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

Data transmission method for sparse matrix vector multiplication and DMA transmission device

A data transmission method and sparse matrix technology, applied in the field of processor micro-architecture, can solve the problems of limited x-vector memory access efficiency, complex hardware structure, multiple hardware resources, etc., and achieve simple flexibility, wide application range, and access speed. quick effect

Pending Publication Date: 2020-04-24
NAT UNIV OF DEFENSE TECH
View PDF7 Cites 4 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0006] However, the above solution has problems such as complex hardware structure, high consumption of hardware resources, and limited efficiency of x vector memory access.

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
  • Data transmission method for sparse matrix vector multiplication and DMA transmission device
  • Data transmission method for sparse matrix vector multiplication and DMA transmission device
  • Data transmission method for sparse matrix vector multiplication and DMA transmission device

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0037] The present invention will be further described below in conjunction with the accompanying drawings and specific preferred embodiments, but the protection scope of the present invention is not limited thereby.

[0038] Such as figure 1 , 2 As shown, the present embodiment is oriented to the data transmission method of sparse matrix-vector multiplication, and the method includes: when x-vector data needs to be accessed during the sparse matrix-vector multiplication solution process, configuring the DMA data channel APip to perform x-vector data according to the SuperGather data transmission mode For irregular memory access, the DMA data channel APip first reads the index of the source address from outside the core, and generates the actual read address of the source data from the base address and the index of the read source address. According to the index of the generated source data The actual read address sends a read data request with a read return address, and move...

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 data transmission method for sparse matrix vector multiplication and a DMA transmission device. The method comprises the following steps: when x vector data needs to be accessed in the sparse matrix vector multiplication solving process, configuring a DMA (Direct Memory Access) data channel APip so as to carry out irregular memory access on the x vector data according toa SuperGather data transmission mode; during memory access, the DMA data channel APip firstly reading the index of the source address from the outside of the core; and generating an actual read address of the source data according to the base address and the index of the read source address, sending a read data request with a read return address according to the generated actual read address of the source data, and moving x vector data required by sparse matrix vector multiplication calculation to the in-core memory bank from the outside of the core. The method has the advantages of being simple in implementation method, high in access speed, high in accuracy, less in occupied resource, flexible in application and the like.

Description

technical field [0001] The invention relates to the technical field of processor micro-architecture, in particular to a data transmission method and a DMA transmission device oriented to sparse matrix-vector multiplication. Background technique [0002] Sparse Multiple-Vector Multiplication (SpMV) is one of the important algorithms for iteratively solving large linear equations, and it is widely used in scientific research and engineering practice fields such as circuit analysis and geological analysis. Since the SpMV algorithm is frequently called in the calculation of large linear equations, and a large number of discrete indirect addressing operations are generated during the calculation process, it will occupy a large amount of hardware resources and reduce the system's computing performance. Therefore, improving the computing performance of SpMV is a widely concerned issue at present. [0003] The system of equations processed by the SpMV algorithm can be expressed as ...

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
IPC IPC(8): G06F17/16G06F13/28
CPCG06F17/16G06F13/28
Inventor 刘胜孙书为陈海燕刘仲杨柳曹亚松张洋刘畅
Owner NAT UNIV OF DEFENSE 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