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

Heterogeneous parallel computing method for sparse matrix-vector multiplication

A sparse matrix and vector multiplication technology, applied in computing, concurrent instruction execution, program control design, etc., can solve problems such as being unsuitable for CPU-GPU heterogeneous computing platform processing, and achieve easy expansion to cluster environments and simple generation methods. Effect

Inactive Publication Date: 2015-11-18
SOUTH CHINA UNIV OF TECH
View PDF10 Cites 8 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

But all these storage methods are not suitable for processing on CPU-GPU heterogeneous computing platforms

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
  • Heterogeneous parallel computing method for sparse matrix-vector multiplication
  • Heterogeneous parallel computing method for sparse matrix-vector multiplication

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0021] The present invention will be further described in detail below in conjunction with the embodiments and the accompanying drawings, but the embodiments of the present invention are not limited thereto.

[0022] A heterogeneous parallel computing method for sparse matrix-vector multiplication, specifically comprising the following steps:

[0023] S1. Determine the value of the adjustable parameter K of the sparse matrix.

[0024] The K value is used to identify the number that can compress the ELL matrix to the minimum. In Hybrid (ELL+COO / CSR), the K value is used to determine how much data is stored in the ELL and how much is stored in the COO / CSR. In this storage structure, it can be considered that the K value determines the calculation amount of the GPU and CPU. The larger the value of K, the greater the calculation amount of the GPU and the smaller the calculation amount of the CPU. When the value of K is optimal, the time for the CPU and GPU to complete the calcul...

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 heterogeneous parallel computing method for sparse matrix-vector multiplication. The method comprises the following steps of: reading a sparse matrix stored in a hard disk by a CPU, determining an adjustable parameter K of the spare matrix, and according to the adjustable parameter K of the spare matrix, applying memory storage space including storage space required for an ELL (ELLPACK) storage structure and a CSR (Compressed Sparse Row) storage structure; at the same time, applying GPU storage space required for the ELL storage structure; filling memory storage space applied by the CPU with matrix data to generate a mixed storage structure; copying data stored in the ELL storage structure in a memory to the GPU storage space for storage; and finally, performing sparse matrix-vector multiplication by using the storage structure after completion of processing. According to the computing method, the computing capabilities of the CPU and a GPU are utilized at the same time when a sparse matrix-vector multiplication computation task is executed by a computer, so that the best computing characteristics of the CPU and the GPU can be exerted separately.

Description

technical field [0001] The invention relates to a data storage method, in particular to a heterogeneous parallel computing method of sparse matrix vector multiplication. technical background [0002] Sparse Matrix-Vector Multiplication (SpMV for short) is one of the most commonly used calculations in scientific computing and engineering applications. In many data mining applications, it is often encountered that the data is extremely sparse, and this type of data is usually represented as a sparse matrix. When the data scale is very large, it is very necessary to use the popular CPU-GPU heterogeneous computing platform to realize SpMV heterogeneous parallel computing. [0003] In general, there are two calculation modes for SpMV calculations on heterogeneous platforms: [0004] The first mode is CPU / GPU collaborative computing. The CPU processes the data and then sends the data to the GPU for calculation. After the GPU calculation is completed, the data is transmitted back...

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): G06F9/38G06F9/302
Inventor 董守斌张铃启陈泽邦
Owner SOUTH CHINA UNIV OF 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