Matrix Multiplication API for Fast Algorithm Selection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Determining appropriate algorithms for matrix operations consumes significant memory, time, and computing resources.

Innovation Solution

A matrix multiplication algorithm library accessible through an API that selects optimal algorithms based on parameters, enabling entities to specify attributes and receive a list of high-performance algorithms suitable for matrix operations.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If traditional methods are used to determine appropriate algorithms for matrix operations, then comprehensive algorithm selection is possible, but significant memory, time, and computing resources are consumed

Engineering Contradiction:

Improvealgorithm selection capabilityVSAvoidmemory and computing resources

Core Design Contradiction:

Adaptability or versatilityVSQuantity of substance

Solution Approach 1:

The system performs preliminary analysis of matrix operation parameters and pre-determines suitable algorithms from a library before actual matrix computation. This advance preparation stores algorithm selections that can be reused, avoiding repeated resource-intensive analysis for similar operations.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

An intermediary algorithm selection system is introduced between the matrix operation request and the computation engine. This intermediary analyzes operation parameters, consults an algorithm library, and selects optimal algorithms, thereby shielding the resource-consuming selection process from repeated execution during actual computations.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Measurement precision

If traditional methods are used to determine appropriate algorithms for matrix operations, then accurate algorithm matching is possible, but significant time is consumed

Engineering Contradiction:

Improvealgorithm matching accuracyVSAvoidalgorithm determination time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system uses parameter-based algorithm selection where operations are categorized by key parameters (matrix dimensions, data types, operation type). By changing the approach from detailed analysis to parameter-matching, the system maintains accurate algorithm matching while significantly reducing determination time.

Inventive Principle:

Principle #35Parameter changes

3Productivity

If comprehensive algorithm libraries are maintained, then optimal algorithms can be selected, but device complexity increases

Engineering Contradiction:

Improvematrix operation efficiencyVSAvoidalgorithm library management complexity

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The algorithm library is segmented into categories based on operation types (multiplication, addition, transformation) and further divided by parameter ranges (matrix sizes, data types). This segmentation allows the system to maintain a comprehensive library while managing complexity through organized structure and targeted selection.

Inventive Principle:

Principle #1Segmentation

Data Source

PatentUS12585726B2Application programming interface to accelerate matrix operations

Publication Date: 2026.03.24 NVIDIA CORP

AI summary

Apparatuses, systems, and techniques to determine a matrix multiplication algorithm for a matrix multiplication operation. In at least one embodiment, a matrix multiplication operation is analyzed to determine an appropriate matrix multiplication algorithm to perform the matrix multiplication algorithm.