CN111695245B - Parallel simulation method for material irradiation damage space resolution random cluster dynamics - Google Patents
Parallel simulation method for material irradiation damage space resolution random cluster dynamics Download PDFInfo
- Publication number
- CN111695245B CN111695245B CN202010429538.XA CN202010429538A CN111695245B CN 111695245 B CN111695245 B CN 111695245B CN 202010429538 A CN202010429538 A CN 202010429538A CN 111695245 B CN111695245 B CN 111695245B
- Authority
- CN
- China
- Prior art keywords
- reaction
- updating
- sector
- defect
- communication data
- 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.)
- Active
Links
Images
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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Geometry (AREA)
- General Physics & Mathematics (AREA)
- Computer Hardware Design (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Computer Graphics (AREA)
- Software Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a material irradiation damage space resolution random cluster dynamics parallel simulation method which can realize large-scale SRSCD simulation. The method comprises the following steps: the simulation volumes of the space resolution random cluster dynamics are distributed to different processes, a three-dimensional Cartesian topological structure of the processes is established, and the area of the processes is divided into a plurality of sectors; determining a communication data structure; traversing each sector in sequence according to the sector number sequence, calculating the time threshold value of the inner circulation of each sector, entering the inner circulation, randomly selecting one reaction in the current sector, updating the defect according to the selected reaction, filling a communication data list and updating the related reaction, when the evolution time of the inner circulation is greater than the time threshold value of the inner circulation, communicating with the neighbor process, synchronizing the corresponding defect information and updating the related reaction. The invention relates to the technical field of nuclear material irradiation damage simulation and the parallel computing field.
Description
Technical Field
The invention relates to the technical field of nuclear material irradiation damage simulation and the parallel computing field, in particular to a method for parallel simulation of material irradiation damage space resolution random cluster dynamics.
Background
The service performance of structural materials in nuclear reactors mainly depends on the dynamic behavior of defects generated by irradiation, such as diffusion, nucleation, growth, annihilation and the like of the defects, and the evolution behavior of the long time scale can cause the change of the size and density distribution of the defects, so that the mechanical properties of the structural materials are degraded, such as irradiation hardening/embrittlement, irradiation swelling and the like. Since these long-scale evolutionary behaviors often exceed the capabilities of microscale simulation methods such as molecular dynamics (Molecular Dynamics, MD), atomic dynamics monte carlo (atomic kinetic Monte Carlo, AKMC), etc., modeling and simulation using mesoscale methods are often required.
The entity Dynamics monte carlo method (object kinetic Monte Carlo, OKMC) and the Cluster Dynamics method (CD) developed based on the average field rate theory are two widely used mesoscale simulation methods. The former simulates the evolution of defects by tracking random diffusion and interactions of each defect within the material; the latter assumes that the defect is in an isotropic, homogeneous medium, and the change in defect concentration over time is tracked by the rate equation of the defect reaction. The OKMC can simulate complex behaviors among various defects and can capture spatial correlation among defects, but is generally limited by the computational time and the computational effort required for the complex behavior simulation among defects. CD methods generally have high computational efficiency and can simulate high irradiation doses and time scales, however, such methods are generally limited to simulating systems containing small amounts of mobile defects, and it is difficult to deal with defect evolution in complex systems.
Spatially resolved random cluster dynamics (Spatially Resolved Stochastic Cluster Dynamics, SRSCD) is a new approach developed in recent years to simulate the behavior of irradiation defects of structural materials. It divides the simulated volume (which may also be referred to as the simulated region) into a plurality of volume elements (cubes), within each of which defects are assumed to be uniformly distributed, which defects can aggregate and decompose, while there is a concentration difference between the volume elements and the resulting diffusion. The reaction rate of the various reactions between defects is deduced from classical CD methods, and the selection and time increment of the reactions are determined by classical KMC algorithms. The SRSCD approach avoids the limitations of CD on defect types and behavior complexity on the one hand, and reduces the computational requirements compared to the KMC on the other hand, but its analog volume is limited by the computational effort brought by the KMC algorithm.
To expand the analog volume of SRSCD, the SRSCD's proposer implements the parallelization of SRSCD, i.e., synchronous parallel SRSCD, using a synchronous parallel KMC algorithm. However, this parallel approach is inefficient in its parallelism because it requires synchronization at every KMC step, and cannot be effectively extended to larger analog volumes.
Disclosure of Invention
The invention aims to provide a material irradiation damage space resolution random cluster dynamics parallel simulation method, which aims to solve the problems that the parallel efficiency is low and large-scale parallel simulation cannot be performed in the prior art.
In order to solve the technical problems, the embodiment of the invention provides a parallel simulation method for material irradiation damage space resolution random cluster dynamics, which comprises the following steps:
the simulation volumes of the space resolution random cluster dynamics are distributed to different processes, a three-dimensional Cartesian topological structure of the processes is established, and the area of the processes is divided into a plurality of sectors;
determining a communication data structure;
traversing each sector in sequence according to the sector number sequence, calculating the time threshold value of the inner circulation of each sector, entering the inner circulation, randomly selecting one reaction in the current sector, updating the defect according to the selected reaction, filling a communication data list and updating the related reaction, when the evolution time of the inner circulation is greater than the time threshold value of the inner circulation, communicating with the neighbor process, synchronizing the corresponding defect information and updating the related reaction.
