CN104298542B - A kind of Laser-Nonlinear transmits parallel simulation method - Google Patents
A kind of Laser-Nonlinear transmits parallel simulation method Download PDFInfo
- Publication number
- CN104298542B CN104298542B CN201410620686.4A CN201410620686A CN104298542B CN 104298542 B CN104298542 B CN 104298542B CN 201410620686 A CN201410620686 A CN 201410620686A CN 104298542 B CN104298542 B CN 104298542B
- Authority
- CN
- China
- Prior art keywords
- cpu
- vector
- time
- laser
- domain
- 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.)
- Expired - Fee Related
Links
Landscapes
- Complex Calculations (AREA)
- Optical Modulation, Optical Deflection, Nonlinear Optics, Optical Demodulation, Optical Logic Elements (AREA)
Abstract
Parallel simulation method is transmitted the present invention relates to a kind of Laser-Nonlinear.Current light laser can be kept more than strong light intensity transmission km in transparent medium, but the low requirement for being extremely difficult to prediction application of computation efficiency and propose.This is based on step Fourier method, significantly improves the simulation velocity of light laser nonlinear transport, and 22 core cpu arithmetic speeds can improve 20 times, and a km is transmitted in emulation light laser in atmosphere only needs the time, meets application requirement.Concrete scheme is that (1) determines light laser light beam parameters and medium parameter, it is determined that the spatial domain and time-domain of emulation, it is determined that the CPU quantity to be dispatched.(2) while it is that each CPU distributes corresponding time arrow and space vector to divide time-domain and spatial domain.(3) each CPU it is determined that space vector and time arrow alternate run, determine CPU in space vector still in time arrow operation according to the minimum principle of data exchange.
Description
Technical field
The present invention relates to computer simulation emulation, more particularly to light laser in atmosphere transmit progress long range height imitate
True technical field.
Background technology
Due to very strong nonlinear effect, light laser is transmitted in transparent medium can break through diffraction limit light intensity with transmission
Distance increases and increased.When light laser is transmitted in transparent medium, diffraction, GVD, Ke Er self-focusings, plasma
Defocus, physical process the vies each other dynamic equilibrium such as high-order Ke Er is defocused, bootstrap propagation phenomenon can be formed in atmosphere.Qiang Ji
The characteristics of optical transport, can be applied to following aspect：
1. remote remote sensing, 2. produce super continuum light spectrum, 3. laser induced lightnings, 4. generation THz waves.
These applications are required for the spatial and temporal distributions for the light intensity transmitted in atmosphere to light laser to carry out Accurate Prediction, and some should
With needing prediction light laser to be transmitted more than km in transparent medium, conventional method is to carry out Crank- in space
It is more than the step Fourier method of Nicholson difference, but its serial approach calculating km that transmission needs the calculating time of several weeks,
Because computational efficiency is low, most transmission ranges for being emulated on current document to light laser are in hundred meters of magnitudes, and propagation of intense laser beam
Using calculating is needed, it is transmitted more than km in atmosphere, and simulation efficiency of the light laser in transparent medium can not meet application
Need.
With the development of computer hardware, computer processor core number is more and more, and hardware is provided for multi-core parallel concurrent
Support.Current step Fourier technique study is in a Dimensional level, its method is simple to Fourier transformation parallelization, process mostly
Need to exchange mass data between (thread), speed-up ratio is low, and light laser can not be represented in the light perpendicular to transmission direction section
Strong distribution situation.
The content of the invention
In view of prior art is not enough above, parallel artificial side is transmitted it is an object of the invention to provide a kind of Laser-Nonlinear
Method, using parallel form, does not divide task with single domain, is allowed to overcome the shortcoming of prior art, improves light laser in non-thread
The simulation efficiency transmitted in property medium.
