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

Accelerating method of linear solver, storage database and GPU system

A database and surface data technology, applied in the direction of electrical digital data processing, special data processing applications, instruments, etc., can solve the problems of low computing efficiency of linear solvers, achieve improved computing speed, easy reading and access, and sufficient hardware resources Effect

Active Publication Date: 2018-01-02
浙江远算科技有限公司
View PDF5 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] The invention provides a linear solver acceleration method, storage database and graphics processor system, which overcomes the problem of low operational efficiency of the linear solver in the prior art

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
  • Accelerating method of linear solver, storage database and GPU system
  • Accelerating method of linear solver, storage database and GPU system
  • Accelerating method of linear solver, storage database and GPU system

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0046] In view of the powerful computing capability of the graphics processor (Graphics Processing Unit, GPU) itself, the linear solver provided by the embodiment of the present invention is introduced into the GPU, so the acceleration method of the linear solver provided by the embodiment of the present invention can be applied to a system with In a computing environment with multiple graphics processors (Graphics Processing Unit, GPU).

[0047] Such as figure 1 As shown, in the embodiment of the present invention, computing tasks can be allocated to multiple GPUs, such as GPU1, GPU2, GPU3, and GPU4, and multiple GPUs can communicate through a message passing interface (Message Passing Interface, MPI). Through the MPI interface, computing tasks can be parallelized at the task level, making full use of the parallel computing capabilities of the GPU. In a preferred solution, in the field of computational fluid dynamics (Computational Fluid Dynamics, CFD), when the created grid...

Embodiment 2

[0074] In order to realize the acceleration of the linear arithmetic unit, the embodiment of the present invention provides a storage database, such as Figure 9 As shown, it includes: a first storage unit 901, a second storage unit 903 and an index unit 902, wherein:

[0075] The first storage unit 901 is used to store the grid unit identification (cell id);

[0076] The second storage unit 903 is used to store the identification (face id) of the internal face and the internal face data corresponding to the face id, the internal face is a plane connecting two grid units, mapped to the same grid The internal surface data corresponding to the internal surface identifier of the unit identifier is stored in an adjacent storage area of ​​the second storage unit;

[0077] The index unit 902 is configured to construct and store an index range of the grid unit ID, where the index range includes a mapping relationship between the grid unit ID and the internal surface ID.

[0078] In...

Embodiment 3

[0088] Such as Figure 10 As shown, the embodiment of the present invention also provides an acceleration method for a linear solver, using the storage database and GPU disclosed in Embodiment 1 and Embodiment 2, the method includes:

[0089] 1001. The GPU divides the task block to be calculated into grid units, and determines the grid unit identifier of the grid unit;

[0090] For example, in the CFD calculation process, when the calculation task of a certain task block is assigned to the GPU, the GPU first divides the task block into grid units, and determines the identification (cell id) of the grid unit according to the unified naming rules;

[0091] 1003. Determine the internal face ID (face id) corresponding to the grid cell ID according to the grid cell ID, and read the internal face data corresponding to the internal face ID according to the internal face ID;

[0092] In the calculation process, according to the description in the second embodiment above, it is necess...

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 embodiment of the invention provides an accelerating method of a linear solver, a storage database and a graphics processing unit system, wherein the storage database comprises a first storage unit, a second storage unit and an index unit; the first storage unit is used for storing a grid cell identifier of a storage task block; the second storage unit is used for storing an identifier of an inner face and inner face data corresponding to the inner face identifier, the inner face is a plane for connecting two grid cell units, and the inner face data corresponding to the inner face identifier mapped to a same grid cell unit is stored in an adjacent storage region of the second storage unit; and the index unit is used for constructing and storing an index range of the grid cell identifier, wherein the index range comprises a mapping relationship between the grid cell identifier and the inner face identifier. According to the accelerating method of the linear solver, the storage database and the graphics processing unit system provided by the invention, accelerated calculation of the linear solver can be achieved, so that the calculating efficiency is improved.

Description

technical field [0001] The present invention relates to the field of linear computing, in particular to an acceleration method for a linear solver, a storage database, and a Graphics Processing Unit (GPU) system. Background technique [0002] In the field of industrial simulation, a common problem is the solution of sparse linear equations, such as the equation Solve, where: A represents the discrete matrix, j represents the vector to be solved, and b represents the source term of the discrete equation system. [0003] In actual calculation, the diagonal, upper triangular and lower triangular elements of the discrete matrix A can be stored in three vectors respectively, where the upper triangular element and the lower triangular element share two index vectors, which are used to obtain the elements in the matrix s position. This storage method is widely used in the field of industrial simulation, and this storage method can also be used with the preconditioned conjugate g...

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): G06F9/50G06F17/12G06F17/30G06F17/50
Inventor 朱兆生童潮臻吴健明
Owner 浙江远算科技有限公司
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