CN105871730B - Network-on-chip router based on network code - Google Patents
Network-on-chip router based on network code Download PDFInfo
- Publication number
- CN105871730B CN105871730B CN201610169324.7A CN201610169324A CN105871730B CN 105871730 B CN105871730 B CN 105871730B CN 201610169324 A CN201610169324 A CN 201610169324A CN 105871730 B CN105871730 B CN 105871730B
- Authority
- CN
- China
- Prior art keywords
- network
- router
- input
- channel
- coding
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L45/00—Routing or path finding of packets in data switching networks
- H04L45/60—Router architectures
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L1/00—Arrangements for detecting or preventing errors in the information received
- H04L1/004—Arrangements for detecting or preventing errors in the information received by using forward error control
- H04L1/0076—Distributed coding, e.g. network coding, involving channel coding
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L43/00—Arrangements for monitoring or testing data switching networks
- H04L43/50—Testing arrangements
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L49/00—Packet switching elements
- H04L49/10—Packet switching elements characterised by the switching fabric construction
- H04L49/109—Integrated on microchip, e.g. switch-on-chip
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The invention discloses a kind of novel compacts based on network code, efficient, quick network-on-chip router, the router includes the input of the road P, output channel, each input channel is equipped with v virtual channel again, it is designed with a routing logic and virtual channel distributor per input channel all the way, is each responsible for the route direction and distribution virtual channel of data transmission;Virtual channel distributor is also connected with switch divider, the switch divider is responsible for the road P input channel distribution crossbar switch, it realizes multiple input channels while requesting the president of the same output channel, network code module NCU is additionally provided between the road P input channel and crossbar switch, the microplate output of virtual channel is by simple coding selection circuit, NCU performs the encoding operation qualified input, by the microplate of coding after the arbitration of switch divider, is sent to crossbar switch.The transmission mode as unit of microplate is supported in invention, realizes real-time coding, and additional hardware and power consumption consumption is effectively reduced.
Description
Technical field
The present invention relates to the research field of network router, in particular to a kind of network-on-chip routing based on network code
Device.
Background technique
With the sustainable development of semiconductor technology, the quantity that one chip accommodates transistor is continuously increased, and integrated circuit steps
Multicore (multiple-core) or many-core (many-core) epoch are entered.Core relies on network-on-chip (Network-on-
Chip or NoC) it links together, the overall performance of chip is improved by concurrent operation.Currently, many-core chip has been carried out
A large amount of business application.Such as: cloud computer single-chip cloud computer " the single-chip cloud of 48 core of Intel
The TILE-Gx72 single-chip of computer " and 72 core of Tilera.These chips be widely used in supercomputing, cloud computing,
In the systems such as big data.
The scale and degree of parallelism of many-core chip constantly increase, and the reliable interconnection and communication between core become increasingly to be stranded
Difficulty, bus and point-to-point interconnection mode are not able to satisfy the performance and expansible needs that on piece interconnects gradually, there has been proposed
Network-on-chip mode based on packet switch.The application of many-core chip is continuously increased, and the concurrent cooperation in chip between core is increasingly
Closely, the data volume transmitted between each other continues to increase, and needs support increasingly complex network behavior.In addition, complicated multiple groups are broadcast
(multi-session multicast) communication is applied also very frequently in many-core chip, such as: Cache consistency and biography
Defeated shared data etc..These factors increase the traffic of network-on-chip, cause network congestion, reduce the property of network-on-chip
Can, increase the power consumption of entire chip.People are badly in need of new network-on-chip and are designed to improve performance and energy consumption efficiency.
The it is proposed of network-on-chip concept is at 2000 or so, and the data for using for reference macroscopical computer network packet switch transmit shape
The frame mode of formula and shared physical channel supports the transmission mode with " microplate (flit) " for unit.1999, masschusetts, U.S.A
The Raw microprocessor of Polytechnics combines static and two sets of communication networks of dynamic, realizes multiple processor cores jointly
Collaboration and communication;Guerrier and Greiner was proposed in 2000 based on fat tree (fat-tree) structure and packet switch SPIN
On-chip network structure.2001, the characteristics of Dally and Towles analyze system on chip hardware resource, propose one it is general
NoC model, to demonstrate network-on-chip in the feasibility for solving SoC interconnection;The same year, Benini and Micheli propose NoC's
Stratification tectonic model.Then, researchers expand extensive research in network-on-chip field, propose a series of online
Network structure, it is more influential to have: the CHAIN and SpiNNaker of Univ Manchester UK, Italian Bologna university and beauty
The XPIPES of Stanford University, state, PHILIPS Co.The NOSTRUM of Sweden's Royal Institute of Technology, French UPMC
SPIN and the MANGO of Denmark Polytechnic University of university etc..
With the development that multi-core processor and shared data are applied, increasingly complex network behavior, such as multicast
(multicast), become the significant consideration of network-on-chip design.The data transmission of network-on-chip mainly has unicast
(unicast) transmission and multicast transmission.Multicast can be realized with multiple unicast operations, but efficiency is very low.Researchers
Numerous studies have been done in terms of multicast network-on-chip, have proposed a variety of NoC frameworks and multicast optimization algorithm for supporting multicast.
Merolla et al. proposes a kind of tree routing device for supporting multicast, can pass through the broadcast (sites) in turn in subtree (subtree)
Mode realizes multicast.Stefan et al. proposes a kind of on-chip network structure of time-sharing multiplex, supports QoS, multicast and efficient
Connection creates a mechanism.Abad et al. proposes a kind of new routing mechanism for supporting multicast, is protected using the method for multicasting adaptively set
Limitation of the correct transmission of multicast without routing is demonstrate,proved, therefore improves the performance of CMP multicast.Samman et al. proposes one kind
It can realize that the multicast of deadlock freedom, the router of NoC support following functions without virtual channel: if certain data packets are hindered
It fills in, the microplate of other data packets is inserted into before them, to avoid deadlock.You Zhiqiang et al. is proposed based on BFT
The network-on-chip of type proposes a kind of multicast routing protocol and nodes encoding design scheme, and the multicast for completing to phase same core is surveyed
Examination, reduces test application time.
The concept of network code is for the first time to be put forward by Ahlswede et al. in 2000.It is a kind of network data
Transmission mode extends the concept and function of traditional routing.In traditional network based on routing, network node (router,
Interchanger etc.) it only carries out the forwarding of data and replicates two operations;And after having used network code, network node can be right
The data received carry out arbitrary encoding operation (such as linear transformation), then coding result is replicated or forwarded.With traditional net
Network is compared, and the advantage of network code is mainly reflected in multicast application, can promote throughput, realizes directed networks
The multicast capacity (multicast capacity) of (directed networks).Network code just causes state once proposition
The extensive concern of border academia, theoretical and application have become the new hot spot of network and communications field research.
The introducing of network code gives Netowrk tape next certain additional consumption, such as router of additional encoding operation, complexity,
Biggish caching.It in network code, needs under the premise of meeting throughput, uses simple calculations as far as possible, it is lesser to have
Confinement, and reduce the quantity of coding nodes.For given network topology structure, the smallest finite field is found to reach network
Capacity has NP difficulty.C.Fragouli and E.Soljanin is in " Information flow decomposition
For network coding " in prove finite field size and destination node number it is closely related, and give one
The upper bound of minimum finite field.Similar, finding least coding nodes number is also to have NP difficulty.M.Langberg,
A.Sprintson, and J.Bruck first will in " The encoding complexity of network coding "
General network is converted to so-called simple network.Using the method for simple network, under acyclic and network that is having ring respectively
The Lower and upper bounds of coding nodes number are gone out.C.Fragouli and E.Soljanin utilizes greedy algorithm, propose one it is new
Network information flow model optimizes to study coding nodes and coding number of edges purpose.Coding nodes, the optimization of coding side can also apply
Genetic algorithm (genetic algorithm).M.Kim, C.W.Ahn, M.M é dard, and M.Effros are in " On
Minimizing network coding resources:An evolutionary approach " in be considered as losing for the first time
Propagation algorithm Optimized Coding Based number of edges.For the coding of chromosome, network G is converted into corresponding line chart G ' first.For each
Encoding scheme, the bit vectors that code coefficient on the corresponding all sides line chart G ' is formed are as chromosome coding.With coding side
Adaptive value of the quantity as chromosome.Preliminary simulation result shows using the result of genetic algorithm better than greedy algorithm.
The research team of Texas A&M University proposes for network code to be applied to integrated electricity for 2006 for the first time
Road interconnects field.Jayakumar, Khatri and Gulati et al. application network coding techniques in ASIC bus reduce energy consumption
7%~8.5%, reduce interconnection bus length and area 6%~10%.Then, they arrive the application extension of network code
FPGA interconnects field, achieves good results.A kind of bidirectional bus for supporting network code, net has also been devised in the research team
Network coding techniques can use the bandwidth that a bidirectional bus realizes two one-way bus, save 49% bus resource and 11%
Bus energy consumption.
2011, Indrusiak for the first time applied to the concept of network code in network-on-chip, inquired into it in NoC group
Broadcast the potential advantages in communication.The angle for the hop count (hops) that article is transmitted from data packet, compares network coding and XY
The sum of hop required for routing is transmitted, tentatively demonstrates the advantage being applied to network coding in NoC.Analytical table
Bright, network code can reduce by 10%~22% transmission hop count.Thuan et al. answers network code in terms of hardware realization
For in CMP, devising the router for supporting network code, it was demonstrated that in the CMP of 9 cores, network code can drop significantly
Low network delay, and improve 1.48 times of handling capacity of network.Shalaby et al. mainly considers butterfly network at 2D-mesh
Optimization problem, propose the algorithms of 5 selections intermediate source node is and intermediate destination node i d, it was demonstrated that network code can be with
Reduce transmission hop count 12%~35%.The optimization problem of network code is expanded to general group from butterfly network by Vonbun et al.
Topological structure is broadcast, the packet transmission range of network-on-chip can be reduced by theoretically demonstrating network code.A.Shalaby,
V.Goulart, and M.-S.Ragab are in " Study of application of network coding on NoCs for
Multicast communications " in document [33], it was demonstrated that it can also be in multicast application effectively using network code
Reduce transmission delay.Hu et al. achieves Preliminary Results, but circuit is set it is also proposed that with the on-chip network structure of network coding technique
Meter and experimental result description are simultaneously unintelligible.
It is rigid to can be seen that the research work in terms of network coding technique is applied to network-on-chip from above document analysis
Ground zero, relevant research achievement be not also comprehensive.These preliminary explorations, which demonstrate network-on-chip, can efficiently use network volume
Code technology improves network throughput, reduces transmission delay, saves transmission energy consumption.But most of technology is all confined to butterfly net
The construction of network, and butterfly network is a special case of network code application.Although butterfly network is expanded to one by some technologies
As multicast topology structure and give theoretical performance, however specific network code construction is not mentioned in text to realize this
A little theoretical performances.But (Intra-session) coding is still confined in multicast.It is compiled by (Inter-session) between multicast
Code, bigger performance gain should may be implemented in we.In addition, the rough theory that these documents are mostly based on network code is commented
Estimate, lack and be accurately delayed and power consumption experimental data under actual circuit environment, optimization algorithm is also based on existing network coding reason
By, for network-on-chip structure and performance characteristics targetedly improved, it is difficult to realization to network code for
The facilitation of network-on-chip carries out research and evaluation comprehensively, comprehensive.
Summary of the invention
The shortcomings that it is a primary object of the present invention to overcome the prior art and deficiency, provide a kind of piece based on network code
Upper network router.
In order to achieve the above object, the invention adopts the following technical scheme:
The present invention provides a kind of network-on-chip router based on network code, which includes that the road P inputs, is defeated
Channel out, each input channel are equipped with v virtual channel again, are designed with a routing logic per input channel all the way
Routing Logic and virtual channel distributor VC allocator, route direction and the distribution for being each responsible for data transmission are virtual
Channel;Virtual channel distributor VC allocator is also connected with switch divider Switch allocator, the switch distribution
Device Switch allocator is a global component, is responsible for the road P input channel distribution crossbar switch crossbar, realizes
Multiple input channels request the president of the same output channel simultaneously, and Crossbar is a crossbar switch, realize that the input of the road P is logical
The rapid data of road to the road P output channel transmits, and network volume is additionally provided between the road P input channel and crossbar switch Crossbar
Code module NCU, the microplate output of virtual channel is by simple coding selection circuit, without what is encoded or do not encode here
Microplate is directly sent to crossbar switch Crossbar;If it is the microplate encoded in this router, then network code mould is sent to
Block NCU, NCU perform the encoding operation qualified input, by the microplate of coding through switch divider Switch
After the arbitration of allocator, it is sent to crossbar switch.
The logic that router passes through modification routing logic Routing Logic component as a preferred technical solution,
Circuit realizes multicast functionality.
The network code module NCU includes input buffer Input buffer and defeated as a preferred technical solution,
Buffer Output buffer out inputs distributor Input allocator according to the sky of input buffer Input buffer
Not busy state is input distribution buffer queue, and output moderator Output arbiter arbitration two-way coding output guarantees to intersect and open
The use for closing Crossbar does not conflict.
The network code module NCU scans input buffer Input buffer as a preferred technical solution, checks
Whether there are effective data to be encoded, and checks this data whether there is or not the paired datas of alignment, if two paths of data is all here, net
Network coding module NCU just carries out data encoding, and modifies packet header, the data packet after forming new coding, is sent to the defeated of free time
Out in buffer Outputbuffer.
The pack arrangement of support network code is defined as a preferred technical solution, as follows:
One data packet is made of several microplates Flit, including a packet header Head, several load datas Payload
It include control information with packet tail a Tail, Head and Tail, what is put in Payload is valid data.
Head includes the key control information for supporting network transmission and coding as a preferred technical solution, by 9 parts
It constitutes, the definition of each part and function are as shown in table 1:
Can be seen that from the definition in above table can just with NC type, Dist to IS and Dist to ID
To determine two data packets are performed the encoding operation in which router, which router multicast transmission is carried out at.
Data packet Dist to IS=2 at the beginning as a preferred technical solution, it is every to pass through a router Dist to
IS subtracts 1, and until is router becomes 0, network code is done in expression in is, and it is both a router that is, which is intermediate source node,
It is an encoder again, the raw data packets that two-way inputs is encoded according to the sequence of microplate;The data packet is sent to
In NCU, after Dist to IS is less than 0, indicates that network code is completed, no longer need to change, Dist to IS holding at this time -1 is not
Become, Distto ID indicates the intermediate destination router that also to arrive by how many a routers, needs data packet carrying out two-way
Multicast transmission.
Compared with the prior art, the invention has the following advantages and beneficial effects:
1, the present invention designs carried out comprehensive study to the network-on-chip based on network code, realize to hardware, power consumption,
The accurate assessment and modeling of performance, propose a kind of novel compact network-on-chip data packet coding structure, with the volume of very little
Outer hardware, delay and energy consumption design the efficient network-on-chip router for supporting microplate transfer mode, so that additional is hard
Part resource at most only accounts for 1/5th of router (for the router of 5 ports.For the router of multiport, additionally
Consumption is less), well solve the alignment and real-time coding problem of microplate in data packet.This is provided for the design of network-on-chip
New method and thinking has theory and practice for further increasing the performance of network-on-chip and many-core chip, reducing power consumption
Novelty.
2, in terms of concrete implementation technological means, the present invention has gathered network code, network-on-chip, on-chip multi-processor
With the newest research results in terms of low-power consumption, merge the related disciplines knowledge such as signal theory, integrated circuit and computer, it is theoretical with
Practice combines, and physical modeling and emulation experiment combine, and finally obtains research achievement.
3, the laboratory facilities of network-on-chip circuit design are related to IC design and Simulation Software Design, such as Fig. 3 institute
Show.The method that NoC circuit design is combined with modeling using IC design eda software tool and FPGA, the specific institute on the left of Fig. 3
Show.The present invention accelerates the efficiency of behavior simulation, the correctness of checking function using FPGA, and is gone out using eda tool emulation testing
The delay and power consumption of each component of new network-on-chip, and carry out accurate delay and power consumption modeling.
Detailed description of the invention
Fig. 1 (a)-Fig. 1 (b) is respectively network code processing figure in butterfly network coding structure figure and network-on-chip;
Fig. 2 is the schematic diagram of the network code bring alignment problem as unit of microplate;
Fig. 3 is the flow chart of the present embodiment experimental method;
Fig. 4 is the structure chart of inventive network router;
Fig. 5 is the schematic diagram that the present invention supports network code pack arrangement.
Specific embodiment
Present invention will now be described in further detail with reference to the embodiments and the accompanying drawings, but embodiments of the present invention are unlimited
In this.
Embodiment
The it is proposed of network code is (such as wireless sensor network, computer network in traditional macro network environment
Deng), there is the router of these networks stronger processing and store function can be enough to answer while supporting routing function
Pay the extra work of network code.Also, for the delay and power consumption of Relative Network transmission, volume consumed by network code
Outer delay and power consumption very little can be ignored.But network-on-chip is one kind for speed and hardware resource requirements pole
Its stringent micronet, processing and caching capabilities are all very weak, it is impossible to which consumption excessive additional hardware and energy consumption resource come
Support network code.
Fig. 1 (a) is a typical butterfly network coding structure: s1, s2 are two multicast source nodes;D1, d2 are two
The destination node of multicast;Is is intermediate source node, is both a router and an encoder, by two-way input
Raw data packets are encoded according to the sequence of flit, are then forwarded;Id is intermediate destination node, by received coding
Data packet afterwards is transmitted to the destination direction of multicast respectively.Intermediate source node is need to receive data packet A from s1, s2,
B makees simple exclusive or processing, is then transferred to intermediate destination node i d.Intermediate destination node i d gives packet delivery
d1,d2.D1, d2 are connected to after data packet A ⊕ B that perhaps B carries out XOR operation and obtains B or A with data packet A.
In macro network in (such as wireless network), the time and energy consumption required for network transmission are far longer than number of nodes
According to the time of processing, is and id are generally processor node, have stronger processing capacity and biggish caching.Data packet A, B
Exclusive or processing is just carried out after completely arriving at is.
However, in the network-on-chip as shown in Fig. 1 (b), the speed that network transmission requires is exceedingly fast, the processing of network code and
Distribution can not pass through processor node (such as nis, nid), otherwise will seriously affect the speed of network-on-chip.These operations must
(is, id) must be completed by network-on-chip router (router).But the recovery operation of network code (such as B=A ⊕ A ⊕ B)
It can be completed in destination node (such as d1, d2).
It is well known that the circuit-level that the router of network-on-chip only supports speed to be exceedingly fast exchanges and routing operations, and delay
Very little is deposited, using microplate (flit) as the unit of transmission, a complete data packet (package) can not be accommodated.Therefore, newly
The network-encoding operation of network-on-chip also must be as unit of flit, and cannot be as unit of package.Since network-on-chip is deposited
Blocking, the transmission of microplate is likely to occur discontinuously, brings huge difficulty to the network code of router, as shown in Figure 2.
The square (being started with c) that grey is filled in Fig. 2 represents the flit after exclusive or coding.Situation 1 is optimal situation,
That is data packet A, B equal length, and at the same time reaching a certain intermediate source node.The intermediate source node by simultaneously reach microplate into
The processing of row exclusive or, forms new microplate (c0-c6) and sends.Situation 2 indicates that data packet A, B two clock cycle of difference arrive
It reaches.A0, a1 and b5, b6 can not carry out network code processing.Situation 3 expression influenced by network congestion, data packet A, B it is micro-
The case where piece reaches intermediate source node intermittently, network code becomes sufficiently complex, so that the recovery of follow-up data is by pole
Big influence.
In addition, as shown in Fig. 1 (b), chip once after the completion of, multicast node position, network characteristic, link bandwidth etc. are all
It can not change.And the optimisation technique of existing network code allows network structure and parameter to be changed according to the requirement of optimization
Become.This generates the new problems of network code optimisation technique.
The technology path of two key cores is as follows:
1) the network-on-chip circuit design of network code is supported
The critical issue of this respect is the design of NoC router.The present invention is on the road NoC of classical support virtual channel
By improving on the basis of device, it is allowed to not only support traditional exchange route function, but also support microplate encoding function, as shown in Figure 4.
The router has P I/O channel, and each input channel has v virtual channel.There is one per input channel all the way
Routing Logic and VC allocator is each responsible for the route direction and distribution virtual channel of data transmission.Switch
Allocator is a global component, is responsible for P input channel distribution crossbar switch crossbar, realizes multiple inputs
The president of the same output channel is requested simultaneously in channel.Crossbar is a crossbar switch, realizes that the road P is input to the output of the road P
Rapid data transmitting.The network-on-chip router of 2D-mesh topological structure has 5 ports (P=5), respectively east, south, west,
It is northern and local.The NoC router of 3D-mesh topological structure has 7 ports (there are also upper and lower).Router can pass through modification
The logic circuit of Routing Logic component realizes multicast functionality.
New router increases network code module (Network Coding Unit or NCU).Virtual channel it is micro-
Piece output encodes selection circuit (according to the definition of table 1 below, checking Dist to IS=0, NC Type=1) by simple,
Microplate without encoding or not encoding here is directly sent to crossbar switch Crossbar;It is compiled if it is in this router
The microplate of code, then be sent to network code component (NCU).NCU performs the encoding operation qualified input.By the micro- of coding
Piece is sent to crossbar switch after the arbitration of Switch allocator.Therefore, the more inputs all the way of NCU of crossbar switch, become
At the crossbar switch of (P+1) xP, the crossbar switch of the PxP of classical router is increased, but ratio very little.
The circuit structure design of NCU is as shown in part under Fig. 4, it is assumed that is the 2D-mesh router of 5 inputs.NCU has 5
Input weaves into two-way coding output wherein be up to 4 circuit-switched data packets need to encode.Therefore, input is 5 in figure, input buffering
Device (Input buffer) is 4, and output buffer (Output buffer) is 2.Input allocator is according to Input
The idle state of buffer is input distribution buffer queue.Output arbiter arbitrates two-way coding output, guarantees
The use of Crossbar does not conflict.
Network coding is the critical component of addressable part.It scans 4 Input buffer, has checked whether there is
The data to be encoded of effect, and check this data whether there is or not the paired datas of alignment.If two paths of data is all here, Network
Coding just carries out data encoding, and modifies packet header (package head), and the data packet after forming new coding is sent to sky
In not busy Output buffer.
Input Buffer plays the role of good for solving the microplate alignment of network code, we can compile in network
Increase the timing that packet issues in code optimization algorithm to calculate, so that data packet reaches the time synchronization of coding nodes.For example, Fig. 1 (b)
In, 1 timing of data packet of the data packet of s2 earlier than s1 is sent, and A and B can reach is simultaneously.If there is slight obstruction,
The caching of A and B may be implemented in Input Buffer, so that they have been aligned carries out coding transmission again.If the Input of a packet
Buffer is full, another packet arrives not yet, does not just do encoding operation according to conventional mode is transmitted.In fact,
It can use the QoS that virtual channel ensures data to be encoded transmission, guarantee that data packet to be encoded is not blocked generally, so that
The phenomenon that being misaligned seldom occurs.
From fig. 4, it can be seen that the structure of NCU is like the circuit for increasing virtual channel all the way, mainly by buffering FIFO, secondary
Device is cut out, crossbar switch is constituted.It can thus be appreciated that additional hardware needed for supporting the NoC router of network code is simultaneously few (for 5 tunnels
NoC router, additional consumption are 1/5th of router;For high base router, additional consumption is smaller).Due to raising property
Can and introduce additional hardware resources can receive.In addition, the size of Input buffer is taken at the data packet of network code
Size and our flit are aligned strategy.It can use with accurate optimization algorithm, more compact data micro sheet structure into one
Step reduces the area of Input buffer.The presence of Output buffer can a smooth lower network transmission speed, but this
Function can also be completed by Input buffer.Output buffer sets or sets much, will be determined by emulation experiment.
And network code is simple xor operation, shared circuit resource is few.To sum up, additional hardware resources needed for NCU are
It is entirely acceptable.
2) compact data packet coding scheme
In order to realize network code in network-on-chip, the packet format transmitted to data is needed to carry out recompiling definition,
Support transmission mode of the network-on-chip with microplate (flit) for basic unit.The present invention defines the pack arrangement of support network code such as
Shown in Fig. 5.One data packet is made of several microplates (Flit), including packet header (Head), several load datas
(Payload) and a packet tail (Tail).Head and Tail includes control information, and what is put in Payload is valid data.
Head includes the key control information for supporting network transmission and coding, is made of 9 parts.The definition of each part
It is as shown in table 1 with function.
The composition and function of 1 Head of table
From defined above as can be seen that being assured that two just with NC type, Dist to IS and Dist to ID
A data packet is performed the encoding operation in which router, which router to carry out multicast transmission at.As shown in Fig. 1 (b), s2 to is
Data packet Dist to IS=2 at the beginning, every to subtract 1 by a router Dist to IS, having arrived is router becomes 0,
Network code is done in expression in is, which is sent in NCU.After Dist to IS is less than 0, indicate that network code is complete
At no longer needing to change, Dist to IS is kept -1 constant at this time.Dist to ID indicates also to arrive by how many a routers
Intermediate destination router needs data packet carrying out two-way multicast transmission.Its operating principle is identical as Dist to IS.
The laboratory facilities of network-on-chip circuit design are related to IC design and Simulation Software Design, as shown in Figure 3.
The method that NoC circuit design is combined with modeling using IC design eda software tool and FPGA, specifically as shown on the left side of figure 3.This
The efficiency of behavior simulation, the correctness of checking function are accelerated in invention using FPGA, and go out new piece using eda tool emulation testing
The delay and power consumption of upper each component of network, and carry out accurate delay and power consumption modeling.
Test assessment is carried out under simulation software environment, as shown in the right side Fig. 3.Delay and power consumption model are input to piece online
In network emulation software tool (according to experience before, secondary development NS2 network law carrys out design and simulation environment), in conjunction with net
The related optimization of network coding is successfully authenticated novelty of the invention with a large amount of real cases.
The above embodiment is a preferred embodiment of the present invention, but embodiments of the present invention are not by above-described embodiment
Limitation, other any changes, modifications, substitutions, combinations, simplifications made without departing from the spirit and principles of the present invention,
It should be equivalent substitute mode, be included within the scope of the present invention.
Claims (2)
1. the network-on-chip router based on network code, which is characterized in that the router include the road P input, output channel,
Each input channel is equipped with v virtual channel again, is designed with a routing logic Routing per input channel all the way
Logic and virtual channel distributor VC allocator is each responsible for the route direction and distribution virtual channel of data transmission;Void letter
Road distributor VC allocator is also connected with switch divider Switch allocator, the switch divider Switch
Allocator is a global component, is responsible for the road P input channel distribution crossbar switch crossbar, realizes multiple inputs
The arbitration of the same output channel is requested in channel simultaneously, and Crossbar is a crossbar switch, realizes the road P input channel to the road P
The rapid data of output channel transmits, and network code module is additionally provided between the road P input channel and crossbar switch Crossbar
NCU, the microplate output of virtual channel is by simple coding selection circuit, without coding or not in the micro- of this router coding
Piece is directly sent to crossbar switch Crossbar;If it is the microplate encoded in this router, then network code module is sent to
NCU, NCU perform the encoding operation qualified input, by the microplate of coding through switch divider Switch
After the arbitration of allocator, it is sent to crossbar switch;
The pack arrangement of support network code is defined as follows:
One data packet is made of several microplates Flit, including a packet header Head, several load datas Payload and one
A packet tail Tail, Head and Tail include control information, and what is put in Payload is valid data;
Head include support network transmission and coding key control information, be made of 10 parts, the definition of each part and
Function is as follows: NC type: showing general data packet NC type=0 or NC data packet NC type=1;Routing
Info: routing iinformation, using fixed route, one route-map of every 3 expressions;After often passing through a router, the information
Move to right 3;Dist to IS: the specified intermediate source router is and how many router has been arrived, 0 indicates to route currently
Device carries out NC operation, and -1 indicates that NC is completed;Dist to ID: the specified intermediate destination road and how many router has been arrived
It indicates to carry out two-way multicast operation in current router by device id, 0, -1 indicates that multicast operation is completed;Package ID: data packet
Serial number;Flit ID: the serial number of microplate;Src1: source node 1;Dest1: destination node 1;Src2: source node 2 works as NC
Type=1 is effective;Dest1: destination node 2, multicast node, regardless of whether being NC can use;It can from above-mentioned definition
To find out, it is assured that two data packets in which router just with NC type, Dist to IS and Dist to ID
It performs the encoding operation, which router to carry out multicast transmission at;
Data packet Dist to IS=2 at the beginning, it is every to subtract 1 by a router Dist to IS, until is router becomes
0, network code is done in expression in is, and is is intermediate source node, is both a router and an encoder, by two-way
The raw data packets of input are encoded according to the sequence of microplate;The data packet is sent in NCU, when Dist to IS is less than
It after 0, indicates that network code is completed, no longer needs to change, Dist to IS is kept -1 constant at this time, and Dist to ID is indicated will also be through
It crosses how many a routers to arrive intermediate destination router, needs data packet carrying out two-way multicast transmission;
The network code module NCU includes input buffer Input buffer and output buffer Output buffer, defeated
Entering distributor Input allocator according to the idle state of input buffer Input buffer is input distribution caching team
Column, output moderator Output arbiter arbitration two-way coding output, guarantee that the use of crossbar switch Crossbar does not conflict;
The network code module NCU scans input buffer Input buffer, checks whether there is effective data to be encoded,
If so, then checking this data, whether there is or not the paired datas of alignment, if two paths of data is all here, network code module NCU just into
Row data encoding, and packet header is modified, the data packet after forming new coding is sent to idle output buffer Output
In buffer.
2. the network-on-chip router according to claim 1 based on network code, which is characterized in that router is by repairing
The logic circuit for changing routing logic Routing Logic component realizes multicast functionality.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610169324.7A CN105871730B (en) | 2016-03-22 | 2016-03-22 | Network-on-chip router based on network code |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610169324.7A CN105871730B (en) | 2016-03-22 | 2016-03-22 | Network-on-chip router based on network code |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105871730A CN105871730A (en) | 2016-08-17 |
CN105871730B true CN105871730B (en) | 2019-03-05 |
Family
ID=56624889
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610169324.7A Active CN105871730B (en) | 2016-03-22 | 2016-03-22 | Network-on-chip router based on network code |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105871730B (en) |
Families Citing this family (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107800700B (en) * | 2017-10-27 | 2020-10-27 | 中国科学院计算技术研究所 | Router and network-on-chip transmission system and method |
CN107894963B (en) * | 2017-11-27 | 2021-07-27 | 上海兆芯集成电路有限公司 | Communication controller and communication method for system-on-a-chip |
CN109766311A (en) * | 2019-01-17 | 2019-05-17 | 上海华测导航技术股份有限公司 | A kind of convenient and fast channel multiplexing implementation method |
CN113032109A (en) * | 2019-12-09 | 2021-06-25 | 北京灵汐科技有限公司 | Data processing method and device and electronic equipment |
CN113312304B (en) * | 2021-06-04 | 2023-04-21 | 海光信息技术股份有限公司 | Interconnection device, mainboard and server |
CN115550235B (en) * | 2022-06-22 | 2024-02-09 | 南京大学 | Single-period router oriented to neural network platform |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101232456A (en) * | 2008-01-25 | 2008-07-30 | 浙江大学 | Distributed type testing on-chip network router |
CN101656681A (en) * | 2009-09-02 | 2010-02-24 | 复旦大学 | Network coding router based on network processor |
CN102629913A (en) * | 2012-04-11 | 2012-08-08 | 浙江大学 | Router device suitable for globally asynchronous locally synchronous on-chip network |
CN104092615A (en) * | 2014-06-10 | 2014-10-08 | 西安电子科技大学 | Network on chip with network coding function, network topology of the network on chip, and route algorithm of the network topology |
-
2016
- 2016-03-22 CN CN201610169324.7A patent/CN105871730B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101232456A (en) * | 2008-01-25 | 2008-07-30 | 浙江大学 | Distributed type testing on-chip network router |
CN101656681A (en) * | 2009-09-02 | 2010-02-24 | 复旦大学 | Network coding router based on network processor |
CN102629913A (en) * | 2012-04-11 | 2012-08-08 | 浙江大学 | Router device suitable for globally asynchronous locally synchronous on-chip network |
CN104092615A (en) * | 2014-06-10 | 2014-10-08 | 西安电子科技大学 | Network on chip with network coding function, network topology of the network on chip, and route algorithm of the network topology |
Also Published As
Publication number | Publication date |
---|---|
CN105871730A (en) | 2016-08-17 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105871730B (en) | Network-on-chip router based on network code | |
Agrawal et al. | Simulation of network on chip for 3D router architecture | |
Zhang et al. | Comparison research between xy and odd-even routing algorithm of a 2-dimension 3x3 mesh topology network-on-chip | |
CN102685017A (en) | On-chip network router based on field programmable gate array (FPGA) | |
Chen et al. | Reducing wire and energy overheads of the SMART NoC using a setup request network | |
CN105844014B (en) | Network-on-chip Encoding Optimization based on chip design cycle and application design cycle | |
CN101808032A (en) | Static XY routing algorithm-oriented two-dimensional grid NoC router optimization design method | |
Wang et al. | Area and power-efficient innovative congestion-aware Network-on-Chip architecture | |
CN105007226A (en) | Multicast router based on package and circuit switching technology and working mode thereof | |
CN109302357A (en) | A kind of on piece interconnection architecture towards deep learning reconfigurable processor | |
Rad et al. | A survey and taxonomy of congestion control mechanisms in wireless network on chip | |
CN105530206B (en) | A kind of Torus network system and its working method with double access infrastructures | |
Duraisamy et al. | Enabling high-performance SMART NoC architectures using on-chip wireless links | |
Wang et al. | An adaptive partition-based multicast routing scheme for mesh-based networks-on-chip | |
Effiong et al. | Scalable and power-efficient implementation of an asynchronous router with buffer sharing | |
Sun et al. | DCBuf: a high-performance wireless network-on-chip architecture with distributed wireless interconnects and centralized buffer sharing | |
Furhad et al. | An analysis of reducing communication delay in network-on-chip interconnect architecture | |
Dehghani et al. | Design and performance evaluation of Mesh-of-Tree-based hierarchical wireless network-on-chip for multicore systems | |
Bourduas et al. | Modeling and evaluation of ring-based interconnects for Network-on-Chip | |
Zhu et al. | BiLink: A high performance NoC router architecture using bi-directional link with double data rate | |
Vinothkumar et al. | Design and Implementation of Router Arbitration in Network on Chip | |
Prasad et al. | High-speed virtual logic network on chip router architecture for various topologies | |
Zhang et al. | A cellular NoC architecture based on butterfly network coding (CBNoC) | |
Aslam et al. | An efficient router architecture and its FPGA prototyping to support junction based routing in NoC platforms | |
Somisetty et al. | Congestion aware negative first routing with fair arbitration for network on chip |
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 |