Method used in the present invention is：
A kind of Laser-Nonlinear transmits parallel simulation method, comprises the following steps：
Laser beam parameter and medium parameter are determined, nonlinear Schrodinger side is set up according to laser parameter and medium parameter
Journey, determines simulation space domain and time-domain, it is determined that the CPU quantity of scheduling；Subdivision grid, calculates inceptive impulse；According to time-domain
The spatial domain vertical with propagation of intense laser beam direction is that CPU distributes time arrow and space vector, each CPU one time of correspondence
Vector sum space vector；Distribution of computation tasks to P CPU；Temporally Vector Parallel calculates nonlinear Schrodinger side to each CPU
The discrete matrix of the journey discrete generation of Crank-Nicholson difference schemes, spatially Vector Parallel solve-temporally vector simultaneously
Row solves times of-spatially Vector Parallel solution, its division time arrow of each each CPU complete independentlies of process or space vector
Business；Wait P CPU to complete simulation calculating, and check that simulation result storage is arrived hard disk by buffering area；
Each CPU emulation mode is as follows in the P CPU：
Temporally Vector Parallel is calculated with Crank-Nicholson difference scheme Discrete Nonlinears Schrodinger side each CPU
Cheng Shengcheng discrete matrix；Spatially vector emulation 1- waits-temporally vectorial solution-wait-spatially vector emulation 2, often
Its division task of the individual each CPU complete independentlies of process；It is non-thread in traditional step Fourier method that spatially vector, which solves module 1,
Property solve the combination of module and Fourier transformation, it is the solution mould in a frequency domain in linear block that temporally vector, which solves module,
Block, it is the inverse Fourier transform in traditional step Fourier method and non-linear solution module group that spatially vector, which solves module 2,
Close；If simulation result is the data needed, simulation result is output to buffering area.
The laser beam parameter includes pulse pattern, carrier frequency, with a tight waist, pulse width, chirp parameter；The laser
Medium parameter includes：GVD parameter, refractive index, Kerr coefficient, high-order Kerr coefficient, multiphoton ionization coefficient.
When distribution of computation tasks to P CPU, while time arrow and space vector are distributed for each CPU, not with list
One time or space are divided, and time and space are divided simultaneously, and correspond in internal memory its time vector sum space vector.
Invention significantly improves the simulation velocity of light laser nonlinear transport, 22 core cpu arithmetic speeds can be improved
20 times, a km is transmitted in emulation light laser in atmosphere only needs the time, meets application requirement.Concrete scheme determines for (1)
Light laser light beam parameters and medium parameter, it is determined that the spatial domain and time-domain of emulation, it is determined that the CPU quantity to be dispatched.(2)
It is that each CPU distributes corresponding time arrow and space vector to divide time-domain and spatial domain simultaneously.(3) each CPU it is determined that
Space vector and time arrow alternate run, determine CPU in space vector still in the time according to the minimum principle of data exchange
Vector operation.Test result shows, speed-up ratio is emulated within 20 cores can emulate a km in one day up to more than 95%, 30 core
More than.
Brief description of the drawings
Fig. 1 is conventional serial simulation process block diagram.
Fig. 2 is the simulation process block diagram of the present invention.
Fig. 3 is i-th of CPU simulation process block diagram.
Embodiment：
The detailed process of the present invention is as shown in Figures 2 and 3.
Embodiment
Determine that the example of one input file form is as follows, is after row # often the problem of needing to solve according to input file
To the annotation of this line：
In order to preferably describe parallel artificial process, Serial Simulation is described first.
Non-linear Schrodinger equation and Ionization Equation that input file creates input file description are read, is asked it is determined that calculating
Domain is solved, according to the mesh parameter subdivision grid in input file.Schrodinger equation and Ionization Equation such as formula (1) and formula (2).
Wherein A is electric field of strong laser complex amplitude, and ρ is that plasma density is the light laser emulation physical quantity to be calculated, and t is
Normalized time, n on the basis of transmission direction2,n4,n6,n8Respectively each non-linear Kerr coefficient of rank, k0For initial laser arteries and veins
Rush wave number, λ0For initial laser wavelength, ω0It is carrier angular frequencies, n0It is λ for wavelength0When linear refractive index, k ＇ ＇ are dispersion systems
Number, ωpeFor plasma frequency, ρ is plasma density ρatFor middle neutral-particle density.
