CN110059378A - A kind of automated manufacturing system Petri network state generation method based on GPU parallel computation - Google Patents
A kind of automated manufacturing system Petri network state generation method based on GPU parallel computation Download PDFInfo
- Publication number
- CN110059378A CN110059378A CN201910261540.8A CN201910261540A CN110059378A CN 110059378 A CN110059378 A CN 110059378A CN 201910261540 A CN201910261540 A CN 201910261540A CN 110059378 A CN110059378 A CN 110059378A
- Authority
- CN
- China
- Prior art keywords
- state
- petri network
- transition
- reachable
- petri
- 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
- 238000000034 method Methods 0.000 title claims abstract description 49
- 238000004519 manufacturing process Methods 0.000 title claims abstract description 15
- 230000007704 transition Effects 0.000 claims abstract description 35
- 238000010276 construction Methods 0.000 claims abstract description 5
- 239000012141 concentrate Substances 0.000 claims abstract description 4
- 239000011159 matrix material Substances 0.000 claims description 10
- 230000008569 process Effects 0.000 claims description 5
- 238000005457 optimization Methods 0.000 abstract description 8
- 238000004458 analytical method Methods 0.000 abstract description 4
- 238000004880 explosion Methods 0.000 abstract description 4
- 230000000694 effects Effects 0.000 abstract description 3
- 238000004422 calculation algorithm Methods 0.000 description 15
- 230000006870 function Effects 0.000 description 7
- 238000013459 approach Methods 0.000 description 5
- 230000001133 acceleration Effects 0.000 description 3
- 230000008901 benefit Effects 0.000 description 3
- 238000003780 insertion Methods 0.000 description 3
- 230000037431 insertion Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- HPTJABJPZMULFH-UHFFFAOYSA-N 12-[(Cyclohexylcarbamoyl)amino]dodecanoic acid Chemical compound OC(=O)CCCCCCCCCCCNC(=O)NC1CCCCC1 HPTJABJPZMULFH-UHFFFAOYSA-N 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 238000005094 computer simulation Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000000151 deposition Methods 0.000 description 1
- 238000003012 network analysis Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
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
- G06F30/22—Design optimisation, verification or simulation using Petri net models
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/02—Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Hardware Design (AREA)
- Evolutionary Computation (AREA)
- Geometry (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Factory Administration (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The automated manufacturing system Petri network state generation method based on GPU parallel computation that the invention discloses a kind of, comprising the following steps: by Petri net model to be solved, Petri network system initial state M0It is converted into input file；The table for initializing a common array or list structure, for indicating newly generated state set OPEN；The table for initializing a red-black tree construction, for indicating generated state set CLOSED；From Petri network system initial state M0Start, all reachable states of search Petri network constitute Reachability state set；Output reachable state concentrates the state number of each reachable state, state vector, specifically which transition obtains the state by transmitting by which state.Method of the invention is by utilizing GPU parallel computation and optimization data structure, the speed for calculating Petri network reachable set can be greatlyd improve, and have the effect of becoming better and better when reachable state number constantly increases, in the analysis to large-scale Petri net model, State-explosion problem can be effectively relieved.
Description
Technical field
It is especially a kind of based on the automatic of GPU parallel computation the invention belongs to automated manufacturing system modeling and analysis field
Manufacture system Petri network state generation method.
Background technique
Reachability graph's analytic approach of Petri network is most basic, one of most common method of Petri network analysis.Usual Petri
The reachability graph of net calculates, and is successively to extend from original state, carries out serial computing in a computer later.But when Petri network
When being on a grand scale, reachable state quantity is sharply increased, and the method for serial computing needs to expend a large amount of calculating time and system money
Source significantly impacts the efficiency of Petri network modeling and analysis.
Traditional serial computing Petri network reachable state set method has following feature and defect:
(1) in program realization, the generation of reachable state uses breadth-first strategy.It is generated by breadth-first strategy
Each layer of successor marking is one of the performance bottleneck for constructing complete set of reachable markings.At state M, the enabled of each transition is judged
Situation, and emit enabled transition, generate new state.To each transition t, will judge whether to enable, and under normal conditions,
The lower only a small number of transition of one mark are enabled.
(2) lookup is newly generated identifies whether to be present in generated reachability graph G (N, M0) it is the most time-consuming place of algorithm.
For transition t each enabled under some state M in current OPEN set, calculates M'=M+ [N] (, t) and obtain several new shapes
State, judge each newly generated state M' whether be present in CLOSED set in, be one occupy the CPU time it is very long appoint
Business, especially when the reachable state quantity of Petri is very big.
For above-mentioned two problems, the advantage that can use GPU parallel computation is optimized: multithreading is on the one hand utilized,
The enabled judgement of several transition and extension new state are calculated simultaneously, greatly reduces and calculates the time；On the other hand in conjunction with GPU's
Hardware structure and feature, for the matrix operation being related to when calculating reachable set, GPU has more preferable higher computational efficiency.As it can be seen that
Have centainly using the method that GPU concurrent operation replaces traditional CPU serial computing Petri network Reachability state set as optimization method
Feasibility.
But if new ask can be exposed when the whole process for calculating Petri network Reachability state set carries out all in GPU
Topic: when due to removal repeat mode, need entirely to have generated the participation of reachability graph, OPEN, CLOSED set will copy to aobvious
In depositing, this makes the communication between CPU memory and GPU video memory that can generate very big time and overhead, while video memory itself
Size, which also restricts algorithm, can calculate the scale of Petri network.
Summary of the invention
It is an object of the invention to the advantage using GPU parallel computation, provide a kind of calculating speed is fast, calculated result compared with
Petri network reachable state generation method excellent, that State-explosion problem can be effectively relieved.
Realize the object of the invention technical solution are as follows: a kind of automated manufacturing system Petri network based on GPU parallel computation
State generation method, comprising the following steps:
Step 1, by Petri net model to be solved, Petri network system initial state M0It is converted into input file；
The table of step 2, one common array of initialization or list structure, for indicating newly generated state set OPEN；
The table for initializing a red-black tree construction, for indicating generated state set CLOSED；
Step 3, from Petri network system initial state M0Start, all reachable states of search Petri network are constituted up to shape
State collection；
Step 4, output reachable state concentrate the state number of each reachable state, state vector, the state specifically by which
A little states are obtained by emitting which transition.
Compared with prior art, the present invention its remarkable advantage are as follows: 1) weighed in GPU and CPU Computing Principle and feature
Pros and cons use based on GPU parallel computation, common to generate Petri network reachable state supplemented by CPU calculating, and state can be effectively relieved
Space Explosion problem；2) the characteristics of making full use of GPU parallel computation to be good at matrix operation improves computational efficiency, especially
In the case that Petri network is larger, the ability of this method analytical calculation reachable state is much higher than existing method；3) in GPU only
It carries out judgement transition to enable to calculate new state with by emitting to change, greatly inhibits the communication overhead between GPU and CPU, guarantor
The efficiency of algorithm is demonstrate,proved；4) traditional data structure is replaced using the data structure of red black tree, so that this link of duplicate removal and original
Method, which is compared, to be had significant optimization and is promoted；5) time complexity of the method for the present invention is by original O (n2) it is reduced to O
(lgn)。
Detailed description of the invention
Fig. 1 is that the present invention is based on the flow charts of the automated manufacturing system Petri network state generation method of GPU parallel computation.
Fig. 2 is that host side of the present invention exchanges data with equipment end and executes program sequential schematic.
Fig. 3 is the corresponding Petri net model figure of plant working system of the embodiment of the present invention.
Specific embodiment
In conjunction with Fig. 1, a kind of automated manufacturing system Petri network state generation method based on GPU parallel computation of the present invention, packet
Include following steps:
Step 1, by Petri net model to be solved, Petri network system initial state M0It is converted into input file；
The table of step 2, one common array of initialization or list structure, for indicating newly generated state set OPEN；
The table for initializing a red-black tree construction, for indicating generated state set CLOSED；
Step 3, from Petri network system initial state M0Start, all reachable states of search Petri network are constituted up to shape
State collection；
Step 4, output reachable state concentrate the state number of each reachable state, state vector, the state specifically by which
A little states are obtained by emitting which transition.
Further, in step 1 input file format are as follows:
The customized incidence matrix for being used for Petri net model evolution, original state sequence are stored in different files
In；Wherein,
(1) incidence matrix is the INTEGER MATRICES N of n row m column, is generated using following formula:
N={ Nij| library institute i is directed toward by transition j if there is an arc, then Nij=1；If there is an arc by library institute i
It is directed toward transition j, then Nij=-1；Otherwise Nij=0 }
In formula, NijFor the element that the i-th row j in matrix N is arranged, 1≤i≤n, 1≤j≤m, n are the library institute number in Petri network, m
To change number；
(2) original state sequence is an integer ordered series of numbers S={ Si| 1≤i≤n }, wherein SiIndicate original state M0In i-th
The number of a library institute Tokken.
Further, from Petri network system initial state M in step 30Start, all reachable states of search Petri network
Reachability state set is constituted, specifically:
Step 3-1, by Petri network system initial state M0OPEN set is added to gather with CLOSED；
Step 3-2, for GPU video memory is distributed and by OPEN aggregate copy to video memory；For each state in OPEN set
Each transition distribute thread；
Step 3-3, new state is concurrently extended using search strategy in GPU, obtains new extended mode collection；
Step 3-4, in conjunction with Fig. 2, the new extended mode collection that step 3-3 is obtained is removed into video memory and OPEN set is added, is released
Put video memory space；
Step 3-5, OPEN is gathered and is compared with CLOSED set, remove repeat mode, i.e., in removal OPEN set
The state in CLOSED set；Later by OPEN gather in state be added to CLOSED set in；
Step 3-6, judge whether OPEN set is sky, if it is empty, terminates search, obtains all reachable states；Otherwise it returns
Return step 3-2.
It is further preferred that the search strategy in step 3-3 specifically uses breadth-first search strategy.
It is further preferred that new state is extended in step 3-3, specifically:
Each transition of each state specifically include following procedure by a GPU thread process in OPEN set:
Step 3-3-1, judge whether some transition of a certain state M enable, if enabled, emit the transition, execute step
3-3-2；Wherein, whether the transition of judging enable, specifically: if transition t meetsThen changing t makes
Can, t indicates the front damming institute of transition t, the quantity of Tokken in M (p) library representation institute p, W (p, t) library representation institute p and transition t it
Between arc weight；
Step 3-3-2, the new state M' that state M is expanded after emitting the transition, formula used are calculated are as follows:
M'=M+ [N] (, t)
In formula, and [N] (, t) and it is the corresponding incidence matrix column vector of transition t.
New state is extended specifically such as algorithm 1:
It is further preferred that repeat mode is removed in step 3-5, specifically:
For each state M in OPEN set, the CLOSED set with red-black tree representation is begun looking for from root node, if looking for
To identical state M, then the state is deleted from OPEN set；If not finding, which is inserted into red black tree.
Red black tree searches node and is inserted into the process of node respectively such as algorithm 2 and algorithm 3:
Function SearchRBT is searched in red black tree first since the root node of tree, if node key value, which is equal to, to search
Node, then return to the pointer of this node.If node key value is less than the node to be searched, continued to search into left subtree,
Otherwise enter right subtree.Until finding the node or reaching the leaf node of red black tree.
It is inserted into the position that function InsertRBT finds insertion in red black tree first in algorithm 3, corresponds to 1 in pseudocode
~7 rows, this is similar with SearchRBT function execution step.After finding corresponding insertion position, insertion operation is executed, and assign
This Node color attribute, 8~16 rows corresponding to pseudocode.Red black tree is possible to no longer to meet after being inserted into new node it originally
The structural property of body needs to be adjusted tree at this time, corresponds to InsertRBT_Fixup function.Due to InsertRBT_
Fixup function is the Tuning function of red-black tree construction, and CL Compare Logic is complex, and this will not be detailed here.
Algorithm 4 is the pseudocode of the automated manufacturing system Petri network reachable set generating algorithm based on GPU, main to include life
At descendant node and removal two steps of duplicate node.For simplicity, host end memory is not provided in algorithm to video memory
The process that duplication and video memory are replicated to host end memory.
Algorithm 4 generates Petri network reachable set using the method for breadth-first.When beginning, OPEN set is gathered with CLOSED
In only original state M0.In the case where OPEN set is not empty, the kernel function GPU- in algorithm 1 is utilized
Successor marking, is then copied in CPU memory by extension successor marking ExpandLayer parallel in GPU, and utilizes calculation
The serial algorithm optimized in method 2,3 completes duplicate checking work.Algorithm 4 also contemplates memory space needed for mark in OPEN set and is greater than
The case where video memory VRAM.When video memory capacity reaches the upper limit, just using the method for extension OPEN set successor marking in batches.With
In upper algorithm, CLOSED set realizes that the data structure of other data is realized using common array or chained list using red black tree.
Below with reference to embodiment, the present invention is described in further detail.
Embodiment
Petri network system model that the present embodiment uses as shown in figure 3, wherein for library institute's quantity for 31, transition quantity is 24,
The initial marking M of system0=2,0,0,0,0,0,0,2,0,0,0,0,0,0,2,0,0,0,0,0,0,2,0,0,0,0,0,0,1,
1,1}.M is identified by changing0Middle library institute Tokken number generates the reachable sets of different scales, to verify the performance of the method for the present invention.
(the P under different initial markings1,P2,P3,P4), the serial approach after being utilized respectively INA software, optimization and the side based on GPU
Three kinds of methods of method (being the method for the present invention in table 1) calculate Reachability state set, and the results are shown in Table 1.
Testing used computer CPU is Intel i5 6500, possesses the clock frequency and 4 processing cores of 3.2GHz
The heart.CPU memory size is 8GB, mechanical hard disk capacity 1TB.GPU uses NVIDIA GTX1050TI type video card, possesses 1.39GHz
Clock frequency and 4GB video memory.Operating system used by testing is window10, and programmed environment is Microsoft Visual
Studio2015 and CUDA 9.0.All experiments are overflowed in CPU memory or duration is more than to terminate for 1 day.
The comparison of 1 Reachability state set formation speed of table
In table 1, o.o.t. indicates that software runing time is more than one day.In the first set of experiments, reachable state quantity is
157360, the INA softwares calculating used time is about 8 times of the method for the present invention used time.When reachable state quantity reaches 6374984, this
Inventive method solving speed is 133 times of INA method.As can be seen from the above table, software I NA phase is analyzed with traditional Petri network
Than whether when Reachability state set quantity is smaller or substantial amounts, serial approach and method of the invention after optimization are all
It has a clear superiority.
The following table 2 further illustrates the performance comparison of the serial approach after the method for the present invention and optimization.
Table 2 uses the acceleration effect after the method for the present invention
In table 2, T1Serial approach after indicating optimization calculates time used in Reachability state set, T2Indicate side of the invention
Method calculates the time used in Reachability state set.When reachable state number is less, the acceleration efficiency of the method for the present invention 20% or so,
With the growth of reachable state quantity, the method for the present invention acceleration efficiency is stablized 30% or so.
Above data shows that method of the invention, can be greatly by utilizing GPU parallel computation and optimization data structure
The speed for calculating Petri network reachable set is improved, and has the effect of becoming better and better when reachable state number constantly increases, to big
When the analysis of type Petri net model, State-explosion problem can be effectively relieved.
Claims (6)
1. a kind of automated manufacturing system Petri network state generation method based on GPU parallel computation, which is characterized in that including with
Lower step:
Step 1, by Petri net model to be solved, Petri network system initial state M0It is converted into input file；
The table of step 2, one common array of initialization or list structure, for indicating newly generated state set OPEN；Initially
The table for changing a red-black tree construction, for indicating generated state set CLOSED；
Step 3, from Petri network system initial state M0Start, all reachable states of search Petri network constitute Reachability state set；
Step 4, output reachable state concentrate the state number of each reachable state, state vector, the state specifically by which shape
State is obtained by emitting which transition.
2. the automated manufacturing system Petri network state generation method according to claim 1 based on GPU parallel computation,
It is characterized in that, the format of input file described in step 1 are as follows:
The customized incidence matrix for being used for Petri net model evolution, original state sequence are stored in different files；Its
In,
(1) incidence matrix is the INTEGER MATRICES N of n row m column, is generated using following formula:
N={ Nij| library institute i is directed toward by transition j if there is an arc, then Nij=1；It is directed toward and is become by library institute i if there is an arc
J is moved, then Nij=-1；Otherwise Nij=0 }
In formula, NijFor the element that the i-th row j in matrix N is arranged, 1≤i≤n, 1≤j≤m, n are the library institute number in Petri network, and m is to become
Move number；
(2) original state sequence is an integer ordered series of numbers S={ Si| 1≤i≤n }, wherein SiIndicate original state M0In i-th of library
The number of institute's Tokken.
3. the automated manufacturing system Petri network state generation method according to claim 1 or 2 based on GPU parallel computation,
It is characterized in that, from Petri network system initial state M described in step 30Start, all reachable state structures of search Petri network
At Reachability state set, specifically:
Step 3-1, by Petri network system initial state M0OPEN set is added to gather with CLOSED；
Step 3-2, for GPU video memory is distributed and by OPEN aggregate copy to video memory；For each of each state in OPEN set
Transition distribution thread；
Step 3-3, new state is concurrently extended using search strategy in GPU, obtains new extended mode collection；
Step 3-4, the new extended mode collection that step 3-3 is obtained is removed into video memory and OPEN set is added, discharge video memory space；
Step 3-5, OPEN is gathered and is compared with CLOSED set, remove repeat mode, i.e., in removal OPEN set
State in CLOSED set；Later by OPEN gather in state be added to CLOSED set in；
Step 3-6, judge whether OPEN set is sky, if it is empty, terminates search, obtains all reachable states；Otherwise return to step
Rapid 3-2.
4. the automated manufacturing system Petri network state generation method according to claim 3 based on GPU parallel computation, step
Rapid 3-3 described search strategy specifically uses breadth-first search strategy.
5. the automated manufacturing system Petri network state generation method according to claim 4 based on GPU parallel computation, step
New state is extended described in rapid 3-3, specifically:
Each transition of each state specifically include following procedure by a GPU thread process in OPEN set:
Step 3-3-1, judge whether some transition of a certain state M enable, if enabled, emit the transition, execute step 3-3-
2；Wherein, whether the transition of judging enable, specifically: if transition t meetsM (p) >=W (p, t) then changes t and enables, t
Indicate the front damming institute of transition t, the quantity of Tokken in M (p) library representation institute p, arc between W (p, t) library representation institute p and transition t
Weight；
Step 3-3-2, the new state M' that state M is expanded after emitting the transition, formula used are calculated are as follows:
M'=M+ [N] (, t)
In formula, and [N] (, t) and it is the corresponding incidence matrix column vector of transition t.
6. the automated manufacturing system Petri network state generation method according to claim 5 based on GPU parallel computation,
It is characterized in that, removes repeat mode described in step 3-5, specifically:
For each state M in OPEN set, the CLOSED set with red-black tree representation is begun looking for from root node, if finding phase
Same state M then deletes the state from OPEN set；If not finding, which is inserted into red black tree.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910261540.8A CN110059378B (en) | 2019-04-02 | 2019-04-02 | Automatic manufacturing system Petri network state generation method based on GPU parallel computing |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910261540.8A CN110059378B (en) | 2019-04-02 | 2019-04-02 | Automatic manufacturing system Petri network state generation method based on GPU parallel computing |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110059378A true CN110059378A (en) | 2019-07-26 |
| CN110059378B CN110059378B (en) | 2022-09-06 |
Family
ID=67318204
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910261540.8A Expired - Fee Related CN110059378B (en) | 2019-04-02 | 2019-04-02 | Automatic manufacturing system Petri network state generation method based on GPU parallel computing |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110059378B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117725348A (en) * | 2024-02-07 | 2024-03-19 | 蓝象智联(杭州)科技有限公司 | Thread management method and system in GPU computing large-scale array summation process |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080320437A1 (en) * | 2007-06-20 | 2008-12-25 | Microsoft Corporation | Constructing Petri Nets from traces for diagnostics |
| CN105652833A (en) * | 2015-12-30 | 2016-06-08 | 南京理工大学 | Bi-directional intelligent search-based manufacturing enterprise shop scheduling optimization method |
-
2019
- 2019-04-02 CN CN201910261540.8A patent/CN110059378B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080320437A1 (en) * | 2007-06-20 | 2008-12-25 | Microsoft Corporation | Constructing Petri Nets from traces for diagnostics |
| CN105652833A (en) * | 2015-12-30 | 2016-06-08 | 南京理工大学 | Bi-directional intelligent search-based manufacturing enterprise shop scheduling optimization method |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117725348A (en) * | 2024-02-07 | 2024-03-19 | 蓝象智联(杭州)科技有限公司 | Thread management method and system in GPU computing large-scale array summation process |
| CN117725348B (en) * | 2024-02-07 | 2024-05-10 | 蓝象智联(杭州)科技有限公司 | Thread management method and system in GPU computing large-scale array summation process |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110059378B (en) | 2022-09-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11630864B2 (en) | Vectorized queues for shortest-path graph searches | |
| CN110659284A (en) | Block sequencing method and system based on tree graph structure and data processing terminal | |
| CN111309976B (en) | GraphX data caching method for convergence graph application | |
| CN110135584A (en) | Extensive Symbolic Regression method and system based on self-adaptive parallel genetic algorithm | |
| CN110187965A (en) | The running optimizatin and data processing method of neural network, equipment and storage medium | |
| CN111915011B (en) | Single-amplitude quantum computing simulation method | |
| CN109471732A (en) | A kind of data distributing method towards CPU-FPGA heterogeneous multi-core system | |
| CN112149808A (en) | Method, system and medium for expanding stand-alone graph neural network training to distributed training | |
| CN105468439A (en) | Adaptive parallel algorithm for traversing neighbors in fixed radius under CPU-GPU (Central Processing Unit-Graphic Processing Unit) heterogeneous framework | |
| CN113313247B (en) | Operation method of sparse neural network based on data flow architecture | |
| CN105556474A (en) | Managing memory and storage space for a data operation | |
| CN111429974B (en) | Molecular dynamics simulation short-range force parallel optimization method on super computer platform | |
| US11222070B2 (en) | Vectorized hash tables | |
| CN105956605B (en) | Three-dimensional structure Similarity-Based Clustering Method based on parallel k-means cluster | |
| CN110059378A (en) | A kind of automated manufacturing system Petri network state generation method based on GPU parallel computation | |
| JPWO2011058657A1 (en) | Parallel computing device, parallel computing method, and parallel computing program | |
| CN104778088A (en) | Method and system for optimizing parallel I/O (input/output) by reducing inter-progress communication expense | |
| CN111832144B (en) | Full-amplitude quantum computing simulation method | |
| CN109522127B (en) | Fluid machinery simulation program heterogeneous acceleration method based on GPU | |
| CN109635473B (en) | Heuristic high-flux material simulation calculation optimization method | |
| Serpa et al. | Flexible use of temporal and spatial reasoning for fast and scalable CPU broad‐phase collision detection using KD‐Trees | |
| CN110109811A (en) | A kind of source tracing method towards GPU calculated performance problem | |
| CN106250327B (en) | One kind hot spot recognition methods and device in key-value storage | |
| CN105573834B (en) | A kind of higher-dimension vocabulary tree constructing method based on heterogeneous platform | |
| Angles et al. | R 3 MAT: A Rapid and Robust Graph Generator |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20220906 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |