ML Runtime Switching Wrappers for Processor Bottlenecks

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing information handling systems face processing bottlenecks when executing machine learning models due to high resource consumption, especially when running multiple software applications like gaming or computer-aided design, which require significant hardware processing resources.

Innovation Solution

A method for runtime switching between machine learning model algorithms using a swappable wrapper generator and inference runtime control module to dynamically select and switch between different hardware processors based on processing resource availability and quality of service metrics.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If machine learning models are executed on information handling systems, then AI productivity tool functionality is provided, but processing bottlenecks occur due to high resource consumption

Engineering Contradiction:

ImproveAI productivity tool functionalityVSAvoidprocessing resource consumption

Core Design Contradiction:

ProductivityVSUse of energy by moving object

Solution Approach 1:

The system dynamically switches between different runtime environments (CPU, GPU, FPGA, ASIC) based on real-time processing requirements and resource availability. The runtime selector module continuously monitors performance metrics and automatically transitions between runtimes to optimize both AI functionality and resource consumption, making the system adaptive rather than static.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The information handling system is designed with multiple runtime environments (CPU, GPU, FPGA, ASIC) that can all execute machine learning models. Each runtime has different strengths for different types of workloads, and the system can universally support all of them through the runtime selector module, allowing the same AI application to leverage different hardware accelerators as needed.

Inventive Principle:

Principle #6Universality (Multi-functionality)

2Adaptability or versatility

If multiple software applications are run simultaneously, then system versatility is improved, but processing bottlenecks worsen due to competing resource demands

Engineering Contradiction:

Improvesoftware application compatibilityVSAvoidprocessing throughput

Core Design Contradiction:

Adaptability or versatilityVSProductivity

Solution Approach 1:

Different runtime environments are assigned to different types of workloads based on their specific requirements. For example, GPU runtimes are used for parallel computing tasks, FPGA for custom acceleration, and CPU for general-purpose operations. This local optimization allows multiple applications to run simultaneously with each getting the most appropriate runtime for its specific needs, reducing resource conflicts.

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

The runtime selector module acts as an intermediary between multiple software applications and the various runtime environments. It manages resource allocation and scheduling, mediating between competing applications to ensure fair and efficient resource distribution, thereby maintaining high productivity even when multiple applications run simultaneously.

Inventive Principle:

Principle #24Intermediary (Mediator)

3Device complexity

If a single runtime is used for machine learning execution, then system complexity is reduced, but adaptability to different processing needs deteriorates

Engineering Contradiction:

Improveruntime management simplicityVSAvoidprocessing optimization flexibility

Core Design Contradiction:

Device complexityVSAdaptability or versatility

Solution Approach 1:

The runtime selector module automatically monitors performance metrics and makes intelligent decisions about which runtime to use without requiring manual intervention or complex configuration. The system self-manages the complexity of coordinating multiple runtimes by implementing its own selection and switching logic, keeping the user interface simple while maintaining high adaptability internally.

Inventive Principle:

Principle #25Self-service

Data Source

PatentUS20260017089A1System and method for dynamically switching machine learning runtimes behind an application interface

Publication Date: 2026.01.15 DELL PROD LP

AI summary

A system and method of runtime switching between machine learning (ML) models during execution of computer-readable program code instructions of an artificial intelligence (AI) productivity tool module with a hardware processor of an information handling system to initiate a request, on behalf of an application being executed on the information handling system, to an AI productivity tool subagent for a first ML model algorithm. Executing a swappable wrapper generator to create a first ML model algorithm wrapper around the first ML model algorithm, and executing an inference runtime control module to monitor the hardware processor resource utilization of runtime associated with the first ML model algorithm. Executing an inference runtime control module to switch to an alternative information handling system hardware processor to execute the first ML model algorithm, and create a second wrapper around the second ML model algorithm to switch to a second ML model algorithm is appropriate.