Equation (1) can be reduced to equationWhereinFor linear operator,For nonlinear operator.
Grid parameter, subdivision grid are determined by input file.
Time-domain Fourier transformation to linear operator obtains frequency-domain linear operator, frequency-domain linear operator Crank-
Nicholson discrete is matrix equation
First subscript represents the r of subdivision grid label in above formula, and second subscript is represented in discrete Fourier transform
The l discrete spectrum component.The matrix of the discrete gained of linear segment is triple diagonal matrix, only the component on diagonal and frequency
Rate component is relevant.For convenience, above matrix equation is written asBoundary condition byDraw.
Non-linear partial is solved in time domain,
It is as follows that traditional simulation method describes process
Light laser nonlinear transport parallel simulation method is described below, it is assumed that CPU number is P, in light beam half
Footpath r directions mesh generation quantity is NR, is NR in time t direction mesh generation quantity, and in z directions, mesh generation quantity is NZ.
To CPU labels, CPU marked as 1,2,3 ..., P.I-th of CPU is designated CPU-i
It is the beginning label of the space vector calculated in CPU-i, Re to define 4 P dimensional vectors Rb, Re, Tb, Te, Rb (i)
(i) it is the termination label of the space vector calculated in CPU-i, Tb (i) is the starting mark of the time arrow calculated in CPU-i
Number, Te (i) is the termination label of the time arrow calculated in CPU-i.
Rb (i)=(i-1) * ceil (NR/P), wherein i=1 ..., P.
Re (i)=i*ceil (NR/P) -1, wherein i=1 ..., P-1.
Tb (i)=(i-1) * ceil (NT/P), wherein i=1 ..., P.
Te (i)=i*ceil (NT/P) -1, wherein i=1 ..., P-1.
Re (P)=NR-1, Te (P)=NR-1.
By non-linear-linear-nonlinear combination of traditional simulation method be converted into spatially vector emulation 1- temporally to
Amount emulation-spatially vectorial emulation 2, spatially vector emulation 1 is in Fu of non-linear partial emulation and linear segment emulation
The combination of leaf transformation.Spatially vector emulation 2 is the inverse Fourier transform that linear segment is emulated and the group that non-linear partial is emulated
Close.
Spatially CPU-i factoids are as follows in vector emulation 1
CPU-i factoids temporally in vector emulation are as follows
For (p=Tb (i):Te(i))
Tridiagonal matrix equation is calculated with chasing method
end for
CPU-i factoids spatially in vector emulation 2 are as follows
The control CPU of light laser nonlinear transport parallel simulation method process is described as follows
CPU-i process description is as follows in light laser nonlinear transport parallel simulation method
The parallel simulation method of the present invention is compared as follows with the conventional serial emulation mode calculating time.
Compare example data volume for NR=1200, NT=1024, NZ=1000, conventional method serial approach calculates the time and is
1466.007s, parallel simulation method calculates time and the speed-up ratio such as following table compared with Serial Simulation.
| CPUs | Time (s) | Speed-up ratio (again) | CPUs | Time (s) | Speed-up ratio (again) |
| 2 | 721.78 | 2 | 14 | 105.808 | 13.86 |
| 4 | 365.351 | 4 | 16 | 95.3445 | 15.38 |
| 6 | 243.998 | 6 | 18 | 83.172 | 17.63 |
| 8 | 183.459 | 7.99 | 20 | 77.065 | 19.02 |
| 10 | 147.1346 | 9.96 | 22 | 71.166 | 20.60 |
| 12 | 123.2416 | 11.9 |
The computational efficiency of parallel simulation method is with the increase near-linear increase of CPU quantity as known from the above, when CPU numbers
Measure for 22 when efficiency improve 20.6 times.
