Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

A memristive memory vector matrix operator and its operation method

A technology of vector matrix and calculation method, which is applied in the field of analog circuits, can solve problems such as the inability to calculate high-precision floating-point numbers, and achieve the effect of balancing calculation accuracy and calculation time, small array area, and small circuit area overhead

Active Publication Date: 2022-05-20
HUAZHONG UNIV OF SCI & TECH
View PDF10 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] Aiming at the above defects or improvement needs of the prior art, the present invention provides a memristive memory vector matrix operator and operation method, the purpose of which is to solve the technical problem that the prior art cannot perform high-precision calculations on floating-point numbers

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • A memristive memory vector matrix operator and its operation method
  • A memristive memory vector matrix operator and its operation method
  • A memristive memory vector matrix operator and its operation method

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0041] A vector-matrix operator in a memristive memory, such as figure 1 As shown, it includes: an input split unit, a decimal multiplication unit and an external accumulation unit; wherein, there are M decimal multiplication units, and all of them include a memristor array; each decimal multiplication unit is connected in parallel, and the output terminals of the input split unit are respectively It is connected with the input end of each decimal multiplication unit, and the output end of each decimal multiplication unit is connected with the input end of the external accumulation unit;

[0042] The input splitting unit is used to unify the floating-point digits in the input floating-point number vector and the floating-point number matrix respectively and split them bit by bit to obtain N decimal vectors and M decimal matrices; the M decimal matrices are respectively After being written into the memristor arrays of M decimal multiplication units, each decimal vector is seque...

Embodiment 2

[0046] A kind of memristive memory vector matrix operation method based on the memristive memory vector matrix operator provided in embodiment 1, such as Figure 5 shown, including the following steps:

[0047] S1. After unifying the digits of each floating-point number in the input floating-point number matrix, split them bit by bit to obtain M decimal matrices, and write them into M memristor arrays respectively;

[0048] Specifically, the memristor array in this embodiment is the same as that in Embodiment 1, and will not be repeated here. According to the corresponding method in Embodiment 1, the digits of each floating-point number in the floating-point number matrix are unified into M bits, and then divided bit by bit. Take the number of digits before the decimal point as 1 (that is, the highest digit is the ones digit) as an example, such as Image 6 As shown, the floating-point number matrix after unifying the number of digits Each floating-point number in is split...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses a memristive memory vector matrix operator and an operation method, comprising: an input splitting unit, a decimal multiplication unit and an external accumulation unit; wherein there are multiple decimal multiplication units, and all of them include a memristor array; The decimal multiplication units are connected in parallel; the output terminals of the input splitting unit are respectively connected with the input terminals of the decimal multiplication units, and the output terminals of each decimal multiplication unit are connected with the input terminals of the external accumulation unit. By splitting the floating-point numbers in the input floating-point number vector and floating-point number matrix bit by bit, the high-precision mapping of floating-point numbers can be realized, and the vector-matrix operation of floating-point numbers can be performed without loss of precision, effectively reducing The influence of the non-ideal effect of the memristive device on the calculation result realizes the high-precision calculation of floating-point numbers. In addition, due to the adoption of the decimal mapping method, the number of devices required by the present invention is small, and the circuit area overhead is also small.

Description

technical field [0001] The invention belongs to the field of analog circuits, and more specifically relates to a memristive memory vector matrix calculator and a calculation method. Background technique [0002] Due to the characteristics of high energy efficiency, low latency, and low computational time complexity, memristor-based in-memory vector-matrix operations have been widely used in data-intensive tasks. Neural networks based on memristor arrays It has already demonstrated the huge advantage of in-memory computing in terms of energy efficiency over its traditional numerical counterparts. However, as a data-intensive task, memristor-based memory-computing integration technology faces considerable challenges when performing numerical calculation acceleration, because the vector-matrix operation performed by the memristor array is an analog Affected by the non-ideal effects of memristive devices and array non-ideal effects, the vector-matrix operation based on a single...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Patents(China)
IPC IPC(8): G06F17/16
CPCH03K19/177G06F17/16Y02D10/00
Inventor 李祎李健聪缪向水
Owner HUAZHONG UNIV OF SCI & TECH
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com