Unlock instant, AI-driven research and patent intelligence for your innovation.

Multicast packet connection circuit and its routing method based on convolutional neural network

A convolutional neural network and connection circuit technology, applied in the field of multicast packet connection circuit on-chip network routers and their routing, can solve the problem of not being able to operate nodes in the same row

Active Publication Date: 2020-09-18
HEFEI UNIV OF TECH
View PDF9 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] However, in order to avoid a blocking situation, the algorithm stipulates that the east-west channel cannot operate on the same row of nodes

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
  • Multicast packet connection circuit and its routing method based on convolutional neural network
  • Multicast packet connection circuit and its routing method based on convolutional neural network
  • Multicast packet connection circuit and its routing method based on convolutional neural network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0051] In the present embodiment, the format of the data packet is as shown in Table 1, and each multicast packet header microchip includes a microchip type flag, a data packet type flag, a microchip type logo, a destination node number, and a source node number;

[0052] Table 1 Network packet format

[0053]

[0054]

[0055] The multicast packet connection circuit used is applied to a 12×16 two-dimensional network, such as figure 1 As shown; the two-dimensional network includes 12×16 multicast packet connection circuits and correspondingly serves as 12×16 routing nodes; the multicast packet connection circuit is composed of an input unit, a priority controller, an address decoder, an arbitrator, and an output unit and a crossbar switch, and contains six transmission directions, such as figure 2 shown;

[0056] In this embodiment, the five input modules in the input unit all include a state control module, a single-multicast selection module, a multicast channel sel...

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 convolutional neural network-based multicast packet connection circuit and its routing method, which is characterized in that the multicast packet connection circuit supports multi-row node data transmission and expands the number of transmission nodes; Unit, priority controller, address decoder, arbiter, output unit and crossbar, where the input unit has five ports: east, south, west, north and local, and the output unit has six ports: east, south, West, North, Local 1 and Local 2. The present invention optimizes and improves the packet connection circuit and its routing method, and can transmit data of different lines simultaneously, thereby improving the transmission efficiency of data. At the same time, it generates two data streams for input transmission, and executes the data stream after the link is established. Transmission, to avoid deadlock and congestion.

Description

technical field [0001] The invention belongs to the communication technology field of integrated circuit on-chip network, and in particular relates to a multicast packet connection circuit on-chip network router and a routing method applied to convolutional neural networks. Background technique [0002] The convolutional neural network performs hundreds of millions or even tens of billions of multiplication and accumulation calculations on the input feature map data and weight data according to instructions to achieve the recognition effect, which requires a design method with high parallelism, high scalability, and high flexibility. Currently, on-chip networks are widely used in convolutional neural networks due to their high parallelism and high scalability. As the number of nodes in the network-on-chip increases, the resource consumption occupied by routing will increase accordingly. For traditional packet connection circuits, only wires and registers are included, and r...

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): H04L12/761H04L12/751H04L12/741H04L12/721H04L12/801H04L45/16H04L45/02H04L45/74
CPCH04L45/08H04L45/14H04L45/16H04L45/74H04L47/12
Inventor 杜高明张华蕾吴继婷张永亮张多利宋宇鲲邢鲲
Owner HEFEI UNIV OF TECH