CN103428804B

CN103428804B - Find mapping scheme and network code location method between network-on-chip task and node

Info

Publication number: CN103428804B
Application number: CN201310330608.6A
Authority: CN
Inventors: 陈亦欧; 胡剑浩; 赵竞; 凌翔
Original assignee: University of Electronic Science and Technology of China
Current assignee: University of Electronic Science and Technology of China
Priority date: 2013-07-31
Filing date: 2013-07-31
Publication date: 2016-03-30
Anticipated expiration: 2033-07-31
Also published as: CN103428804A

Abstract

The invention discloses mapping scheme and network code location method between a kind of searching network-on-chip task and node, it is applied in the mapping on wireless NoC for multicast, when the distribution of search optimal node and routing scheme, not only consider that clean culture task total power consumption and response time are as the preferential index of assessed form, ensure that the throughput of wherein multicast task is maximum and Internet Transmission that is multicast task is least in power-consuming simultaneously.Thus make the performance of this preferred conceptual design network out out of employing reach best, and and other is to the Measures compare finding the priority scheme mapped between task and node under multiple target function, and there is complexity, realize simple feature.

Description

Find mapping scheme and network code location method between network-on-chip task and node

Technical field

The invention belongs to integrated circuit (IC) design field, special in the task matching of wireless network-on-chip and the search procedure of network code path and encoded point position, be specifically related to a kind of method of mapping scheme and network code position between task for finding wireless network-on-chip and node.

Background technology

On information market, along with the continuous expansion of microchip technology scale, extensive SOC (system on a chip) is made to become possibility.Network-on-chip, as a kind of brand-new communication construction designing multi core chip, can tackle chip chamber global communication problem complicated in system-on-chip designs well.But the metal owing to network-on-chip two dimension connects, when network-on-chip carries out multi-hop transmission, there is high time delay and high energy consumption problem, when processor popularization, network-on-a-chip reliability can not be protected.Wireless network-on-chip (WirelessNetwork-on-Chip) replaces wired connection with wireless connections, wireless transmission channel is faster inserted between network-on-chip node, increase the transmission range of network-on-chip node, reduce the jumping figure of information transmission path in multiple nucleus system, system delay and energy consumption is reduced to reach, increase the object of system throughput, improve entire system performance further.

Network code has a kind ofly merged the message-switching technique of coding and route, and its core concept allows intermediate node in a network to carry out information processing, by improving code efficiency, reaches the theoretical upper bound of information flow.Network code can improve the throughput bandwidth availability ratio of network, network robustness and fail safe, and under having the condition of broadcast in a communication network, uses network code to be the theoretical upper bound that necessarily can reach the information flow speed that Shannon's theorems proposes.And wireless network-on-a-chip wireless transmission node disposes on-chip antenna usually, the requirement that multicast transmission network needs node can be broadcasted can be adapted to well.

By Cha Xin with extensively collect documents and materials, we find disclosedly have following a few class for finding on network-on-chip the method for mapping scheme between task and node:

Document " Zhou Ganmin; Yin Yongsheng; Hu Yonghua etc. the NoC based on ant colony optimization algorithm maps. computer engineering with application .2005; 41 (18): 7-10. " adopt one effectively ant colony optimization algorithm achieve NoC mapping, while by duty mapping to network-on-chip node, decrease the communication realize of system as much as possible.The target function of the method is power consumption, and the method is roughly as follows:

According to the number of ant cyclically for each unappropriated node according to probability assignments task (this probability be by a kind of mode calculate by the probability of a peer distribution to certain task), every circulation primary all will make all nodes all be assigned to corresponding task, then adopt two exchange processes to carry out Local Search, find out the optimal solution that power consumption is minimum, and the pheromones of ant corresponding to optimal solution is upgraded by the power consumption number pheromones degree of decay of optimal solution and the value of current information element, by to after said process successive ignition, finally select the allocative decision making power consumption minimum.

The mapping scheme that this method obtains has less communication realize, but does not consider the delay performance of system.

Document " Yang Shengguang; Li Li; brilliant human relations etc. towards the NoC mapping method of energy consumption and time delay. electronic letters, vol .2008; 36 (5): 937-942 " based on lattice structure NoC platform, establish the unified target function being intended to optimization system communication energy consumption and link load variance based on ant group algorithm.The calculating formula of the target function cost of the method is:

cost=λ×E+(1-λ)×VAR

Wherein, λ is proportionality coefficient, and E is energy consumption, and VAR is link load variance.The method is roughly as follows:

According to the number of ant cyclically for each unappropriated node according to probability assignments task (probability be by a kind of mode calculate by the probability of a peer distribution to certain task), every circulation primary all will make all nodes all be assigned to corresponding task, then adopt two exchange processes to carry out Local Search, find out power consumption minimal solution, and the pheromones of ant corresponding to optimal solution is upgraded by the power consumption number pheromones degree of decay of optimal solution and the value of current information element, by to after said process successive ignition, finally select the allocative decision making target function cost minimum.

This method does not consider that communication energy consumption and time of implementation are diverse physical quantitys, and the mapping scheme of acquisition not necessarily has excellent performance.

Document Tanglei, ShashiKumar.Atwo-stepgeneticalgorithmformappingtaskgraph stoanetworkonchiparchitecture.ProceedingsoftheEuromicroS ymposiumonDigitalSystemDesign2003.Antalya, Turkey:IEEE, 2003:180-187. propose the method that a kind of employing two step genetic algorithms realize duty mapping problem.The method is roughly as follows:

(1) first supposing that the time delay on all limits in Flow chart task is equal, is a constant, i.e. average; For each task matching functional unit (as processor, memory etc.), make total time delay of system minimum

(2) use the limit time delay of real data flow diagram, optimize further, by functional unit allocation on the node of NoC.

The whole optimizing phase be divide into two steps by the method, thus reduces the complexity of calculating, shortens computing time, but the energy consumption of the mapping scheme obtained and time delay can not be simultaneously minimum.

Document YiouChen, JianhaoHu, GengshengChen, XiangLing. " EnergyandDelay-AwareMappingforReal-TimeDigitalProcessing SystemonNetworkonChipplatforms ", inProceedingsofIEEEInternationalSystemonChipConference, LasVegas, USA, 2010, pp.375-378. propose a kind of mapping method MM-Map from data flow diagram to NoC network topology newly, realize the how point-to-multipoint mapping of different scales.Propose in literary composition:

(1) data flow diagram of autonomous definition, utilizes the weights of data flow diagram and annexation to draw complete link information, can reflect performance parameters such as mapping front digital signaling system time delay, power consumption objectively.

(2) select basic genetic algorithmic as the prototype of mapping algorithm, enable algorithm adapt to NoC mapping problems, to obtain good algorithmic statement characteristic by some improvement; Set up power consumption model, delay model and multi-objective Model for mapping objects simultaneously, compare the Mathematical Modeling of traditional power consumption and time delay, the performance change of system before and after mapping can be represented more accurately.

This mapping method principle is simple, step less, have good using value.But this mapping method for the application with clean culture task, in the face of comprising the application of multicast task, owing to not considering the correlated performance after multicast duty mapping, still can only show shortcoming.

Summary of the invention

The object of the invention is to overcome above-mentioned defect of the prior art, a kind of method of mapping scheme and network code position between task for finding wireless network-on-chip and node is provided, the wireless network-on-chip for multitask multinode propose a kind of find the while of making clean culture task total power consumption and response time minimum.

Further aim of the present invention is to provide a kind of method of mapping scheme and network code position between task for finding wireless network-on-chip and node, ensures that the throughput of multicast task is maximum and Internet Transmission that is multicast task is least in power-consuming.

In order to realize foregoing invention object, the technical solution used in the present invention is:

Mapping scheme and network code location method between a kind of searching network-on-chip task and node, comprise the steps:

1). stochastic generation one comprises the scheme group of K allocative decision and be each allocative decision X=(x ₁, x ₂..., x _n) set a Probability p of intersecting _taskwith the probability q of variation _task, and duty mapping iterations T _task, and make the variable t of the number of times of logger task iteration of maps _task=0 and the scheme number Z=1 that obtained of initialization, maximal solution number Z is set _max;

2). obtain scheme group in wireless network-on-chip clean culture task power consumption and response time under often kind of scheme, and record maximum power dissipation and maximum response time is the poorest power consumption number and the poorest response time value;

3). calculate scheme group in the fitness value of each scheme, the priority of this fitness value size reflection scheme, it is higher be worth larger priority, so-called priority level be by clean culture task total power consumption and the response time little be excellent;

4). by scheme group in K scheme be divided at random individual scheme group, the scheme number wherein in L Shi Mei scheme group, selects prepreerence scheme in every scheme group according to the dominance relation determined in the 3rd step and forms a priority scheme group

5). by priority scheme group in scheme press random pair, described pairing refers to divides into groups according to two schemes one group, and with the Probability p of the intersection of setting in step 1 _ithe numbering of the node in any one same position in the scheme of two in often pair of scheme is exchanged, and then all to be combined by often pair of scheme after exchanging is obtained scheme group

6). with the mutation probability q of setting in step 1 _ichange scheme group in the numbering of node in each scheme on any one position, obtain scheme group

7). the fitness value order from small to large calculated by the 3rd step, from scheme group middle selection individual scheme, with scheme group scheme combine, form and comprise the new scheme group of K scheme

8). by scheme group with be merged into scheme group

9). numerical procedure group in wireless network-on-chip clean culture total power consumption and response time under often kind of scheme.During calculating, first judge whether the multicast task source and destination node of each scheme in scheme group is mapped on different wireless network-on-chip nodes, if, then direct by the poorest power consumption number and the poorest corresponding time value assignment to the power consumption of this programme and response time value, if not, then numerical procedure group in wireless network-on-chip clean culture total power consumption and response time under often kind of scheme.

10). by scheme group in all 2K scheme be divided into multiple group, each group represents a boundary set, and to little group #, the priority of the scheme of the group that the scheme in less group of numbering is larger than numbering is high; So-called priority level is little by power consumption and response time is excellent.

11). the scheme group of scheme that what initialization one was new do not have then from the group that the 10th step has been divided, by the numbering order from big to small of group, successively the scheme in group is joined in, until scheme group in scheme number individual more than K, then the scheme of the group finally added all is taken out.

12). the 10th) in step, determine that the concrete grammar step of the order of priority of each scheme in each group is as follows:

12-1 makes the distance value of each scheme in group be 0, and described distance value represents whether contacting of the program and other scheme be tight.

12-2 is according to the 9th) size of the power consumption of each scheme that calculates of step carries out inverted order arrangement to scheme, and the scheme that power consumption is less comes before the larger scheme of power consumption,

12-3 calculates the power consumption distance of each scheme, the value that the power consumption distance of each scheme obtains for power consumption that the power consumption coming that scheme after it deducts that scheme come before it,

12-4 is according to the 9th) the response time size of each scheme that calculates of step carries out inverted order arrangement to scheme, and response time less scheme comes before response time larger scheme,

12-5 calculates the response time distance of each scheme, the value that the response time distance of each scheme obtained for response time that the response time coming that scheme after it deducts that scheme come before it,

The response time Distance geometry power consumption distance of each scheme is added the distance value of the scheme that obtains by 12-6,

12-7 comes foremost scheme minimum for power consumption and response time minimum scheme, then according to distance size, remaining scheme is sorted, and come after the minimum scheme of power consumption and response time minimum scheme, the scheme that distance value is larger comes before the less scheme of distance value, and the priority that the priority ratio coming scheme above like this comes scheme is below high.

13). from the 12nd) according to priority height selection scheme adds scheme group in the group that takes out step until scheme group in scheme number be K.

14). by scheme group in K scheme be divided at random individual scheme group, the scheme number wherein in L Shi Mei scheme group, according to the 10th) step and the 12nd) dominance relation determined of step selects prepreerence scheme in every scheme group and forms a priority scheme group

15). by priority scheme group in scheme random pair again, with step 1) in the Probability p of intersection of setting _taskthe numbering of the node in any one same position in the scheme of two in often pair of scheme is exchanged, and then all to be combined by often pair of scheme after exchanging is obtained scheme group

16). with step 1) in the probability q of variation of setting _taskchange scheme group in the numbering of node in each scheme on any one position, obtain scheme group

17). according to the 10th) the scheme group determined of step the order of priority of middle scheme, according to priority order from high to low, from scheme group middle selection individual scheme, with scheme group scheme combine, form and comprise the new scheme group of K scheme

18. if t _task<T _task, then t _task=t _task+ 1, then turn back to step 2), continue iteration; Otherwise enter step 19);

19). calculate scheme group in the fitness value of each scheme, by scheme maximum for fitness value, i.e. power consumption and response time simultaneously minimum scheme J _minas the task of wireless network-on-chip to internodal duty mapping scheme.

Wherein, under the 2nd) calculating often kind of mapping scheme in step, the method for wireless network-on-chip clean culture task power consumption and response time task is as follows:

The power consumption of the wireless network-on-chip described in duty mapping is that the power consumption of all processing nodes and internodal data transmit power consumption sum; Response time described in duty mapping is the longest path time delay from being input to output, comprises the transmission time of exchanges data between the process time delay of all tasks in path and task.

According to the embodiment of the present invention, the inventive method also comprises the steps:

I. according to the scheme J that duty mapping obtains _min, determine that multicast task originating task and object task are in the mapping node position of wireless network-on-chip with according to relation between the multicast task in Flow chart task, from X _sourcereject the node of the wireless network-on-chip that isolated multicast task source node maps, remaining there is glue connection relation multicast task originating task and the node set that maps of object task, with vector with represent.Describedly isolated refer to that this multicast source node and other multicast source nodes do not have identical information destination node or its transmission path with other multicast transmission without any conflicting, then claim this multicast source node to be isolated.Described glue connection relation, refer to that there is several Multicast source has the identical stay of two nights, and transmission path has the situation of intersection each other, so then claims these Multicast source and the stay of two nights to have glue to join relation.

Ii. X is calculated _source' in each source node to X _des' in all paths of each destination node, record all paths with matrix total_path, a=j*m, path_ID=[b ₁, b ₂... b _a] record the total number of paths of each source to each destination node;

Iii. vectorial Scheme=(y is used ₁, y ₂... y _a) carry out a kind of scheme of presentation code Path selection; y _krepresent the numbering of the paths in this coding path selection scheme, namely this paths is on total_path paths; Stochastic generation one comprises the scheme group of P allocative decision and be each coding path scheme Scheme=(y ₁, y ₂... y _a) set a Probability p of intersecting _codingwith the probability q of variation _coding, and coding mapping iterations T _coding, and the variable t of the number of times of order record iteration _coding=0;

Iv. scheme group is obtained in the throughput of multicast task and Internet Transmission power consumption under often kind of scheme;

V. the non-dominated ranking method of Deb (Science Press published in 2007, and author introduces in " multi-objective Evolutionary Algorithm and the application thereof " of Zheng Jinhua) is adopted to calculate scheme group in the fitness value of each scheme, the priority of this fitness value size reflection scheme, it is higher be worth larger priority, so-called priority level be large by multicast multiplexed transport throughput and network power consumption little be excellent;

Vi. by scheme group in P scheme be divided at random individual scheme group, the scheme number wherein in N Shi Mei scheme group, selects prepreerence scheme in every scheme group according to the dominance relation determined in v step and forms a priority scheme group

Vii. by priority scheme group in scheme press random pair, described pairing refers to divides into groups according to two schemes one group, and with the Probability p of the intersection set in step I _codingthe numbering of the node in any one same position in the scheme of two in often pair of scheme is exchanged, and then all to be combined by often pair of scheme after exchanging is obtained scheme group

Viii. with the mutation probability q set in step I _codingchange scheme group in the numbering of node in each scheme on any one position, obtain scheme group

Ix. the fitness value order from small to large calculated is walked by v, from scheme group middle selection individual scheme, with scheme group scheme combine, form and comprise the new scheme group of K scheme

X. by scheme group with be merged into scheme group

Xi. according to the method numerical procedure group of step I v in multicast task throughput and network power consumption under often kind of scheme;

Xii. the boundary set adopting Deb to introduce in the article AFastElitistNon-dominatedSortingGeneticAlgorithmforMulti-objectiveOptimization:NSGA-II to be published on magazine IEEETransactionsonEvolutionaryComputation the 6th volume for 2002 produces method by scheme group in 2P scheme be divided into multiple group, each group represents a boundary set, and to little group #, the priority of the scheme of the group that the scheme in less group of numbering is larger than numbering is high; So-called priority level be by throughput large little with network power consumption be excellent;

Xiii. the scheme group not having scheme that initialization one is new then from the group that xth ii step has been divided, by the numbering order from big to small of group, successively the scheme in group is joined in, until scheme group in scheme number individual more than P, then the scheme of the group finally added all is taken out.

Xiv. determine the order of priority of each scheme in each group in xth ii step, concrete grammar is as follows:

Xiv-1 makes the distance value of each scheme in group be 0, and described distance value represents whether contacting of the program and other scheme be tight;

Xiv-2 walks the network power consumption of each scheme calculated size according to xth i carries out inverted order arrangement to scheme, and the scheme that power consumption is less comes before the larger scheme of power consumption;

Xiv-3 calculates the network power consumption distance of each scheme, the value that the power consumption distance of each scheme obtains for power consumption that the power consumption coming that scheme after it deducts that scheme come before it;

Xiv-4 walks each scheme calculated throughput size according to xth i arranges scheme, and the scheme that throughput is larger comes before the less scheme of throughput;

Xiv-5 calculates the throughput distance of each scheme, the value that the throughput distance of each scheme obtains for throughput that the throughput coming that scheme after it deducts that scheme come before it;

The throughput Distance geometry power consumption of each scheme distance is added the distance value of the scheme that obtains by xiv-6;

Xiv-7 comes foremost scheme maximum for throughput and the minimum scheme of network power consumption; then according to distance size, remaining scheme is sorted; and come after the maximum scheme of throughput and the minimum scheme of network power consumption; the scheme that distance value is larger comes before the less scheme of distance value, and the priority that the priority ratio coming scheme above like this comes scheme is below high;

Xv. in the group that have passed through taking-up xth iii step, according to priority height selection scheme adds scheme group until scheme group in scheme number be P;

Xvi. by scheme group in P scheme be divided at random individual scheme group, the scheme number wherein in L Shi Mei scheme group, walks according to xth ii step and xiv the dominance relation determined and selects prepreerence scheme in every scheme group and form a priority scheme group V _t';

Xvii. by priority scheme group V _t' in scheme random pair again, with the Probability p of the intersection set in step I ii _codingthe numbering of the node in any one same position in the scheme of two in often pair of scheme is exchanged, and then all to be combined by often pair of scheme after exchanging is obtained scheme group

Xviii. with the probability q of the variation set in step I ii _codingchange scheme group in the numbering of node in each scheme on any one position, obtain scheme group

Xix. the scheme group determined is walked according to xth ii the order of priority of middle scheme, according to priority order from high to low, from scheme group middle selection individual scheme, with scheme group scheme combine, form and comprise the new scheme group of P scheme

If xx. t _coding<T _coding, then t _coding=t _coding+ 1, then turn back to step I v, otherwise enter step xxi;

Xxi. the non-dominated ranking method of Deb (Science Press published in 2007, and author introduces in " multi-objective Evolutionary Algorithm and the application thereof " of Zheng Jinhua) is adopted to calculate scheme group in the fitness value of each scheme, by scheme maximum for fitness value, the scheme that namely Internet Transmission power consumption is minimum and throughput is simultaneously maximum is as a mapping scheme of wireless network-on-chip coding mapping;

Xxii. judge whether the coding mapping scheme throughput now obtained is the minimum that the i-th v step specifies; and whether power consumption is the maximum of the i-th v step regulation, if throughput is not minimum and power consumption is not maximum, then record this mapping scheme; Z=Z+1, then enters step xxiii simultaneously; Otherwise, return step 2);

If xxiii. Z≤Z _max, then step 2 is returned), otherwise, enter step xxiv;

Xxiv. the non-dominated ranking method of Deb (Science Press published in 2007, and author introduces in " multi-objective Evolutionary Algorithm and the application thereof " of Zheng Jinhua) is adopted to calculate Z _maxthe fitness value of individual mapping scheme, selects the scheme that fitness value is maximum, i.e. the scheme that multicast task throughput is maximum and multicast multiplexed transport power consumption is minimum is as finally mapping selection scheme, and record coding maps the numbering of path and the encoded point selected.

In above-mentioned i-th v step, solve scheme group in under often kind of scheme the throughput of multicast task and the method for Internet Transmission power consumption as follows:

Described Internet Transmission power consumption refers to the power consumption that wireless routing node in network carries out transfer of data and brings comprise the arbitration power consumption of node, coding power consumption and wireless channel transmission power consumption; The parameter value proposed in paper " DesignofMulti-ChannelWirelessNoCtoImproveOn-ChipCommunic ationCapacity " literary composition that the reference of correlation computations parameter DanZhao, YiWang, JianLi and TakamaroKikkawa deliver.

When calculating throughput under often kind of scheme and Internet Transmission power consumption, first judge whether the combination of paths of this scheme meets the condition that multicast multiplexed transport can reach maximum multicast rate; If meet the condition that can reach maximum multicast rate, obtain the encoded point of this combination of paths, and calculate throughput and the Internet Transmission power consumption of this mapping scheme; If can not reach the condition of maximum multicast rate, then the throughput of this mapping scheme multicast task is a minimum, and this minimum can be 0.The network power consumption of this mapping scheme multicast multiplexed transport is a maximum.

Described multicast multiplexed transport throughput (the multicast information amount transmitted in network in the unit interval is weighed) is calculated as follows:

R = \underset{s &Element; S}{Σ} \frac{f_{s} \cdot n_{s}}{T_{M}}

Wherein, S is multicast source node set, and s is one of them source node, f _srepresent that in multicast figure, source node s is to the traffic of single destination node, n _srepresent the destination node number that source node s is corresponding, T _mfor the transmission time of multicast task.

In above-mentioned i-th i step, the Mesh topological structure that the computational methods of total_path adopt with wireless NoC and routing algorithm associated, concrete steps are:

Ii-1X _source' there is j source node, X _des' there is m destination node; b ₀=0; Make label x=1, y=1; To an a=j*m dimension group path_ID=[b ₁, b ₂... b _a] assignment, be specially:

(1) p=1 is made, q=1;

(2) X is selected _source' in p source node (make its node serial number be n _s) and X _des' q destination node (make its node serial number be n _d), obtaining their coordinates in Mesh topology according to their node serial number is (x _s, y _s) and (x _d, y _d).Wherein,

x_{s} = n_{s} - 1 - \sqrt{M} \cdot floor (\frac{n_{s} - 1}{\sqrt{M}}),

y_{s} = floor (\frac{n_{s} - 1}{\sqrt{M}}),

x_{d} = n_{d} - 1 - \sqrt{M} \cdot floor (\frac{n_{d} - 1}{\sqrt{M}}),

y_{d} = floor (\frac{n_{d} - 1}{\sqrt{M}}),

Floor () is downward bracket function;

(3) calculate from node n _sto node n _dthe number in shortest route path be

Max_path_num = \frac{(| x_{s} - x_{d} | + | y_{s} - y_{d} |)!}{| x_{s} - x_{d} |! | y_{s} - y_{d} |!},

By these path number consecutivelies, represent path number value with e, each paths is expressed as successively make c=(q-1) j+p, b _c=Max_path_num; Then p=p+1;

(4) if p>j, then, q=q+1, enters step (5); Otherwise, get back to step (2);

(5) if q>m, then complete assignment, enter step I i-2; Otherwise make p=1, get back to step (2);

Ii-2 initialization oK the full remainder group total_path of row;

Ii-3 selects X _source' in an xth source node and X _des' y destination node, make k=(y-1) j+x;

Ii-4 is from of total_path row starts to the oK, current source is recorded to current destination node b altogether _keach node serial number on paths; Initialization g=1, g'=1, e=1; Detailed step is as follows:

I. e paths is selected, make r be node number on this paths, source node is the 1st node on this path, and that node that path is connected with source node is the 2nd node on this path, all the other nodes are analogized, and destination node is r node on this path;

II. by the node serial number assignment on g' position in current path to the element on the capable g column position of h of array total_path; Then g'=g'+1, g=g+1;

If III. g'≤r, then return Step II, otherwise enter step IV;

IV.total_path line number h=h+1, path number e=e+1, enter step V;

If V. e≤b _k, then step I is entered; Otherwise enter step I i-5;

Ii-5 now, from X _source' in an xth source node to X _des' all shortest route paths of y destination node all recorded; Label x=x+1, enters step I i-6;

If ii-6 is x>j, then enter ii-7.Otherwise return ii-3;

Ii-7 label y=y+1, if y<=m, then makes x=1, returns ii-3; Otherwise, represent the assignment completed total_path.

Compared with prior art, technique effect of the present invention:

The present invention is directed to multicast and be applied in mapping on wireless NoC, when the distribution of search optimal node and routing scheme, not only consider that clean culture task total power consumption and response time are as the preferential index of assessed form, ensure that the throughput of wherein multicast task is maximum and Internet Transmission that is multicast task is least in power-consuming simultaneously.Thus make the performance of this preferred conceptual design network out out of employing reach best, and and other is to the Measures compare finding the priority scheme mapped between task and node under multiple target function, there is complexity, realize simple feature, method fast convergence rate of the present invention can also be drawn by emulation.

Accompanying drawing explanation

Fig. 1 is the task image be made up of 7 tasks that the embodiment of the present invention is enumerated.

Fig. 2 is the network-on-chip be made up of 16 nodes that the embodiment of the present invention is enumerated.

In figure, 1 is clean culture task node, and 2 is the traffics between task, and 3 is multicast source task nodes, and 4 is multicast task objective task nodes, and 5 is processing nodes of wireless NoC, and 6 is routing nodes of wireless NoC.

Embodiment

Below in conjunction with test example and embodiment, the present invention is described in further detail.But this should be interpreted as that the scope of the above-mentioned theme of the present invention is only limitted to following embodiment, all technology realized based on content of the present invention all belong to scope of the present invention.

The inventive method is mainly towards the Flow chart task that there is multicast task.There are clean culture and multicast two kinds of task nodes in this kind of Flow chart task, the transfer of data between clean culture task node adopts unicast transmission mode, and the topological structure that its routing algorithm is adopted by wireless NoC determines.Transfer of data between multicast task node adopts broadcast transmission methods, adopts complete trails routing algorithm.Described complete trails routing algorithm refer to routed path can select from source node to destination node arbitrarily forward link (along source node to the transmission direction of destination node in the routing algorithm of forward i.e. topology regulation herein) consider that this type of Flow chart task is when the mapping of wireless network-on-chip, not only need the performance considering clean culture duty mapping, also need the performance considering multicast duty mapping.The present invention be the wireless network-on-chip of multitask multinode propose a kind of searching make clean culture task total power consumption and the response time simultaneously minimum and ensure that the throughput of wherein multicast task Internet Transmission that is maximum and multicast task is least in power-consuming time task and node between map and select the mapping method of corresponding multicast node-routing path configuration network code.The mapping scheme performance adopting the inventive method to obtain is high, and fast convergence rate, complexity is low.

7 tasks are had in task image shown in Fig. 1, network-on-chip shown in Fig. 2 has 16 nodes, and the present embodiment is made the present invention based on this and being illustrated, and is not limited to the present invention, network-on-chip can have other node numbers, and task image also can have the task of other numbers.The present embodiment is exactly a kind ofly 16 nodes of the wireless network-on-chip shown in the task matching to Fig. 2 of 7 shown in Fig. 1 will design routed path to find, and after making to map, the total power consumption of network-on-chip and response time reach minimum value simultaneously and ensure the mapping scheme that the throughput of wherein multicast task is maximum and multicast Task Network is least in power-consuming.For convenience of description, we are with vectorial X=(x ₁, x ₂, x ₃, x ₄, x ₅, x ₆, x ₇) represent that 7 tasks of network-on-chip are to a kind of scheme of 16 node mappings, the x on n-th position of vectorial X _n(1≤x _n≤ 16) numbering of the n-th duty mapping to node is represented, i.e. x _nillustrate and give xth by the task matching being numbered n _nnode.

Minimum the while of searching total power consumption and response time, ensure that the step of the scheme that the throughput of wherein multicast task is maximum and multicast Task Network is least in power-consuming is as follows simultaneously:

The step of duty mapping is as follows:

1. stochastic generation one comprises the scheme group of 20 allocative decisions each allocative decision vector representation in scheme group is: (11151071632), (810814101112), (12121014847), (166255412), (9811105615), (54113841), (1316916493), (48121511614), (4987377), (841035107), (10241341016), (101112101213), (2147314716), (2132613144), (5471413152), (121551012137), (7114101459), (14111015131310), (15169104316), (71711698).The probability setting the intersection of each scheme is p _task=0.4, the probability of variation is q _task=0.1, duty mapping iterations is T _task=100, and the variable t of the number of times of order record iteration _task=0.The scheme number that initialization has obtained is Z=1, and arranging maximal solution number is Z _max=2.

2. obtain scheme group in the power consumption of wireless network-on-chip and response time under often kind of mapping scheme

The calculating parameter of 2-1 power consumption is with reference to DanZhao, the simulation parameter proposed in paper " DesignofMulti-ChannelWirelessNoCtoImproveOn-ChipCommunic ationCapacity " literary composition that YiWang, JianLi and TakamaroKikkawa deliver.Select the UWB of 32nm specification to connect, data transfer rate is 20Gbps, and every bit power consumption is 2.7pJ.In this example, the total power consumption of each scheme is respectively: 0.2444, and 0.2445,0.2445,0.2446,0.2445,0.2445,0.2445,0.2445,0.2445,0.2445,0.2445,0.2445,0.2444,0.2444,0.2446,0.2445,0.2445,0.2444,0.2445,0.2445.

The 2-2 response time is the longest path time delay from being input to output, comprise all processing node task processing times in wireless input node to output node path and node and inter-node link transmission time and.In this example, the response time of each scheme is respectively: 0.1058, and 0.1067,0.1067,0.1072,0.1067,0.1066,0.1066,0.1064,0.1072,0.1062,0.1069,0.1067,0.1064,0.1064,0.1072,0.1069,0.1064,0.1061,0.1067,0.1069.

3. adopt the non-dominated ranking method of Deb (Science Press published in 2007, and author introduces in " multi-objective Evolutionary Algorithm and the application thereof " of Zheng Jinhua) to calculate scheme group in the fitness value of each scheme, be respectively: 2.0000,1.2632,1.0526,0.0526,1.1579,0.9474,0.6842,1.4211,0.2105,1.5789,0.6842,0.6842,1.7368,1.7368,0.0526,0.6842,1.4211,1.8947,0.3684,0.3684.

4. by scheme group in 20 schemes be divided into 10 scheme groups at random, 2 schemes are had in every scheme group, select the prepreerence scheme of every scheme group according to the dominance relation determined in the 3rd step, then the prepreerence scheme selected from all scheme groups is formed a new scheme group new scheme group in scheme be respectively: (11151071632), (12121014847), (9811105615), (48121511614), (841035107), (10241341016), (2147314716), (121551012137), (14111015131310), (15169104316).

5. by scheme group in scheme press random pair, described pairing refers to divides into groups according to two schemes one group, and with the Probability p of the intersection of setting in step 1 _task=0.4 exchanges the numbering of the node between often pair of scheme on random site, and then by all exchanged after often pair of scheme be combined and obtain scheme group scheme group in scheme be respectively: (11151071632), (12121014847), (9811105615), (48121511614), (8410341016), (1024135107), (2147314716), (121551012137), (14111015131310), (15169104316)

6. with the mutation probability q of setting in step 1 _task=0.1 changes scheme group in the numbering of node in each scheme on random site, obtain scheme group scheme group in scheme be respectively: (11151071632), (12121014847), (9811105615), (48121511614), (8410341016), (1024135107), (2143314716), (121551012137), (13111015131310), (15169104316)

7. press fitness order from small to large, from scheme group middle selection 10 schemes, with scheme group scheme combine, form and comprise the new scheme group of 20 schemes for (11151071632), (12121014847), (9811105615), (48121511614), (8410341016), (1024135107), (2143314716), (121551012137), (13111015131310), (15169104316), (12121014847), (9811105615), (810814101112), (48121511614), (7114101459), (841035107), (2147314716), (2132613144), (14111015131310), (11151071632).

8. by scheme group with be merged into the scheme group comprising 40 schemes i.e. (11151071632), (810814101112), (12121014847), (166255412), (9811105615), (54113841), (1316916493), (48121511614), (4987377), (841035107), (10241341016), (101112101213), (2147314716), (2132613144), (5471413152), (121551012137), (7114101459), (14111015131310), (15169104316), (71711698), (11151071632), (12121014847), (9811105615), (48121511614), (8410341016), (1024135107), (2143314716), (121551012137), (13111015131310), (15169104316), (12121014847), (9811105615), (810814101112), (48121511614), (7114101459), (841035107), (2147314716), (2132613144), (14111015131310), (11151071632).

9. numerical procedure group in wireless network-on-chip power consumption and response time under often kind of scheme, then judge whether the multicast task source and destination node of each scheme in scheme group is mapped on different wireless network-on-chip nodes, if so, then uses scheme group in the poorest power consumption number (0.2446) and the poorest response time value (0.1072) power consumption of the program and response time value are upgraded, total power consumption corresponding to each scheme respectively is:

0.2444，0.2445，0.2445，0.2446，0.2445，0.2445，0.2445，0.2445，0.2445，0.2445，0.2445，0.2445，0.2444，0.2444，0.2446，0.2445，0.2445，0.2444，0.2445，0.2445，0.2444，0.2446，0.2446，0.2446，0.2446，0.2446，0.2446，0.2446，0.2446，0.2446，0.2445，0.2445，0.2445，0.2445，0.2445，0.2445，0.2444，0.2444，0.2444，0.2444，。

The response time of each mapping scheme respectively is:

0.1058，0.1067，0.1067，0.1072，0.1067，0.1066，0.1066，0.1064，0.1072，0.1062，0.1069，0.1067，0.1064，0.1064，0.1072，0.1069，0.1064，0.1061，0.1067，0.1069，0.1058，0.1072，0.1072，0.1072，0.1072，0.1072，0.1072，0.1072，0.1072，0.1072，0.1067，0.1067，0.1067，0.1064，0.1064，0.1062，0.1064，0.1064，0.1061，0.1058。

10. the boundary set adopting Deb to introduce in the article AFastElitistNon-dominatedSortingGeneticAlgorithmforMulti-objectiveOptimization:NSGA-II to be published on magazine IEEETransactionsonEvolutionaryComputation the 6th volume for 2002 produces method by scheme group in 40 schemes be divided into 13 groups, each group represents a boundary set, and to little group #, the priority of the scheme of the group that the scheme in less group of numbering is larger than numbering is high; So-called priority level is little by total power consumption and response time is excellent.

The scheme of the group obtained as stated above 1 li has scheme 1, scheme 21, scheme 40; Scheme inside group 2 has scheme 18, scheme 39; The scheme that group is 3 li has scheme 13, scheme 14, scheme 37, scheme 38; Scheme inside group 4 has scheme 10, scheme 36; The scheme that group is 5 li has scheme 8, scheme 17, scheme 34, scheme 35; The scheme that group is 6 li has scheme 2, scheme 33; Scheme inside group 7 has scheme 5, scheme 32; Scheme 3 is had, scheme 31 inside group 8; The scheme that group is 9 li has scheme 6; Scheme inside group 10 has scheme 7, scheme 11, scheme 16; Scheme inside group 11 has scheme 19, scheme 20; Scheme 9 is had inside group 12; The scheme that group is 13 li has scheme 4, scheme 15, scheme 22, scheme 23, scheme 24, scheme 25, scheme 26, scheme 27, scheme 28, scheme 29, scheme 30.

The scheme group not having scheme that 11. initialization one are new then from the group that the 10th step has been divided, by the numbering order from small to large of group, from group 1 to group 13, successively the scheme in group is joined in, until scheme group in scheme number more than 20, now the scheme of the group finally added (i.e. group 8) is all taken out.

12. orders of priority determining each scheme in each group obtained in the 10th step, concrete grammar is as follows:

The size of the power consumption of each scheme that 12-2 calculates according to the 9th step carries out inverted order arrangement to scheme, and the scheme that power consumption is less comes before the larger scheme of power consumption,

12-4 carries out inverted order arrangement according to the response time size of each scheme that the 9th step calculates to scheme, and response time less scheme comes before response time larger scheme,

13. from the 12nd step sequence after group according to priority height selection scheme add scheme group until scheme group in scheme number be 20, after completing, scheme group in scheme be: (11151071632), (11151071632), (11151071632), (14111015131310), (14111015131310), (2147314716), (2132613144), (2147314716), (2132613144), (841035107), (841035107), (48121511614), (7114101459), (48121511614), (7114101459), (810814101112), (810814101112), (9811105615), (9811105615), (12121014847).

14. by scheme group in 20 schemes be divided into 10 scheme groups at random, 2 schemes are had in every scheme group, the dominance relation determined according to the 10th and the 12nd step selects the prepreerence scheme of every scheme group, then the prepreerence scheme selected from all scheme groups is formed a new scheme group scheme group in scheme comprise (11151071632), (11151071632), (14111015131310), (2132613144), (2132613144), (841035107), (7114101459), (7114101459), (810814101112), (9811105615)

15. by scheme group in scheme random pair again, with the Probability p of intersection of setting in step 1 _task=0.4 exchanges the numbering of the node between often pair of scheme on random site, and then all being combined is obtained scheme group i.e.: (11151071632), (11151071632), (14112613144), (2131015131310), (2132613144), (841035107), (7114101459), (7114101459), (810814101112), (9811105615)

16. with the probability q of the variation of setting in step 1 _task=0.1 changes scheme group in the numbering of node in each scheme on random site, obtain scheme group i.e.: (11151071632), (11151071632), (14112612144), (2121015131310), (2132613144), (841034107), (7114101459), (7114101459), (810814101112), (9911105715).

17. press fitness order from small to large, from scheme group middle selection 10 schemes, with scheme group scheme combine, form and comprise the new scheme group of 20 schemes scheme is (11151071632), (11151071632), (14112612144), (2121015131310), (2132613144), (841034107), (7114101459), (7114101459), (810814101112), (9911105715), (48121511614), (7114101459), (810814101112), (810814101112), (9811105615), (9811105615), (12121014847), (11151071632), (11151071632), (11151071632).

18. if iterations t _task<100, then t _task=t _task+ 1, then turn back to step 2, carry out new round interative computation; Otherwise enter step 19;

19. adopt the non-dominated ranking method of Deb (Science Press published in 2007, and author introduces in " multi-objective Evolutionary Algorithm and the application thereof " of Zheng Jinhua) to calculate scheme group in the fitness value of each scheme, by scheme minimum for fitness value, i.e. total power consumption and response time simultaneously minimum scheme J _min(11,15,10,7,16,3,2) as the task of network-on-chip to internodal duty mapping scheme.

The step of coding mapping is as follows:

I. according to the scheme J that duty mapping obtains _min, determine multicast task originating task and the object task mapping node position X at wireless network-on-chip _source(15,10,7), object duty mapping position X _des(16,3,2).According to relation between the multicast task in Fig. 1, reject isolated multicast source node 7 and the destination node 2 of its correspondence, the remaining multicast task source and destination node set with glue connection relation, for: X _source'=(10,15) and X _des'=(3,16).

Ii. X is calculated _source' in each source node to X _des' in all paths of each destination node, account form is as follows:

Ii-1X _source' there are 2 source nodes, X _des' there are 2 destination nodes.b ₀=0。Make label x=1, y=1.To 4 dimension group path_ID=[b ₁, b ₂... b ₄] assignment, method is as follows:

(1) p=1 is made, q=1;

(2) X is selected _source' in the 1st source node (node serial number is 10) and X _des' the 1st destination node (node serial number is 3), obtain its coordinate by node serial number and be respectively (1,2) and (2,0).

(3) number in the shortest route path calculated from this source node to this destination node is Max_path_num=3, and by these path number consecutivelies, represent path number value with e, each paths is expressed as successively make c=1, b ₁=1.Then p=p+1;

(4) if p>j, then, q=q+1, enters step (5); Otherwise, get back to step (2);

(5) if q>m, then complete assignment, enter step I i-2; Otherwise make p=1, get back to step (2).

The full remainder group total_path of ii-2 initialization 8 row 8 row;

Ii-3 selects X _source' in the 1st source node (node 10) and X _des' the 1st destination node (node 3), make k=1.

Ii-4 is capable to h'=3 from the h=1 of total_path is capable, records current source to each node serial number in current destination node altogether 3 paths.Initialization g=1, g'=1, e=1.Detailed step is as follows:

I. the 1st paths is selected, node number on this paths is 4, and source node (node 10) is the node of the 1st on this path, and node 11 is the 2nd node on this path, node 7 is the 3rd node on this path, and destination node (node 3) is the node of the 4th on this path.

II. by numbering (10) assignment of the node of the 1st in current path to the element on the 1st row the 1st column position of array total_path.Then g'=g'+1, g=g+1;

If III. g'≤4, then return Step II, otherwise enter step IV;

IV.total_path line number h=h+1, path number e=e+1, enter step V;

If V. e≤3, then enter step I; Otherwise enter step I i-5;

Ii-5 now, from X _source' in the 1st source node (node 10) to X _des' all shortest route paths of the 1st destination node (node 3) all recorded, be [10,11,7,3,0,0,0,0].Label x=x+1.Enter step I i-6;

If ii-6 is x>2, then enter ii-7.Otherwise return ii-3; Ii-7 label y=y+1, if y<=2, then makes x=1, returns ii-3; Otherwise complete the assignment to total_path, obtaining matrix total_path is:

total_path = [\begin{matrix} 10 & 11 & 7 & 3 & 0 & 0 & 0 & 0 \\ 10 & 6 & 7 & 3 & 0 & 0 & 0 & 0 \\ 10 & 6 & 2 & 3 & 0 & 0 & 0 & 0 \\ 15 & 11 & 7 & 3 & 0 & 0 & 0 & 0 \\ 10 & 11 & 12 & 16 & 0 & 0 & 0 & 0 \\ 10 & 11 & 15 & 16 & 0 & 0 & 0 & 0 \\ 10 & 14 & 15 & 16 & 0 & 0 & 0 & 0 \\ 15 & 16 & 0 & 0 & 0 & 0 & 0 & 0 \end{matrix}]

Path_ID is [3131].

Iii. stochastic generation one comprises the scheme group of 20 allocative decisions be respectively: (1321), (1131), (3131), (3121), (1111), (2111), (3121), (2131), (3111), (1111), (2121), (2121), (2121), (2111), (2131), (3111), (2131), (2121), (2131), (3111).For each coding path design of scheme crossover probability p _coding=0.5 and mutation probability q _coding=0.1, and coding mapping iterations T _coding=100, and the variable t of the number of times of order record iteration _coding=0;

Iv. scheme group is judged in combination of paths under often kind of scheme whether meet the condition that multicast multiplexed transport can reach maximum multicast rate, if can not reach, then its throughput is set to 0, power consumption is set to 10, and obtains scheme group in under all the other various schemes the throughput of multicast task be: 0,0,0,0,0,0,0,0,26.7,0,0,0,0,0,0,26.7,0,0,0,26.7.

Power consumption is: 10,10,10,10,10,10,10,10,0.0001,10,10,10,10,10,10,0.0001,10,10,10,0.0001.

V. adopt the non-dominated ranking method of Deb, calculate scheme group in the fitness value of each scheme be: 0.8421,0.8421,0.8421,0.8421,0.8421,0.8421,0.8421,0.8421,1.8947,0.8421,0.8421,0.8421,0.8421,0.8421,0.8421,1.8947,0.8421,0.8421,0.8421,1.8947.

Vi. by scheme group in 20 schemes be divided into 10 scheme groups at random, wherein have 2 schemes in every scheme group.Select prepreerence scheme in every scheme group according to the dominance relation determined in v step and form a priority scheme group for: (1321), (3131), (1111), (3121), (3111), (2121), (2121), (3111), (2131), (3111).

Vii. by priority scheme group in scheme press random pair, described pairing refers to divides into groups according to two schemes one group, and with the Probability p of the intersection set in step I ii _codingthe numbering of the node in any one same position in the scheme of two in often pair of scheme is exchanged by=0.5, and then all to be combined by often pair of scheme after exchanging is obtained scheme group for: (1131), (3321), (1111), (3121), (3121), (2111), (2111), (3121), (2131), (3111).

Vii. with the mutation probability q set in step I _coding=0.1 changes scheme group in the numbering of node in each scheme on any one position, obtain scheme group for: (1131), (3121), (1121), (3121), (3121), (2111), (2111), (3121), (3131), (3111).

Ix. the fitness value order from small to large calculated is walked by v, from scheme group middle selection 10 schemes, with scheme group scheme combine, form and comprise the new scheme group of 20 schemes for: (1131), (3121), (1121), (3121), (3121), (2111), (2111), (3121), (3131), (3111), (2121), (2121), (2111), (2131), (2131), (2121), (2131), (3111), (3111), (3111).

X. by scheme group with be merged into scheme group for: (1321), (1131), (3131), (3121), (1111), (2111), (3121), (2131), (3111), (1111), (2121), (2121), (2121), (2111), (2131), (3111), (2131), (2121), (2131), (3111), (1131), (3121), (1121), (3121), (3121), (2111), (2111), (3121), (3131), (2121), (2121), (2111), (2131), (2131), (2121), (2131), (0000), (0000), (0000), (0000).

Xi. scheme group is judged in combination of paths under often kind of scheme whether meet the condition that multicast multiplexed transport can reach maximum multicast rate, if can not reach, then its throughput is set to 0, power consumption is set to 10, and obtains scheme group in under all the other various schemes the throughput of multicast task be respectively 0,0,0,0,0,0,0,0,26.7,0,0,0,0,0,0,26.7,0,0,0,26.7,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0.

Network power consumption is respectively: 10, and 10,10,10,10,10,10,10,0.0001,10,10,10,10,10,10,0.0001,10,10,10,0.0001,10,10,10,10,10,10,10,10,10,10,10,10,10,10,10,10,10,10,10,10;

Xii. the boundary set adopting Deb to introduce in the article AFastElitistNon-dominatedSortingGeneticAlgorithmforMulti-objectiveOptimization:NSGA-II to be published on magazine IEEETransactionsonEvolutionaryComputation the 6th volume for 2002 produces method by scheme group in 40 schemes be divided into 2 groups, each group represents a boundary set, and to little group #, the priority of the scheme of the group that the scheme in less group of numbering is larger than numbering is high; So-called priority level be by throughput large little with network power consumption be excellent.The scheme of the group obtained as stated above 1 li has scheme 9, scheme 16, scheme 20; Scheme in remaining scheme Dou Shi group 2.

The scheme group of scheme that what Xiii initialization one was new do not have then from the group that xth ii step has been divided, by the numbering of group from the order of 1 to 2, successively the scheme in group is joined in, until scheme group in scheme number more than 20, then the scheme of the group finally added (group 2) is all taken out.

Xiv-1 makes the distance value of each scheme in group be 0, and described distance value represents whether contacting of the program and other scheme be tight.

Xiv-2 walks the network power consumption of each scheme calculated size according to xth i carries out inverted order arrangement to scheme, and the scheme that power consumption is less comes before the larger scheme of power consumption,

Xiv-3 calculates the network power consumption distance of each scheme, the value that the power consumption distance of each scheme obtains for power consumption that the power consumption coming that scheme after it deducts that scheme come before it,

Xiv-4 walks each scheme calculated throughput size according to xth i arranges scheme, and the scheme that throughput is larger comes before the less scheme of throughput,

Xiv-5 calculates the throughput distance of each scheme, the value that the throughput distance of each scheme obtains for throughput that the throughput coming that scheme after it deducts that scheme come before it,

The throughput Distance geometry power consumption of each scheme distance is added the distance value of the scheme that obtains by xiv-6,

Xiv-7 comes foremost scheme maximum for throughput and the minimum scheme of network power consumption, then according to distance size, remaining scheme is sorted, and come after the maximum scheme of throughput and the minimum scheme of network power consumption, the scheme that distance value is larger comes before the less scheme of distance value, and the priority that the priority ratio coming scheme above like this comes scheme is below high.

Xv. scheme group is added from have passed through in the group that takes out xth iii step according to priority height selection scheme until scheme group in scheme number be 20. for: (3111), (3111), (3111), (1321), (2131), (1131), (3131), (3121), (1111), (2111), (3121), (2131), (1111), (2121), (2121), (2121), (2111), (2131), (2131), (2121).

Xvi. by scheme group in 20 schemes be divided into 10 scheme groups at random, have 2 schemes in every scheme group, walk according to xth ii the dominance relation determined and select prepreerence scheme in every scheme group and form a priority scheme group V _t'.V _t' be (3111), (3111), (2131), (3131), (1111), (3121), (1111), (2121), (2111), (2131).

Xvii. by priority scheme group V _t' in scheme random pair again, with the Probability p of the intersection set in step I ii _codingthe numbering of the node in any one same position in the scheme of two in often pair of scheme is exchanged by=0.5, and then all to be combined by often pair of scheme after exchanging is obtained scheme group for: (3111), (3111), (2131), (3131), (1111), (3121), (1121), (2111), (2131), (2111).

Xviii. with the probability q of the variation set in step I ii _coding=0.1 changes scheme group in the numbering of node in each scheme on any one position, obtain scheme group for: (3111), (3111), (2131), (3131), (1111), (3121), (1121), (2111), (2131), (2111).

Xix. the scheme group determined is walked according to xth ii the order of priority of middle scheme, according to priority order from high to low, from scheme group middle selection 10 schemes, with scheme group scheme combine, form and comprise the new scheme group of 20 schemes for (3111), (3111), (2131), (3131), (1111), (3121), (1121), (2111), (2131), (2111), (2121), (2121), (2121), (2111), (2131), (2131), (2121), (3111), (3111), (3111).

Xxi. scheme group is calculated according to the method described in step I v in the fitness value of each scheme, by scheme minimum for fitness value, namely scheme is as a mapping scheme of wireless network-on-chip coding mapping.Scheme is (3111).

Xxii. judge whether the coding mapping scheme throughput now obtained is not 0, and whether power consumption is 10, and because scheme throughput is 26.7 for this reason, power consumption is 0.0001, therefore records this mapping scheme, Z=Z+1, then enters step xxiii simultaneously.Otherwise, return the step 2 of duty mapping part.

If xxiii. Z≤2, then return the step 2 of duty mapping part, otherwise, enter step xxiv.

Xxiv. 2 mapping schemes are obtained.Scheme 1 multicast task throughput is 26.7, and power consumption is 0.0001.Scheme 2 multicast task throughput is 20, and power consumption is 0.0006.Adopt non-dominated ranking method (Science Press's publication in 2007 of Deb, introduce in " multi-objective Evolutionary Algorithm and application thereof " that author is Zheng Jinhua) obtain being respectively 1 and 0 based on the fitness of multicast task throughput and multicast multiplexed transport power consumption in 2 mapping schemes, select the maximum scheme of fitness, i.e. the scheme that multicast task throughput is maximum and multicast multiplexed transport power consumption is minimum is as finally mapping selection scheme.The scheme finally obtained is scheme 1: duty mapping scheme is (11151071632), and coding mapping scheme is (3111), and the path namely selected is

[\begin{matrix} 10 & 6 & 2 & 3 \\ 15 & 11 & 7 & 3 \\ 10 & 11 & 12 & 16 \\ 15 & 16 & 0 & 0 \end{matrix}]

Corresponding encoded point is routing node 11, namely corresponding processing node PE11.

By reference to the accompanying drawings the specific embodiment of the present invention is described in detail above, but the present invention is not restricted to above-mentioned execution mode, in the spirit and scope situation of claim not departing from the application, those skilled in the art can make various amendment or remodeling.

Claims

1. find mapping scheme and a network code location method between network-on-chip task and node, it is characterized in that, comprise the steps:

2). obtain scheme group J _tin wireless network-on-chip clean culture task power consumption and response time under often kind of scheme, and record maximum power dissipation and maximum response time is the poorest power consumption number and the poorest response time value;

3). calculate scheme group J _tin the fitness value of each scheme, the priority of this fitness value size reflection scheme, it is higher be worth larger priority, so-called priority level be by clean culture task total power consumption and the response time little be excellent;

4). by scheme group J _tin K scheme be divided at random individual scheme group, the scheme number wherein in L Shi Mei scheme group, selects prepreerence scheme in every scheme group according to the dominance relation determined in the 3rd step and forms a priority scheme group

7). the fitness value order from small to large calculated by the 3rd step, from scheme group J _tmiddle selection individual scheme, with scheme group scheme combine, form and comprise the new scheme group of K scheme

8). by scheme group with be merged into scheme group

9). numerical procedure group in wireless network-on-chip clean culture total power consumption and response time under often kind of scheme, during calculating, first judge whether the multicast task source and destination node of each scheme in scheme group is mapped on different wireless network-on-chip nodes, if, then direct by the poorest power consumption number and the poorest corresponding time value assignment to the power consumption of this programme and response time value, if not, then numerical procedure group in wireless network-on-chip clean culture total power consumption and response time under often kind of scheme;

10). by scheme group R _tin all 2K scheme be divided into multiple group, each group represents a boundary set, and to little group #, the priority of the scheme of the group that the scheme in less group of numbering is larger than numbering is high; So-called priority level is little by power consumption and response time is excellent;

11). the scheme group of scheme that what initialization one was new do not have then from the group that the 10th step has been divided, by the numbering order from big to small of group, successively the scheme in group is joined in, until scheme group in scheme number individual more than K, then the scheme of the group finally added all is taken out;

12-1 makes the distance value of each scheme in group be 0, and described distance value represents whether contacting of the program and other scheme be tight;

12-2 is according to the 9th) size of the power consumption of each scheme that calculates of step carries out inverted order arrangement to scheme, and the scheme that power consumption is less comes before the larger scheme of power consumption;

12-3 calculates the power consumption distance of each scheme, the value that the power consumption distance of each scheme obtains for power consumption that the power consumption coming that scheme after it deducts that scheme come before it;

12-4 is according to the 9th) the response time size of each scheme that calculates of step carries out inverted order arrangement to scheme, and response time less scheme comes before response time larger scheme;

12-5 calculates the response time distance of each scheme, the value that the response time distance of each scheme obtained for response time that the response time coming that scheme after it deducts that scheme come before it;

The response time Distance geometry power consumption distance of each scheme is added the distance value of the scheme that obtains by 12-6;

12-7 comes foremost scheme minimum for power consumption and response time minimum scheme, then according to distance size, remaining scheme is sorted, and come after the minimum scheme of power consumption and response time minimum scheme, the scheme that distance value is larger comes before the less scheme of distance value, and the priority that the priority ratio coming scheme above like this comes scheme is below high;

13). from the 12nd) according to priority height selection scheme adds scheme group in the group that takes out step until scheme group in scheme number be K;

18. if t _task< T _task, then t _task=t _task+ 1, then turn back to step 2), continue iteration; Otherwise enter step 19);

19). calculate scheme group in the fitness value of each scheme, by scheme maximum for fitness value, i.e. power consumption and response time simultaneously minimum scheme J _minas the task of wireless network-on-chip to internodal duty mapping scheme;

Wherein, K, L, be positive integer.

2. mapping scheme and network code location method between searching network-on-chip task according to claim 1 and node, it is characterized in that, the 2nd) calculate often kind of scheme in step under wireless network-on-chip clean culture task power consumption and the method for response time as follows: the wireless network-on-chip clean culture task power consumption described in duty mapping is that the power consumption of all processing nodes and internodal data transmit power consumption sum; Response time described in duty mapping is the longest path time delay from being input to output, comprises the transmission time of exchanges data between the process time delay of all tasks in path and task.

3. mapping scheme and network code location method between searching network-on-chip task according to claim 1 and 2 and node, is characterized in that, also comprise the steps:

I. according to the scheme J that duty mapping obtains _min, determine that multicast task originating task and object task are in the mapping node position of wireless network-on-chip with according to relation between the multicast task in Flow chart task, from X _sourcereject the node of the wireless network-on-chip that isolated multicast task source node maps, remaining there is glue connection relation multicast task originating task and the node set that maps of object task, use vector respectively with represent;

V. scheme group is calculated in the fitness value of each scheme, the priority of this fitness value size reflection scheme, it is higher be worth larger priority, so-called priority level be large by multicast multiplexed transport throughput and network power consumption little be excellent;

X. by scheme group with be merged into scheme group

Xii. by scheme group in 2P scheme be divided into multiple group, each group represents a boundary set, and to little group #, the priority of the scheme of the group that the scheme in less group of numbering is larger than numbering is high; So-called priority level be by throughput large little with network power consumption be excellent;

Xiii. the scheme group not having scheme that initialization one is new then from the group that xth ii step has been divided, by the numbering order from big to small of group, successively the scheme in group is joined in, until scheme group in scheme number individual more than P, then the scheme of the group finally added all is taken out;

Xiv-7 comes foremost scheme maximum for throughput and the minimum scheme of network power consumption, then according to distance size, remaining scheme is sorted, and come after the maximum scheme of throughput and the minimum scheme of network power consumption, the scheme that distance value is larger comes before the less scheme of distance value, and the priority that the priority ratio coming scheme above like this comes scheme is below high;

If xx. t _coding< T _coding, then t _coding=t _coding+ 1, then turn back to step I v, otherwise enter step xxi;

Xxi. scheme group is calculated in the fitness value of each scheme, by scheme maximum for fitness value, the scheme that namely Internet Transmission power consumption is minimum and throughput is simultaneously maximum is as a mapping scheme of wireless network-on-chip coding mapping;

Xxii. judge whether the coding mapping scheme throughput now obtained is the minimum that the i-th v step specifies, and whether power consumption is the maximum that the i-th v step specifies, if throughput is not minimum and power consumption is not maximum, then record this mapping scheme, Z=Z+1, then enters step xxiii simultaneously; Otherwise, return step 2);

If xxiii. Z≤Z _max, then step 2 is returned), otherwise, enter step xxiv;

Xxiv. Z is calculated _maxthe fitness value of individual mapping scheme, selects the scheme that fitness value is maximum, i.e. the scheme that multicast task throughput is maximum and multicast multiplexed transport power consumption is minimum is as finally mapping selection scheme, and record coding maps the numbering of path and the encoded point selected;

Wherein, P, N, be positive integer.

4. mapping scheme and network code location method between searching network-on-chip task according to claim 3 and node, is characterized in that, in above-mentioned i-th v step, solves scheme group in under often kind of scheme the throughput of multicast task and the method for Internet Transmission power consumption as follows:

When calculating throughput and the Internet Transmission power consumption of the multicast task under often kind of scheme, first judge whether the combination of paths of this scheme meets the condition that multicast multiplexed transport can reach maximum multicast rate; If meet the condition that can reach maximum multicast rate, obtain the encoded point of this combination of paths, and calculate throughput and the Internet Transmission power consumption of the multicast task of this mapping scheme; If can not reach the condition of maximum multicast rate, then the throughput of this mapping scheme multicast task is a minimum, and the network power consumption of this mapping scheme multicast multiplexed transport is a maximum; Described Internet Transmission power consumption refers to the power consumption that wireless routing node in network carries out transfer of data and brings comprise the arbitration power consumption of node, coding power consumption and wireless channel transmission power consumption;

Described multicast multiplexed transport throughput is calculated as follows:

5. mapping scheme and network code location method between searching network-on-chip task according to claim 4 and node, it is characterized in that, in above-mentioned i-th i step, the Mesh topological structure that the computational methods of total_path adopt with wireless NoC and routing algorithm associated, concrete steps are:

(1) p=1 is made, q=1;

(2) X is selected _source' in p source node, make its node serial number be n _s, and X _des' q destination node, make its node serial number be n _d, obtaining their coordinates in Mesh topology according to their node serial number is (x _s, y _s) and (x _d, y _d); Wherein, floor () is downward bracket function;

(3) calculate from node n _sto node n _dthe number in shortest route path be by these path number consecutivelies, represent path number value with e, each paths is expressed as successively make c=(q-1) j+p, b _c=Max_path_num; Then p=p+1;

(4) if p>j, then, q=q+1, enters step (5); Otherwise, get back to step (2);

Ii-2 initialization oK the full remainder group total_path of row;

If III. g'≤r, then return Step II, otherwise enter step IV;

IV.total_path line number h=h+1, path number e=e+1, enter step V;

If V. e≤b _k, then step I is entered; Otherwise enter step I i-5; ;

If ii-6 is x>j, then enter ii-7, otherwise return ii-3;