CN108363865A - The asynchronous transport simulation method and system that PARTICLE TRANSPORT FROM domain decomposition parallel calculates - Google Patents
The asynchronous transport simulation method and system that PARTICLE TRANSPORT FROM domain decomposition parallel calculates Download PDFInfo
- Publication number
- CN108363865A CN108363865A CN201810135323.XA CN201810135323A CN108363865A CN 108363865 A CN108363865 A CN 108363865A CN 201810135323 A CN201810135323 A CN 201810135323A CN 108363865 A CN108363865 A CN 108363865A
- Authority
- CN
- China
- Prior art keywords
- particle
- subdivision
- node
- regional
- simulation
- 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.)
- Pending
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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Hardware Design (AREA)
- Evolutionary Computation (AREA)
- Geometry (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
A kind of asynchronous transport simulation method and system that PARTICLE TRANSPORT FROM domain decomposition parallel calculates, this method include：The domain mesh for combining geometry is obtained multiple subregions by host process using the subdivision algorithm of more subdivision nodes；Host process is the subregion and particle source of each regional process distribution simulation；Regional process obtains particle from the particle source, and simulation particle transports；The particle that regional process will be moved to subzone boundaries is sent to corresponding another regional process, or, the particle that other regional process are sent is received, and updates the particle migration record of the subregion according to the number of particles for sending or receiving between other regional process；When not when the particle simulated, the particle migration record of the subregion is sent to host process by regional process in subregion；The particle migration that host process receives each regional process is recorded and is checked, and judges whether end simulation.This method can avoid global communication, reduce the process stand-by period of entire algorithm, ensure higher parallel efficiency.
Description
Technical field
This application involves Monte Carlo PARTICLE TRANSPORT FROM analogue technique field, more particularly to one kind are defeated for Monte Carlo particle
Transport the asynchronous transport simulation method and system that domain decomposition parallel calculates.
Background technology
In nuclear reactor designs, criticality safety analysis and SHIELDING CALCULATION, it is several that particle (neutron and photon), which transports solution,
The core that value simulation calculates.Particle transport theory be research microcosmic particle (neutron, photon etc.) in medium movement (including production
Nuclear reaction occurs for raw, migration etc.) process and rule theory.Monte Carlo (Monte Carlo, abbreviation MC) PARTICLE TRANSPORT FROM
Method is also referred to as stochastic simulation method or statistics curvaturev, is a kind of calculating side for carrying out stochastic simulation to particle behavior on computers
Method.Monte Carlo method is a kind of common method for numerical simulation in particle transport theory, and first, it establishes true medium (packet
Include geometry and medium nucleic composition) simulation model, then simulate a large amount of particles in the medium movement (including particle production
Raw, migration, collide etc. physical processes with the atomic nucleus of medium) and record appropriate is carried out, finally all particles are generated
Record it is for statistical analysis, obtain numerical result.
Due to Monte Carlo method using parsing combination geometric description mode establish medium geometry and without
It is discrete, therefore geometry approximation very little；Using the energy process mode in Continuous Energy point section, particle is determined using random number
With medium nucleic reaction type etc., physical model mechanism improvement, therefore the application in nuclear reactor physical analysis field increasingly
Paid attention to by researcher, is known as the ultimate method of reactor numerical simulation.
Monte Carlo method needs to simulate a large amount of particles (ten million or even hundred million magnitudes), and serial program is all to simulate one by one, grain
The simulated time that sub- simulation process occupies is very long.But actually particle is independent from each other each other, has naturally parallel
Property, it can be solved by parallel computation：Equipped with N number of process, each process replicates entire dielectric model, then each process point
It is simulated with 1/N particle, record result is finally integrated together for statistical analysis obtain as a result, can so save big
The simulated time of amount.Good parallel efficiency would generally be obtained based on the parallel Monte Carlo method of particle, is Monte Carlo grain
The most traditional parallel method of sub- transport simulation.
However, as numerical simulation is in the requirement refined with physical modeling that deepens continuously, nuclear reactor field proposes
Whole core Careful Numerical Simulation, dielectric model contain up to ten million a solids, include hundreds of nucleic in each solid, entirely
Dielectric model memory requirements has reached tens of G or more, far more than the memory size that is matched of CPU core of current computer, one
The memory size of course allocation can not replicate entire dielectric model, lead to not realize using the parallel Monte Carlo method of particle
It calculates.
Therefore the domain decomposition parallel computational methods for needing development Monte Carlo PARTICLE TRANSPORT FROM, entire dielectric model is carried out
Domain mesh is divided into multiple subregions, and distribution is stored on the memory headroom of multiple process proportionings, and multiple processes realize parallel meter
It calculates.Some such process just will appear particle in simulation particle movement in the medium and leave local subregion and enter it
The case where his subregion, this just needs the process that particle is sent to other processes.The Region Decomposition of Monte Carlo PARTICLE TRANSPORT FROM
Parallel computation necessarily introduces the communication of the transmission particle during simulation particle, if processing is bad to substantially reduce parallel effect
Rate.
Currently, in the implicit Monte Carlo that heat radiation transports (Implicit Monte Carlo, abbreviation IMC) method,
Have and transported algorithm and ripe program for what domain decomposition parallel calculated, such as the KULL programs and Milagro programs in the U.S..
KULL programs are the radiation transport simulation programs of U.S.'s LLNL development in laboratory, and IMC software packages can realize region
Parallel computation is decomposed, PARTICLE TRANSPORT FROM algorithm is after the particle simulation that all processes complete local subregion, between carry out process
Transmission particle communication, be synchronous communication algorithm.
Milagro programs are the radiation transport simulation journeys for using asynchronous communication algorithm of U.S.'s LANL development in laboratory
Sequence, the termination condition of asynchronous communication algorithm are to judge whether the particle number for completing simulation is equal to source specified when simulation starts
Particle number.This is because in radiation transport simulation process, particle not will produce secondary, so particle number will not be sent out
Changing；And in PARTICLE TRANSPORT FROM simulation, either neutron or photon can all occur nuclear reaction and generate secondary neutron or secondary
Grade photon, therefore the particle number simulated will be more than source particle number specified when starting.Therefore the knot of the asynchronous communication algorithm
Beam condition cannot be used for PARTICLE TRANSPORT FROM simulation.
Invention content
To solve the above problem in the prior art, the purpose of the application is to propose a kind of PARTICLE TRANSPORT FROM region point
The asynchronous transport simulation method and system for solving parallel computation realizes the transmission particle between multiple processes using point-to-point communication
Communication, has been overlapped simulation and call duration time, it is especially set out it is asynchronous transport termination condition, avoid global communication, reduce entire
The process stand-by period of algorithm, to improve parallel efficiency.
In order to achieve the above objectives, the asynchronous of PARTICLE TRANSPORT FROM domain decomposition parallel calculating that the embodiment of the present application proposes transports mould
Quasi- method, including：Host process carries out subdivision using the subdivision algorithm of more subdivision nodes to the region for combining geometry, obtains multiple sons
Region；Host process is the subregion and particle source of each regional process distribution simulation；Regional process obtains grain from the particle source
Son, simulation particle transport；The particle that regional process will be moved to subzone boundaries is sent to corresponding another regional process, or
Person receives the particle that other regional process are sent, and according to the number of particles for sending or receiving between other regional process
Update the particle migration record of the subregion；When not when the particle simulated, regional process is by the sub-district in subregion
The particle migration record in domain is sent to host process；The particle migration that host process receives each regional process is recorded and is checked, and judgement is
No end simulation.
In order to achieve the above objectives, what the embodiment of the present application proposed calculates for Monte Carlo PARTICLE TRANSPORT FROM domain decomposition parallel
Asynchronous transport simulation system, including：Host process, for being cutd open the region for combining geometry using the subdivision algorithm of more subdivision nodes
Get multiple subregions, is moved for the subregion and particle source, the particle for receiving each regional process of each regional process distribution simulation
It moves and records and check, judge whether end simulation；Regional process, for obtaining particle from the particle source, simulation particle is defeated
Fortune, the particle that will be moved to subzone boundaries are sent to corresponding another regional process, alternatively, receiving the transmission of other regional process
The particle come, and the particle migration of the subregion is updated according to the number of particles for sending or receiving between other regional process
Record, and, when not when the particle simulated, the particle migration of the subregion is recorded and sent by regional process in subregion
To host process.
By the above technical solution provided by the embodiments of the present application as it can be seen that being handled between multiple processes by point-to-point communication
Particle transmits, and the communication between two regional process does not influence the simulation task of other regional process, when being overlapped simulation and communication
Between, the termination condition of asynchronous transport simulation is limited by particle migration record, avoids global communication, reduces entire algorithm
The process stand-by period, ensure that higher parallel efficiency.
The additional aspect of the application and advantage will be set forth in part in the description, and will partly become from the following description
It obtains obviously, or recognized by the practice of the application.
Description of the drawings
In order to illustrate the technical solutions in the embodiments of the present application or in the prior art more clearly, to embodiment or will show below
There is attached drawing needed in technology description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
Some embodiments of application for those of ordinary skill in the art without creative efforts, can be with
Obtain other attached drawings according to these attached drawings.
Fig. 1 is that being calculated for Monte Carlo PARTICLE TRANSPORT FROM domain decomposition parallel for one embodiment of the application proposition is asynchronous defeated
Transport the flow diagram of analogy method；
Fig. 2 is subdivision root node and subdivision node schematic diagram on one embodiment topology information of the application；
Fig. 3 is that one embodiment subdivision node subdivision of the application is two area schematics；
Fig. 4 is that the asynchronous transport simulation termination condition that one embodiment of the application proposes judges schematic diagram；
Fig. 5 is that the asynchronous of Monte Carlo PARTICLE TRANSPORT FROM domain decomposition parallel calculating that be used for of one embodiment of the application transports mould
The structural schematic diagram of quasi- system；
Fig. 6 is asynchronous the transporting for the calculating of Monte Carlo PARTICLE TRANSPORT FROM domain decomposition parallel of another embodiment of the application
The flow diagram of analogy method.
Specific implementation mode
The embodiment of the present application provides a kind of asynchronous transporting mould for what Monte Carlo PARTICLE TRANSPORT FROM domain decomposition parallel calculated
Quasi- method and system.
In order to make those skilled in the art better understand the technical solutions in the application, below in conjunction with the application reality
The attached drawing in example is applied, technical solutions in the embodiments of the present application is clearly and completely described, it is clear that described implementation
Example is merely a part but not all of the embodiments of the present application.Based on the embodiment in the application, this field is common
The every other embodiment that technical staff is obtained without creative efforts should all belong to the application protection
Range.
Fig. 1 is that being calculated for Monte Carlo PARTICLE TRANSPORT FROM domain decomposition parallel for one embodiment of the application proposition is asynchronous defeated
The flow diagram for transporting analogy method, as shown in Figure 1, this method includes：
Step 101, host process carries out subdivision using the subdivision algorithm of more subdivision nodes to the region for combining geometry, obtains more
Sub-regions；
Step 102, host process is the subregion and particle source of each regional process distribution simulation；
Step 103, regional process obtains particle from the particle source, and simulation particle transports；
Step 104, regional process will be moved to the particles of subzone boundaries and be sent to corresponding another regional process, or
Person receives the particle that other regional process are sent, and according to the number of particles for sending or receiving between other regional process
Update the particle migration record of the subregion；
Step 105, when not when the particle simulated, regional process remembers the particle migration of the subregion in subregion
Record is sent to host process；
Step 106, host process receives the particle migration of each regional process and records and check, and judges whether end simulation.
More subdivision nodes refer under subdivision root node, child node are set to subdivision node, in subdivision knot
Subdivision is carried out in point.
Wherein, the total number of particles for being sent to other processes that local process is recorded and the particle received from other processes are total
Number, referred to as " particle migration record ".Without loss of generality, it is assumed that all there is particle communication between any two process, then " particle moves
The length (referring to array length) of shifting record " is exactly 2 (P-1), and wherein P is the regional process for executing domain decomposition parallel simulation
Number.Particle migration record is that the asynchronous algorithm that transports terminates the important evidence judged.
Receive the particle library that particle library is used to store the particle that local process is sent received from other processes.
In order to avoid the particle accumulation for receiving particle library is excessive, the simulation priority for being preferably arranged to receive particle library is higher than
The extraction Particle tracking simulation first from reception particle library of particle source, i.e. local process, when there is no particle in reception particle library,
Particle is just extracted from particle source (set of the particle i.e. in original partition subregion).In simulation process, regional process note
The heading of particle, movement track and various change are recorded, such as particle is absorbed, fly out subregion, according to certain condition
Give up particle (particle is dead) etc..Particle source is the initial distribution for the extraction particle being arranged when simulation, generates the source of simulation particle
Head.
To meet the memory requirements of PARTICLE TRANSPORT FROM calculating, when carrying out high-resolution simulation calculating, calculated using Region Decomposition
Method (algorithms of more subdivision nodes).For realization Region Decomposition, need to carry out subdivision to the region for combining geometry, before subdivision,
Firstly the need of one subdivision root node of selection.Subdivision root node refers to the knot on the solid correspondence combination topology information for carry out subdivision
Point, subdivision node determine that it is in subdivision node that the region to combining geometry, which carries out subdivision, in the child node of subdivision root node
It is carried out in child node.Subdivision root node and subdivision node are typically all to be served as by the node with many child nodes and offspring, such as
Shown in Fig. 2.
Using the algorithm of more subdivision nodes, by the way that entire dielectric model is carried out domain mesh to the region for combining geometry,
It is divided into multiple subregions.Subdivision root node and subdivision node can automatically be searched for by more subdivision Node Algorithms or be referred to by user
It is fixed.
Specifically, when host process specifically carries out subdivision, subdivision root node is selected first, under the subdivision root node of selection
Multiple subdivision nodes are determined in multiple child nodes, subdivision goes out multiple subregions from the child node of determining multiple subdivision nodes,
As shown in Figure 3.
As one embodiment of the invention, the selection gist of subdivision root node is：Using subdivision root node as the son of root node
Total first setting ratio of higher than whole tree node of the node sum of tree；And using the child node of subdivision root node as the son of root node
Total second certainty ratio of less than whole tree node of the node sum of tree.First setting ratio can be 70%, and the second certainty ratio can
Think 20%, invention is not limited thereto.
As one embodiment of the invention, multiple subdivision knots are determined from multiple child nodes under the subdivision root node of selection
Point foundation be：It is higher than using subdivision root node as the subtree of root node by the node sum of the subtree of root node of subdivision node
Node sum third setting ratio；And it is less than with subdivision root by the node sum of the subtree of root node of the child node of subdivision node
Node is total 4th certainty ratio of node of the subtree of root node.Second setting ratio can be 80%, and the second certainty ratio can be with
It is 30%, invention is not limited thereto.
After searching suitable subdivision node, the child node of subdivision node can be carried out by bounding volume method uniform
Subdivision, and then each process will replicate portion for subdivision node and its sibling, father node and ancestors' node.By with top
Method can reach preferable load balance effect to combination geometric areas decomposition.
Multiple subregion distributions are stored on the memory headroom of multiple process proportionings, and as multiple regions course allocation wants mould
Quasi- subregion obtains particle respectively by multiple regions process from the particle source of corresponding sub-region and is simulated, realized simultaneously
Row calculates.During in each region, process carries out particle simulation, the case where simulation particle moves in the medium, and particle exists
It is moved in the medium that each smaller subregion is simulated, it is likely that the boundary of subregion can be moved to, and generate and move to phase
The trend of other adjacent subregions.At this point, regional process will be moved to the particle of the subzone boundaries of oneself institute's management and control, according to grain
The factors such as the position of son and the direction of motion judge the particle will move to which adjacent subregion, have also determined that this
Which regional process particle is sent to and carries out subsequent simulation.Equally, the particle of other regional process simulation also will produce in this way
Trans-regional movement, then local process be also possible to receive the particle sent from other processes.When local particle source
It is all modeled end with the particle received, can not continue the grain of simulation in the subregion that regional process is in charge of
Son, at this point, regional process can come and go particle to receive is sent to host process with the record, that is, particle migration record sent, etc.
New particle to be received is simulated, or until the END instruction of host process can formal end simulation.Host process receives each
The particle migration of regional process is recorded and is checked, the trend and quantity of the particle recorded in being recorded according to particle migration, and judgement is
The quantity forwarded of no particle is corresponding with received quantity, that is, verifies whether that all particles all " are carried it through to the end ", to judge to be
No end simulation.
In this application, for regional process is opposite host process, for a sub-regions, with this sub-regions
Corresponding regional process is alternatively referred to as local process, is hereinafter no longer explained to this.
Embodiments herein handles the particle between multiple processes by point-to-point communication and transmits, between two regional process
Communication do not influence the simulation tasks of other regional process, avoid global communication, be overlapped simulation and call duration time, ensure that
Higher parallel efficiency.
According to one embodiment of the application, the host process is a specified regional process, the particle migration note
Record include the regional process be sent to other regional process population and the regional process received from other regions into
The population of journey, wherein the particle migration that the host process receives each regional process is recorded and checked, and judges whether to terminate mould
It is quasi-, it specifically includes：Host process summarizes the particle migration record of each regional process, counts the total number of particles sent and received；Work as hair
When the total number of particles sent is equal with the total number of particles of reception, end of identification is sent to each regional process.
According to one embodiment of the application, the host process is recorded according to the particle migration of each regional process, and judgement is
No end simulation, further comprises：Host process, which records the particle migration received, to be merged, and is generated sending metrix and is received square
Battle array, wherein the matrix element in the sending metrix and receiving matrix is particle quantity forwarded or the grain between two regional process
Sub- received quantity；Judge whether the sending metrix is equal to receiving matrix；If so, end simulation, otherwise continues to simulate, and wait
New particle migration record to be received.
According to one embodiment of the application, when regional process sends particle to another regional process, the method is also
Including：The particle that will be moved to corresponding subzone boundaries is stored to transmission buffer area；When to be sent in the transmission buffer area
When reaching preset quantity to the particle of another regional process, the particle in the transmission buffer area is sent to corresponding another area
Domain process.
According to one embodiment of the application, the particle by the transmission buffer area is sent to corresponding another area
Domain process, specifically includes：Ready message is sent to corresponding another regional process, wherein ready message includes being sent
Particle number；The confirmation signal that corresponding another regional process returns is received, point-to-point communication is established with it；It will be to be sent
The analog information of particle be sent to corresponding another regional process.
It is described when regional process receives the particle that other regional process are sent according to one embodiment of the application
Method further includes：It detects whether to receive ready message；If receiving, received according to the particle number application in ready message
Spatial cache；Confirmation signal is replied, point-to-point communication is established with other described regional process；Other described regions are received to send
Particle analog information, and be stored in and receive in particle library.
According to one embodiment of the application, the analog information of the particle received is stored in and receives it in particle library
Afterwards, further include：Particle is obtained from the reception particle library and carries out transport simulation, wherein the simulation for receiving particle library is excellent
First grade is higher than the simulation priority of the particle source.
According to one embodiment of the application, the host process is a regional process being previously designated, for example, in order to
The regional process that wherein process number is 0 after distribution region process can be specified as host process for the sake of convenient.Host process is removed
Outside the function of realizing host process, particle simulation equally is carried out to the subregion particle library being locally-assigned, sends and receives particle
Process it is also essentially identical with common regional process, only record in regional process and host process transmission particle migration and wait for knot
The process of beam has obvious difference, because itself is exactly host process, is just not necessarily to retransmit particle after itself completing simulation
Migration record.
The specific embodiment of the application is described in detail below.
Local process is in simulation particle, (the divided sub-district of dielectric model when Particles Moving is to subzone boundaries
The boundary in domain), transmission buffer area is stored it in, has expired (such as N number of particle) until sending buffer area, has sent jointly to correspond to
Process：Local process sends process of the particle number to target area first, then waits for the return signal of other side always, until
Target process, which has sent, to be got out received signal and comes, and just the particle of buffer area is sent；Since local process is waiting
During the return signal for waiting for target process, there may be other processes (including target process) to send particle and come, so waiting for
While local process to check whether there is particle and send over, if so, then receiving particle.After being sent, local process note
Record the particle number sent to this target area process, while the area that empties the cache.
Whether local process can periodically check has particle to need to receive, such as the simulation in every completion to M particle
Afterwards, the information (such as information encoding, particle number, particle data etc.) for whether receiving other processes and sending checked.
If it was found that when receiving particle number (such as integer) for some process transmission, local process has been known that other side's process has
Particle will send over, and apply for order caching area at once, and then return signal is gone over, and point-to-point communication is established with other side's process,
Receive real particle information.After successful reception, the particle received is stored in and is received in particle library, while particle moves
Move the cumulative particle number received of record.It, can be by the sequence for receiving particle information when there is multiple processes transmission particles to come
It handles successively.It completes after receiving particle, process is simulated from extraction particle in particle library is received.
It all simulates and is over when the particle in local process discovery reception particle library and particle source, just in all transmission buffer areas
The particle still having all is sent, even if number deficiency is N number of, makes itself to be in idle condition.Since regional process oneself may be used also
To be host process (Master processes), so next needing to be divided into following two situation processing：
1, the process is not Master processes (host process)：Particle migration record be sent to Master processes, then into
Enter wait state, is simulated if the particle for receiving the transmission of other processes in waiting time, also more after simulation
Particle migration record after new, which retransmits, gives Master processes, continues waiting for；If receiving end of identification, exit, terminates
Transport simulation；
2, the process is Master processes：Particle migration record is received, verification particle migration record judges whether to terminate.
Specifically, the verification of Master processes receives, sends population, the number of particles that sends and receives is equal just represent it is all into
The particle of journey, which is simulated, to be terminated, and sending end of identification if terminating gives other regional process, if not terminating, Master processes
It into wait state, is simulated if receiving particle, receives new particle migration record and check again, it is correct until checking
Until sending out end of identification.
Specifically verification algorithm is, for example,：Master processes, which record the particle migration of each region process received, to be merged
At sending and receiving two matrixes：S-matrix and R matrixes judge that whether equal the two matrixes are (as shown in Figure 4).Wherein, S squares
The matrix element s of battle arrayi,jExpression process i is sent to the total number of particles of process j, the particle migration record sent from process i；R matrixes
Matrix element ri,jTotal number of particles of the expression process j received from process i, the particle migration record sent from process j, it is clear that S squares
Element in battle array and R matrix leading diagonals is all 0.The time complexity of matrix audit process is O (P2), P be Region Decomposition simultaneously
Traveling journey number.
As shown in the above, embodiments herein handles the particle between multiple processes by point-to-point communication and passes
It passs, the communication between two regional process does not influence the simulation task of other regional process, has been overlapped simulation and call duration time, has passed through
Particle migration records to limit the termination condition of asynchronous transport simulation, avoids global communication, reduces the process of entire algorithm
Stand-by period ensure that higher parallel efficiency.
Based on same inventive concept, the embodiment of the present application also provides one kind being used for Monte Carlo PARTICLE TRANSPORT FROM Region Decomposition
The asynchronous transport simulation system of parallel computation, can be used to implement method described in above-described embodiment, such as the following examples
It is described.The principle solved the problems, such as due to the asynchronous transport simulation system calculated for Monte Carlo PARTICLE TRANSPORT FROM domain decomposition parallel
It is similar to the asynchronous transport simulation method calculated for Monte Carlo PARTICLE TRANSPORT FROM domain decomposition parallel, it is accordingly used in Monte Carlo
The implementation for the asynchronous transport simulation system that PARTICLE TRANSPORT FROM domain decomposition parallel calculates may refer to be used for Monte Carlo PARTICLE TRANSPORT FROM
The implementation for the asynchronous transport simulation method that domain decomposition parallel calculates, overlaps will not be repeated.It is used below, term
The combination of the software and/or hardware of predetermined function may be implemented in " unit " or " module ".Although described in following embodiment
System is preferably realized with software, but the realization of the combination of hardware or software and hardware is also that may and be contemplated.
Fig. 5 is that the asynchronous of Monte Carlo PARTICLE TRANSPORT FROM domain decomposition parallel calculating that be used for of one embodiment of the application transports mould
The structural schematic diagram of quasi- system.The system of the present embodiment can be to realize that the logical block of corresponding function is constituted, or fortune
Row has the electronic equipment of corresponding function software.
As shown in figure 5, this is used for the asynchronous transport simulation system packet that Monte Carlo PARTICLE TRANSPORT FROM domain decomposition parallel calculates
It includes：Host process 10 and regional process 11.
Specifically, host process 10, for for each regional process distribution simulation subregion and particle source, receive each region into
The particle migration of journey is recorded and is checked, and judges whether end simulation；
Regional process 11, for obtaining particle from the particle source, simulation particle transports, and will be moved to subzone boundaries
Particle be sent to corresponding another regional process, alternatively, receive the particle that other regional process are sent, and according to other
The number of particles for sending or receiving between regional process updates the particle migration record of the subregion, and, when in subregion
Not when the particle simulated, the particle migration record of the subregion is sent to host process by regional process.
Using the algorithm of more subdivision nodes, by the way that entire dielectric model is carried out domain mesh to the region for combining geometry,
It is divided into multiple subregions.Subdivision root node and subdivision node can automatically be searched for by more subdivision Node Algorithms or be referred to by user
It is fixed.
Specifically, when host process specifically carries out subdivision, subdivision root node is selected first, under the subdivision root node of selection
Multiple subdivision nodes are determined in multiple child nodes, subdivision goes out multiple subregions from the child node of determining multiple subdivision nodes,
As shown in Figure 3.
As one embodiment of the invention, the selection gist of subdivision root node is：Using subdivision root node as the son of root node
Total first setting ratio of higher than whole tree node of the node sum of tree；And using the child node of subdivision root node as the son of root node
Total second certainty ratio of less than whole tree node of the node sum of tree.First setting ratio can be 70%, and the second certainty ratio can
Think 20%, invention is not limited thereto.
As one embodiment of the invention, multiple subdivision knots are determined from multiple child nodes under the subdivision root node of selection
Point foundation be：It is higher than using subdivision root node as the subtree of root node by the node sum of the subtree of root node of subdivision node
Node sum third setting ratio；And it is less than with subdivision root by the node sum of the subtree of root node of the child node of subdivision node
Node is total 4th certainty ratio of node of the subtree of root node.Second setting ratio can be 80%, and the second certainty ratio can be with
It is 30%, invention is not limited thereto.
After searching suitable subdivision node, the child node of subdivision node can be carried out by bounding volume method uniform
Subdivision, and then each process will replicate portion for subdivision node and its sibling, father node and ancestors' node.By with top
Method can reach preferable load balance effect to combination geometric areas decomposition.
Specifically, multiple subregion distributions are stored on the memory headroom of multiple process proportionings, as multiple regions process
Subregion to be simulated is distributed, particle is obtained from the particle source of corresponding sub-region respectively by multiple regions process carries out mould
It is quasi-, realize parallel computation.During in each region, process carries out particle simulation, feelings that the particle of simulation moves in the medium
Condition, and particle moves in the medium that each smaller subregion is simulated, it is likely that the boundary of subregion can be moved to, and is produced
The raw trend for moving to other adjacent subregions.At this point, regional process will be moved to the subzone boundaries of oneself institute's management and control
Particle judges the particle will move to which adjacent subregion, also according to factors such as the position of particle and the directions of motion
It determines and this particle is sent to which regional process carries out subsequent simulation.Equally, the particle of other regional process simulation
Also such trans-regional movement is will produce, then local process is also possible to receive the particle sent from other processes.When
Local particle source and the particle received are all modeled end, can be in the subregion that regional process is in charge of
Continue the particle of simulation, is sent out with the record, that is, particle migration record sent at this point, regional process can come and go particle to receive
Host process is given, new particle to be received is waited to be simulated, or until the END instruction of host process can formal end simulation.
The particle migration that host process receives each regional process is recorded and is checked, the trend of the particle recorded in being recorded according to particle migration and
Quantity judges whether that the quantity forwarded of particle with received quantity is corresponding, that is, verifies whether that all particles all " are carried it through to the end ",
To judge whether end simulation.
Wherein, the host process is a specified regional process.For example, can be specified for convenience's sake in distribution area
The regional process that wherein process number is 0 after the process of domain is as host process.Host process is other than realizing the function of host process, equally
Particle simulation is carried out to the subregion particle library that is locally-assigned, send and receive the process of particle also with common regional process base
This is identical, is only recorded in regional process and host process transmission particle migration and the process of end is waited for have obvious difference, because
Itself to be exactly host process, just it is not necessarily to retransmit particle migration record after itself completing simulation.
This system handles the particle between multiple processes by point-to-point communication and transmits, and the communication between two regional process is not
The simulation task for influencing other regional process, avoids global communication, has been overlapped simulation and call duration time, ensure that it is higher simultaneously
Line efficiency.
In a specific embodiment, the particle migration record includes that regional process is sent to other regional process
The population of population and regional process received from other regional process, the host process are specifically used for summarizing each regional process
Particle migration records, and counts the total number of particles sent and received, and, when the total number of particles phase of the total number of particles and reception of transmission
Whens equal, end of identification is sent to each regional process.
In a specific embodiment, the system also includes the corresponding transmission buffer areas of each regional process and reception grain
Word bank, the transmission buffer area is for preserving the particle for moving to corresponding subzone boundaries, and the reception particle library is for protecting
Deposit the particle received from other regional process.
The regional process is additionally operable to obtain particle progress transport simulation from the reception particle library, wherein described to connect
The simulation priority for receiving particle library is higher than the simulation priority of the particle source.
Embodiments herein handles the particle between multiple processes by point-to-point communication and transmits, between two regional process
Communication do not influence the simulation tasks of other regional process, be overlapped simulation and call duration time, limited by particle migration record
The termination condition of fixed asynchronous transport simulation, avoids global communication, reduces the process stand-by period of entire algorithm, ensure that compared with
High parallel efficiency.
During the specific implementation of the embodiment of the present application, the entire algorithm of asynchronous analog can be divided into 3 parts, sent
Particle algorithm receives particle algorithm and waits for and terminates algorithm, is packaged into 3 functions：Send particle function, receive particle function and
It waits for and terminates function.
Specifically, wherein sending particle function includes：Particle is stored to buffer area is sent, until sending buffer area completely
(such as be ready for sending and had N number of particle to the buffer area of adjacent area process C), send jointly to corresponding region into
Journey：Local process sends particle number to other side first, then waits for the return signal of other side always, until target process is sent
Getting out received signal comes, and just the particle of buffer area is sent in the past (due to being simulation process, so actually sending
It is the analog information of particle)；Since local process is during the return signal for waiting for target process, there may be other regional process
(including local process is sending the target area process of particle) sends particle and comes, so local process will check whether there is
Particle sends over, if so, then receiving particle.After being sent, the particle number that local process record is sent, while emptying slow
Deposit area.
Receiving particle function includes：According to the integer (particle number) received, then application order caching area is returned
Complex signal is gone over, and is established point-to-point communication with other side's process, is received real particle information.After successful reception, reception
To particle be stored in and receive in particle library, while the particle number that particle migration record receives.Grain is sent when there are multiple processes
When son comes, handled successively by information sequence is received.
Start to wait for after the completion of all particle simulations in the zone and terminate function, waiting terminates function and includes：All hairs
The particle still having in buffer area is sent all to send, even if number deficiency is N number of.If this process is not Master processes, grain
Son migration record is sent to Master processes, subsequently into wait state, if receiving the transmission of other processes in waiting time
Particle then calls and receives particle function reception particle, is then back to particle number, exits；If receiving end of identification, return
- 1 is returned, is exited；If this process is Master processes, particle migration record is received, two matrixes is merged into, compares two matrixes
It is whether equal, if two matrixes are equal, send end of identification and give other processes, return to -1, exit；If two matrixes differ,
Then remain waiting for state；If receiving the particle of other processes transmission in waiting time, calls and receive particle function reception grain
Son receives particle function and returns to the particle number received, exits.
In conventional Monte Carlo PARTICLE TRANSPORT FROM algorithm, increase the calling of above three function, so that it may be applicable in realize
In the asynchronous transport simulation algorithm that Monte Carlo PARTICLE TRANSPORT FROM domain decomposition parallel calculates.As shown in fig. 6, specific implementation mode is such as
Under, place is repeated with previous embodiment, is not being repeated：
Step 1, Master processes are the subregion and particle source of each regional process distribution simulation, start simulation particle.
Step 2, regional process obtains a particle from local.If obtaining successfully, step 3 is executed, if failure, illustrates
Local particle has been simulated, step 9 is executed.
Step 3, simulation particle.
Step 4, judge whether particle reaches subzone boundaries.If so, 5 are thened follow the steps, if particle does not reach sub-district
Domain boundary just completes to simulate, and thens follow the steps 6.
Step 5, it calls and sends particle function, which is sent to other processes.
Step 6, the simulation to the particle is completed.
Step 7, the simulation for often completing M particle, checks whether there is particle and sends over, 8 are entered step if having；If not yet
Have, is then directly entered step 2；
Step 8, it calls and receives particle function, preserve the particle received.Subsequently into step 2, continue simulation particle.
Step 9, call waiting terminates function, judges whether to terminate according to return value.
Step 10, whether return value is more than 0.If return value is more than 0,2 are entered step；If return value is less than 0, terminate
Simulation.
The present embodiment handles the particle between multiple processes by point-to-point communication and transmits, the communication between two regional process
The simulation task for not influencing other regional process, has been overlapped simulation and call duration time, is limited by particle migration record asynchronous
The termination condition of transport simulation, avoids global communication, reduces the process stand-by period of entire algorithm, ensure that it is higher simultaneously
Line efficiency.
It should be noted that in the description of the present application, term " first ", " second " etc. are used for description purposes only, without
It can be interpreted as indicating or implying relative importance.In addition, in the description of the present application, unless otherwise indicated, the meaning of " multiple "
It is two or more.
Any process described otherwise above or method description are construed as in flow chart or herein, and expression includes
It is one or more for realizing specific logical function or process the step of executable instruction code module, segment or portion
Point, and the range of the preferred embodiment of the application includes other realization, wherein can not press shown or discuss suitable
Sequence, include according to involved function by it is basic simultaneously in the way of or in the opposite order, to execute function, this should be by the application
Embodiment person of ordinary skill in the field understood.
It should be appreciated that each section of the application can be realized with hardware, software, firmware or combination thereof.Above-mentioned
In embodiment, software that multiple steps or method can in memory and by suitable instruction execution system be executed with storage
Or firmware is realized.It, and in another embodiment, can be under well known in the art for example, if realized with hardware
Any one of row technology or their combination are realized：With the logic gates for realizing logic function to data-signal
Discrete logic, with suitable combinational logic gate circuit application-specific integrated circuit, programmable gate array (PGA), scene
Programmable gate array (FPGA) etc..
Those skilled in the art are appreciated that realize all or part of step that above-described embodiment method carries
Suddenly it is that relevant hardware can be instructed to complete by program, the program can be stored in a kind of computer-readable storage medium
In matter, which includes the steps that one or a combination set of embodiment of the method when being executed.
In the description of this specification, reference term " one embodiment ", " some embodiments ", " example ", " specifically show
The description of example " or " some examples " etc. means specific features, structure, material or spy described in conjunction with this embodiment or example
Point is contained at least one embodiment or example of the application.In the present specification, schematic expression of the above terms are not
Centainly refer to identical embodiment or example.Moreover, particular features, structures, materials, or characteristics described can be any
One or more embodiments or example in can be combined in any suitable manner.
Although embodiments herein has been shown and described above, it is to be understood that above-described embodiment is example
Property, it should not be understood as the limitation to the application, those skilled in the art within the scope of application can be to above-mentioned
Embodiment is changed, changes, replacing and modification.
Claims (19)
1. a kind of asynchronous transport simulation method that PARTICLE TRANSPORT FROM domain decomposition parallel calculates, which is characterized in that including：
Host process carries out subdivision using the subdivision algorithm of more subdivision nodes to the region for combining geometry, obtains multiple subregions；
Host process is the subregion and particle source of each regional process distribution simulation；
Regional process obtains particle from the particle source, and simulation particle transports；
The particle that regional process will be moved to subzone boundaries is sent to corresponding another regional process, alternatively, receiving other areas
The particle that domain process is sent, and the subregion is updated according to the number of particles for sending or receiving between other regional process
Particle migration record；
When in subregion not when the particle simulated, regional process by the particle migration of subregion record be sent to it is main into
Journey；
The particle migration that host process receives each regional process is recorded and is checked, and judges whether end simulation.
2. according to the method described in claim 1, it is characterized in that, more subdivision nodes are from multiple under subdivision root node
The multiple subdivision nodes selected in child node；Host process carries out the region for combining geometry using the subdivision algorithm of more subdivision nodes
Subdivision obtains multiple subregions, including：
Select subdivision root node；
Multiple subdivision nodes are determined from multiple child nodes under the subdivision root node of selection；
Subdivision is carried out from the child node of determining multiple subdivision nodes by bounding volume method, generates multiple subregions.
3. according to the method described in claim 2, it is characterized in that, the selection gist of subdivision root node is：With subdivision root node
For total first setting ratio of higher than whole tree node of node sum of the subtree of root node；
And compare calmly by less than whole tree node sum second of the node sum of the subtree of root node of the child node of subdivision root node
Example.
4. according to the method described in claim 3, it is characterized in that, from multiple child nodes under the subdivision root node of selection really
Determining the foundation of multiple subdivision nodes is：It is higher than using subdivision root node as root by the node sum of the subtree of root node of subdivision node
The node sum third setting ratio of the subtree of node；
And it is less than using subdivision root node as the son of root node by the node sum of the subtree of root node of the child node of subdivision node
Total 4th certainty ratio of the node of tree.
5. described according to the method described in claim 1, it is characterized in that, the host process is a specified regional process
Particle migration record include the regional process be sent to other regional process population and the regional process received from
The population of other regional process, wherein the particle migration that the host process receives each regional process is recorded and checked, and judgement is
No end simulation, specifically includes：
Host process summarizes the particle migration record of each regional process, counts the total number of particles sent and received；
When the total number of particles of transmission is equal with the total number of particles of reception, end of identification is sent to each regional process.
6. according to the method described in claim 5, it is characterized in that, the host process is remembered according to the particle migration of each regional process
Record, judges whether end simulation, further comprises：
Host process, which records the particle migration received, to be merged, and generates sending metrix and receiving matrix, wherein the sending metrix
And the matrix element in receiving matrix is the particle quantity forwarded or particle received quantity between two regional process；
Judge whether the sending metrix is equal to receiving matrix；
If so, end simulation, otherwise continues to simulate, and new particle migration to be received is waited to record.
7. according to the method described in claim 1, it is characterized in that, when regional process to another regional process send particle when,
The method further includes：
The particle that will be moved to corresponding subzone boundaries is stored to transmission buffer area；
When the particle for being sent to another regional process in the transmission buffer area reaches preset quantity, the transmission is cached
Particle in area is sent to corresponding another regional process.
8. the method according to the description of claim 7 is characterized in that the particle by the transmission buffer area is sent to pair
Another regional process answered, specifically includes：
Ready message is sent to corresponding another regional process, wherein ready message includes the particle number to be sent；
The confirmation signal that corresponding another regional process returns is received, point-to-point communication is established with it；
The analog information of particle to be sent is sent to corresponding another regional process.
9. according to the method described in claim 1, it is characterized in that, when regional process receives the grain that other regional process are sent
The period of the day from 11 p.m. to 1 a.m, the method further include：
It detects whether to receive ready message；
If receiving, according to the particle number application order caching space in ready message；
Confirmation signal is replied, point-to-point communication is established with other described regional process；
The analog information for the particle that other described regions are sent is received, and is stored in and receives in particle library.
10. according to the method described in claim 9, it is characterized in that, the analog information of the particle received is stored in reception
After in particle library, further include：
Particle is obtained from the reception particle library and carries out transport simulation, wherein the simulation priority for receiving particle library is high
In the simulation priority of the particle source.
11. according to claim 1-10 any one of them methods, which is characterized in that the host process is one and is previously designated
Regional process.
12. a kind of asynchronous transport simulation system that the PARTICLE TRANSPORT FROM domain decomposition parallel based on more subdivision nodes calculates, feature
It is, including：
The domain mesh for combining geometry is obtained multiple subregions, is by host process, the subdivision algorithm for the more subdivision nodes of use
The subregion and particle source, the particle migration for receiving each regional process of each regional process distribution simulation are recorded and are checked, and judgement is
No end simulation；
Regional process, for obtaining particle from the particle source, simulation particle transports, and will be moved to the particle of subzone boundaries
Be sent to corresponding another regional process, alternatively, receive the particle that other regional process are sent, and according to other regions into
The number of particles for sending or receiving between journey updates the particle migration record of the subregion, and, when not waited in subregion
When the particle of simulation, the particle migration record of the subregion is sent to host process by regional process.
13. system according to claim 12, which is characterized in that more subdivision nodes are from more under subdivision root node
The multiple subdivision nodes selected in a child node；The host process is specifically used for：Subdivision root node is selected, from the subdivision root of selection
Multiple subdivision nodes are determined in multiple child nodes under node, are tied from the son of determining multiple subdivision nodes by bounding volume method
Subdivision goes out multiple subregions in point.
14. system according to claim 13, which is characterized in that the selection gist of subdivision root node is：With subdivision root knot
Point is total first setting ratio of higher than whole tree node of node sum of the subtree of root node；
And compare calmly by less than whole tree node sum second of the node sum of the subtree of root node of the child node of subdivision root node
Example.
15. system according to claim 14, which is characterized in that from multiple child nodes under the subdivision root node of selection
Determining the foundation of multiple subdivision nodes is：It is higher than with subdivision root node as the node sum of the subtree of root node using subdivision node and is
The node sum third setting ratio of the subtree of root node；
And it is less than using subdivision root node as the son of root node by the node sum of the subtree of root node of the child node of subdivision node
Total 4th certainty ratio of the node of tree.
16. system according to claim 12, which is characterized in that subdivision root node and subdivision node pass through more subdivision nodes
Algorithm is searched for or is specified by user automatically；The particle migration record includes the grain that regional process is sent to other regional process
The population of subnumber and regional process received from other regional process, the host process are specifically used for summarizing the grain of each regional process
Son migration record, counts the total number of particles sent and received, and, when the total number of particles of transmission and the total number of particles of reception are equal
When, end of identification is sent to each regional process.
17. system according to claim 12, which is characterized in that further include the corresponding transmission buffer area of each regional process
With reception particle library, the transmission buffer area is for preserving the particle for moving to corresponding subzone boundaries, the reception particle
Library is for preserving the particle received from other regional process.
18. system according to claim 17, which is characterized in that the regional process is additionally operable to from the reception particle library
Middle acquisition particle carries out transport simulation, wherein the simulation priority for receiving particle library is excellent higher than the simulation of the particle source
First grade.
19. according to the system described in any one of claim 12-18, which is characterized in that the host process is one specified
Regional process.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810135323.XA CN108363865A (en) | 2018-02-09 | 2018-02-09 | The asynchronous transport simulation method and system that PARTICLE TRANSPORT FROM domain decomposition parallel calculates |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810135323.XA CN108363865A (en) | 2018-02-09 | 2018-02-09 | The asynchronous transport simulation method and system that PARTICLE TRANSPORT FROM domain decomposition parallel calculates |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108363865A true CN108363865A (en) | 2018-08-03 |
Family
ID=63005609
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810135323.XA Pending CN108363865A (en) | 2018-02-09 | 2018-02-09 | The asynchronous transport simulation method and system that PARTICLE TRANSPORT FROM domain decomposition parallel calculates |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108363865A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112130998A (en) * | 2020-09-23 | 2020-12-25 | 中国核动力研究设计院 | Optimization method of nuclear reactor system analysis program suitable for multi-loop pressurized water reactor |
| CN116049036A (en) * | 2022-10-31 | 2023-05-02 | 中国人民解放军国防科技大学 | Processor cache configuration method for Monte Carlo particle transport |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103955567A (en) * | 2014-04-10 | 2014-07-30 | 中国科学院近代物理研究所 | GPU-base (Graphics Processing Unit-based) multi-particle transport simulation method |
| CN105138778A (en) * | 2015-08-31 | 2015-12-09 | 上海联影医疗科技有限公司 | Simulation method and system for particle transport |
| CN106355017A (en) * | 2016-08-30 | 2017-01-25 | 中国科学院合肥物质科学研究院 | Continuous-energy-point Monte Carlo particle transport method based on multi-group virtual sections |
| CN106354946A (en) * | 2016-08-30 | 2017-01-25 | 中国科学院合肥物质科学研究院 | Transition region based Monte Carlo and certainty theory coupled particle transport method |
| CN107239352A (en) * | 2017-05-31 | 2017-10-10 | 北京科技大学 | The communication optimization method and its system of a kind of dynamics Monte Carlo Parallel Simulation |
-
2018
- 2018-02-09 CN CN201810135323.XA patent/CN108363865A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103955567A (en) * | 2014-04-10 | 2014-07-30 | 中国科学院近代物理研究所 | GPU-base (Graphics Processing Unit-based) multi-particle transport simulation method |
| CN105138778A (en) * | 2015-08-31 | 2015-12-09 | 上海联影医疗科技有限公司 | Simulation method and system for particle transport |
| CN106355017A (en) * | 2016-08-30 | 2017-01-25 | 中国科学院合肥物质科学研究院 | Continuous-energy-point Monte Carlo particle transport method based on multi-group virtual sections |
| CN106354946A (en) * | 2016-08-30 | 2017-01-25 | 中国科学院合肥物质科学研究院 | Transition region based Monte Carlo and certainty theory coupled particle transport method |
| CN107239352A (en) * | 2017-05-31 | 2017-10-10 | 北京科技大学 | The communication optimization method and its system of a kind of dynamics Monte Carlo Parallel Simulation |
Non-Patent Citations (1)
| Title |
|---|
| 李刚: "蒙特卡罗粒子输运的组合几何区域分解并行算法研究与应用", 《中国博士学位论文全文数据库 工程科技II辑》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112130998A (en) * | 2020-09-23 | 2020-12-25 | 中国核动力研究设计院 | Optimization method of nuclear reactor system analysis program suitable for multi-loop pressurized water reactor |
| CN112130998B (en) * | 2020-09-23 | 2022-02-01 | 中国核动力研究设计院 | Optimization method of nuclear reactor system analysis program suitable for multi-loop pressurized water reactor |
| CN116049036A (en) * | 2022-10-31 | 2023-05-02 | 中国人民解放军国防科技大学 | Processor cache configuration method for Monte Carlo particle transport |
| CN116049036B (en) * | 2022-10-31 | 2023-07-21 | 中国人民解放军国防科技大学 | Processor cache configuration method for Monte Carlo particle transport |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| KR102611938B1 (en) | Generate integrated circuit floorplans using neural networks | |
| Li et al. | An improved tabu search algorithm for the stochastic vehicle routing problem with soft time windows | |
| Marsan | Stochastic Petri nets: an elementary introduction | |
| Calvet et al. | Solving the multidepot vehicle routing problem with limited depot capacity and stochastic demands | |
| CN106020941A (en) | Selecting Resource Allocation Policies and Resolving Resource Conflicts | |
| CN108122032A (en) | A kind of neural network model training method, device, chip and system | |
| Barbalho et al. | A hybrid data mining GRASP with path-relinking | |
| CN109918199A (en) | Distributed figure processing system based on GPU | |
| Latorre-Biel et al. | Combining simheuristics with Petri nets for solving the stochastic vehicle routing problem with correlated demands | |
| CN111859027A (en) | Graph calculation method and device | |
| CN108363865A (en) | The asynchronous transport simulation method and system that PARTICLE TRANSPORT FROM domain decomposition parallel calculates | |
| Sun et al. | A light-robust-optimization model and an effective memetic algorithm for an open vehicle routing problem under uncertain travel times | |
| de Souza et al. | Distributed MPC for urban traffic networks: A simulation‐based performance analysis | |
| Goudarzi et al. | Design of a universal logic block for fault-tolerant realization of any logic operation in trapped-ion quantum circuits | |
| Qi et al. | Intersection traffic deadlock formation and its probability: A petri net‐based modeling approach | |
| Peacock et al. | A distributed approach to queueing network simulation | |
| CN111935005A (en) | Data transmission method, device, processing equipment and medium | |
| Mitropoulou et al. | Evolution strategies-based metaheuristics in structural design optimization | |
| CN107426012A (en) | A kind of fault recovery method and its device based on super fusion architecture | |
| Seghir et al. | A new discrete imperialist competitive algorithm for QoS-aware service composition in cloud computing | |
| CN106293936A (en) | A kind of determination method and device of the locally optimal solution of physical host virtual memory | |
| CN112988403B (en) | Integrated circuit simulation multithread management parallel method and device with security function | |
| Youness et al. | A behavior and delay equivalent petri net model for performance evaluation of communication protocols | |
| Borissov et al. | Propagating spin modes in canonical quantum gravity | |
| CN112948114A (en) | Edge computing method and edge computing platform |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180803 |