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

Matrix processing device, method and processing equipment

A processing device and matrix technology, applied in the field of data processing, can solve problems such as tediousness, numerous steps, and complex operation logic of the device, and achieve the effects of simplifying operation logic, improving processing capacity, and simple wiring

Active Publication Date: 2020-01-17
中昊芯英(杭州)科技有限公司
View PDF11 Cites 1 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, in the known prior art, the method for matrix reconfiguration has many steps, and the device for realizing the reconfiguration is complicated and the operation logic is cumbersome; especially in the disclosed systolic array design, it has not been found that any The scheme achieves the real efficient and convenient matrix reconstruction

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
  • Matrix processing device, method and processing equipment
  • Matrix processing device, method and processing equipment
  • Matrix processing device, method and processing equipment

Examples

Experimental program
Comparison scheme
Effect test

example 1

[0079] See Image 6 , is a schematic structural diagram of the matrix processing device 600 provided by the embodiment of the present invention. The matrix processing device 600 includes a node network 620 , a data memory 630 , a decoder 640 and a data cache array 650 . The node network 620 is coupled in the form of a two-dimensional matrix by the node units 610 along the horizontal direction and the vertical direction. The node unit 610 includes a first register, a second register, a reconfiguration controller, and a reconfiguration actuator. The reconfiguration controller is connected to the first register, the second register and the reconfiguration actuator respectively; the first register is coupled along a straight line in the horizontal direction to form a first register chain, and the second register is coupled to a second register chain along a straight line in a vertical direction. The register chain; the reconfiguration actuators are coupled to each other along st...

example 2

[0100] See Figure 12 , is a schematic structural diagram of the matrix processing device 1200 provided by the embodiment of the present invention. The matrix processing device 1200 includes a node network 1220 , a data memory 1230 , a decoder 1240 and a data cache array 1250 . The node network 1220 is coupled in the form of a two-dimensional matrix by the node units 1210 along the horizontal direction and the vertical direction. The node unit 1210 includes a first register, a second register, a reconfiguration controller, and a reconfiguration actuator. The reconfiguration controller is connected to the first register, the second register and the reconfiguration actuator respectively; the first register is coupled along a straight line in the horizontal direction to form a first register chain, and the second register is coupled to a second register chain along a straight line in a vertical direction. The register chain; the reconfiguration actuators are coupled to each oth...

example 3

[0114] See Figure 16, is a schematic structural diagram of the matrix processing device 1600 provided by the embodiment of the present invention. The matrix processing device 1600 includes a node network 1620 , a data memory 1630 , a decoder 1640 and a data cache array 1650 . The node network 1620 is coupled in the form of a two-dimensional matrix by the node units 1610 along the horizontal direction and the vertical direction. The node unit 1610 includes a first register, a second register, and a reconfiguration controller. The reconfiguration controller is respectively connected to the first register and the second register; the first register is coupled along a straight line in the horizontal direction to form a first register chain, and the second register is coupled to form a second register chain along a straight line in the vertical direction.

[0115] The decoder 1640 is connected to the node network 1620 and the data storage 1630 respectively. The data storage 163...

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 matrix processing device, method and processing equipment. The deep neural network often needs to load a weight matrix in model training, and also often needs to load a matrix which is subjected to reconstruction processing relative to an original weight matrix through plane rotation, reverse order, transposition and the like. According to the invention, the reconstruction controller is used as a repetitive control unit in the pulsation array, and matrix reconstruction is carried out by changing the transmission direction of data on a specific node, so that a series of matrix processing taking matrix reconstruction as a core, such as matrix plane rotation and matrix transposition, in the pulsation array by taking a vector as a basic unit, is efficiently and conveniently realized.

Description

technical field [0001] The invention relates to the field of data processing, in particular to a matrix processing device, a matrix processing method and processing equipment. Background technique [0002] The exponential growth of digital data available on the Internet and IoT is driving the need for high-performance data analysis algorithms. Among them, the deep neural network (Deep Neural Network) has broad application prospects in the fields of target recognition, automatic driving and drone navigation. However, the deep neural network needs to perform a large amount of repetitive data calculations, and the traditional processing architecture cannot achieve balanced and efficient operation in data transmission and calculation, especially when dealing with a large number of multiplication and addition operations. bring a huge load. [0003] Systolic Array (Systolic Array) structure is a network formed by the coupling of multiple processing units. Each repeated processin...

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 Applications(China)
IPC IPC(8): G06F7/78G06F15/78G06N3/063
CPCG06F7/78G06F15/7867G06N3/063
Inventor 闯小明杨龚轶凡郑瀚寻张斌
Owner 中昊芯英(杭州)科技有限公司
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