The description of the specification of disclosure provided above is exemplary, rather than restricted.According to above-mentioned teaching,
Many modifications and changes of the present invention are all possible.Therefore, it is to preferably explain this hair to select and describe embodiment
Bright principle and its application, and understand those of ordinary skill in the art, on the premise of essence of the present invention is not departed from, Suo Youxiu
Change and change is each fallen within protection scope of the present invention defined by the claims.
Claims (3)
1. a kind of Laser-Nonlinear transmits parallel simulation method, it is characterised in that the described method comprises the following steps：
Laser beam parameter and medium parameter are determined, non-linear Schrodinger equation is set up according to laser parameter and medium parameter,
Simulation space domain and time-domain are determined, it is determined that the CPU quantity of scheduling；Subdivision grid, calculates inceptive impulse；According to time-domain and with
The vertical spatial domain in propagation of intense laser beam direction is that CPU distributes time arrow and space vector, one time arrow of each CPU correspondences
And space vector；Distribution of computation tasks to P CPU；Temporally Vector Parallel calculates non-linear Schrodinger equation use to each CPU
The discrete matrix of the discrete generation of Crank-Nicholson difference schemes, spatially Vector Parallel solve 1- temporally Vector Parallels
Solve times of-spatially Vector Parallel solution, its division time arrow of each each CPU complete independentlies of process or space vector
Business；Wait P CPU to complete simulation calculating, and check that simulation result storage is arrived hard disk by buffering area；
Each CPU emulation mode is as follows in the P CPU：
Temporally Vector Parallel calculates the non-linear Schrodinger equation discrete life of Crank-Nicholson difference schemes to each CPU
Into discrete matrix；Spatially vector emulation 1- waits-temporally vectorial solution-wait-spatially vector emulation 2, each mistake
Its division task of each CPU complete independentlies of journey；Spatially vector solves module 1 and asked to be non-linear in traditional step Fourier method
The combination of module and Fourier transformation is solved, it is the solution module in a frequency domain in linear block that temporally vector, which solves module, is pressed
It is the inverse Fourier transform in traditional step Fourier method and non-linear solution block combiner that space vector, which solves module 2,；Such as
Fruit simulation result is the data needed, then simulation result is output into buffering area.
2. the parallel simulation method of Laser-Nonlinear transmission according to claim 1, it is characterised in that the laser beam
Parameter includes pulse pattern, carrier frequency, with a tight waist, pulse width, chirp parameter；The laser medium parameter includes：Group velocity
Dispersion parameters, refractive index, Kerr coefficient, high-order Kerr coefficient, multiphoton ionization coefficient.
3. the parallel simulation method of Laser-Nonlinear transmission according to claim 1, it is characterised in that calculating task
When distributing to P CPU, while distributing time arrow and space vector for each CPU, do not divided with single time or space, when
Between and space divide simultaneously, and correspond in internal memory its time vector sum space vector.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410620686.4A CN104298542B (en) | 2014-11-06 | 2014-11-06 | A kind of Laser-Nonlinear transmits parallel simulation method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410620686.4A CN104298542B (en) | 2014-11-06 | 2014-11-06 | A kind of Laser-Nonlinear transmits parallel simulation method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104298542A CN104298542A (en) | 2015-01-21 |
| CN104298542B true CN104298542B (en) | 2017-09-08 |
Family
ID=52318282
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410620686.4A Expired - Fee Related CN104298542B (en) | 2014-11-06 | 2014-11-06 | A kind of Laser-Nonlinear transmits parallel simulation method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104298542B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112084621B (en) * | 2020-07-27 | 2024-04-09 | 北京空间机电研究所 | Method, medium and equipment for simulating long-distance transmission of optical fiber laser |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103559383A (en) * | 2013-09-30 | 2014-02-05 | 上海交通大学 | Method for predicting and evaluating strong earthquake resistance of nuclear power station breakwater |
| CN103605872A (en) * | 2013-12-03 | 2014-02-26 | 北京航空航天大学 | Multilevel parallel simulation method for quasi-three-dimensional equation in complex electromagnetic environment |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7881621B2 (en) * | 2006-03-02 | 2011-02-01 | Emcore Corporation | Optical transmission system with directly modulated laser and feed forward noise cancellation |
-
2014
- 2014-11-06 CN CN201410620686.4A patent/CN104298542B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103559383A (en) * | 2013-09-30 | 2014-02-05 | 上海交通大学 | Method for predicting and evaluating strong earthquake resistance of nuclear power station breakwater |
| CN103605872A (en) * | 2013-12-03 | 2014-02-26 | 北京航空航天大学 | Multilevel parallel simulation method for quasi-three-dimensional equation in complex electromagnetic environment |
Non-Patent Citations (2)
| Title |
|---|
| "基于分步傅里叶变换法对非线性薛定谔方程的数值仿真";李莹，崔庆丰;《长春理工大学学报（自然科学版）》;20110331;第34卷(第1期);第43-45页 * |
| "超快激光光丝阵列产生机理研究";高慧;《中国博士学位论文全文数据库 基础科学辑》;20140615(第06期);第33-36页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104298542A (en) | 2015-01-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Janssen | Progress in ocean wave forecasting | |
| Choromański et al. | Scale-free graph with preferential attachment and evolving internal vertex structure | |
| Chalikov et al. | Modeling extreme waves based on equations of potential flow with a free surface | |
| Chalikov et al. | Numerical modeling of 3D fully nonlinear potential periodic waves | |
| Zeng et al. | A novel interval grey prediction model considering uncertain information | |
| Vincenti et al. | Detailed analysis of the effects of stencil spatial variations with arbitrary high-order finite-difference Maxwell solver | |
| Bülbül et al. | A Taylor matrix method for the solution of a two-dimensional linear hyperbolic equation | |
| Cannon et al. | A GPU-accelerated finite-difference time-domain scheme for electromagnetic wave interaction with plasma | |
| Nielsen | Transformation of a wave energy spectrum from encounter to absolute domain when observing from an advancing ship | |
| CN104253650B (en) | The estimation unit of intrachannel nonlinear damage and method | |
| Xin et al. | High order nonlocal symmetries and exact interaction solutions of the variable coefficient KdV equation | |
| CN104794289A (en) | Implementation method for complete matching of absorbing boundary under expansion rectangular coordinate system | |
| Wu et al. | Instruction on the construction of coherent structures based on variable separation solutions of (2+ 1)-dimensional nonlinear evolution equations in fluid mechanics | |
| Li et al. | A new numerical scheme for the nonlinear Schrödinger equation with wave operator | |
| Garayshin et al. | An analytical and numerical study of long wave run-up in U-shaped and V-shaped bays | |
| CN104298542B (en) | A kind of Laser-Nonlinear transmits parallel simulation method | |
| Miao et al. | Nonlocal symmetries and explicit solutions of the AKNS system | |
| Xiao et al. | Hybrid time-domain model for ship motions in nonlinear extreme waves using HOS method | |
| Bruzón et al. | On the similarity solutions and conservation laws of the Cooper‐Shepard‐Sodano equation | |
| Yucheng et al. | Incremental learning method of least squares support vector machine | |
| CN109905151A (en) | Multi-modal irregular OAM electromagnetic wave generation, separation method and device based on UCA antenna | |
| Zhang et al. | Lump and Lump‐Type Solutions of the Generalized (3+ 1)‐Dimensional Variable‐Coefficient B‐Type Kadomtsev‐Petviashvili Equation | |
| Korotkevich | Influence of the condensate and inverse cascade on the direct cascade in wave turbulence | |
| Neirameh | New soliton solutions to the fractional perturbed nonlinear Schrodinger equation with power law nonlinearity | |
| Di Cerbo | On Fujita’s log spectrum conjecture |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20170908 Termination date: 20201106 |