CN104375806A - Parallel computing component and method and corresponding parallel software development method and system - Google Patents
Parallel computing component and method and corresponding parallel software development method and system Download PDFInfo
- Publication number
- CN104375806A CN104375806A CN201410662947.9A CN201410662947A CN104375806A CN 104375806 A CN104375806 A CN 104375806A CN 201410662947 A CN201410662947 A CN 201410662947A CN 104375806 A CN104375806 A CN 104375806A
- Authority
- CN
- China
- Prior art keywords
- parallel computation
- numerical
- parallel
- data
- computation component
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to the technical field of parallel computing, in particular to a parallel computing component and method and a corresponding parallel software development method and system. The parallel computing component comprises a parallel computing component interface, a data distribution storage unit, a data parallel communication unit and a numerical parallel computing unit, wherein the parallel computing component interface is used for providing numerical data for other units and feeding back numerical data subjected to numerical computation; the data distribution storage unit is used for storing a data field and the numerical data received from the parallel computing component interface in a private memory space of a progress; the data parallel communication unit is used for copying data fields from private memory spaces of other progresses and storing the data fields to a local progress; the numerical parallel computing unit is used for computing the numerical data of the local progress independently according to a numerical analog computation process of a numerical algorithm. By means of the method and system, the reusability, maintainability and extendibility of a software module can be improved remarkably, and the computing efficiency of application software is improved remarkably.
Description
Technical field
The present invention relates to parallel computing field, particularly about a kind of parallel computation component, method and corresponding parallel software development method and system.
Background technology
Current, serial program parallelization is the conventional research and development method of numerical simulation Parallel application software, point two steps: the first, under the condition calculated on a small scale, for actual application problem, and design data structure and numerical algorithm, development serial program; Second, based on high-performance calculation machine architecture and Parallel Programming Environment, by serial program parallelization, gained concurrent program can obtain the analog result consistent with serial program on the one hand under small-scale design conditions, expands calculating scale on the other hand by increasing processor check figure.
The advantage of serial program parallelization is directly simple, but is difficult to adapt to the changes in demand of practical application and the complicated of high-performance calculation machine architecture.The changes in demand of practical application finds expression in many physical processes or multiple space and time scales coupling, and the sustainable growth of code complexity, needs multiple team collaborative design and development; The complicated of high-performance calculation machine architecture finds expression in usefulness wall, and practical application area expert is difficult to directly write high-effect computed program code.Parallel subqueries is embedded between the code line of serial program by serial program parallelization, do not change data structure, numerical algorithm and modular structure, and serial program designs towards calculating on a small scale usually, practical application is relatively simple, the performance requirement of less consideration large-scale calculations, also less consideration module is multiplexing, presents that counting yield is low, reusable degree is low, the deficiency of maintainable and poor expandability.Thus, along with the lasting lifting of high-performance computer peak performance, increasingly sophisticated and various along with practical application, the research and development method of serial program parallelization has been difficult to the quick research and development adapting to numerical simulation Parallel application software of new generation, urgently proposes new research and development method and system.
Summary of the invention
In order to solve in prior art parallel computation for the inefficient problem of different application software development reusability difference, the present invention proposes a kind of parallel computation component, method and corresponding parallel software development method and system, the reusability of remarkable lifting software module, maintainability and extensibility, the counting yield of remarkable lifting application software, make application expert can under the prerequisite without the need to understanding computing machine complex system structure, serial programming ground research and development Parallel application software, adapt to the numerical simulation changes in demand of complicated practical application and the complicated of high-performance calculation machine architecture.
Embodiments provide a kind of parallel computation component, comprise,
Parallel computation component interface, for providing numeric data for distributed data storage unit, data parallel communication unit and numerical parallel computation unit and returning through the numeric data after numerical evaluation process, described numeric data mainly comprises data fields and calculating parameter;
Described distributed data storage unit, for the numeric data storing data fields at the privately owned memory headroom of process and receive from parallel computation component interface;
Described data parallel communication unit, copy in the data fields of the privately owned memory headroom from other processes numerical parallel computation unit rely on grid cell and grid cell on magnitude of physical quantity, be stored into local process, meet the data dependence demand of numerical parallel computation unit;
Described numerical parallel computation unit, for the numerical simulation calculation flow process according to numerical algorithm, independently carries out numerical evaluation to the numeric data of local process, and sends the numeric data after numerical evaluation to parallel computation component interface.
The embodiment of the present invention additionally provides a kind of generation method of parallel computation component, comprises,
Determine the input and output of parallel computation interface, there is provided numeric data for distributed data storage unit, data parallel communication unit and numerical parallel computation unit and return through the numeric data after numerical evaluation process, described numeric data mainly comprises data fields and calculating parameter;
The data fields stored at the privately owned memory headroom of process according to the format determination data distribution memory cells of numerical algorithm and the numeric data received from parallel computation component interface;
Determine the magnitude of physical quantity on the grid cell that data parallel communication unit copies from the data fields of the privately owned memory headroom of other processes and grid cell, be stored into local process, meet the data dependence demand of numerical parallel computation unit;
Determine the computational logic of numerical parallel computation unit, write the serial subroutine that process performs, be registered to numerical parallel computation unit, independently numerical evaluation carried out to the numeric data of local process, and send the numeric data after numerical evaluation to parallel computation component interface.
The embodiment of the present invention additionally provides a kind of numerical simulation Parallel application software development methodology applying parallel computation component, comprises,
According to the application of numerical simulation, generate the magnitude of physical quantity data on discrete grid block and discrete grid block, form numerical simulation data field, distributed store is to the privately owned memory headroom of each process;
According to the numerical simulation function that the application of described numerical simulation is to be achieved, determine the performing step of multiple parallel computation function, determine corresponding parallel computation component according to described parallel computation function, and determine the annexation between described parallel computation component according to described performing step;
Assemble described parallel computation component according to all parallel computation components and annexation thereof, form the Parallel application software consistent with numerical simulation function to be achieved.
The embodiment of the present invention additionally provides a kind of numerical simulation Parallel application software development system applying parallel computation component, comprises,
Data fields distribution memory cells, for the application according to numerical simulation, generate the physical quantity data on discrete grid block and discrete grid block, form numerical simulation data field, distributed store is to the privately owned memory headroom of each process;
Numerical simulation calculation flow process determining unit, according to the numerical simulation function that the application of described numerical simulation is to be achieved, determines the performing step of multiple parallel computation function;
Parallel computation component determining unit, for determining corresponding parallel computation component according to described parallel computation function, and determines the annexation between described parallel computation component according to described performing step;
Parallel computation component installaiton unit, for assembling described parallel computation component according to all parallel computation components and annexation thereof, forms the Parallel application software consistent with numerical simulation function to be achieved.
By the method and system of above-described embodiment, significantly can promote the reusability of software module, maintainability and extensibility, the counting yield of remarkable lifting application software, make application expert can under the prerequisite without the need to understanding computing machine complex system structure, serial programming ground research and development Parallel application software, adapts to the numerical simulation changes in demand of complicated practical application and the complicated of high-performance calculation machine architecture.
Accompanying drawing explanation
In order to be illustrated more clearly in the technical scheme in the embodiment of the present invention, below the accompanying drawing used required in describing embodiment is briefly described, apparently, accompanying drawing in the following describes is only some embodiments of the present invention, for those of ordinary skill in the art, under the prerequisite not paying creative work, other accompanying drawing can also be obtained according to these accompanying drawings.In the accompanying drawings:
Figure 1 shows that the structural representation of a kind of parallel computation component of the embodiment of the present invention;
Figure 2 shows that the process flow diagram of the generation method of a kind of parallel computation component of the embodiment of the present invention;
Figure 3 shows that a kind of process flow diagram applying the numerical simulation Parallel application software development methodology of parallel computation component of the embodiment of the present invention;
Figure 4 shows that a kind of structural representation applying the numerical simulation Parallel application software development system of parallel computation component of the embodiment of the present invention;
Fig. 5 gives the schematic diagram of a zoning;
Figure 6 shows that for " windward discrete scheme numerical parallel computation component ", the embodiment of the present invention illustrates that neighbours exchange the generative process process flow diagram of general numerical evaluation component;
Figure 7 shows that a kind of another process flow diagram applying the numerical simulation Parallel application software development methodology of parallel computation component of the embodiment of the present invention;
Figure 8 shows that the calculation process topology that the embodiment of the present invention aforementioned linear convection equation parallel numerical solves.
Embodiment
For making the object of the embodiment of the present invention, technical scheme and advantage clearly understand, below in conjunction with accompanying drawing, the embodiment of the present invention is described in further details.At this, schematic description and description of the present invention is for explaining the present invention, but not as a limitation of the invention.
Be illustrated in figure 1 the structural representation of a kind of parallel computation component of the embodiment of the present invention.
Comprise parallel computation component interface 101, distributed data storage unit 102, data parallel communication unit 103, numerical parallel computation unit 104.
Described parallel computation component interface 101, for providing numeric data for distributed data storage unit 102, data parallel communication unit 103 and numerical parallel computation unit 104 and returning through the numeric data after numerical evaluation process, described numeric data mainly comprises data fields and calculating parameter.
Wherein, data fields comprises discrete grid block and is defined in the magnitude of physical quantity on grid, and according to the domain mesh of discrete grid block between each process, distributed store is at the privately owned memory headroom of corresponding process; Calculating parameter comprises controling parameters independent of discrete grid block and physical parameter, and calculating parameter is shared by each process usually.
Described distributed data storage unit 102, for the numeric data storing data fields at the privately owned memory headroom of process and receive from parallel computation component interface 101.
Described data parallel communication unit 103, copy in the data fields of the privately owned memory headroom from other processes numerical parallel computation unit 104 rely on grid cell and unit on magnitude of physical quantity, be stored into local process, meet the data dependence demand of numerical parallel computation unit 104.
These action needs and other processes are carried out data communication and just may be completed, and the data communication between all processes can executed in parallel.
Described numerical parallel computation unit 104, for the numerical simulation calculation flow process according to numerical algorithm, independently carries out numerical evaluation to the numeric data of local process, and sends the numeric data after numerical evaluation to parallel computation component interface.
Wherein, numerical parallel computation unit 104 is that independent parallel calculates between each process.
As a preferred embodiment of the present invention, described parallel computation component interface comprises further, input interface and output interface, described input interface is used for providing numeric data for described distributed data storage unit 102, data parallel communication unit 103 and numerical parallel computation unit 104; Described output interface is for exporting the numeric data after the process of described parallel computation component.
As a preferred embodiment of the present invention, described calculating parameter is shared by each process, and data fields is decomposed by net region, and in units of mesh flake, distributed store is at the privately owned memory headroom of each process.
As a preferred embodiment of the present invention, the parallel communications type of described data parallel communication unit 103 comprises: neighbours' switch type, overall protocol type, sparse protocol type and oriented scan type.Include but not limited to above-mentioned communication type in embodiments of the present invention, communication type in the art is all appreciated that and can be applied in technical solution of the present invention.
As a preferred embodiment of the present invention, the parallel computation type of described numerical parallel computation unit 104 comprises: physical quantity adjusts internal memory, physical quantity initialize, general numerical evaluation, time step to calculate and physical quantity deposits new value.Include but not limited to above-mentioned parallel computation type in embodiments of the present invention, parallel computation type in the art is all appreciated that and can be applied in technical solution of the present invention.
As a preferred embodiment of the present invention, described numerical parallel computation unit 104 be can separate configurations to complete the Numerical Computation Function of different numerical algorithms.
By the component in above-described embodiment, different parallel computation components can be generated by for different numerical algorithms, parallel computation component interface in parallel computation component, distributed data storage unit, data parallel communication unit all can be general or carry out small-scale amendment, numerical parallel computation unit by the individual demand of application software R&D expert according to practical application, can carry out instantiation according to numerical algorithm serial programming.Numerical algorithm is different, and numerical parallel computation unit also can be different, but numeric data, distributed data storage unit are consistent with data parallel communication unit.Thus by multiplexing number Value Data, distributed data storage unit and data parallel communication unit, parallel computation component can adapt to the fast integration of dissimilar numerical algorithm and different numerical algorithm software module, promote the reusable ability of software module.
Be illustrated in figure 2 the process flow diagram of the generation method of a kind of parallel computation component of the embodiment of the present invention.
Comprise step 201, determine the input and output of parallel computation interface, there is provided numeric data for distributed data storage unit, data parallel communication unit and numerical parallel computation unit and return through the numeric data after numerical evaluation process, described numeric data mainly comprises data fields and calculating parameter.
Step 202, the data fields stored at the privately owned memory headroom of process according to the format determination data distribution memory cells of numerical algorithm and the numeric data received from parallel computation component interface.
Step 203, determines the magnitude of physical quantity on the grid cell that data parallel communication unit copies from the data fields of the privately owned memory headroom of other processes and described grid cell, is stored into local process, meets the data dependence demand of numerical parallel computation unit.
Step 204, determine the computational logic of numerical parallel computation unit, write the serial subroutine that process performs, be registered to numerical parallel computation unit, independently numerical evaluation is carried out to the numeric data of local process, and send the numeric data after numerical evaluation to parallel computation component interface.
As a preferred embodiment of the present invention, described calculating parameter is shared by each process, and data fields is decomposed by net region, and in units of mesh flake, distributed store is at the privately owned memory headroom of each process.
As a preferred embodiment of the present invention, the parallel communications type in described data parallel communication unit comprises: neighbours' switch type, overall protocol type, sparse protocol type and oriented scan type.Include but not limited to above-mentioned communication type in embodiments of the present invention, communication type in the art is all appreciated that and can be applied in technical solution of the present invention.
As a preferred embodiment of the present invention, the parallel computation type in described numerical parallel computation unit comprises: physical quantity adjusts internal memory, physical quantity initialize, general numerical evaluation, time step to calculate and physical quantity deposits new value.Include but not limited to above-mentioned parallel computation type in embodiments of the present invention, parallel computation type in the art is all appreciated that and can be applied in technical solution of the present invention.
By the component that method in above-described embodiment generates, different parallel computation components can be generated by for different numerical algorithms, parallel computation component interface in parallel computation component, distributed data storage unit, data parallel communication unit all can be general or carry out small-scale amendment, numerical parallel computation unit by the individual demand of application software R&D expert according to practical application, can carry out instantiation according to numerical algorithm serial programming.Numerical algorithm is different, and numerical parallel computation unit also can be different, but numeric data, distributed data storage unit are consistent with data parallel communication unit.Thus by multiplexing number Value Data, distributed data storage unit and data parallel communication unit, parallel computation component can adapt to the fast integration of dissimilar numerical algorithm and different numerical algorithm software module, promote the reusable ability of software module.
Be illustrated in figure 3 a kind of process flow diagram applying the numerical simulation Parallel application software development methodology of parallel computation component of the embodiment of the present invention.
Comprise step 301, according to the application of numerical simulation, generate the magnitude of physical quantity data on discrete grid block and discrete grid block, form numerical simulation data field, distributed store is to the privately owned memory headroom of each process.
Step 302, according to the numerical simulation function that the application of described numerical simulation is to be achieved, determine the performing step of multiple parallel computation function, determine corresponding parallel computation component according to described parallel computation function, and determine the annexation between described parallel computation component according to described performing step.
Wherein, described parallel computation function refers to that a parallel computation function is merged in the once communication of numerical simulation function and calculating.
Step 303, assembles described parallel computation component according to all parallel computation components and annexation thereof, forms the Parallel application software consistent with numerical simulation function to be achieved.
As a preferred embodiment of the present invention, described discrete grid block comprises structured grid or unstrctured grid, and described data fields comprises the selection of coordinate system, the generation of discrete grid block, the statement of physical quantity; Physical quantity data on described generation discrete grid block comprise, and generate described physical quantity data in the different geometric positions of discrete grid block.
As a preferred embodiment of the present invention, described different geometric position comprises, the element sides heart of the unit center of discrete grid block, the cell node of discrete grid block, discrete grid block, limit, the mesh flake appearance limit heart, the mesh flake appearance rib rib heart, the mesh flake outside surface center of area.
As a preferred embodiment of the present invention, determine that corresponding parallel computation component comprises further according to described parallel computation function, the parallel computation component that the form of numerical algorithm is corresponding with described parallel computation function is selected from parallel computation component base, if parallel computation component not corresponding with described parallel computation function in described parallel computation component base, the form then generating numerical algorithm corresponds to the parallel computation component of described parallel computation function, and joins in described parallel computation component base.
As a preferred embodiment of the present invention, the parallel computation component that the form of described generation numerical algorithm corresponds to described parallel computation function comprises further, determine the input and output of parallel computation interface, there is provided numeric data for distributed data storage unit, data parallel communication unit and numerical parallel computation unit and return through the numeric data after numerical evaluation process, described numeric data mainly comprises data fields and calculating parameter; The data fields stored at the privately owned memory headroom of process according to the format determination data distribution memory cells of numerical algorithm and the numeric data received from parallel computation component interface; Determine the magnitude of physical quantity on the grid cell that data parallel communication unit copies from the data fields of the privately owned memory headroom of other processes and unit, be stored into local process, meet the data dependence demand of numerical parallel computation unit; Determine the computational logic of numerical parallel computation unit, write the serial subroutine that process performs, be registered to numerical parallel computation unit, independently numerical evaluation carried out to the numeric data of local process, and send the numeric data after numerical evaluation to parallel computation component interface.
As a preferred embodiment of the present invention, assemble described parallel computation component according to described all parallel computation components and annexation thereof to comprise further, according to described all parallel computation components and annexation thereof, all parallel computation components are connected, and distributed data storage unit, data parallel communication unit, the numerical parallel computation unit in parallel computation component described in instantiation, described parallel computation component is configured and is defined.
By the method for the invention described above embodiment, significantly can promote the reusability of software module, maintainability and extensibility, the counting yield of remarkable lifting application software, make application expert can under the prerequisite without the need to understanding computing machine complex system structure, serial programming ground research and development Parallel application software, adapts to the numerical simulation changes in demand of complicated practical application and the complicated of high-performance calculation machine architecture.
Be illustrated in figure 4 a kind of structural representation applying the numerical simulation Parallel application software development system of parallel computation component of the embodiment of the present invention.
Comprise data fields distribution memory cells 401, for the application according to numerical simulation, generate the physical quantity data on discrete grid block and discrete grid block, form numerical simulation data field, distributed store is to the privately owned memory headroom of each process.
Numerical simulation calculation flow process determining unit 402, according to the numerical simulation function that the application of described numerical simulation is to be achieved, determines the performing step of multiple parallel computation function.
Parallel computation component determining unit 403, for determining corresponding parallel computation component according to described parallel computation function, and determines the annexation between described parallel computation component according to described performing step.
Parallel computation component installaiton unit 404, for assembling described parallel computation component according to all parallel computation components and annexation thereof, forms the Parallel application software consistent with numerical simulation function to be achieved.
As a preferred embodiment of the present invention, parallel computation component installaiton unit 404 comprises further, configuration subelement 4041, assembling subelement 4042, code building subelement 4043;
Described configuration subelement 4041, for being configured each parallel computation component and defining;
Assembling subelement 4042, for according to described all parallel computation components and annexation thereof, connects all parallel computation components;
Code building subelement 4043, for the distributed data storage unit in parallel computation component described in instantiation, data parallel communication unit, numerical parallel computation unit.
By the system of the invention described above embodiment, significantly can promote the reusability of software module, maintainability and extensibility, the counting yield of remarkable lifting application software, make application expert can under the prerequisite without the need to understanding computing machine complex system structure, serial programming ground research and development Parallel application software, adapts to the numerical simulation changes in demand of complicated practical application and the complicated of high-performance calculation machine architecture.
In embodiment as shown in Figure 1, parallel computation component interface 101 is distributed data storage unit 102, data parallel communication unit 103 and numerical parallel computation unit 104 provide numeric data and return through the numeric data after numerical parallel computation process.Here, numeric data mainly comprises data fields and calculating parameter, and wherein, data fields comprises discrete grid block and is defined in magnitude of physical quantity on grid, and calculating parameter comprises controling parameters independent of discrete grid block and physical parameter.
Parallel computation component interface is parallel computation component extraction numeric data or carries out the window of information interaction with user interface.
Described user interface exponential quantity simulates the mathematical calculation model and numerical evaluation parameter that input or output in current value parallel computation step.
Described parallel computation component interface comprises parallel computation component input interface and parallel computation component output interface, described parallel computation component input interface is distributed data storage unit 102, data parallel communication unit 103 and numerical parallel computation unit 104 provide numeric data, and parallel computation component output interface provides the numeric data after the process of parallel computation component.
Distributed data storage unit 102, stores data fields for the privately owned memory headroom in process, and from the numeric data that parallel computation component interface 101 receives.In numeric data, calculating parameter is shared by each process usually, and data fields is decomposed by net region, and in units of mesh flake, distributed store is at the privately owned memory headroom of each process.Fig. 5 gives the schematic diagram of a zoning, by discrete be the grid of 20x20 unit, grid is 7 mesh flakes by subdivision, wherein, 0th, 1,3, No. 4 mesh flake distributed store is to the privately owned memory headroom of No. 0 process, the 2nd, 5, No. 6 mesh flake distributed store to the privately owned memory headroom of No. 1 process, wherein, the subregion that mesh flake is formed after representing mesh generation, all grid cells comprising this subregion and all magnitudes of physical quantity be defined on grid cell.
Data parallel communication unit 103, copies data for the private room from other processes, and these data need by communication usually, transmits and is stored into the privately owned memory headroom of local process, meeting the data dependence demand of numerical parallel computation unit 104.The communication of data parallel communication unit 103 between each process can executed in parallel.For No. 3 mesh flake of No. 0 process in Fig. 5, numerical parallel computation needs the 0th, 1,4,2, of neighbours the No. 6 mesh flake to provide the magnitude of physical quantity of border grid cell usually, wherein, 0th, the data of 1, No. 4 mesh flake can be obtained by local internal storage access, and the data of the 2nd, No. 6 mesh flake can only by obtaining with the data communication of No. 1 process.
Numerical parallel computation unit 104, according to the numerical algorithm of application individual character, carries out numerical evaluation to the numeric data of local process, and sends the numeric data after numerical evaluation process to parallel computation component interface 101.For the mesh flake of 7 in Fig. 5, No. 0 process travels through the 0th, 1,3, No. 4 mesh flake successively, No. 1 process travels through the 2nd, 5, No. 6 mesh flake similarly, each mesh flake calls numerical evaluation subroutine calculate, process grid and magnitude of physical quantity on corresponding mesh flake, two processes are that complete independent parallel calculates.Here, described numerical evaluation subroutine refers to the numerical algorithm serial subroutine of writing according to application individual character.Numerical parallel computation unit 104 needs data parallel communication unit 103 to perform a data communication before performing usually.
Data parallel communication and numerical parallel computation can be expressed by the mode of computing machine pseudo-language, and no matter also can express by natural language form, be the mode of pseudo-language, or the form of natural language, has strict parallel computation grammar request.
Parallel computation component expresses concrete Computed-torque control by specific data parallel communication type and numerical parallel computation type.In parallel computation member designs process, everyly relate to data parallel communication or numerical parallel computation part, all need the type in type of service set.Preferably, in the embodiment of the present invention, described data parallel communication unit 103 uses following data parallel communication type:
Neighbours' switch type: along the border grid cell of mesh flake, copies mesh node coordinate and grid cell magnitude of physical quantity from adjacent mesh sheet, is stored into the correspondence memory address of border grid cell;
Global reduction type: all grid cells traveling through all mesh flakes, for physical quantity asks reduction value, comprises maximal value, minimum value, cumulative sum etc.;
Sparse reduction type: around certain regional area, traversal covers the grid cell of the mesh flake in this region, for physical quantity asks reduction value, comprises maximal value, minimum value, cumulative sum etc.;
Oriented scan type: by the oriented data dependence relation between grid cell, travels through all grid cells of all mesh flakes, quotes mesh node coordinate and the magnitude of physical quantity of neighboring grid cells, completes the numerical evaluation of current grid unit.
Preferably, in the embodiment of the present invention, numerical parallel computation unit 104 uses following numerical parallel computation type:
Physical quantity adjusts internal memory: traversal mesh flake, and be mesh node coordinate or the application of grid cell magnitude of physical quantity or releasing memory space, memory headroom is used for the floating point values of save mesh node coordinate, or the floating-point of physical quantity or fixed-point value; Period, all processes can perform by independent parallel.
Physical quantity initialize: traversal mesh flake, be mesh node coordinate or grid cell magnitude of physical quantity initialize one by one, initial value comprises floating point values and fixed-point value grid cell; Period, all processes can perform by independent parallel.
General numerical evaluation: traversal mesh flake, according to the actuating logic of the numerical algorithm of application individual character, carries out floating-point or fixed-point arithmetic to mesh node coordinate or magnitude of physical quantity one by one grid cell; All processes can perform by independent parallel.
Time step calculates: traversal mesh flake, and according to the actuating logic of the time step numerical algorithm of application individual character, grid cell ground step-length computing time, asks the minimum value of all grid cell time steps and return one by one; Period, all processes can perform by independent parallel.
Physical quantity deposits new value: traversal mesh flake, and grid cell ground renewal mesh node coordinate or grid cell magnitude of physical quantity are current up-to-date numerical result one by one, for numerical parallel computation is prepared next time.
Parallel computation component is simulated the abstract of the analog functuion of the numerical algorithm that Parallel application software adopts by logarithm value and formed.For numeric datas such as the discrete grid block of distributed store and magnitudes of physical quantity, parallel computation component, by the configuration of numerical value parallel communications and numerical parallel computation, logarithm value data can process independently, complete corresponding Numerical Computation Function.Thus, in the different step of numerical algorithm, can create and configure different parallel computation components independently, numerical simulation Parallel application software can be built by these components.
Linear convection equation is the quasi-representative equation needing in numerical simulation to solve.The numerical algorithm determining to solve linear convection equation is discrete scheme windward, according to this form, determine that data parallel communication type is for neighbours' switch type, numerical parallel computation type is general numerical evaluation type, thus determine the type of parallel computation component for neighbours exchange general numerical evaluation component.Be illustrated in figure 6 the embodiment of the present invention and illustrate that neighbours exchange the generative process process flow diagram of general numerical evaluation component for " windward discrete scheme numerical parallel computation component ".
Step 601, to determine in numerical parallel computation the input and output of parallel computation interface in discrete scheme windward, determine that numeric data to be entered was magnitude of physical quantity and the calculating parameter in a upper discrete physics moment of discrete scheme dependence windward, determine that numeric data to be output is that discrete scheme performs magnitude of physical quantity and the calculating parameter in the present physical moment of rear feedback windward.
Step 602, is stored into distributed data storage unit by discrete grid block and the magnitude of physical quantity be defined on discrete grid block.
Step 603, according to the data dependence relation between mesh flake in discrete scheme windward, for mesh flake is determined to need the magnitude of physical quantity from the upper physics moment that neighbours' mesh flake grid cell copies, and physical quantity is registered in the data parallel communication unit of parallel computation component.The data parallel communication of all processes can perform by independent parallel.
Step 604, according to the numerical algorithm logic of discrete scheme windward, for mesh flake writes serial subroutine, is registered to the numerical parallel computation unit of parallel computation component.The numerical parallel computation of all processes can perform by independent parallel.
The numerical parallel computation of the discrete scheme windward structure manufacture of linear convection equation is complete, can be stored in parallel computation component base.According to the numerical algorithm flow process of numerical simulation Parallel application software and the demand of function, carry out concentrating preservation and management to the parallel computation component created by described parallel computation component base.
Be illustrated in figure 7 a kind of another process flow diagram applying the numerical simulation Parallel application software development methodology of parallel computation component of the embodiment of the present invention.
Comprise step 701, according to the numerical simulation application of Parallel application software, generate the magnitude of physical quantity data on discrete grid block and discrete grid block, form numerical simulation data field, distributed store is to the privately owned memory headroom of each process.
Concrete, generate discrete grid block and carry out net region subdivision, all subregions (all grid cells containing on subregion) after subdivision are assigned to each process; Each process, on the subregion had separately, for grid cell generates magnitude of physical quantity, forms respective numerical simulation data field, is stored into privately owned memory headroom; The numerical simulation data field of the privately owned memory headroom of all processes is brought together, the common numerical simulation data field forming distributed store, the i.e. data fields of numeric data described in aforementioned data distribution memory cells.
Data fields comprises the selection of coordinate system, the generation of discrete grid block, the statement of physical quantity, and wherein, discrete grid block can be structured grid, also can be unstrctured grid; Physical quantity can be defined in the different geometric positions of discrete grid block, comprises unit center, cell node, the element sides heart, limit, the mesh flake appearance limit heart, the mesh flake appearance rib rib heart, the mesh flake outside surface center of area etc.
Step 702, according to the numerical simulation function of the numerical algorithm that Parallel application software adopts, determines all parallel computation functions to be achieved, determines the calculation process topology be made up of all parallel computation functions.
Described parallel computation functional realiey step refers in this step, data parallel communication performs at the most once between each process, numerical parallel computation is independent execution between each process, and the once communication in algorithm flow and corresponding calculating merging thereof are become a parallel computation function.Connect these parallel computation functions, form calculation process topology.
A calculation process topology comprises three partial contents, the initialization of first numeric data, and it two is numerical evaluation of numeric data, and it three is storages of numeric data.Be initialized as numeric data initialize, numerical evaluation processes numeric data successively according to the analog functuion of numerical algorithm, stores and preserves the numeric data after numerical evaluation process.Wherein, numerical evaluation can according to the analog functuion of numerical algorithm, refine to concrete parallel computation function, until every parallel computation function needs to perform a numerical value parallel communications at the most between process, thus a parallel computation component can be built independently.
Step 703, determines corresponding parallel computation component according to described parallel computation function, and determines the annexation between described parallel computation component according to described performing step.
Concrete, after the calculation process topology of Parallel application software is determined, all parallel computation functions just determine, and every parallel computation function can be used as a step, then the actuating logic of Binding number value-based algorithm, and the performing step of Parallel application software just determines.What Fig. 8 showed is the calculation process topology that aforementioned linear convection equation parallel numerical solves.Determined by calculation process topology, the parallel numerical that realize equation solves, and needs execution three steps, comprising: initialization, time stepping and reception numerical solution.Wherein, initialization is subdivided into two sub-steps of physical quantity application memory headroom (step 803) and physical quantity initialize (step 804); Time stepping is subdivided into calculating (step 805), temporally step-length integration time step (step 808), three sub-steps of step update time (step 809) of time step; Receive numerical solution and be included in important physical moment storage numerical solution to facilitate two sub-steps of physical analysis (step 810) and physical quantity releasing memory space (step 811).Described " on time spacer step integration time step " can be decomposed into again computing grid numerical flux (step 806), upgrade two sub-steps of grid physical value (step 807).Thus, the calculation process topology that parallel numerical solves linear convection equation comprises 8 parallel computation functions, then the actuating logic of Binding number value-based algorithm, and 8 performing steps of Parallel application software determine.
The performing step of Parallel application software can determine the annexation between parallel computation component and each parallel computation component after determining.
First, according to the performing step in calculation process topology, determine the parallel computation component corresponding to each step, then determine the annexation between each parallel computation component according to logical relation between each step.
Still for the calculation process of aforementioned linear convection equation topology.Physical quantity application memory headroom (step 803) can adjust the parallel computation component of the numerical parallel computation type of internal memory to complete by physical quantity, is labeled as A1; Physical quantity initialize (step 804) can be completed by the parallel computation component of the numerical parallel computation type of physical quantity initialize, is labeled as A2; The parallel computation component of the numerical parallel computation type that the calculating (step 805) of time step can be calculated by time step completes, and is labeled as A3; Computing grid numerical flux (step 806) can have been come by the aforementioned numerical parallel computation of discrete scheme windward component, is labeled as A4; Upgrade grid physical amount (step 807) to be completed by the parallel computation component of general numerical evaluation type, be labeled as A5; Step update time (step 809) is deposited the numerical evaluation type of new value parallel computation component by physical quantity completes, and is labeled as A6; Receive the parallel computation component that numerical solution (step 810) can deposit the parallel computation component of the numerical evaluation type of new value and the numerical evaluation type of physical quantity tune internal memory by physical quantity jointly to complete, be labeled as A7 and A8 respectively.Physical quantity releasing memory space (step 811) can adjust the parallel computation component of the numerical evaluation type of internal memory to complete by physical quantity, is labeled as A9.Wherein step 803, step 810 and step 811 multiplexed physical amount can adjust the parallel computation component of numerical evaluation type of internal memory.
According to the logical relation of the performing step of aforementioned calculation process topology, the annexation between 8 parallel computation components can be determined.The numeric data that parallel computation component A1 is formed with step 701 is input, for physical quantity applies for memory headroom on mesh flake, stores the value of physical quantity; The numeric data that parallel computation component A2 exports using parallel computation component A1 as input, for possessing the physical quantity initialize of memory headroom; Parallel computation component A3 using the numeric data after parallel computation component A2 process as input, according to solving the numerical algorithm of time step, computing time step-length; Parallel computation component A4 using the numeric data after parallel computation component A3 process as input, according to the numerical algorithm of discrete scheme windward, computing grid numerical flux; Parallel computation component A5, using the numeric data after parallel computation component A4 process and grid values flux as input, according to the numerical algorithm upgrading physical quantity, upgrades the value of physical quantity; Parallel computation component A6, using the numeric data after parallel computation component A5 process as input, stores the value of the physical quantity in present physical moment; Parallel computation component A7 and A8 memory parallel calculate the numeric data in the present physical moment that component A6 stores to external space, and parallel computation component A9 discharges the memory headroom of parallel computation component A2 application.
Parallel computation component can be chosen from the parallel computation component base set up in advance.If parallel computation component cannot obtain from parallel computation component base, then can create interim parallel computation component, concrete creation method can with reference to the method for aforementioned generation component, for interim parallel computation component carries out Numerical Computation Function description.Use after maturation until interim parallel computation component, transferred to ripe parallel computation component, add in parallel computation component base, for the research and development of follow-up Parallel application software.
Step 704, according to described annexation assembling parallel computation component to form the Parallel application software corresponding with numerical simulation function to be achieved.
As previously mentioned, now determine the annexation between parallel computation component needed for analog functuion to be achieved and parallel computation component, according to described annexation, each parallel computation component has been carried out assembling and can generate the application software corresponding with analog functuion to be achieved.
Assembling parallel computation component refers to that parallel computation component clear and definite in calculation process topology is determined complete, and the parallel computation component do not had in parallel computation component base also make complete after, annexation between each parallel computation component determined according to calculation process topology, each parallel computation component is carried out to configuration and the definition of technology essential factor, comprising: the amendment, the definition of input/output argument value etc. of attribute data, doing technique preparation for completing application software generation.
After completing parallel computation component installaiton, perform " generating code " operation, the rule that can generate according to system code and key element, according to parallel computation component, automatically generate the parallel source code corresponding with LINUX operating system.Code building for other system should be understood to the part in technical solution of the present invention too.
In one embodiment of this invention, the code of generation can according to address configuration information, under the physical pathway being directly generated to the server of specifying.
Preferably, in another embodiment of the invention, the code of generation can run by direct compilation.
So far, the numerical simulation application software corresponding with analog functuion to be achieved is generated.
During parallel computation, data fields distributed store at the privately owned memory headroom of each process; Distributed data storage unit is the data fields that parallel computation component management is stored in process private room, data parallel communication unit copies from other process private rooms the grid cell and magnitude of physical quantity that non-local process private room that numerical parallel computation relies on stores for parallel computation component, numerical parallel computation unit performs the numerical algorithm of practical application individual character independently, the data fields for the treatment of progress private room; The data fields of parallel computation component interface external only input and output calculating parameter and distributed store.Thus, parallel computation component shields the distributed store of data fields, encapsulate the data parallel communication needed for numerical parallel computation, support that domain expert is under the prerequisite without the need to understanding parallel computation realization, realizes the parallel computation functional realiey step in numerical simulation calculation flow process topology serial programming.
Discrete grid block subregion and the data fields such as physical values being defined in grid cell can be optimized design in conjunction with the performance characteristics of high-performance calculation machine architecture, data parallel communication and corresponding numerical evaluation load balance can be realized by the high-performance calculation expert of specialty, thus, parallel computation component has higher counting yield, is more adapted to Large-scale parallel computing.
High-performance calculation machine architecture changes, only need redesign or distributed store structure, the data parallel communication unit of optimal design data fields, without the need to changing parallel computation component interface, data parallel storage unit and numerical parallel computation unit, and the numerical simulation calculation flow process topology of application software, thus calculating component and application software can adapt to the complicated of high-performance calculation machine architecture fast.
In the inside of parallel computation component, numerical parallel computation unit by the individual demand of application software R&D expert according to practical application, can carry out instantiation according to numerical algorithm serial programming.Numerical algorithm is different, and numerical parallel computation unit also can be different, but numeric data, distributed data storage unit are consistent with data parallel communication unit.Such as, in the generation of the aforementioned numerical parallel computation of discrete scheme windward component, discrete scheme can be different windward, but the parallel computation interface of these forms, distributed data storage communicates with data parallel can be consistent.Thus by multiplexing number Value Data, distributed data storage unit and data parallel communication unit, parallel computation component can adapt to the fast integration of dissimilar numerical algorithm and different numerical algorithm software module, promote the reusable ability of software module.
Numerical simulation flow process topology can according to the numerical simulation customizing functions to be achieved of application software, all parallel computation components can dynamic construction independently, even if in the inside of parallel computation component, distributed data storage unit, data parallel communication unit and numerical parallel computation unit also can Design and implementations independently, the internal part of all parallel computation components and parallel computation component all can by different individuals or team's complete independently, therefore, Parallel application research and development of software method based on parallel computation component significantly can promote safeguarding and expandability of code, adapt to the changes in demand of practical application better.
Above-described specific embodiment; object of the present invention, technical scheme and beneficial effect are further described; be understood that; the foregoing is only specific embodiments of the invention; the protection domain be not intended to limit the present invention; within the spirit and principles in the present invention all, any amendment made, equivalent replacement, improvement etc., all should be included within protection scope of the present invention.
Claims (9)
1. a parallel computation component, is characterized in that comprising,
Parallel computation component interface, for providing numeric data for distributed data storage unit, data parallel communication unit and numerical parallel computation unit and returning through the numeric data after numerical evaluation process, described numeric data mainly comprises data fields and calculating parameter;
Described distributed data storage unit, for the numeric data storing data fields at the privately owned memory headroom of process and receive from parallel computation component interface;
Described data parallel communication unit, copy in the data fields of the privately owned memory headroom from other processes numerical parallel computation unit rely on grid cell and grid cell on magnitude of physical quantity, be stored into local process, meet the data dependence demand of numerical parallel computation unit;
Described numerical parallel computation unit, for the numerical simulation calculation flow process according to numerical algorithm, independently carries out numerical evaluation to the numeric data of local process, and sends the numeric data after numerical evaluation to parallel computation component interface.
2. a generation method for parallel computation component, is characterized in that comprising,
Determine the input and output of parallel computation interface, there is provided numeric data for distributed data storage unit, data parallel communication unit and numerical parallel computation unit and return through the numeric data after numerical evaluation process, described numeric data mainly comprises data fields and calculating parameter;
The data fields stored at the privately owned memory headroom of process according to the format determination data distribution memory cells of numerical algorithm and the numeric data received from parallel computation component interface;
Determine the magnitude of physical quantity on the grid cell that data parallel communication unit copies from the data fields of the privately owned memory headroom of other processes and grid cell, be stored into local process, meet the data dependence demand of numerical parallel computation unit;
Determine the computational logic of numerical parallel computation unit, write the serial subroutine that process performs, be registered to numerical parallel computation unit, independently numerical evaluation carried out to the numeric data of local process, and send the numeric data after numerical evaluation to parallel computation component interface.
3. apply a numerical simulation Parallel application software development methodology for parallel computation component, it is characterized in that comprising,
According to the application of numerical simulation, generate the magnitude of physical quantity data on discrete grid block and discrete grid block, form numerical simulation data field, distributed store is to the privately owned memory headroom of each process;
According to the numerical simulation function that the application of described numerical simulation is to be achieved, determine the performing step of multiple parallel computation function, determine corresponding parallel computation component according to described parallel computation function, and determine the annexation between described parallel computation component according to described performing step;
Assemble described parallel computation component according to all parallel computation components and annexation thereof, form the Parallel application software consistent with numerical simulation function to be achieved.
4. a kind of numerical simulation Parallel application software development methodology applying parallel computation component according to claim 3, it is characterized in that, determine that corresponding parallel computation component comprises further according to described parallel computation function, the parallel computation component that the form of numerical algorithm is corresponding with described parallel computation function is selected from parallel computation component base, if parallel computation component not corresponding with described parallel computation function in described parallel computation component base, the form then generating numerical algorithm corresponds to the parallel computation component of described parallel computation function, and join in described parallel computation component base.
5. a kind of numerical simulation Parallel application software development methodology applying parallel computation component according to claim 4, it is characterized in that, the parallel computation component that the form of described generation numerical algorithm corresponds to described parallel computation function comprises further, determine the input and output of parallel computation interface, there is provided numeric data for distributed data storage unit, data parallel communication unit and numerical parallel computation unit and return through the numeric data after numerical evaluation process, described numeric data mainly comprises data fields and calculating parameter; The data fields stored at the privately owned memory headroom of process according to the format determination data distribution memory cells of numerical algorithm and the numeric data received from parallel computation component interface; Determine the magnitude of physical quantity on the grid cell that data parallel communication unit copies from the data fields of the privately owned memory headroom of other processes and unit, be stored into local process, meet the data dependence demand of numerical parallel computation unit; Determine the computational logic of numerical parallel computation unit, write the serial subroutine that process performs, be registered to numerical parallel computation unit, independently numerical evaluation carried out to the numeric data of local process, and send the numeric data after numerical evaluation to parallel computation component interface.
6. a kind of numerical simulation Parallel application software development methodology applying parallel computation component according to claim 3, is characterized in that, described parallel computation function refers to that a parallel computation function is merged in the once communication of numerical simulation function and calculating.
7. a kind of numerical simulation Parallel application software development methodology applying parallel computation component according to claim 3, it is characterized in that, assemble described parallel computation component according to described all parallel computation components and annexation thereof to comprise further, according to described all parallel computation components and annexation thereof, all parallel computation components are connected, and distributed data storage unit, data parallel communication unit, the numerical parallel computation unit in parallel computation component described in instantiation, described parallel computation component is configured and is defined.
8. apply a numerical simulation Parallel application software development system for parallel computation component, it is characterized in that comprising,
Data fields distribution memory cells, for the application according to numerical simulation, generate the physical quantity data on discrete grid block and discrete grid block, form numerical simulation data field, distributed store is to the privately owned memory headroom of each process;
Numerical simulation calculation flow process determining unit, according to the numerical simulation function that the application of described numerical simulation is to be achieved, determines the performing step of multiple parallel computation function;
Parallel computation component determining unit, for determining corresponding parallel computation component according to described parallel computation function, and determines the annexation between described parallel computation component according to described performing step;
Parallel computation component installaiton unit, for assembling described parallel computation component according to all parallel computation components and annexation thereof, forms the Parallel application software consistent with numerical simulation function to be achieved.
9. a kind of numerical simulation Parallel application software development system applying parallel computation component according to claim 8, it is characterized in that, parallel computation component installaiton unit comprises further, configuration subelement, assembling subelement, code building subelement;
Described configuration subelement, for being configured each parallel computation component and defining;
Assembling subelement, for according to described all parallel computation components and annexation thereof, connects all parallel computation components;
Code building subelement, for the distributed data storage unit in parallel computation component described in instantiation, data parallel communication unit, numerical parallel computation unit.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410662947.9A CN104375806B (en) | 2014-11-19 | 2014-11-19 | A kind of parallel computation component, method and corresponding parallel software development method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410662947.9A CN104375806B (en) | 2014-11-19 | 2014-11-19 | A kind of parallel computation component, method and corresponding parallel software development method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104375806A true CN104375806A (en) | 2015-02-25 |
| CN104375806B CN104375806B (en) | 2015-12-09 |
Family
ID=52554756
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410662947.9A Active CN104375806B (en) | 2014-11-19 | 2014-11-19 | A kind of parallel computation component, method and corresponding parallel software development method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104375806B (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106407005A (en) * | 2016-08-31 | 2017-02-15 | 中国科学院计算机网络信息中心 | Concurrent process merging method and system based on multi-scale coupling |
| CN108228364A (en) * | 2017-12-27 | 2018-06-29 | 中国航天空气动力技术研究院 | For data storing seal method, system and the medium of high performance parallel computation |
| CN108509220A (en) * | 2018-04-02 | 2018-09-07 | 厦门海迈科技股份有限公司 | Revit engineering calculation amounts method for parallel processing, device, terminal and medium |
| CN109542406A (en) * | 2018-11-05 | 2019-03-29 | 清华大学 | Parallel implementation method and system for mode development |
| CN110716751A (en) * | 2018-07-12 | 2020-01-21 | 赛灵思公司 | High-parallelism computing platform, system and computing implementation method |
| CN111984916A (en) * | 2020-10-09 | 2020-11-24 | 北京应用物理与计算数学研究所 | Mathematical equation solving component and parallel software research and development method and system |
| CN112783476A (en) * | 2021-01-15 | 2021-05-11 | 中国核动力研究设计院 | Easily-extensible software system, calling method and terminal for reactor core numerical solver |
| CN113157806A (en) * | 2021-04-19 | 2021-07-23 | 清华大学 | Grid data distributed storage service system, method, device, equipment and medium |
| CN113420466A (en) * | 2021-08-12 | 2021-09-21 | 北京应用物理与计算数学研究所 | Cross-platform automatic performance optimization oriented unit computing component and method |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101661514A (en) * | 2008-05-21 | 2010-03-03 | 中国石化股份胜利油田分公司地质科学研究院 | Oil deposit black oil model numerical simulation system |
| CN102314342A (en) * | 2010-06-18 | 2012-01-11 | 微软公司 | Be used for the counterfoil that calls that the compiler of data parallel programming model generates |
-
2014
- 2014-11-19 CN CN201410662947.9A patent/CN104375806B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101661514A (en) * | 2008-05-21 | 2010-03-03 | 中国石化股份胜利油田分公司地质科学研究院 | Oil deposit black oil model numerical simulation system |
| CN102314342A (en) * | 2010-06-18 | 2012-01-11 | 微软公司 | Be used for the counterfoil that calls that the compiler of data parallel programming model generates |
Non-Patent Citations (1)
| Title |
|---|
| 尹九阳 等: "并行计算构件技术在过程模拟中的应用", 《计算机与应用化学》 * |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106407005B (en) * | 2016-08-31 | 2019-09-27 | 中国科学院计算机网络信息中心 | A kind of concurrent process merging method and system based on multi-scale coupling |
| CN106407005A (en) * | 2016-08-31 | 2017-02-15 | 中国科学院计算机网络信息中心 | Concurrent process merging method and system based on multi-scale coupling |
| CN108228364B (en) * | 2017-12-27 | 2020-12-18 | 中国航天空气动力技术研究院 | Data storage packaging method, system and medium for high-performance parallel computing |
| CN108228364A (en) * | 2017-12-27 | 2018-06-29 | 中国航天空气动力技术研究院 | For data storing seal method, system and the medium of high performance parallel computation |
| CN108509220A (en) * | 2018-04-02 | 2018-09-07 | 厦门海迈科技股份有限公司 | Revit engineering calculation amounts method for parallel processing, device, terminal and medium |
| CN110716751A (en) * | 2018-07-12 | 2020-01-21 | 赛灵思公司 | High-parallelism computing platform, system and computing implementation method |
| CN109542406A (en) * | 2018-11-05 | 2019-03-29 | 清华大学 | Parallel implementation method and system for mode development |
| CN111984916A (en) * | 2020-10-09 | 2020-11-24 | 北京应用物理与计算数学研究所 | Mathematical equation solving component and parallel software research and development method and system |
| CN111984916B (en) * | 2020-10-09 | 2021-01-12 | 北京应用物理与计算数学研究所 | Mathematical equation solving component and parallel software research and development method and system |
| CN112783476A (en) * | 2021-01-15 | 2021-05-11 | 中国核动力研究设计院 | Easily-extensible software system, calling method and terminal for reactor core numerical solver |
| CN112783476B (en) * | 2021-01-15 | 2022-02-22 | 中国核动力研究设计院 | Easily-extensible software system, calling method and terminal for reactor core numerical solver |
| CN113157806A (en) * | 2021-04-19 | 2021-07-23 | 清华大学 | Grid data distributed storage service system, method, device, equipment and medium |
| CN113420466A (en) * | 2021-08-12 | 2021-09-21 | 北京应用物理与计算数学研究所 | Cross-platform automatic performance optimization oriented unit computing component and method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104375806B (en) | 2015-12-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104375806B (en) | A kind of parallel computation component, method and corresponding parallel software development method and system | |
| Bellm et al. | Herwig 7.0/Herwig++ 3.0 release note | |
| Neic et al. | Automating image-based mesh generation and manipulation tasks in cardiac modeling workflows using meshtool | |
| Cheung et al. | NeuroFlow: a general purpose spiking neural network simulation platform using customizable processors | |
| Ruiz-Montiel et al. | Design with shape grammars and reinforcement learning | |
| CN105531688B (en) | The service of resource as other services is provided | |
| CN102831139A (en) | Co-range partition for query plan optimization and data-parallel programming model | |
| EP3163436A1 (en) | Visual software modeling method based on software meta-view for constructing software view | |
| CN101710286A (en) | Parallel programming model system of DAG oriented data driving type application and realization method | |
| Hüning et al. | Modeling & simulation as a service with the massive multi-agent system MARS | |
| CN111915011B (en) | Single-amplitude quantum computing simulation method | |
| Protalinskii et al. | Analysis and modelling of complex engineering systems based on the component approach | |
| CN111708641A (en) | Memory management method, device and equipment and computer readable storage medium | |
| CN102780583B (en) | Method for service description, service combination and service quality assessment of Internet of Things | |
| CN112100564A (en) | Master-slave game robust energy management method for community multi-microgrid system | |
| CN110457817A (en) | The method and operating platform of architectural design scheme are automatically generated based on computer logic algorithm | |
| Soncini-Sessa et al. | A DSS for planning and managing water reservoir systems | |
| CN109491571A (en) | A kind of association of configuration item and its associated diagram methods of exhibiting and system | |
| Sokolinsky | BSF-skeleton: A template for parallelization of iterative numerical algorithms on cluster computing systems | |
| CN108764576A (en) | A kind of equipment support task multiple target based on resource capability describes method | |
| Mansutti et al. | Towards distributed bigraphical reactive systems | |
| EP3163437A1 (en) | System element view-based visual modelling method for constructing system view | |
| CN101599012A (en) | A kind of graphical development method that is used for grid computing | |
| CN114756216A (en) | High-expandability integrated modeling simulation method | |
| CN110377934A (en) | The design and calculation method and device of electronic product peculiar to vessel |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant |