Memristor Crossbar Array for Linear Transformation Acceleration

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Linear transformations in computer applications are computation-intensive and resource-hungry, making them inefficient to implement optimally with general processors.

Innovation Solution

A hardware implementation using a crossbar array of memristors, where memory cells are programmed according to a linear transformation matrix, allowing for vector-matrix multiplication to calculate linear transformations efficiently.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If linear transformations are implemented using general processors, then computational flexibility is maintained, but processing speed and resource efficiency deteriorate

Engineering Contradiction:

Improvecomputational flexibilityVSAvoidprocessing speed

Core Design Contradiction:

Adaptability or versatilityVSProductivity

Solution Approach 1:

The patent replaces the mechanical computation system of general processors with an electrical/physical system using memristor crossbar arrays. The linear transformation computation is performed through physical electrical signals flowing through the memristor network, where the conductance values directly represent matrix elements, enabling parallel computation of all matrix-vector multiplication operations simultaneously. This physical substitution achieves orders of magnitude speedup while maintaining computational flexibility through programmable memristor conductance values.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Adaptability or versatility

If linear transformations are implemented using general processors, then programming versatility is maintained, but resource consumption and computational overhead increase

Engineering Contradiction:

Improveprogramming versatilityVSAvoidresource consumption

Core Design Contradiction:

Adaptability or versatilityVSUse of energy by moving object

Solution Approach 1:

The patent substitutes the software-based computation approach of general processors with a hardware-based physical system. Energy-consuming operations such as sequential arithmetic calculations, memory access, and control logic are replaced by passive electrical signal propagation through the memristor crossbar array. The computation emerges from the physical properties of the circuit rather than active processing, dramatically reducing energy consumption while maintaining versatility through reconfigurable memristor programming.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

3Adaptability or versatility

If linear transformations are implemented using general processors, then general-purpose computing capability is maintained, but computational intensity and time requirements increase

Engineering Contradiction:

Improvegeneral-purpose computing capabilityVSAvoidcomputation time

Core Design Contradiction:

Adaptability or versatilityVSLoss of time

Solution Approach 1:

The patent segments the linear transformation computation into spatially distributed operations across the memristor crossbar array. Each intersection or region of the crossbar performs a portion of the matrix-vector multiplication simultaneously, with results naturally aggregating through electrical signal summation. This spatial segmentation enables parallel execution of what would otherwise be sequential operations in software, achieving massive acceleration while maintaining general-purpose capability through reconfigurable programming.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent transitions computation from the temporal dimension (sequential processing in time) to the spatial dimension (parallel processing across physical space). The crossbar array's two-dimensional structure allows simultaneous engagement of multiple computation paths, converting time-consuming sequential operations into space-parallel operations. This dimensional transformation fundamentally changes the computational paradigm from sequential to parallel execution.

Inventive Principle:

Principle #17Another dimension (Dimensionality change)

Applied Scientific Principles

This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.

Function Achieved in This Case

This approach enables fast and accurate hardware calculations of linear transformations, reducing resource usage and improving processing efficiency in applications like data, image, and video processing.

Implementation Method 1

Memristors are devices that can be programmed to different resistive states by applying a programming energy, such as a voltage

Methodology Applied

Scientific EffectMemristance: Magnetoresistance

Implementation Method 2

Large crossbar arrays of memory devices with memristors can be used in a variety of applications, including memory, programmable logic, signal processing control systems, pattern recognition

Methodology Applied

Scientific EffectElectrical Conduction: Conduction (electrical)

Data Source

PatentUS10529418B2Linear transformation accelerators

Publication Date: 2020.01.07 HEWLETT PACKARD ENTERPRISE DEV LP

AI summary

Examples herein relate to linear transformation accelerators. An example linear transformation accelerator may include a crossbar array programmed to calculate a linear transformation. The crossbar array has a plurality of words lines, a plurality of bit lines, and a memory cell coupled between each unique combination of one word line and one bit line, where the memory cells are programmed according to a linear transformation matrix. The plurality of word lines are to receive an input vector, and the plurality of bit lines are to output an output vector representing a linear transformation of the input vector.