CN109635473A - A kind of heuristic high-throughput material simulation calculation optimization method - Google Patents
A kind of heuristic high-throughput material simulation calculation optimization method Download PDFInfo
- Publication number
- CN109635473A CN109635473A CN201811556665.5A CN201811556665A CN109635473A CN 109635473 A CN109635473 A CN 109635473A CN 201811556665 A CN201811556665 A CN 201811556665A CN 109635473 A CN109635473 A CN 109635473A
- Authority
- CN
- China
- Prior art keywords
- carried out
- shared
- executed
- ulation
- throughput
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
Abstract
The present invention proposes a kind of heuristic high-throughput material simulation calculation optimization method, belongs to materials science field.This method determines the shared execution pattern of all operations in high-throughput material computer sim- ulation first;Then it selects one to be not carried out operation comprising being not carried out in the most model of number of jobs from high-throughput material computer sim- ulation and the operation is executed using isolated execution mode;Next execution operation is obtained using heuristic information, until all job executions finish in high-throughput material computer sim- ulation.The present invention can increase substantially the efficiency of high-throughput job execution, greatly shorten the time of new material design using the atomic radius of neighbouring relations and different elements between elements different in the periodic table of elements as heuristic information.
Description
Technical field
The invention belongs to materials science fields, and in particular to a kind of heuristic high-throughput material simulation calculation optimization method.
Background technique
Currently, the acquisition of new material finds that the mode of new material is gone to by a large amount of by many experiments from traditional
Simulation calculation design the high-throughput calculating mode of new material, can greatly promote the efficiency of new material acquisition.
So-called high throughput material calculates, and is exactly the computing resource powerful by ability, high-volume can once be rapidly completed
Calculating task, by the analysis to calculated result, filter out meet the requirements candidate material design.And such calculate is appointed
Business exists in the form of operation on high performance computing system on concrete form, and high-throughput material of being known as calculates operation.By
It is usually quite greatly in its calculation amount, therefore how to optimize and improve its performance just and become a challenge.Mainstream at present
Performance optimization method calculate what operation carried out both for single high-throughput material.As long as can be improved the property an of operation
Can, this method also has identical effect of optimization to other similar operations.But the shortcomings that such methods be they only from
The local message of single operation sets out to optimize, and does not account for internal association physically between different work, from
The possible more significant optimization potentiality of macroscopic view and overall angle analysis institute are ignored, thus the effect optimized it is easy by
The limitation of various local factors can not more fully increase substantially the optimization performance of overall operation.
It is calculated by the high-throughput material of representative of material genetic engineering, it is each that current groundwork is placed on realization calculating
In the automation linking in stage, typical project is international AFlow and MP project, but how homework book is greatly lowered
The execution time of body especially optimizes the execution time of computation-intensive operation, still not no good solution.
Summary of the invention
The purpose of the present invention is for overcome the shortcomings of prior art be unable to fully optimization the computation-intensive job execution time
Place proposes a kind of heuristic high-throughput material simulation calculation optimization method.The present invention combines high-throughput material computer sim- ulation to appoint
The characteristics of business, opens a brand-new path to realize that performance optimizes, is not only easy to implement, but also can obtain highly significant
Effect.The efficiency that high-throughput job execution can be increased substantially greatly shortens the time of new material design.