Further, the assigning the simulated volumes of spatially resolved random cluster dynamics to different processes and establishing a three-dimensional cartesian topology of the processes, and dividing the area of the processes into a plurality of sectors comprises:
the simulation volumes of the space resolution random cluster dynamics are distributed to different processes in a space division mode, and a three-dimensional Cartesian topological structure of the processes is established;
each process creates respective local volume elements according to the given volume element number, and initializes the association relationship between the volume elements on each process;
dividing the area of each process into a plurality of sectors and numbering;
and establishing a mapping relation between the volume element and the sector.
Further, each sector has at least two volume elements in each dimension, i.e., each process has at least four volume elements in each dimension.
Further, the products of the injection reaction are cascade defects, and the reactants and products of the non-injection reaction are common defects;
the determining a communication data structure includes:
if the defect is a common defect, constructing a communication data structure commDefectSwap;
if the defect is a cascade defect, constructing a communication data structure cascadeDefectSwap.
Further, the communication data structure commDefectSwap includes: sendProc, numSend, sendBuff, recvProc, numRecv and recvBuff;
where sendProc represents a destination process number of transmission, numSend represents the number of transmission data, sendBuff represents a transmission data buffer, recvProc represents a source process number of reception, numRecv represents the number of reception data, and recvBuff represents a reception data buffer.
Further, the communication data structure cascadedeffectswap includes: cascadeCell, numCells, sendProc, numSend, sendIndex, sendBuff, recvProc, numRecv and recvBuff;
wherein cascadeCell represents a volume element identifier where an injection reaction is located, numCells represents the number of volume elements where an injection reaction occurs, sendProc represents the destination process number of transmission, numSend represents the number of transmission data, sendIndex represents the index of a transmission data buffer, sendBuff represents a transmission data buffer, recvProc represents the source process number of reception, numRecv represents the number of reception data, and recvBuff represents a reception data buffer.
Further, traversing each sector in sequence according to the sector number sequence, calculating a time threshold value of the inner loop of each sector, entering the inner loop, randomly selecting a reaction in the current sector, updating the defect according to the selected reaction, filling the communication data list and updating the relevant reaction, when the evolution time of the inner loop is greater than the time threshold value of the inner loop, communicating with the neighbor process, synchronizing the corresponding defect information and updating the relevant reaction, wherein the method comprises the following steps:
initializing the evolution time t of the outer loop, and entering the outer loop:
traversing each sector in turn according to the sequence of the sector numbers, and calculating the time threshold tau of each inner cycle;
initializing the evolution time kmc_time of the inner loop, and entering the inner loop of the current sector:
calculating a time increment delta t according to the total reaction rate on the current sector;
randomly selecting a reaction in the current sector;
updating defects and filling a communication data list according to types of reactants and products involved in the reaction, and updating corresponding reactions according to the defects involved in the reaction;
the evolution time kmc-time of the inner loop accumulates the time increment delta t, and whether the current kmc-time is larger than the time threshold tau of the inner iteration is judged;
if yes, synchronizing the common defects and updating the related reactions, synchronizing the cascade defects and updating the related reactions; if not, returning to execute the operation of calculating the time increment delta t according to the total reaction rate on the current sector;
accumulating tau by the evolution time T of the outer loop, and judging whether the current evolution time T is greater than an outer loop time threshold T or not;
if yes, the outer circulation is ended; otherwise, returning to execute the operation of traversing each sector in sequence according to the sector number sequence.
Further, updating the defect according to the types of reactants and products involved in the reaction, filling the communication data list, and updating the corresponding reaction according to the defect involved in the reaction comprises:
updating corresponding defects according to the types of reactants and products involved in the reaction;
if the reaction occurs in the boundary volume element, judging whether the selected reaction is an injection reaction or not;
if the reaction is injected, judging whether the product is in a Boundary region or not; the Boundary region is a region formed by volume elements which are positioned in a neighbor process and adjacent to the current process;
if yes, setting the value of the corresponding volume element number in the cascadeeContell array of the communication data structure cascadeDefectSwap to be 1;
if not, the relevant reaction is updated according to the defect related to the reaction.
Further, the method further comprises:
if not, determining if the reactant/product is in the Boundary/Ghost region; the Boundary region is a region formed by volume elements which are positioned in the current process and adjacent to the neighbor process; the Ghost area is an area formed by volume elements which are positioned in the neighbor process and adjacent to the current process;
if yes, filling the defect related to the reaction into a communication data structure commDefectSwap;
if not, the relevant reaction is updated according to the defect related to the reaction.
Further, the synchronizing common defects and updating the related reactions includes:
the current process and the neighbor process establish non-blocking communication, and exchange defect information through a communication data structure commDefectSwap;
updating the received defect information into the corresponding volume element;
updating a response associated therewith based on the received defect information;
the synchronizing cascade defect and updating related reactions comprise:
filling all defects in volume elements with the median of '1' in the cascadeCell array into a communication data structure cascadeDefectSwap;
the current process and the neighbor process establish non-blocking communication, and defect information is exchanged through a communication data structure cascadeDefectSwap;
updating the received defect information into the corresponding volume element;
and updating the response related to the received defect information.
The technical scheme of the invention has the following beneficial effects:
in the scheme, the simulation volumes of the space-resolved random cluster dynamics are distributed to different processes, a three-dimensional Cartesian topological structure of the processes is built, and the area of the processes is divided into a plurality of sectors; determining a communication data structure; sequentially traversing each sector according to the sequence of the sector numbers, calculating a time threshold value of the inner circulation of each sector, entering the inner circulation, randomly selecting a reaction in the current sector, updating defects according to the selected reaction, filling a communication data list and updating related reactions, when the evolution time of the inner circulation is greater than the time threshold value of the inner circulation, communicating with a neighbor process, synchronizing corresponding defect information and updating related reactions; in this way, the SL algorithm is introduced into the SRSCD, so that the SRSCD parallelism is realized, the SRSCD parallelism efficiency is high, the SRSCD can be effectively expanded into large-volume simulation, and the large-scale SRSCD simulation is realized.
Drawings
FIG. 1 is a schematic flow chart of a method for parallel simulation of material irradiation damage space-resolved random cluster dynamics provided by an embodiment of the invention;
fig. 2 is a detailed flow chart of a parallel simulation method for material irradiation damage space resolution random cluster dynamics provided by the embodiment of the invention;
FIG. 3 is a schematic diagram of simulated volume partitioning and sector partitioning of SL_SRSCD in two dimensions according to an embodiment of the invention;
fig. 4 is a schematic diagram of 6 communication directions according to an embodiment of the present invention;
fig. 5 is a schematic diagram of a communication data structure commDefectSwap in sl_srscd according to an embodiment of the present invention;
fig. 6 is a schematic diagram of a communication data structure cascadedeffectswap in a sl_srscd according to an embodiment of the present invention;
fig. 7 is a communication schematic diagram of the sl_srscd method according to an embodiment of the present invention under a two-dimensional situation.
Detailed Description
In order to make the technical problems, technical solutions and advantages to be solved more apparent, the following detailed description will be given with reference to the accompanying drawings and specific embodiments.
Aiming at the problems that the existing parallel efficiency is low and large-scale parallel simulation cannot be performed, the invention provides a material irradiation damage space resolution random cluster dynamics parallel simulation method.
In order to better understand the parallel simulation method of the material irradiation damage space resolution random cluster dynamics, which is provided by the embodiment of the invention, the following description is firstly made on some terms:
1. for each type of reaction, the defects involved in the reaction are called reactants (reactants), the defects produced by the reaction are products (products), and the reactions between the defects mainly comprise five types:
(1) Injection reaction (Implantation): indicating the occurrence of an initial defect, without reactants. If the electron irradiation is carried out, the product is two point defects (vacancy and self-interstitial atoms), and if the neutron irradiation is carried out, the product is a plurality of (more than 2) defects including point defects and small clusters;
(2) Decomposition reaction (association): the defective cluster emits a point defect. The reactant is one defect, the product is two defects, and one of the defects is a point defect;
(3) Trap absorption reaction (sink_remove): defects disappear by absorption by intrinsic defects (such as grain boundaries, dislocations, etc.) within the material, these intrinsic defects being called "traps"; the reactant is a defect, and the product is zero;
(4) Aggregation reaction (managing): growth and recombination of defects (recombination of vacancy defects and self-interstitial defects); the reactants are two defects and the product is generally one defect;
(5) Diffusion reaction (Diffusion): the defect migrates from one voxel to another voxel; both reactants and products are a drawback;
wherein only diffusion reactions occur between two volume elements, and the remaining reactions occur within the volume elements.
2. A synchronized sub-lattice (Synchronous Sublattice, SL) algorithm which, in the case of an allowable error, in order to avoid collisions between processors, further divides the area on each processor (corresponding to a process hereinafter) into sub-areas (corresponding to a sector hereinafter) and numbers the sub-areas, each processor simultaneously and independently executing KMC events on the sub-areas corresponding to the same number, and does not communicate for the processor having a boundary event occurred until the execution time of the next event exceeds a time threshold τ. Therefore, the parallel efficiency of the SL algorithm is generally much higher than the synchronous parallel KMC algorithm, which is synchronized every step.
In the embodiment, the SL algorithm with higher parallel efficiency is introduced into the SRSCD, so that the material irradiation damage space resolution random cluster dynamics parallel simulation method disclosed by the embodiment of the invention is actually an SRSCD massive parallel simulation method-SL_SRSCD method based on the SL algorithm, and the SRSCD can realize three-dimensional massive parallel simulation under the condition of introducing a certain error. This error of the SL algorithm is allowed and can be physically interpreted as a smaller dependence of an event somewhere on a further distance in the system, so that the event can be spatially decoupled. Therefore, it is feasible for the SL algorithm to extend to SRSCD.
As shown in fig. 1, the method for parallel simulation of material irradiation damage space-resolved random cluster dynamics provided by the embodiment of the invention comprises the following steps:
s101, distributing the simulation volumes of the space resolution random cluster dynamics to different processes, establishing a three-dimensional Cartesian topological structure of the processes, and dividing the area of the processes into a plurality of sectors;
s102, determining a communication data structure;
s103, traversing each sector in sequence according to the sector number sequence, calculating the time threshold value of the inner cycle of each sector, entering the inner cycle, randomly selecting one reaction in the current sector, updating the defect according to the selected reaction, filling a communication data list and updating the related reaction, and when the evolution time of the inner cycle is greater than the time threshold value of the inner cycle, communicating with the neighbor process, synchronizing the corresponding defect information and updating the related reaction.
According to the material irradiation damage space resolution random cluster dynamics parallel simulation method, simulation volumes of space resolution random cluster dynamics are distributed to different processes, a three-dimensional Cartesian topological structure of the processes is built, and a process area is divided into a plurality of sectors; determining a communication data structure; sequentially traversing each sector according to the sequence of the sector numbers, calculating a time threshold value of the inner circulation of each sector, entering the inner circulation, randomly selecting a reaction in the current sector, updating defects according to the selected reaction, filling a communication data list and updating related reactions, when the evolution time of the inner circulation is greater than the time threshold value of the inner circulation, communicating with a neighbor process, synchronizing corresponding defect information and updating related reactions; in this way, the SL algorithm is introduced into the SRSCD, so that the SRSCD parallelism is realized, the SRSCD parallelism efficiency is high, the SRSCD can be effectively expanded into large-volume simulation, and the large-scale SRSCD simulation is realized.
In a specific embodiment of the foregoing method for parallel simulation of material irradiation damage space-resolved random cluster dynamics, further, the assigning simulated volumes of the space-resolved random cluster dynamics to different processes and establishing a three-dimensional cartesian topology of the processes, and dividing a region of the processes into a plurality of sectors (S101) includes:
the simulation volumes of the space resolution random cluster dynamics are distributed to different processes in a space division mode, and a three-dimensional Cartesian topological structure of the processes is established;
each process creates respective local volume elements according to the given volume element number, and initializes the association relationship between the volume elements on each process;
dividing the area of each process into a plurality of sectors (sectors) and numbering; typically, two-dimensionally divided into 4 sectors; three-dimensionally dividing into 8 sectors;
and establishing a mapping relation between the volume element and the sector.
In this embodiment, initializing the association relationship between the volume elements on each process refers to storing the local numbers of the neighbor volume elements in the left, right, front, back, up and down directions for each volume element.
In this embodiment, as shown in fig. 2, the purpose of S101 is to initialize the analog volume, and fig. 3 is a schematic diagram of region (analog volume) division and sector division of the sl_srscd in two dimensions, where P0 to P8 are process numbers; 1 to 4 are sector numbers; coarsely realizing as a process boundary; the dashed line is the sector boundary; the fine implementation is a voxel boundary. Taking the process P4 as an example, the dash-dot line is a communication area of P4, where light gray is a Ghost area (the Ghost area is an area formed by volume elements located in the neighboring process and adjacent to the current process (P4)), and dark gray is a Boundary area (the Boundary area is an area formed by volume elements located in the current process (P4) and adjacent to the neighboring process).
In this embodiment, the actual simulation space is three-dimensional, so that a three-dimensional cartesian topological structure needs to be established for the process, the area in the process is further divided into 8 sectors, the sector numbers are 1-8, two in the x and Y, Z directions respectively, and then the local numbers of the neighbor processes, sectors and volume elements in the 6 communication directions shown in fig. 4 are stored for each process, sector and volume element. Fig. 4 is a schematic diagram of 6 communication directions, which are also 6 directions of defect diffusion. Since diffusion in SRSCD is caused by concentration gradients, only the direction of diffusion of the surface neighbors is needed, and only communication with the surface neighbors is needed.
In this embodiment, in order to avoid a conflict caused by deleting and adding the same defect in the same volume element, when the sl_srscd method is used, a constraint needs to be made: it is ensured that each dimension of each sector contains at least two volume elements, i.e. each dimension of each process contains at least four volume elements.
In the specific implementation mode of the material irradiation damage space resolution random cluster dynamics parallel simulation method, further, products of injection reaction are cascade defects, and reactants and products of non-injection reaction are common defects;
the determining a communication data structure includes:
if the defect is a common defect, constructing a communication data structure commDefectSwap;
if the defect is a cascade defect, constructing a communication data structure cascadeDefectSwap.
In this embodiment, taking the example of simulated neutron irradiation, the injection reaction may generate a plurality of defects in the SRSCD simulation system, and in order to distinguish these defects from defects participating in other reactions, the product of the injection reaction is referred to as "cascade defects" in the sl_srscd, and the reactants and products of other reactions are referred to as "normal defects".
In a specific embodiment of the foregoing method for parallel simulation of material irradiation damage spatially resolved random cluster dynamics, further, the communication data structure commDefectSwap includes: sendProc, numSend, sendBuff, recvProc, numRecv and recvBuff;
where sendProc represents a destination process number of transmission, numSend represents the number of transmission data, sendBuff represents a transmission data buffer, recvProc represents a source process number of reception, numRecv represents the number of reception data, and recvBuff represents a reception data buffer.
FIG. 5 is a schematic diagram of a communication data structure, commDefectSwap, in SL_SRSCD, where send 1-3 represent neighbor processes sending data to 3 surface neighbors; recv 1-3 represent receiving data from 3 face-adjacent neighbor processes in the opposite direction; the defect type in sendBuff represents the defect type of the defect, num represents the number of defects to be updated (num >0 represents the defect as a product, num <0 represents the defect as a reactant), cell represents the local number of the volume element where the defect is located, if the defect is in the Boundary region, dir and neighbor are both equal to 0, if the defect is in the Boundary region, it is indicated that the cell is the volume element in the Boundary region, dir is equal to the direction in which the Boundary region is located, and neighbor is equal to the local volume element number adjacent to the cell; the significances of the deltatype [ ], num and cell in recvBuff are the same as those of the transmission buffer, recvDir is the opposite direction of dir, and recvNeighbor is the volume element number in the Ghost region.
In a specific embodiment of the foregoing method for parallel simulation of material irradiation damage spatially resolved random cluster dynamics, further, the communication data structure cascadeDefectSwap includes: cascadeCell, numCells, sendProc, numSend, sendIndex, sendBuff, recvProc, numRecv and recvBuff;
wherein cascadeCell indicates a volume element identifier where an injection reaction (execution) is located, numCells indicates the number of volume elements where the injection reaction occurs, sendProc indicates the destination process number of transmission, numSend indicates the number of transmission data, sendIndex indicates the index of a transmission data buffer, sendBuff indicates the transmission data buffer, recvProc indicates the source process number of reception, numRecv indicates the number of reception data, and recvBuff indicates the reception data buffer.
FIG. 6 is a schematic diagram of a communication data structure cascadeDefectSwap in SL_SRSCD, where send 1-3 represent sending data to 3 neighbor processes; recv 1-3 represent the receipt of data from 3 neighbor processes in the opposite direction; when data is sent, all defects in all volume elements with injection reaction in a Boundary region in a certain direction are required to be filled into sendBuff; for each volume element with the median of '1' in the cascadeCell array, firstly filling an index (index) of the volume element in sendBuff, then sequentially filling defects in the volume element, then filling the next volume element, and sequentially proceeding; index includes: volume element number (cell), neighbor volume element number (neighbor cell) located in the Ghost region, total number of defect types in volume element (numDefects), total number of boundary volume elements (numCells) where injection reactions occur; after the index is filled, the defects (defects) in the volume element are sequentially filled. When receiving data, the next index can be located according to the numCells and the numDefects in the previous index, and then the defect is taken out and updated to the defect with the number equal to the neighbor cell.
In a specific embodiment of the foregoing method for parallel simulation of material irradiation damage space-resolved random cluster dynamics, further, as shown in fig. 2, the steps of traversing each sector sequentially according to a sequence of sector numbers, calculating a time threshold of each intra-sector cycle into the intra-cycle, randomly selecting a reaction in the current sector, updating a defect according to the selected reaction, filling a communication data list, and updating a relevant reaction, when an evolution time of the intra-cycle is greater than the time threshold of the intra-cycle, communicating with a neighbor process, synchronizing corresponding defect information, and updating the relevant reaction include:
initializing the evolution time t of the outer loop, e.g., t=0, and entering the outer loop:
traversing each sector in turn according to the sequence of the sector numbers, and calculating the time threshold tau of each inner cycle;
initializing an evolution time kmc_time of the inner loop, for example kmc_time=0, into the inner loop of the current sector:
calculating a time increment delta t according to the total reaction rate on the current sector;
randomly selecting a reaction in the current sector;
updating defects and filling a communication data list according to types of reactants and products involved in the reaction, and updating corresponding reactions according to the defects involved in the reaction;
the evolution time kmc-time of the inner loop accumulates the time increment delta t, and whether the current kmc-time is larger than the time threshold tau of the inner iteration is judged;
if yes, synchronizing the common defects and updating the related reactions, synchronizing the cascade defects and updating the related reactions; if not, returning to execute the operation of calculating the time increment delta t according to the total reaction rate on the current sector;
accumulating tau by the evolution time T of the outer loop, and judging whether the current evolution time T is greater than an outer loop time threshold T (namely the time of the whole simulation process);
if yes, the outer circulation is ended; otherwise, returning to execute the operation of traversing each sector in sequence according to the sector number sequence.
In this embodiment, the time threshold τ is determined by the maximum average reaction rate in all sectors over all processes for each sector.
In this embodiment, after entering the outer loop, the calculation on 8 sectors is sequentially performed according to the sequence of the sector numbers, one sector corresponds to one inner loop, if the time threshold T of the outer loop is not reached after all the 8 sectors are performed, the 8 sectors are traversed again according to the sequence of the sector numbers, and if the inner loop of one sector is not finished and the current evolution time T reaches the time threshold T, the inner loop is jumped out, the corresponding defect information is synchronized, and the whole loop process is ended.
In a specific embodiment of the foregoing method for parallel simulation of material irradiation damage space-resolved random cluster dynamics, further, updating the defect and filling the communication data list according to the types of reactants and products involved in the reaction, and updating the corresponding reaction according to the defect involved in the reaction includes:
updating corresponding defects according to the types of reactants and products involved in the reaction;
if the reaction occurs in the boundary volume element, judging whether the selected reaction is an injection reaction or not;
if the reaction is injected, judging whether the product is in a Boundary region or not;
if yes, setting the value of the corresponding volume element number in the cascadeeContell array of the communication data structure cascadeDefectSwap to be 1, thereby recording the volume element number where the reaction is located;
if not, the reaction related to the defect is updated according to the defect related to the reaction (refer to the five types of reactions).
In a specific embodiment of the foregoing method for parallel simulation of material irradiation damage spatially resolved random cluster dynamics, the method further includes:
if not, determining if the reactant/product is in the Boundary/Ghost region;
if yes, filling the defect related to the reaction into a communication data structure commDefectSwap;
if not, the reaction related to the defect is updated according to the defect related to the reaction (refer to the five types of reactions).
In a specific embodiment of the foregoing method for parallel simulation of material irradiation damage space-resolved random cluster dynamics, further, the synchronizing common defects and updating related reactions includes:
the current process and the neighbor process establish non-blocking communication, and exchange defect information through a communication data structure commDefectSwap, specifically: establishing non-blocking communication with the neighbor process by using a non-blocking communication function MPI_Isend/MPI_Irecv of MPI, exchanging "common defect" through a communication data structure commDefectSwap, see FIG. 7 (b);
updating the received defect information (i.e. defects in the receiving buffer) into corresponding volume elements;
updating the response associated with the received defect information (referring to the five types of responses above);
the synchronizing cascade defect and updating related reactions comprise:
filling all defects in volume elements with the median of '1' in the cascadeCell array into a communication data structure cascadeDefectSwap according to the structure of FIG. 6;
the current process and the neighbor process establish non-blocking communication, and exchange defect information through a communication data structure cascadeDefectSwap, specifically: establishing non-blocking communication with the neighbor process by using a non-blocking communication function MPI_Isend/MPI_Irecv of MPI, exchanging "cascade defect" through a communication data structure cascadeDefectSwap, see FIG. 7 (c);
updating the received defect information into the corresponding volume element, specifically: according to the received cell number, deleting the existing defect, and then adding the received defect into the cell;
the response associated therewith (referred to as the five types of responses above) is updated based on the received defect information.
In this embodiment, fig. 7 is a communication schematic diagram of the sl_srscd method in the two-dimensional case, P0 to P8 are process numbers, 1 to 4 are sector numbers, and a dotted line indicates an area requiring communication. Taking communication of sector 1 on process P4 as an example, fig. 7 (a) is a schematic diagram of communication between process P4 and four neighbor processes; FIG. 7 (b) is a schematic diagram showing the communication between P4 and the neighbor process through commDefectSwap exchanging the data of the Ghost area and the bound area, and combining the two to send/receive; fig. 7 (c) is a communication schematic diagram of the P4 exchanging "cascade defect" with the neighbor process through cascades defect swap, and only needs to send the Boundary region data and receive the Ghost region data.
Compared with the prior art, the invention has the beneficial effects that:
(1) The SL_SRSCD method has higher parallel efficiency, and can effectively expand the SRSCD into large-volume simulation to realize large-scale SRSCD simulation;
(2) Through the constructed communication data structure and the communication mode (each sector is used for communication after the calculation (namely, the operation of selecting reaction, updating defect, filling communication data list and updating related reaction) is completed, and is non-blocking communication), redundant communication is effectively reduced, calculation and communication are overlapped through non-blocking communication, synchronous waiting time caused by communication is further reduced, and parallel efficiency is improved.
While the foregoing is directed to the preferred embodiments of the present invention, it will be appreciated by those skilled in the art that various modifications and adaptations can be made without departing from the principles of the present invention, and such modifications and adaptations are intended to be comprehended within the scope of the present invention.
Claims (10)
1. A method for parallel simulation of material irradiation damage space resolution random cluster dynamics is characterized by comprising the following steps:
the simulation volumes of the space resolution random cluster dynamics are distributed to different processes, a three-dimensional Cartesian topological structure of the processes is established, and the area of the processes is divided into a plurality of sectors;
determining a communication data structure;
traversing each sector in sequence according to the sector number sequence, calculating the time threshold value of the inner circulation of each sector, entering the inner circulation, randomly selecting one reaction in the current sector, updating the defect according to the selected reaction, filling a communication data list and updating the related reaction, when the evolution time of the inner circulation is greater than the time threshold value of the inner circulation, communicating with the neighbor process, synchronizing the corresponding defect information and updating the related reaction.
2. The method for parallel simulation of spatially resolved random cluster dynamics of irradiation damage to materials according to claim 1, wherein the steps of assigning simulated volumes of spatially resolved random cluster dynamics to different processes and establishing a three-dimensional cartesian topology of the processes and dividing regions of the processes into a plurality of sectors comprise:
the simulation volumes of the space resolution random cluster dynamics are distributed to different processes in a space division mode, and a three-dimensional Cartesian topological structure of the processes is established;
each process creates respective local volume elements according to the given volume element number, and initializes the association relationship between the volume elements on each process;
dividing the area of each process into a plurality of sectors and numbering;
and establishing a mapping relation between the volume element and the sector.
3. The method for parallel simulation of material irradiation damage spatially resolved random cluster dynamics according to claim 1, wherein each dimension of each sector comprises at least two volume elements, namely each dimension of each process comprises at least four volume elements.
4. The method for parallel simulation of material irradiation damage space-resolved random cluster dynamics according to claim 1, wherein the products of the injection reaction are cascade defects, and the reactants and products of the non-injection reaction are common defects;
the determining a communication data structure includes:
if the defect is a common defect, constructing a communication data structure commDefectSwap;
if the defect is a cascade defect, constructing a communication data structure cascadeDefectSwap.
5. The method for parallel simulation of material irradiation damage spatially resolved random cluster dynamics according to claim 4, wherein the communication data structure commDefectSwap comprises: sendProc, numSend, sendBuff, recvProc, numRecv and recvBuff;
where sendProc represents a destination process number of transmission, numSend represents the number of transmission data, sendBuff represents a transmission data buffer, recvProc represents a source process number of reception, numRecv represents the number of reception data, and recvBuff represents a reception data buffer.
6. The method for parallel simulation of material irradiation damage spatially resolved random cluster dynamics according to claim 4, wherein the communication data structure cascadeDefectSwap comprises: cascadeCell, numCells, sendProc, numSend, sendIndex, sendBuff, recvProc, numRecv and recvBuff;
wherein cascadeCell represents a volume element identifier where an injection reaction is located, numCells represents the number of volume elements where an injection reaction occurs, sendProc represents the destination process number of transmission, numSend represents the number of transmission data, sendIndex represents the index of a transmission data buffer, sendBuff represents a transmission data buffer, recvProc represents the source process number of reception, numRecv represents the number of reception data, and recvBuff represents a reception data buffer.
7. The method for parallel simulation of material irradiation damage space resolution random cluster dynamics according to claim 1, wherein the steps of traversing each sector in sequence according to the sector number order, calculating a time threshold of each sector inner loop to enter the inner loop, randomly selecting a reaction in the current sector, updating a defect according to the selected reaction, filling a communication data list and updating a related reaction, communicating with a neighbor process when the evolution time of the inner loop is greater than the time threshold of the inner loop, synchronizing corresponding defect information and updating the related reaction include:
initializing the evolution time t of the outer loop, and entering the outer loop:
traversing each sector in turn according to the sequence of the sector numbers, and calculating the time threshold tau of each inner cycle;
initializing the evolution time kmc_time of the inner loop, and entering the inner loop of the current sector:
calculating a time increment delta t according to the total reaction rate on the current sector;
randomly selecting a reaction in the current sector;
updating defects and filling a communication data list according to types of reactants and products involved in the reaction, and updating corresponding reactions according to the defects involved in the reaction;
the evolution time kmc-time of the inner loop accumulates the time increment delta t, and whether the current kmc-time is larger than the time threshold tau of the inner iteration is judged;
if yes, synchronizing the common defects and updating the related reactions, synchronizing the cascade defects and updating the related reactions; if not, returning to execute the operation of calculating the time increment delta t according to the total reaction rate on the current sector;
accumulating tau by the evolution time T of the outer loop, and judging whether the current evolution time T is greater than an outer loop time threshold T or not;
if yes, the outer circulation is ended; otherwise, returning to execute the operation of traversing each sector in sequence according to the sector number sequence.
8. The parallel simulation method of material irradiation damage space-resolved random cluster dynamics according to claim 7, wherein updating defects according to types of reactants and products involved in the reaction, filling a communication data list, and updating corresponding reactions according to defects involved in the reaction comprises:
updating corresponding defects according to the types of reactants and products involved in the reaction;
if the reaction occurs in the boundary volume element, judging whether the selected reaction is an injection reaction or not;
if the reaction is injected, judging whether the product is in a Boundary region or not; the Boundary region is a region formed by volume elements which are positioned in a neighbor process and adjacent to the current process;
if yes, setting the value of the corresponding volume element number in the cascadeeContell array of the communication data structure cascadeDefectSwap to be 1;
if not, the relevant reaction is updated according to the defect related to the reaction.
9. The method of parallel simulation of spatially resolved random cluster dynamics for irradiation damage to a material of claim 8, further comprising:
if not, determining if the reactant/product is in the Boundary/Ghost region; the Boundary region is a region formed by volume elements which are positioned in the current process and adjacent to the neighbor process; the Ghost area is an area formed by volume elements which are positioned in the neighbor process and adjacent to the current process;
if yes, filling the defect related to the reaction into a communication data structure commDefectSwap;
if not, the relevant reaction is updated according to the defect related to the reaction.
10. The method for parallel simulation of material irradiation damage spatially resolved random cluster dynamics according to claim 7, wherein said synchronizing common defects and updating related reactions comprises:
the current process and the neighbor process establish non-blocking communication, and exchange defect information through a communication data structure commDefectSwap;
updating the received defect information into the corresponding volume element;
updating a response associated therewith based on the received defect information;
the synchronizing cascade defect and updating related reactions comprise:
filling all defects in volume elements with the median of '1' in the cascadeCell array into a communication data structure cascadeDefectSwap;
the current process and the neighbor process establish non-blocking communication, and defect information is exchanged through a communication data structure cascadeDefectSwap;
updating the received defect information into the corresponding volume element;
and updating the response related to the received defect information.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010429538.XA CN111695245B (en) | 2020-05-20 | 2020-05-20 | Parallel simulation method for material irradiation damage space resolution random cluster dynamics |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010429538.XA CN111695245B (en) | 2020-05-20 | 2020-05-20 | Parallel simulation method for material irradiation damage space resolution random cluster dynamics |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111695245A CN111695245A (en) | 2020-09-22 |
| CN111695245B true CN111695245B (en) | 2023-04-28 |
Family
ID=72478037
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010429538.XA Active CN111695245B (en) | 2020-05-20 | 2020-05-20 | Parallel simulation method for material irradiation damage space resolution random cluster dynamics |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111695245B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112668156B (en) * | 2020-12-14 | 2023-06-20 | 北京遥感设备研究所 | Method for calculating storage life of nonmetallic part |
| CN112885414B (en) * | 2021-03-31 | 2022-11-08 | 中国核动力研究设计院 | Irradiation damage simulation system and method based on rate theory and cluster dynamics |
| CN113239522B (en) * | 2021-04-20 | 2022-06-28 | 四川大学 | Atmospheric pollutant diffusion simulation method based on computer cluster |
| CN114490047B (en) * | 2022-01-12 | 2022-09-09 | 北京科技大学 | Heterogeneous data transmission method for nuclear fuel fission gas cluster dynamic simulation |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109727647A (en) * | 2018-12-29 | 2019-05-07 | 中国原子能科学研究院 | A kind of analogy method and system that irradiation causes material structure to develop |
| CN110024047A (en) * | 2016-12-15 | 2019-07-16 | 西屋电气有限责任公司 | Permit the real-time measurement of nuclear component and the integration of atom modeling |
| CN110459269A (en) * | 2019-08-07 | 2019-11-15 | 中国原子能科学研究院 | A kind of multi-scale coupling analogy method of nuclear reactor material irradiation damage |
| CN110457810A (en) * | 2019-08-07 | 2019-11-15 | 中国原子能科学研究院 | The Rate Theory Parallel Simulation method that reactor critical material vacancy, gap develop |
-
2020
- 2020-05-20 CN CN202010429538.XA patent/CN111695245B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110024047A (en) * | 2016-12-15 | 2019-07-16 | 西屋电气有限责任公司 | Permit the real-time measurement of nuclear component and the integration of atom modeling |
| CN109727647A (en) * | 2018-12-29 | 2019-05-07 | 中国原子能科学研究院 | A kind of analogy method and system that irradiation causes material structure to develop |
| CN110459269A (en) * | 2019-08-07 | 2019-11-15 | 中国原子能科学研究院 | A kind of multi-scale coupling analogy method of nuclear reactor material irradiation damage |
| CN110457810A (en) * | 2019-08-07 | 2019-11-15 | 中国原子能科学研究院 | The Rate Theory Parallel Simulation method that reactor critical material vacancy, gap develop |
Non-Patent Citations (2)
| Title |
|---|
| Dandan Chen et al..An overview: multiscale simulation in understanding the radiation damage accumulation of reactor materials.Simulation: Transactions of the Society for Modeling and Simulation International.2019,全文. * |
| Jaime Marian et al..Stochastic cluster dynamics method for simulations of multispecies irradiation damage accumulation.Journal of Nuclear Materials.2011,全文. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111695245A (en) | 2020-09-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111695245B (en) | Parallel simulation method for material irradiation damage space resolution random cluster dynamics | |
| CN107688999B (en) | Block chain-based parallel transaction execution method | |
| US9092901B2 (en) | Parallel grid population | |
| US20080234990A1 (en) | Computation of multiple body interactions | |
| US20040024578A1 (en) | Discrete event simulation system and method | |
| CN110543663B (en) | Coarse-grained MPI + OpenMP hybrid parallel-oriented structural grid area division method | |
| CN112149808A (en) | Method, system and medium for expanding stand-alone graph neural network training to distributed training | |
| CN115878729B (en) | Node block storage allocation optimization method and system based on alliance chain | |
| Bandic et al. | Mapping quantum circuits to modular architectures with QUBO | |
| Checkaraou et al. | Hybrid mpi+ openmp implementation of extended discrete element method | |
| Allmaier et al. | Parallel graph generation algorithms for shared and distributed memory machines | |
| Moore et al. | Internal diffusion-limited aggregation: parallel algorithms and complexity | |
| CN111695243B (en) | Communication method for synchronous parallel space resolution random cluster dynamics | |
| CN111695244B (en) | Material irradiation defect storage method suitable for multi-element SRSCD simulation | |
| Chen et al. | A massively parallel spatially resolved stochastic cluster dynamics method for simulations of irradiated materials | |
| Chen | The merge phase of parallel divide-and-conquer scheme for 3D Delaunay triangulation | |
| CN110704693A (en) | Distributed graph calculation system and distributed graph calculation method | |
| CN108599173B (en) | Method and device for solving batch power flows | |
| JPH07140289A (en) | Neutron transport simulator | |
| Algis et al. | Efficient GPU Implementation of Particle Interactions with Cutoff Radius and Few Particles per Cell | |
| Chung et al. | A dynamic firing speculation to speedup distributed symbolic state-space generation | |
| Li et al. | ANT-MOC: Scalable Neutral Particle Transport Using 3D Method of Characteristics on Multi-GPU Systems | |
| Couchman | Gravitational N-body simulation of large-scale cosmic structure | |
| Kalgin | Parallel implementation of asynchronous cellular automata on a 32-core computer | |
| CN116432389A (en) | Three-dimensional neutron transport characteristic line parallel solving method for heterogeneous clusters |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |