CN103763171B - Construction method for interconnection network of massively parallel computing system - Google Patents
Construction method for interconnection network of massively parallel computing system Download PDFInfo
- Publication number
- CN103763171B CN103763171B CN201310743767.9A CN201310743767A CN103763171B CN 103763171 B CN103763171 B CN 103763171B CN 201310743767 A CN201310743767 A CN 201310743767A CN 103763171 B CN103763171 B CN 103763171B
- Authority
- CN
- China
- Prior art keywords
- interference networks
- nodes
- oct
- node
- octagon
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
Landscapes
- Data Exchanges In Wide-Area Networks (AREA)
- Computer And Data Communications (AREA)
Abstract
The invention discloses a construction method for an interconnection network of a massively parallel computing system. The OCT interconnection network is formed by 8*2k*2m nodes. Firstly, every eight nodes are connected to form one Octagon interconnection network, 2k*2m Octagon interconnection networks are obtained in total, and the nodes in each Octagon interconnection network are encoded through four-digit Johnson codes; secondly, the 2k*2m nodes which are the same in node code in each Octagon interconnection network are connected to form one Torus interconnection network, wherein each line of each Torus interconnection network has 2m nodes, each column of each Torus interconnection network has 2k nodes, and eight Torus interconnection networks are obtained in total; the nodes in each Torus interconnection network are encoded through (k+m)-digit Johnson codes, and namely the OCT interconnection network can be recorded as the OCT (k, m) interconnection network, wherein k and m are natural numbers and are parameters of the number of the nodes of the interconnection network. According to the construction method, expansion of the interconnection network is performed under the condition that node degree is kept unchanged; the OCT (k, m) interconnection network is a symmetrical and regular interconnection network, the Johnson coding method is adopted for node coding, codes of any adjacent nodes only have a one-digit difference, and therefore the routing algorithm is simple and efficient.
Description
Technical field
The invention belongs to parallel computing field, more particularly to a kind of massive parallel processing interference networks construction
Method.
Background technology
With the continuous development of hardware technology, the particularly development of super large-scale integration technique so that comprising thousands of
The extensive multicomputer system of processors up to ten thousand is possibly realized.Such as Milky Way -1A have 7168 calculate nodes, Fujitsu it is super
Computing system more than 80,000 calculate node.At following several years, new application promoted the number of single-chip processor core with algorithm
Amount is up to the extensive super computer system number of nodes of the foundation eighties in 20th century.We are moving towards the calculating of E levels
(Exascale Computing) epoch, it is contemplated that 2018, super computer system is up to 1exaFLOPS (1018FLOPS), enters
Enter E levels and calculate (Exascale Computing) epoch.
The epoch are calculated into E levels, multicomputer system scale is up to millions of to up to ten million processor cores, Internet
Network has important impact to the performance of such large-scale multicomputer system, will determine the meter of following large-scale parallel application
Calculate and storage performance.In order to improve the communication efficiency of parallel computation, people always research structure it is simple, node degree is little, network
Diameter is little, routing policy is simple and good extendible interference networks.
The topological structure of Torus interference networks has normality, symmetry, fault-tolerance, short diameter, Residuated Lattice etc. special
Property, thus it is deep welcome by researcher and practitioners, be that a kind of mostly important and most attraction parallel computer is mutual
Connecting network topology structure.However, in millions of interference networks to up to ten million processor cores, traditional Torus interference networks
Diameter become very large, be not suitable for the interconnection of following parallel system, simultaneously as many concurrent programs will be in one group of section
Frequently communicated (i.e. local communication) in point, therefore, it is proposed to various hierarchical interconnection networks based on Torus (HINs,
Hierarchical Interconnection Networks).In these hierarchical interconnection networks, by calculate node constitute compared with
Low layer interference networks carry out local communication, and the higher level interference networks being made up of cluster are used for telecommunication.These layering Internets
The diameter of network is all that, by the product of every layer network diameter, relative diameter is still larger.Octagon interference networks are by F.Karim et al.
For on-chip interconnection network, the topological structure of the interference networks has the properties such as normality, symmetry, short diameter.
2 dimensions Torus interference networks (referred to as T (k, m)) of 2k × 2m are a kind of interference networks with following properties：1)
It is made up of 2k × 2m node and 8k × m bar direct link；2) with the abscissa of 2m Johnson's code identification node of m positions,
With the ordinate of 2k Johnson's code identification node of k positions, using the ordinate of node as high-order abscissa as low level group
One nodes encoding of synthesis, so, any one node can be identified with the binary coding of k+m positions；3) rule of nodes encoding
It is then：There is and during an only difference coding of two nodes in and if only if T (k, m), two nodes be it is adjacent, i.e., this two
There is a direct link between individual node.
Fig. 1 gives the topological structure and nodes encoding (k=2, m=3) of T (k, m) interference networks.T (k, m) interference networks
With following good nature：1. the nodes encoding of each row and column is all binary unit Distance Cyclic Codes；2. any one is saved
Point has and only four adjacent nodes (when nodes encoding digit is more than or equal to 5,2 dimension Gray codes are unsatisfactory for this characteristic),
Naturally form Torus structures；If 3. k or m increases by one, corresponding node number only increases 4m or 4k (2 dimension lattice
It is 2k × 2m that thunder encodes the nodes to be formed, if k or m increases by one, it is original interstitial content that corresponding node number increases
2 times)；4. the minimum range number of any two nodes encoding XOR gained " 1 " being between two nodes；5. have
The routing mechanism of Hypercube simple structures.
Octagon interference networks are a kind of interference networks with following properties：1) by 8 nodes and 12 direct link
Composition；2) with the coordinate of Johnson's code identification node of 4；3) when in interference networks two nodes encodings have and only one
When each of different or two nodes encoding XOR results is " 1 ", two nodes are adjacent, i.e., between the two nodes
There is a direct link.
Fig. 2 is the topological structure and nodes encoding of Octagon interference networks.As shown in Figure 2, Octagon interference networks tool
There is following good property：1) Connected degree of arbitrary node is 3, a diameter of the 2 of whole network in network, and network has regular
The good characteristic such as property, symmetry, short diameter, low Connected degree；2) there are 3 in network between any two node without the link handed over,
If two nodes are joined directly together, the length of this 3 links is respectively Isosorbide-5-Nitrae, 4, it is otherwise 2, and 3,3, therefore, with good appearance
Mistake and concurrency；3) when any two nodes encoding XOR result is 1 " 1 " or 4 " 1 ", two nodes are adjacent, arbitrarily
When two nodes encoding XOR results are 2 " 1 " or 3 " 1 ", the distance of two nodes is 2.It has the disadvantage that network does not possess
Extensibility.
The content of the invention
The present invention proposes a kind of massive parallel processing construction method for interconnection network, and the interference networks of construction are combined
The advantage of Torus interference networks and Octagon interference networks, with short diameter, normality, symmetry and good autgmentability.
In construction method for interconnection network interconnection network nodes are encoded using Johnson's coding method so that Design of Routing Algorithm letter
It is single, separately design based on the unicast of hybrid coding, broadcast routing protoco.
The embodiment of the present invention is achieved in that a kind of massive parallel processing construction method for interconnection network, the party
Method includes：
First, an Octagon interference networks will be connected into per 8 nodes, 2k × 2m Octagon Internet will be obtained
Network, each Octagon network is referred to as a piece of；
Secondly, to per a piece of employing identical nodes encoding method, i.e., every interior joint is all compiled with 4 johnoson codes
Code；
Again, then by each piece interior joint coding identical 2k × 2m node 2m node of every row and each column 2k are connected into
The Torus interference networks of individual node, can be designated as T (k, m) interference networks, and 8 T (k, m) interference networks are obtained；
Then, identical nodes encoding method, i.e. each T (k, m) interference networks are adopted to each T (k, m) interference networks
Interior joint is all encoded with k+m positions johnoson code；
The interference networks being constructed such that are referred to as OCT interference networks, can be designated as OCT (k, m) interference networks, wherein k, and m is certainly
So number is the parameter of interconnection network nodes quantity.
Further, OCT (k, m) interference networks adopt following nodes encoding method, and each nodes encoding is by two parts (At,
Ao) composition, wherein Ao(4 johnoson codes) is each Octagon nodes coding, At(k+m positions johnoson code) is each
Nodes encoding in the coding of Octagon pieces, namely each T (k, m) network.
Further, OCT (k, m) the interconnection network nodes scale of construction can be expanded by extension T (k, m) interference networks
Exhibition, as long as coding digit is increased into one, i.e. m or k increases 1, two rows is increased by T (k, m) interference networks or two row is (corresponding
Node number be increased by 4k or 4m), i.e., increase by 8 × 4k or 8 × 4m node, formation on OCT (k, m) network foundation
OCT (k, m+1) or OCT (k+1, m) interference networks；Originally each Octagon interconnection network nodes connection in OCT (k, m) network
Relation is not changed in, and the Connected degree of node is not changed in.T (k, m+1) or T (k+1, m) in interference networks, except with newly-increased section
Outside the connected node of point, other nodes do not have any variation with annexation.
The present invention is one by the short diameter combined structure of the extensibility of Torus interference networks and Octagon interference networks
Plant simple extendible OCT (k, m) interference networks.The interference networks be a kind of node degree be 7 it is regular symmetrical extendible
Interference networks, can keep the constant extension for carrying out network size of node degree, network node coding to adopt Johnson's coding staff
Method so that routing algorithm is simply efficient.Analysis and test result indicate that, OCT (k, m) interference networks have good communication performance,
Fault-tolerant ability, extensibility, are a kind of interference networks of suitable Large-scale parallel computing.
Description of the drawings
Fig. 1 is T (k, m) interference networks provided in an embodiment of the present invention and nodes encoding (k=2, m=3)；
Fig. 2 is Octagon interference networks provided in an embodiment of the present invention and nodes encoding；
Fig. 3 is massive parallel processing construction method for interconnection network flow chart provided in an embodiment of the present invention；
Fig. 4 is OCT (k, m) interference networks (k=2, m=2) provided in an embodiment of the present invention.
Specific embodiment
In order that the purpose of the present invention, technological invention and advantage become more apparent, with reference to embodiments, to the present invention
It is further elaborated.It should be appreciated that specific embodiment described herein is not used to only to explain the present invention
Limit the present invention.
Fig. 3 proposes a kind of massively parallel computer system interference networks construction side that case study on implementation of the present invention is provided
Method, the method includes：
In step S101, first, one is connected into according to above-mentioned Octagon interference networks describing mode per 8 nodes
Octagon interference networks, are obtained 2k × 2m Octagon interference networks, and each Octagon interference networks is referred to as a piece of；
In step s 102,2k × 2m piece is connected into into as follows Torus networks：2k × 2m piece interior joint
Encoded with 4 johnoson codes, each piece interior joint coding identical node is described according to above-mentioned T (k, m) interference networks
Mode connect into T (k, m) interference networks；
In step s 103, the nodes encoding of OCT (k, m) interference networks：OCT (k, m) adopts following coding method, each
Nodes encoding is by two parts (At, Ao) composition, wherein Ao(4 johnoson codes) is that the node in each Octagon interference networks is compiled
Code, At(k+m positions johnoson code) is the coding of each Octagon piece, namely the nodes encoding in each T (k, m) network；
In step S104, OCT (k, m) interconnection network topological structure is connected into.
Fig. 4 is OCT (k, m) interference networks (k=2, m=2) provided in an embodiment of the present invention, and solid line is represented in figure
Octagon interconnected network links, dotted line represents T (k, m) interconnected network links, and small circle represents network node, marks up and down
There is direct link to connect between number identical end points.OCT (k, m) interconnection network nodes scale can by extend T (k, m) net
Network and extend, as long as will coding digit increase by one, i.e. m or k increase 1, be increased by T (k, m) network two rows or two row (phases
The node number answered is increased by 4k or 4m), i.e., increase by 8 × 4k or 8 × 4m node on OCT (k, m) network foundation, formed
OCT (k, m+1) or OCT (k+1, m) interference networks.Originally each Octagon piece network connection relation does not have in OCT (k, m) network
Change, the Connected degree of node is not changed in.In T (k, m+1) or T (k+1, m) in interference networks, except being connected with newly-increased node
Node outside, other nodes and annexation do not have any variation.
The property of OCT (k, the m) interference networks of construction：
Property 1.OCT (k, m) interference networks are regular interference networks, and the Connected degree of arbitrary node is 7.
Because each Octagon piece is regular interference networks and node connectivity is 3, according to the structure of OCT (k, m) network
The process of making is apparent from, and Octagon pieces are regarded as a node, and the network is exactly T (k, m) interference networks and node connectivity is 4, institute
So that OCT (k, m) network is regular Network, and node connectivity is 3+4=7.
In property 2.OCT (k, m) interference networks between any two node apart from maximum (network diameter) be k+m+2.
Due to the diameter and the diameter sum of 2k node cycle of a diameter of 2m node cycle of T (k, m) interference networks, as
M+k, is apparent from according to the construction process of OCT (k, m) network, and T (k, m) interference networks are regarded as a node, and the network is exactly
Octagon interference networks, its a diameter of 2.So, OCT (k, m) network diameter is Torus diameters and Octagon diameter sums,
That is m+k+2.
Property 3.OCT (k, m) interference networks are symmetrical networks.
It is apparent from according to the construction process of OCT (k, m) interference networks, any node identification is all isomorphic to for origin in the network
Itself, i.e., observe what network was just as from any node.Simplify the realization of routing algorithm, i.e. routing algorithm and node location
It is irrelevant.
The number of links of property 4.OCT (k, m) interference networks is 112 × k × m.
The number of links of OCT (k, m) interference networks is the Torus chains of 2k × 2m Octagon number of links 12 and 8 2k × 2m
2 × 2k of way × 2m sums, 2k × 2m × 12+8 × 2 × 2k × 2m=112 × k × m.
The partial width that waits of property 5.OCT (k, m) interference networks is 24 × k × m.
When the partial widths such as network are that network is divided into two equal networks, it is necessary to the minimal communications number of links left out.OCT
(k, m) interference networks decile be by the partial widths such as 2k × 2m Octagon interference networks decile, Octagon be 6, so decile
Width is 6 × 2k × 2m=24 × k × m.
In order to further illustrate the good characteristic of OCT (k, m) interference networks, table 1 give OCT (k, m) interference networks,
(2k, 2m, 3)-OMMH (also referred to as Torus Embedded Hypercube) interference networks and 2 dimension Torus interference networks it is right
Than wherein N=8 × 2k × 2m=4k × 8m.
The performance characteristic of 1 three kinds of static networks of table
Johnson's coding method is adopted in interference networks so that OCT (k, m) interference networks have the property that：To OCT
Any two node A (Am+k+3 ... Am+k, Am+k-1 ... AmAm-1 ... A1A0), B (Bm+k+3 ... Bm+ in (k, m) interference networks
K, Bm+k-1 ... BmBm-1 ... B1B0), Ai, Bi, Aj, Bj ∈ { 0,1 }, i ∈ { 4 ..., m+k+3 }, j ∈ { 0,1,2,3 } then appoint
The distance between two node A of meaning, B are
Because each nodes encoding is Johnson's coding in each nodes encoding in Octagon pieces and T (k, m), by OCT
(k, m) interference networks construction process understands, the distance between any two node be the distance of two nodes in T (k, m) with
Two nodes are in Octagon apart from sum.There is and only one difference the distance between two nodes coding in Octagon pieces
Or each of two nodes encoding XOR results is adjacent node when being " 1 ", i.e., the distance between any two node is
The two nodes encodings difference number of position or the number of two nodes encoding identical bits add 1；Two nodes encodings have and only in T (k, m)
It is adjacent node when having a difference, i.e., the distance between any two node is the number of two nodes encodings difference position.The property
Facilitate the design of routing algorithm.
The Unicast Routing Algorithm of OCT (k, m) interference networks and performance evaluation：
1) Unicast Routing Algorithm
Assume A (Am+k+3...Am+k, Am+k-1...AmAm-1...A1A0) node is to B (Bm+k+3...Bm+k, Bm+k- 1...BmBm-1...B1B0) node sends data, the Hamming distance of A nodes and B node (D)Wherein "" represent A and B and carry out step-by-step XOR, " Hamming " function
Represent the number sum operation " 1 " after A and B XORs.Coding method and OCT (k, m) by Octagon and T (k, m) node
The construction process of interference networks understands that the Shortest path routing process of OCT (k, m) is as follows：Routing procedure is as follows：
If 1. A and B is in same Octagon pieces, then compiled from the T (k, m) of 2.2 sections, A nodes and B node
Code is identical, i.e.,As long as entering
Row Ao=A3...A0, Bo=B3...B0Octagon pieces in route, the distance of source node A and destination node B is 1 or 2.Such as
ReallyOr 4, then node 4 directly sends data to node B；Otherwise, A is calculatedo
Three adjacent nodes be Ao1, Ao2, Ao3With BoDistance, by node AoData is activation to destination node distance for 1 phase
Neighbors, then from adjacent node is sent to destination node B again.
If 2. A and B is in same T (k, m), then saved from 2.2, the Octagon's of A nodes and B node
Nodes encoding is identical, i.e.,As long as carrying out At=Am+k+ 3...AmAm-1...A5A4Node is to Bt=Bm+k+3...BmBm-1...B5B4T (k, m) network route.T (k, m) network
Left and right adjacent node coding only has abscissa to differ one, and neighbouring node only has ordinate to differ one, node At
Left adjacent node be encoded toRight adjacent node is encoded toUpper adjacent node is encoded to Lower adjacent node is encoded to
So AtAdjacent node and BtDistance be 。Hmin=min { Hl, Hr, Hu, Hd, bag is sent out
It is sent to HminCorresponding adjacent node and by AtThe coding of the adjacent node is revised as, is then calculated Value, if H ≡ 0, then AtThe node of mark is destination node, otherwise repeats mistake above
Journey.
It is arbitrary also not in same T (k, m) network if 3. A and B are neither in same Octagon networks
Two nodes, then packet is routed to into node A ' (A in same Octagon networks first in the way of 1.m+k- 1...AmAm-1...B3B2B1B0), A ' and B in same T (k, m), then in the way of 2. by packet in T (k, m) Road
By to destination node B.
2) algorithm performance analysis
The routing algorithm major advantage of OCT (k, m) is that T (k, m) network adopts Johnson's coding method so that T (k, m)
The number of any two nodes encoding XOR gained " 1 " is the minimum range between two nodes in network, and the coding is hidden
The relation between the routing iinformation and adjacent node of the overall situation is contained.So that during target node data, as long as storage present node
Just can correct route data with the coding of destination node.Octagon networks are also adopted by Johnson's coding method, any two
When nodes encoding XOR result is 1 " 1 " or 4 " 1 ", two nodes are adjacent, and any two nodes encoding XOR result is 2
When individual " 1 " or 3 " 1 ", the distance of two nodes is 2 so that network routing variable is obtained simply.
According to OCT (k, m) Unicast Routing Algorithm, data are propagated in Octagon needs 2 wheels communication behaviour under worst case
Make, propagating in same T (k, m) needs k+m to take turns traffic operation under worst case, therefore, need k+m under worst case altogether
+ 2 wheel traffic operations.Data can be sent to destination node, the communication efficiency of algorithm by algorithm along shorter path from source node
It is higher.The data that forward each time of above Unicast Routing Algorithm are carried out by shortest path, so, under worst case, road
By path not over network diameter k+m+2, the communication efficiency of algorithm is 1/ (k+m+2).
The broadcast routing protoco of OCT (k, m) interference networks and performance evaluation：
1) broadcast routing protoco
Hypothesis sends data to other all nodes by node A.First, by the data of node A in the Octagon pieces being located
All nodes of place piece are inside broadcast to, the node for then receiving data message spreads data to each using recurrence method for doubling
All nodes within the T (k, m) being located.
2) algorithm performance analysis
Broadcast is carried out by this way, and node A needs data broadcasting to all nodes in the Octagon pieces of place
2 wheel traffic operations, then data carry out broadcast within T (k, m) needs m+k to take turns traffic operation.So, whole broadcast needs k+
M+2 takes turns traffic operation, and the communication efficiency of algorithm is 1/ (k+m+2).
The present invention combines the extensibility of the short diameter of Octagon interference networks and Torus interference networks, constructs
OCT (k, m) interference networks：First, an Octagon network is connected into according to the mode for defining 6 per 8 nodes, 2k is obtained
× 2m Octagon network, each Octagon network is referred to as a piece of；2k × 2m piece is connected into into as follows Torus
Network：2k × 2m piece is encoded with 4 johnoson codes, by each piece interior joint coding identical node according to definition 5
Mode connects into T (k, m) interference networks；The coding of OCT (k, m)：OCT (k, m) adopts following coding method, each nodes encoding
By two parts (At, Ao) composition, wherein Ao(4 johnoson codes) is the nodes encoding of each Octagon network, At(k+m positions are about
The inferior code of writing brush) be each Octagon piece coding, namely the nodes encoding in each T (k, m) network.Devise again based on OCT
The unicast of (k, m) interconnection network architecture, broadcast routing protoco
Presently preferred embodiments of the present invention is the foregoing is only, not to limit the present invention, all essences in the present invention
Any modification, equivalent and improvement made within god and principle etc., should be included within the scope of the present invention.
Claims (3)
1. a kind of massive parallel processing construction method for interconnection network, it is characterised in that the interference networks of construction are referred to as
Octagon connects Torus interference networks, and Octagon connection Torus interference networks are referred to as OCT interference networks, OCT Internets
Network is made up of 8 × 2k × 2m node, and the building method includes：First, an Octagon interconnection will be connected into per 8 nodes
Network, is obtained 2k × 2m Octagon interference networks, and each Octagon interference networks interior joint is carried out with 4 johnoson codes
Coding；Secondly, then 2k × 2m Octagon interconnection network nodes are encoded into identical node and connects into 2m node of every row and often
The Torus interference networks of 2k node of row, can be designated as T (k, m) interference networks, and 8 T (k, m) interference networks, each T is obtained
(k, m) interference networks interior joint is encoded with k+m positions johnoson code, i.e., the OCT interference networks can be designated as OCT (k, m) interconnection
Network, wherein k, m is parameter that natural number is interconnection network nodes quantity.
2. massive parallel processing construction method for interconnection network as claimed in claim 1, it is characterised in that：OCT (k, m)
Each nodes encoding of interference networks is made up of two parts (At, Ao), and wherein Ao is compiled for each Octagon interference networks interior nodes
Code, Ao is 4 johnoson codes；At is each Octagon interference networks coding of itself, namely in each T (k, m) interference networks
Nodes encoding, At is k+m positions johnoson code.
3. massive parallel processing construction method for interconnection network as claimed in claim 1, it is characterised in that：Connect into
OCT (k, m) interconnection network topological structure, OCT (k, m) interconnection network nodes scale can be expanded by extension T (k, m) network
Exhibition, as long as coding digit is increased into one, i.e. m or k increases 1, is increased by two rows or two row, corresponding section in T (k, m) network
Point number is increased by 4k or 4m, i.e., increase by 8 × 4k or 8 × 4m node on OCT (k, m) network foundation, forms OCT (k, m
+ 1) or OCT (k+1, m) interference networks；In OCT (k, m+1) or OCT, (k+1, m) original each Octagon is mutual in interference networks
Even network node annexation is not changed in, each T (k, m+1) or T (k+1, m) in interference networks, except with newly-increased node phase
Outside node even, the annexation of other nodes does not have any variation.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310743767.9A CN103763171B (en) | 2013-12-31 | 2013-12-31 | Construction method for interconnection network of massively parallel computing system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310743767.9A CN103763171B (en) | 2013-12-31 | 2013-12-31 | Construction method for interconnection network of massively parallel computing system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103763171A CN103763171A (en) | 2014-04-30 |
| CN103763171B true CN103763171B (en) | 2017-05-10 |
Family
ID=50530334
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310743767.9A Expired - Fee Related CN103763171B (en) | 2013-12-31 | 2013-12-31 | Construction method for interconnection network of massively parallel computing system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103763171B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109561034B (en) * | 2018-12-25 | 2021-06-04 | 中科曙光信息产业成都有限公司 | Three-dimensional network topology structure and routing algorithm thereof |
| CN112039786B (en) * | 2019-06-04 | 2021-11-19 | 清华大学 | Torus network-based broadcasting method |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101420355A (en) * | 2008-11-28 | 2009-04-29 | 西安邮电学院 | Star-cluster double-loop on chip network topology construction |
| CN102780936A (en) * | 2012-07-17 | 2012-11-14 | 西安电子科技大学 | Optical on-chip network system of non-blocking communication and communication method thereof |
| CN103297305A (en) * | 2013-06-20 | 2013-09-11 | 西安邮电大学 | Network-on-chip system for layered MCC |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8407660B2 (en) * | 2007-09-12 | 2013-03-26 | Neal Solomon | Interconnect architecture in three dimensional network on a chip |
-
2013
- 2013-12-31 CN CN201310743767.9A patent/CN103763171B/en not_active Expired - Fee Related
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101420355A (en) * | 2008-11-28 | 2009-04-29 | 西安邮电学院 | Star-cluster double-loop on chip network topology construction |
| CN102780936A (en) * | 2012-07-17 | 2012-11-14 | 西安电子科技大学 | Optical on-chip network system of non-blocking communication and communication method thereof |
| CN103297305A (en) * | 2013-06-20 | 2013-09-11 | 西安邮电大学 | Network-on-chip system for layered MCC |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103763171A (en) | 2014-04-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102804712B (en) | Method and apparatus for selecting between multiple equal cost paths | |
| Zahavi | D-Mod-K routing providing non-blocking traffic for shift permutations on real life fat trees | |
| Huang et al. | On the fault-tolerant hamiltonicity of faulty crossed cubes | |
| CN101420355B (en) | Star-cluster double-loop on chip network topology construction | |
| CN103399902A (en) | Generation and search method for reachability chain list of directed graph in parallel environment | |
| CN101414952A (en) | Expandable parallel calculation interconnection network topological structure based on hypercube | |
| CN106779225A (en) | A kind of optimal path method for solving comprising Dominator collection | |
| Feder et al. | A local switch Markov chain on given degree graphs with application in connectivity of peer-to-peer networks | |
| CN105450741A (en) | Supply server selection and restoration data transmission method in distributed storage system | |
| Fu | Fault-free Hamiltonian cycles in twisted cubes with conditional link faults | |
| Hung | Constructing two edge-disjoint Hamiltonian cycles and two-equal path cover in augmented cubes | |
| CN103763171B (en) | Construction method for interconnection network of massively parallel computing system | |
| Hsieh | Some edge‐fault‐tolerant properties of the folded hypercube | |
| CN109547160A (en) | A kind of cyclic shift network code construction method | |
| Maniscalco et al. | Binary and m-ary encoding in applications of tree-based genetic algorithms for QoS routing | |
| CN103428087B (en) | The longest path preset protection P using depth of round first traversal encloses generation method | |
| CN117408587A (en) | Method for mining maximum propagation point set of large-scale express logistics network | |
| Yang et al. | Fault-tolerant pancyclicity of the Mobius cubes | |
| Liu et al. | Wl-align: Weisfeiler-lehman relabeling for aligning users across networks via regularized representation learning | |
| El-Mesady et al. | [Retracted] Decompositions of Circulant‐Balanced Complete Multipartite Graphs Based on a Novel Labelling Approach | |
| Hung et al. | Embedding two edge-disjoint Hamiltonian cycles and two equal node-disjoint cycles into twisted cubes | |
| CN107453997A (en) | A kind of optimization method for routing based on double costs | |
| Jiang et al. | Toward Lower Repair Bandwidth of Piggybacking Codes via Jointly Design for Both Data and Parity Nodes | |
| CN107241222A (en) | A kind of network topology calibration method | |
| Das et al. | Stirling networks: a versatile combinatorial topology for multiprocessor systems |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20170510 Termination date: 20171231 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |