Executable Graph Reuse Through Common-Subgraph Modification

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing executable graphs are typically usable for only a single workload of a task graph from which they were created, limiting their versatility in performing workloads associated with new task graphs.

Innovation Solution

Techniques for modifying an executable graph to adapt it for performing workloads associated with new task graphs, allowing reuse across multiple task graphs.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If an executable graph is created from a task graph, then the workload can be performed by configuring computing resources according to the executable graph, but the executable graph is typically usable for only a single workload and cannot be reused for new task graphs

Engineering Contradiction:

Improvereusability of executable graphVSAvoidconfiguration overhead

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The executable graph is modified to perform multiple functions by adding a determination module that identifies common subgraphs between the executable graph and new task graphs, and a modification module that adapts the executable graph to perform workloads from different task graphs. This allows a single executable graph to serve multiple workloads rather than being limited to one specific task graph.

Inventive Principle:

Principle #6Universality (Multi-functionality)

Solution Approach 2:

The executable graph is divided into identifiable subgraphs or components that can be independently analyzed and matched against task graphs. The determination module segments the graph to find commonalities, and the modification module selectively adjusts specific portions rather than requiring complete reconfiguration, reducing overall complexity.

Inventive Principle:

Principle #1Segmentation

2Productivity

If an executable graph is created specifically for a single task graph, then it can perform that workload efficiently, but it requires creating new executable graphs for each new task graph, increasing configuration overhead

Engineering Contradiction:

Improveworkload execution efficiencyVSAvoidtime for creating executable graphs

Core Design Contradiction:

ProductivityVSLoss of time

Solution Approach 1:

The system performs preliminary analysis by determining common subgraphs between the existing executable graph and new task graphs before full configuration. This preliminary identification of reusable components allows for faster adaptation, as the common subgraph structure has already been optimized and validated, reducing the time required to create executable graphs for new workloads.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The modification module adjusts parameters of the executable graph such as computing resource allocations, data transfer configurations, and task scheduling based on the specific requirements of new task graphs while preserving the underlying graph structure. This allows efficient adaptation without complete recreation of the executable graph.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS12443449B2Techniques for modifying an executable graph to perform a workload associated with a new task graph

Publication Date: 2025.10.14 NVIDIA CORP

AI summary

Techniques to modify executable graphs to perform different workloads. In at least one embodiment, an executable version of a first task graph is modified by applying a non-executable version of a second task graph to executable version of first task graph so that executable version of first task graph can perform a second workload of non-executable version of second task graph.