The present invention proposes a kind of heuristic high-throughput material simulation calculation optimization method, which is characterized in that this method includes
Following steps:
1) the shared execution pattern of all operations in high-throughput material computer sim- ulation is determined;Specific step is as follows:
Two operations, the doping member of described two operations 1-1) are selected in any model of high-throughput material computer sim- ulation
Element is adjacent element, and two operations are denoted as X and Y respectively;
X 1-2) is executed using isolated execution mode;
1-3) Y shares X using different shared execution patterns and executes, and chooses execution time shortest shared execution pattern and makees
The shared execution pattern used for all operations;
2) it is not carried out from high-throughput material computer sim- ulation comprising being not carried out selection one in the most model of number of jobs
Operation simultaneously executes the operation using isolated execution mode;Specific choice method is as follows:
Comprising being not carried out in the most model of number of jobs, operation is not carried out to each, counting it adjacent can share
The number of operation, selection can adjacent shared number of jobs be most is not carried out operation and executes the work using isolated execution mode
Industry;If there is it is multiple be not carried out operation have it is equal in number can adjacent shared operation, then randomly select one of them and do not hold
Then row operation executes the operation with isolated execution mode;Wherein, each operation can adjacent shared operation be doped chemical with
The adjacent operation of the disastrously miscellaneous element of the operation;
3) next execution operation is obtained using heuristic information, until all operations are held in high-throughput material computer sim- ulation
Row finishes;Specific step is as follows:
3-1) determine that all share executes operation pair, described share executes operation to the following requirement of satisfaction: Mei Gezuo
Two operations of industry centering belong to same model, and one of operation has executed, another operation has not carried out;
3-2) operation is executed to determining to can share:
If there is can share execute operation pair, then calculate two doped chemicals of each operation pair atomic radius it
Difference selects the wherein the smallest operation pair of absolute difference, executes the operation pair with the shared execution pattern that step 1) determines
In the operation that is not carried out, then return to step 2);Operation pair is executed if there is no that can share, and high-throughput material meter
The operation being also not carried out in emulation is calculated, then returns to step 2);Until all operations are held in high-throughput material computer sim- ulation
Row finishes, and method terminates.
The features of the present invention and beneficial effect are:
(1) the method for the present invention is carried out by analyzing the internal association in high-throughput material calculating between numerous emulation jobs
Optimization, is a kind of optimization method for having overall operation in mind.It is entirely different with the performance optimization method thinking of single operation before, but
The local optimization methods for single operation before being can be insusceptibly integrated among this method, and the two is a kind of mutual
The relationship mended and be superimposed.
(2) optimization method overall effect highly significant proposed by the present invention, although the possible effect of a small number of operations is unobvious,
It is that the performance of most of operations has and is obviously improved.The efficiency that high-throughput job execution can be increased substantially greatly shortens new
The time of design of material
(3) this method only needs considerably less well-known information, i.e., the phase in the periodic table of elements between different elements
The atomic radius of adjacent relationship and different elements can carry out the similarity degree between different work to analyze and do on this basis
Optimal Decision-making out, decision-making technique were not only simple but also efficient.
Specific embodiment
The present invention proposes a kind of heuristic high-throughput material simulation calculation optimization method, combined with specific embodiments below into one
Detailed description are as follows for step.
The present invention proposes a kind of heuristic high-throughput material simulation calculation optimization method, it is assumed that a high-throughput material calculates
Emulation (HTCS, High Throughput Computing Simulation) is made of M model, and expression formula is as follows:
Wherein M indicates the sum of high-throughput material computation model, ModeliIndicate i-th of model.
Wherein i-th of model can be by NiA operation composition, therefore can be further model M odeliIt indicates are as follows:
Wherein, Si,jIndicate j-th of operation of i-th of model;
Different work corresponding for the same model, doped chemical is different, and other aspects (refer to all inputs
The initial setting up of file) it is all just the same.
It is illustrated below based on widely used VASP software and how operation is calculated and optimized.Here first
Define two kinds of executive modes of high-throughput operation: isolated execution mode and shared execution pattern.
INCAR, the POSCAR independently generated when so-called isolated execution mode is exactly the operation based on design for the operation,
POTCAR and KPOINT file input file, is executed on concurrent computational system, does not have any pass with other operations
System.
So-called shared execution pattern, it is related to two operations A and B, and wherein A is the operation completed, and B is to hold
Capable operation, B optimize its operational efficiency by sharing the result that A is executed.It is specific to divide three kinds of different shared execution patterns again,
It is that shared CONTCAR is executed respectively, shares CHGCAR execution and shared CONTCAR and CHGCAR is executed simultaneously.So-called B shares A
CONTCAR execute, refer to POSCAR that the CONTCAR that obtains after B has been executed using A replaces the B original design phase to generate into
Row executes.The CHGCAR that B shares A is executed, and refers to that the CHGCAR that B is exported using A obtains its electricity as additional input file
Lotus distribution, this needs while modifying its original INCAR file, sets 1 for ICHARG therein label.B shares A's
CONTCAR and CHGCAR are executed, and refer to that B not only uses the CONTCAR of A as its POSCAR, but also using the CHGCAR of A as
Additional input is executed to obtain its distribution of charges, while setting 1 for the ICHARG label in its INCAR file.Three
The effect of the shared executive mode of kind can be different because of different operations.
We define the concept of adjacent element below, for two elements EA and EB in the periodic table of elements, if they
Same a line in the periodic table of elements, and direct neighbor are then mutually known as row adjacent element.If they are in the periodic table of elements
Same row, and direct neighbor is then mutually known as column adjacent element.Two elements either go it is adjacent or column it is adjacent, all claim
For adjacent element.
The concept that can share operation is defined further below.To the operation based on same model, since they are only adulterated
Element is different, and defining the adjacent operation of doped chemical each other can adjacent shared operation.
The present invention proposes a kind of heuristic high-throughput material simulation calculation optimization method, comprising the following steps:
1) the shared execution pattern of all operations in high-throughput material computer sim- ulation is determined;Specific step is as follows:
Two operations, the doping member of described two operations 1-1) are selected in any model of high-throughput material computer sim- ulation
Element is adjacent element, and two operations are denoted as X and Y respectively;
X 1-2) is executed using isolated execution mode, all results being finished are stored in the output text of the operation
In part, which can describe some fundamental characteristics of given design of material.
1-3) three kinds of different shared execution patterns of Y are shared X and are executed, and choose and execute time shortest shared execution pattern
The shared execution pattern used as all operations.
If workload is very big, the different X and Y operation of multiple groups can be chosen from different models and is surveyed in advance
Examination carries out selecting shared execution pattern according to the quality of average effect.
2) it is not carried out from high-throughput material computer sim- ulation comprising being not carried out selection one in the most model of number of jobs
Operation simultaneously executes the operation using isolated execution mode;Specific choice method is as follows:
Comprising being not carried out in the most model of number of jobs, operation is not carried out to each, counting it adjacent can share
The number of operation, selection can adjacent shared number of jobs be most is not carried out operation and executes the work using isolated execution mode
Industry;If there is it is multiple be not carried out operation can adjacent shared number of jobs it is identical.It then randomly selects one of them and is not carried out work
Then industry executes the operation with isolated execution mode;
3) next execution operation is obtained using heuristic information, until all operations are held in high-throughput material computer sim- ulation
Row finishes;Specific step is as follows:
3-1) determine that all share executes operation pair, described share executes operation to the following requirement of satisfaction: Mei Gezuo
Two operations of industry centering belong to same model, and one of operation has executed, another operation has not carried out;Will institute
Have and meets the operation of above-mentioned condition to<A1, B1>...,<Am, Bm>... it records.
3-2) operation is executed to determining to can share:
If there is can share execute operation pair, then calculate two doped chemicals of each operation pair atomic radius it
Difference selects the wherein the smallest operation pair of absolute difference, executes the operation pair with the shared execution pattern that step 1) determines
In the operation that is not carried out, then return to step 2);Operation pair is executed if there is no that can share, and high-throughput material meter
The operation being also not carried out in emulation is calculated, then returns to step 2);Until all operations are held in high-throughput material computer sim- ulation
Row finishes, and method terminates.
Claims (2)
1. a kind of heuristic high-throughput material simulation calculation optimization method, which is characterized in that method includes the following steps:
1) the shared execution pattern of all operations in high-throughput material computer sim- ulation is determined;Specific step is as follows:
Two operations 1-1) are selected in any model of high-throughput material computer sim- ulation, the doped chemical of described two operations is
Two operations are denoted as X and Y by adjacent element respectively;
X 1-2) is executed using isolated execution mode;
1-3) Y shares X using different shared execution patterns and executes, and chooses and executes time shortest shared execution pattern as institute
The shared execution pattern for thering is operation to use;
2) operation is not carried out comprising being not carried out selection one in the most model of number of jobs from high-throughput material computer sim- ulation
And the operation is executed using isolated execution mode;Specific choice method is as follows:
Comprising being not carried out in the most model of number of jobs, operation is not carried out to each, counting it can adjacent shared operation
Number, selection can adjacent shared number of jobs be most is not carried out operation and executes the operation using isolated execution mode;Such as
Fruit there are it is multiple be not carried out operation have it is equal in number can adjacent shared operation, then randomly select one of them and be not carried out work
Then industry executes the operation with isolated execution mode;Wherein, each operation can adjacent shared operation be doped chemical and the work
The adjacent operation of the disastrously miscellaneous element of industry;
3) next execution operation is obtained using heuristic information, until all job executions are complete in high-throughput material computer sim- ulation
Finish;Specific step is as follows:
3-1) determine that all share executes operation pair, described share executes operation to the following requirement of satisfaction: each operation pair
In two operations belong to same model, one of operation has executed, another operation has not carried out;
3-2) operation is executed to determining to can share:
Operation pair is executed if there is that can share, then calculates the difference of the atomic radius of two doped chemicals of each operation pair, is selected
The smallest operation pair of wherein absolute difference is selected, which is executed with the shared execution pattern that step 1) determines and is not held
Capable operation, then returns to step 2);Operation pair is executed if there is no that can share, and high-throughput material computer sim- ulation
In the operation that is also not carried out, then return to step 2);Until all job executions finish in high-throughput material computer sim- ulation,
Method terminates.
2. the method as described in claim 1, which is characterized in that the shared execution pattern includes: that shared CONTCAR executes mould
Formula shares CHGCAR execution pattern and shares CONTCAR and CHGCAR execution pattern simultaneously;The shared CONTCAR executes mould
Formula refers to that being not carried out operation is held using having executed the CONTCAR substitution that operation obtains and be not carried out the original POSCAR of operation
Row;The shared CHGCAR execution pattern, refer to be not carried out operation using executed operation output CHGCAR as additionally
Input file obtains corresponding distribution of charges, and modification is not carried out operation original I NCAR file simultaneously, and ICHARG therein is marked
Note is set as 1;The shared CONTCAR and CHGCAR execution pattern refers to that being not carried out operation utilization has executed what operation obtained
CONTCAR substitution is not carried out the original POSCAR of operation and is executed, and is not carried out operation utilization and has executed operation output
CHGCAR obtains corresponding distribution of charges as additional input file, and modification is not carried out operation original I NCAR file simultaneously,
1 is set by ICHARG therein label.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811556665.5A CN109635473B (en) | 2018-12-19 | 2018-12-19 | Heuristic high-flux material simulation calculation optimization method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811556665.5A CN109635473B (en) | 2018-12-19 | 2018-12-19 | Heuristic high-flux material simulation calculation optimization method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109635473A true CN109635473A (en) | 2019-04-16 |
CN109635473B CN109635473B (en) | 2022-08-02 |
Family
ID=66075503
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811556665.5A Active CN109635473B (en) | 2018-12-19 | 2018-12-19 | Heuristic high-flux material simulation calculation optimization method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109635473B (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110334055A (en) * | 2019-06-06 | 2019-10-15 | 苏州迈高材料基因技术有限公司 | A method of it obtaining material and calculates data |
CN113127973A (en) * | 2021-04-16 | 2021-07-16 | 湖南大学 | CAE simulation technology-based multi-material intelligent material selection method and system and electronic equipment |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1319761A (en) * | 2000-03-15 | 2001-10-31 | 清华大学 | Apparatus and method for high-flux electric rotation detection |
CN103631988A (en) * | 2013-10-22 | 2014-03-12 | 芜湖大学科技园发展有限公司 | Multi-user simulation data management platform for electric power system simulation |
US20140380320A1 (en) * | 2013-06-20 | 2014-12-25 | International Business Machines Corporation | Joint optimization of multiple phases in large data processing |
CN104992258A (en) * | 2012-07-05 | 2015-10-21 | 爱利门图供应链管理(开曼)有限公司 | Method and system for controlling supply chains |
CN108123109A (en) * | 2016-11-28 | 2018-06-05 | 华为技术有限公司 | Lithium cobaltate cathode material and preparation method thereof and lithium rechargeable battery |
-
2018
- 2018-12-19 CN CN201811556665.5A patent/CN109635473B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1319761A (en) * | 2000-03-15 | 2001-10-31 | 清华大学 | Apparatus and method for high-flux electric rotation detection |
CN104992258A (en) * | 2012-07-05 | 2015-10-21 | 爱利门图供应链管理(开曼)有限公司 | Method and system for controlling supply chains |
US20140380320A1 (en) * | 2013-06-20 | 2014-12-25 | International Business Machines Corporation | Joint optimization of multiple phases in large data processing |
CN103631988A (en) * | 2013-10-22 | 2014-03-12 | 芜湖大学科技园发展有限公司 | Multi-user simulation data management platform for electric power system simulation |
CN108123109A (en) * | 2016-11-28 | 2018-06-05 | 华为技术有限公司 | Lithium cobaltate cathode material and preparation method thereof and lithium rechargeable battery |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110334055A (en) * | 2019-06-06 | 2019-10-15 | 苏州迈高材料基因技术有限公司 | A method of it obtaining material and calculates data |
CN110334055B (en) * | 2019-06-06 | 2023-04-11 | 苏州迈高材料基因技术有限公司 | Method for acquiring material calculation data |
CN113127973A (en) * | 2021-04-16 | 2021-07-16 | 湖南大学 | CAE simulation technology-based multi-material intelligent material selection method and system and electronic equipment |
CN113127973B (en) * | 2021-04-16 | 2022-05-10 | 湖南大学 | CAE simulation technology-based multi-material intelligent material selection method and system and electronic equipment |
Also Published As
Publication number | Publication date |
---|---|
CN109635473B (en) | 2022-08-02 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Wang et al. | A two-phase differential evolution for uniform designs in constrained experimental domains | |
CN106779372A (en) | Based on the agricultural machinery dispatching method for improving immune Tabu search algorithm | |
CN104166903B (en) | Mission planning method and system based on process division | |
CN109815537A (en) | A kind of high-throughput material simulation calculation optimization method based on time prediction | |
Reiter et al. | Allocating indivisible resources affording external economies or diseconomies | |
CN104820636A (en) | Ant colony algorithm based parallel test case sequence generation method | |
Darbandi et al. | Tasks mapping in the network on a chip using an improved optimization algorithm | |
CN109635473A (en) | A kind of heuristic high-throughput material simulation calculation optimization method | |
CN103226728B (en) | High density polyethylene polymerization cascade course of reaction Intelligent Measurement and yield optimization method | |
Hendrix et al. | A scalable algorithm for single-linkage hierarchical clustering on distributed-memory architectures | |
Zhang et al. | A collaborative service group-based fuzzy QoS-aware manufacturing service composition using an extended flower pollination algorithm | |
Zhang | An immune genetic algorithm for simple assembly line balancing problem of type 1 | |
CN106033442A (en) | Parallel breadth-first search method based on shared memory system structure | |
Kazem et al. | A modified simulated annealing algorithm for static task scheduling in grid computing | |
Tang et al. | A parallel shortest path algorithm based on graph-partitioning and iterative correcting | |
Xue et al. | The analysis and research of parallel genetic algorithm | |
Ruchkin et al. | Parallelism in embedded microprocessor systems based on clustering | |
Eklund | A massively parallel GP engine in VLSI | |
Abdolazimi et al. | Connected components of big graphs in fixed mapreduce rounds | |
Ahmed et al. | Solving the Travelling Salesman Problem Using an Ant Colony System Algorithm | |
CN111240988B (en) | Data generation method and generation system for non-equivalent connection load | |
Bożejko et al. | Parallel scatter search algorithm for the flow shop sequencing problem | |
Rudy et al. | GACO: a parallel evolutionary approach to multi-objective scheduling | |
Qing et al. | External topological sorting in large graphs | |
Wever et al. | Automatic machine learning: Hierarchical planning versus evolutionary optimization |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |