiencyVSAvoiddevice complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The dedicated accelerator management cores are designed with multi-functionality, handling both interrupt reception and operation result acquisition tasks. This universality allows the system to improve CPU utilization through core specialization without proportionally increasing overall device complexity, as the management cores perform multiple functions within the accelerator management subsystem.

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

Data Source

PatentUS20250181401A1Accelerator offload device, accelerator offload system and accelerator offload method
Publication Date: 2025.06.05 NT T INC
  • US20250181401A1 patent drawing
  • US20250181401A1 patent drawing
  • US20250181401A1 patent drawing

AI summary

An accelerator offload device that offloads specific processing of an application program to an accelerator includes: a request-related processing part; a request I/O part; a response I/O part; and a response-related processing part. The request-related processing part is configured to perform predetermined processing required before performing offloading to the accelerator and then notify the request I/O part of a request to perform offloading. The request I/O part is configured to operate on a first CPU core to perform request processing of notifying the accelerator of an offload request. The response I/O part is configured to operate on a second CPU core different from the first CPU core to perform response processing of notifying the response-related processing part of operation completion of the accelerator. The response-related processing part is configured to perform an operation described in the application program by using an operation result of the accelerator.