WO2013011653A1 - 過負荷を回避する超低消費電力化データ駆動ネットワーキング処理装置 - Google Patents
過負荷を回避する超低消費電力化データ駆動ネットワーキング処理装置 Download PDFInfo
- Publication number
- WO2013011653A1 WO2013011653A1 PCT/JP2012/004420 JP2012004420W WO2013011653A1 WO 2013011653 A1 WO2013011653 A1 WO 2013011653A1 JP 2012004420 W JP2012004420 W JP 2012004420W WO 2013011653 A1 WO2013011653 A1 WO 2013011653A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- data
- unit
- transmitted
- driven processor
- packet
- Prior art date
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L47/00—Traffic control in data switching networks
- H04L47/70—Admission control; Resource allocation
- H04L47/74—Admission control; Resource allocation measures in reaction to resource unavailability
- H04L47/745—Reaction in network
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F1/00—Details not covered by groups G06F3/00 - G06F13/00 and G06F21/00
- G06F1/26—Power supply means, e.g. regulation thereof
- G06F1/28—Supervision thereof, e.g. detecting power-supply failure by out of limits supervision
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F1/00—Details not covered by groups G06F3/00 - G06F13/00 and G06F21/00
- G06F1/26—Power supply means, e.g. regulation thereof
- G06F1/32—Means for saving power
- G06F1/3203—Power management, i.e. event-based initiation of a power-saving mode
- G06F1/3234—Power saving characterised by the action undertaken
- G06F1/3296—Power saving characterised by the action undertaken by lowering the supply or operating voltage
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L47/00—Traffic control in data switching networks
- H04L47/10—Flow control; Congestion control
- H04L47/12—Avoiding congestion; Recovering from congestion
- H04L47/122—Avoiding congestion; Recovering from congestion by diverting traffic away from congested entities
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/50—Reducing energy consumption in communication networks in wire-line communication networks, e.g. low power modes or reduced link rate
Definitions
- the present invention relates to an ultra-low power consumption data-driven networking processing apparatus that avoids overload using a data-driven processor.
- a data driven processor capable of executing parallel processing without overhead has been developed.
- the data-driven processor is designed based on a data-driven principle that enables execution of all instructions in a program when all necessary input data is available.
- techniques using such a data driving principle techniques described in Patent Documents 1 and 2 below are conventionally known.
- Patent Document 1 Japanese Patent Laid-Open No. 2004-13602 describes a configuration in which a data driven processor has a plurality of processing elements (PE), and a packet in which data is divided is input to each processing element (PE).
- the processing element (PE) includes a firing control unit (FC), a function processing unit (FP), a program storage unit (PS), and the like, and firing control is performed based on an input packet. It is determined whether or not all packets necessary for instruction execution are prepared in the unit (FC: Firing Control).
- the packet is transmitted from the firing control unit (FC) to the function processing unit (FP: Functional Processor), the calculation is performed, and the packet having the result is stored in the program memory.
- PS Program Storage
- the program storage unit (PS) generates a packet according to the new command in accordance with the transmitted packet and a new command specified in the command memory (IM), and sends another packet as the specified destination.
- the packet is transmitted to the processing element (PE).
- Patent Document 1 describes that each processing element (PE) has a super pipeline structure using a self-synchronous elastic pipeline. That is, each functional block such as an ignition control unit (FC), a function processing unit (FP), and a program storage unit (PS) in the processing element (PE) is configured by a plurality of pipeline stages. Each pipeline stage has a data latch (DL: Data Latch) that holds the packet, a logic circuit (LC: Logic Circuit) that processes the held packet, and a synchronization signal (clock signal, trigger) to the data latch.
- DL Data Latch
- LC Logic Circuit
- STCM Self-timed Transfer Control Mechanism
- Patent Document 2 Japanese Patent Laid-Open No. 2005-108086 describes a technique related to a data driven processor.
- PE processing element
- SW1 first switch
- an instruction is issued based on a data dependency relationship and processed in parallel, and based on a program counter
- a configuration is described in which control drive packets that are sequentially issued and sequentially processed can be processed on the same pipeline.
- Each processing element (PE) described in Patent Document 2 decodes an instruction fetch unit (IF: Instruction Fetch) that fetches an instruction based on an input packet and an instruction fetched and issued by the instruction fetch unit (IF).
- the firing condition is determined based on the instruction decode unit (ID: Instruction Decode), the second switch (SW2) for branching the instruction output from the instruction decode unit (ID), and the packet transmitted from the second switch (SW2).
- execution unit (EX: Execution) to be executed and the instruction processed by the execution unit (EX) are control drive instructions such as branch calculation And a write back unit (WB: Write Back) for executing writing to the register (REG).
- Patent Document 3 Japanese Patent Laid-Open No. 2010-20598
- an autonomous distributed communication network (ad hoc network) is used, and a data driven processor is used.
- a data driven processor In the pipeline stage, if processing is not performed in the downstream pipeline stage and data latch (data retention) is not performed, power supply is stopped and processing is performed in the downstream pipeline stage.
- Patent Document 4 Japanese Patent Application Laid-Open No. 2011-30210 discloses a transmission function for realizing a flooding function in which each node constituting a network grasps the load state of the node and efficiently executes data retransmission. Count the number of packets waiting for transmission corresponding to the queue, compare the number of packets waiting for transmission with a threshold, set the probability of resending the message, and set the waiting time to wait for retransmission based on the comparison result. A technique for setting and performing retransmission processing based on the set probability and standby time is described.
- JP 2004-13602 ("0019” to “0035") Japanese Patent Laying-Open No. 2005-108086 (“0025” to “0057”) JP 2010-20598 A (“0025” to “0034”) JP 2011-30210 A (“0002" to "0011", abstract)
- Patent Documents 1 and 2 and other conventional inventions research and development have been conducted with a focus on speeding up information processing, but research based on the viewpoint of low power consumption. Development was inadequate.
- a processor mounted on a terminal is provided with a low power consumption standby state. When a packet is not transmitted or received, the standby state is set. Further, power supply to a circuit block that is not used is stopped. It was the extent which aimed at reduction of. However, since the return from the standby state to the operation state is performed by an interrupt process, overhead for recovery occurs, and the effect of reducing power consumption is limited when switching between the standby state and the operation state occurs frequently.
- an ad hoc network capable of transmitting and receiving data between terminals driven by a simple battery and to function as a communication network by transmitting / receiving, relaying, and transferring (retransmitting) necessary information.
- information such as earthquake early warnings and tsunami information can be received by as many terminals as possible even when the communication infrastructure is not functioning.
- the power consumption is large, the battery of the terminal increases, and if the battery runs out, the number of nodes that make up the ad hoc network decreases and the number of terminals to which information is transmitted is limited. Less is desirable.
- the present invention has a technical problem of reducing power consumption in a networking system.
- an ultra-low power consumption data driven networking processing apparatus for avoiding overload of the invention of claim 1 is provided.
- An ignition control unit that determines whether or not an ignition condition is satisfied by determining whether or not all packets necessary for instruction execution have been received for a packet into which data to be processed is divided, and the ignition condition is
- a data-driven processor having at least a data processing unit that executes processing according to the packet transmitted from the firing control unit when established,
- Load avoidance means It is provided with.
- an ultra-low power consumption data driven networking processing apparatus for avoiding an overload according to the first aspect.
- Output control means for stopping the output of the retransmission packet when it is determined that the data processing load in the data driven processor is higher than a preset high load state based on the current consumption in the data driven processor When, It is provided with.
- an ultra-low power consumption data driven networking processing apparatus for avoiding an overload according to the first or second aspect.
- Voltage setting means for setting a voltage to be applied to the data driven processor based on a current consumption in the data driven processor to a voltage corresponding to a data processing load in the data driven processor;
- Power control means for controlling the power supply circuit to control the voltage applied to the data driven processor to a voltage set by the voltage setting means; It is provided with.
- an ultra-low power consumption data driven networking processing apparatus for avoiding an overload according to any one of the first to third aspects.
- a first merging unit in which a packet input to the data driven processor and data processed by the data processing unit merge;
- the firing control unit for receiving a packet or data transmitted from the first joining unit;
- a second merging unit where data transmitted from the ignition control unit and data processed by the data processing unit merge;
- An instruction fetch unit that fetches an instruction based on data transmitted from the second merging unit;
- An instruction decode unit for analyzing an instruction transmitted from the instruction fetch unit;
- the data processing unit for processing the instruction analyzed by the instruction decoding unit;
- a memory access unit for writing data to the memory processed by the data processing unit and reading from the memory; Based on the data written by the memory access unit, when the data is a unary operation and needs to be processed again by the data processing unit, the data is transmitted to the second merging unit, and the data is A second diversion unit that transmits to the first diversion unit on the downstream side when it is
- An ultra-low power consumption data driven networking processing apparatus for avoiding overload of the invention according to claim 5 is provided.
- An ignition control unit that determines whether or not an ignition condition is satisfied by determining whether or not all packets necessary for instruction execution have been received for a packet into which data to be processed is divided, and the ignition condition is
- a data-driven processor having at least a data processing unit that executes processing according to the packet transmitted from the firing control unit when established,
- a power supply circuit for supplying power to the data driven processor;
- Voltage setting means for setting a voltage to be applied to the data driven processor based on a current consumption in the data driven processor to a voltage corresponding to a data processing load in the data driven processor;
- Power control means for controlling the power supply circuit to control the voltage applied to the data driven processor to a voltage set by the voltage setting means; It is provided with.
- an ultra-low power consumption data driven networking processing apparatus for avoiding overload of the invention according to claim 6 is provided:
- An ignition control unit that determines whether or not an ignition condition is satisfied by determining whether or not all packets into which data to be processed are divided are received; and from the ignition control unit when the ignition condition is satisfied
- a data processing processor having at least a data processing unit that executes processing according to the transmitted data;
- a first merging unit where a packet input to the data-driven processor and data processed by the data processing unit merge; and a firing control unit that receives a packet or data transmitted from the first merging unit
- a second merging unit where the data transmitted from the firing control unit and the data processed by the data processing unit merge, and an instruction to fetch an instruction based on the data transmitted from the second merging unit
- a fetch unit an instruction decode unit for analyzing an instruction transmitted from the instruction fetch unit; the data processing unit for processing an instruction analyzed by the instruction decode unit; and data processed by the data processing unit
- a memory access unit for writing to and reading from the memory
- the data is transmitted to the second merging unit, and when the data is a polynomial operation and the data needs to be processed by the data processing unit.
- the data processing unit needs to process the data again by the data processing unit based on the second branching unit transmitted to the first branching unit on the downstream side and the data transmitted from the second branching unit.
- the data is data
- the first data is transmitted to the first merge unit, and the data driven processor outputs the data when the data is not required to be processed by the data processing unit.
- a pipeline stage having a diversion section; It is provided with.
- overload can be avoided and power consumption in the networking system can be reduced compared to a configuration in which a packet is accepted even in the case of overload.
- the second aspect of the present invention it is possible to suppress the traffic of the entire network even when the load is high, as compared with the case where the packet is retransmitted, and to reduce the power consumption by efficient processing.
- the processing time can be shortened in the case of a single term operation as compared with the configuration without the second merging portion and the second branching portion, and the processing time can be reduced conventionally.
- the applied voltage can be reduced while maintaining or improving the power consumption of the networking system to reduce power consumption in the networking system.
- FIG. 1 is an overall explanatory diagram of Embodiment 1 of a networking system according to the present invention.
- FIG. 2 is a functional explanatory diagram of the terminal of the first embodiment, so-called functional block diagram.
- FIG. 3 is a functional block diagram of the chip multiprocessor according to the first embodiment.
- FIG. 4 is an explanatory diagram of the architecture of the data driven processor according to the first embodiment, and is a functional block diagram.
- FIG. 5 is an explanatory diagram of the implementation of the data driven processor of the first embodiment in a circuit, and is an explanatory diagram of a self-synchronous elastic pipeline.
- FIG. 6 is an explanatory diagram of an example of a voltage setting table according to the first embodiment.
- FIG. 1 is an overall explanatory diagram of Embodiment 1 of a networking system according to the present invention.
- the ultra-low power consumption networking system S according to the first embodiment of the present invention includes a plurality of terminals T1 to T4.
- an autonomous decentralized communication network formed by wireless communication of terminals T1 to T4 so-called an ad hoc network N, is adopted, and as an example of data Audio data and video data (for example, live broadcast video of a disaster site) can be streamed.
- Each of the terminals T1 to T4 of the first embodiment includes an information processing apparatus (so-called PC: personal computer, notebook PC or desktop PC) having an ad hoc network communication function, or a portable information terminal (so-called PDA: Personal Digital Assistant). ), A portable music player, a cellular phone, or a portable game machine having a communication function.
- PC personal computer, notebook PC or desktop PC
- PDA Personal Digital Assistant
- FIG. 2 is a functional explanatory diagram of the terminal of the first embodiment, so-called functional block diagram.
- the mobile phone 11 as an example of the ultra-low power consumption data-driven networking processing device that is an example of the terminals T1 to T4 of the first embodiment and that avoids overloading is connected to other terminals T1 to T4
- a communication unit 12 for transmitting / receiving data between them, an offloading interface 13 for transmitting / receiving a packet in which data is divided by being connected to the communication unit 12, and a mobile phone electrically connected to the offloading interface 13 11 is connected to the application processor 14 for controlling the camera built in 11 and the display screen 11a, processing for input from the input key 11b, etc., and the offloading interface 13, and performs data processing such as communication processing.
- control signals are input to the communication unit 12 from an acceptance control unit 18 as an example of an overload avoidance unit and a retransmission control unit 19 as an example of an output control unit.
- the chip multiprocessor 16 is composed of an ultra-low power consumption data driven chip multiprocessor, ULP-DDCMP (Ultra Low Power-Data Drive Chip Multiprocessor).
- UDP / IP User Datagram Protocol / Internet Protocol
- processing as an example of a communication protocol is performed.
- the present invention is not limited to this, and a conventionally known communication protocol (for example, TCP / IP: Transmission Control Protocol / Internet Protocol) or the like can also be used.
- FIG. 3 is a functional block diagram of the chip multiprocessor according to the first embodiment.
- the chip multiprocessor 16 according to the first embodiment is an example of a unit that connects a processor core 21 configured by a data driven processor and a core that connects the processor cores 21 and transmits and receives information.
- the inter-core connection network 22 (Interconnection Network).
- the inter-core connection network 22 according to the first embodiment includes a token router that transfers a packet transmitted to and received from each processor core 21 to a packet transmission destination.
- FIG. 4 is an explanatory diagram of the architecture of the data driven processor according to the first embodiment, and is a functional block diagram.
- the processor core 21 of the first embodiment has a so-called cyclic pipeline structure.
- a packet output from the inter-core connection network 22 is input to a first merge unit (M: Merge) 31.
- M: Merge first merge unit
- the packet input to the first merging unit 31 is transmitted to the firing control unit (FC) 32, and the firing control unit 32 determines whether or not all the packets necessary for command execution have been received. To determine whether or not the ignition condition is satisfied.
- a second merge unit (MB: Merge for Bypass) 33.
- the packet input to the second merging unit 33 is input to an instruction fetch unit (IF: Instruction Fetch) 34, and the instruction of the input packet is fetched, that is, the instruction is read.
- the packet processed by the instruction fetch unit 34 is input to an instruction decode unit (ID: Instruction Decode) 36, where an instruction is analyzed and a control signal is generated.
- the packet processed by the instruction decoding unit 36 is processed by a data processing unit (arithmetic unit, EX: EXecution) 37. Packets processed by the data processing unit 37 are written to and read from a memory (register) in a memory access unit (MA: Memory Access / WB: Write Back) 38.
- the data processed by the memory access unit 38 is determined in a second diversion unit (BB: Branch for Bypass) 39 as to whether the data is a target of a unary operation or a target of a multinomial operation. If the data processing unit 37 needs to process the data again (for example, if the data processing is not completed after n times of data processing), the data is transmitted to the second merging unit 33.
- identification information for determining whether data is a target of unary operation or multinomial operation is added to the header of the packet. For example, identification information “0” is a unary operation and “1” is When the multinomial operation is set, it can be determined whether or not it is a single operation based on 1-bit identification information.
- an input variable such as increment (a process for adding +1 to a variable n), decrement (a process for subtracting -1 from a variable n), power or power, etc. has one operation process.
- increment a process for adding +1 to a variable n
- decrement a process for subtracting -1 from a variable n
- power or power etc.
- polynomial operation for two input values n and m, there is an operation process in which two or more input variables such as n + m and nm are input.
- the packet that has not been transmitted to the second merging unit 33 in the second diversion unit 39 is data of a polynomial operation that requires data processing again in the first diversion unit (B: Branch) 41. Is transmitted to the inter-core connection network 22 from the processor core 21 when the data processing in the processor core 21 is completed, and is transmitted to the first merge unit 31 when the data processing is necessary. Is done. That is, in the processor core 21 of the first embodiment, the second merging unit 33 and the second diverting unit 39 which are not provided in the conventional data driven type processor are provided, and regarding the unary operation, the firing control unit The processing in the instruction fetch unit 34 is executed without going through 32.
- FIG. 5 is an explanatory diagram of the implementation of the data driven processor of the first embodiment in a circuit, and is an explanatory diagram of a self-synchronous elastic pipeline.
- each processor core 21 of the first embodiment has a pipeline structure with a self-synchronous elastic pipeline.
- the processor core 21 of the first embodiment has a plurality of pipeline stages 51 corresponding to the pipeline stages 31 to 41 on the functional blocks (architecture) of the firing control unit 32 and the like.
- each pipeline stage 51 is a logic circuit (LC: Logic) Circuit) that executes processing of each pipeline stage 51 based on the packet transmitted from the upstream pipeline stage 51 along the packet flow.
- LC Logic
- DL data latch
- STCM Self-timed Transfer Control Mechanism
- the self-synchronous transfer control mechanism 54 according to the first embodiment is configured in the same manner as that of Patent Document 3 (Japanese Patent Application Laid-Open No. 2010-20598), and data transfer control provided corresponding to each data latch 53.
- the circuit includes a C element (Coincidence Element) 54 a as an example of a circuit and a delay element (Delay Element) 54 b that guarantees a packet processing time in each logic circuit 52.
- the operation and control of the self-synchronous transfer control mechanism 54 according to the first embodiment are known as described in Patent Document 3 (Japanese Patent Laid-Open No. 2010-20598), and thus detailed description thereof is omitted. To do.
- a power supply line 56 for supplying a driving voltage for driving is connected to the logic circuit 52 and the data latch 53.
- the power supply line 56 is a drive voltage line 56a that supplies a drive voltage (positive voltage, drain voltage, drive power) Vdd required when the logic circuit 52 executes processing, and the logic circuit 52 does not execute processing.
- a power switch (PS) 56d is provided. That is, by controlling the power switch 56d, the drive voltage Vdd and the minimum voltage Vmin can be supplied to each logic circuit 52 and the data latch 53, or the voltage supply can be turned off.
- a power supply circuit 57 is connected to the power supply line 56 and supplies the voltages Vdd, Vmin, and Vss.
- the power supply control unit 17 is connected to the power supply circuit 57, and the drive voltage Vdd is controlled to a variable voltage value by the power supply control unit 17.
- a galvanometer 58 is connected to the power supply line 56, and a consumption current Iss in the power supply line 56 is detected by a consumption current detection means 17a of the power supply control means 17 through an ADC (analog / digital converter) (not shown). Is done.
- the current consumption Iss of each processor core 21 is detected by the power supply control means 17.
- the power supply control means 17 of the first embodiment has a QoS setting storage means 17b, and the QoS setting storage means 17b stores the setting of QoS: Quality of Service.
- the processing in the processor core 21 is “power saving priority” that gives priority to power saving even if the processing speed is reduced, or “speed that gives priority to processing speed over power saving” It is memorized whether it is "priority”.
- the QoS setting can be set and changed from the application processor 14 or the processor core 21 of the terminal 11.
- FIG. 6 is an explanatory diagram of an example of a voltage setting table according to the first embodiment.
- the voltage setting table storage unit 17c of the power control unit 17 stores a voltage setting table to be referred to when setting the drive voltage Vdd applied to the processor core 21 based on the current consumption Iss.
- the optimum drive voltage Vdd is derived and selected for the current consumption Iss and Qos by experiments or the like in advance, and the relationship between the current consumption Iss and the drive voltage Vdd is stored as a table. .
- FIG. 1 In FIG.
- V1 is set as the drive voltage Vdd and “speed priority” is set.
- V1 ′ (> V1) is set.
- V2 (> V1 ′ is set in the case of “power saving priority”
- V2 ′ (> V2) in the case of “speed priority”. Is set.
- the voltage setting means 17d of the power supply control means 17 sets the drive voltage Vdd to a voltage corresponding to the data processing load (processing capability, throughput) in the processor core 21 based on the consumption current Iss. .
- the voltage setting unit 17 according to the first embodiment uses the voltage setting table storage unit 17c based on the consumption current Iss acquired by the consumption current detection unit 17a and the QoS setting stored in the QoS storage unit 17b.
- the drive voltage Vdd is set with reference to the voltage setting table stored in (1).
- the power supply control unit 17 according to the first embodiment has a PID as an example of feedback control from the current drive voltage toward the target drive voltage Vdd set when the drive voltage Vdd is set by the voltage setting unit 17d.
- the power supply circuit 57 is controlled by the control (Proportional Integral Differential) so as to converge to the target drive voltage value Vdd.
- the acceptance control means 18 of the first embodiment includes a consumption current detection means 18a, an acceptance threshold value storage means 18b, and an overload determination means 18c configured in the same manner as the consumption current detection means 17a.
- the acceptance threshold storage unit 18b stores a threshold I1 for determining whether to accept or reject the input of a packet from the offloading interface 13.
- the current consumption Iss when the load on each processor core 21 is an overload is measured and set in advance through experiments or the like.
- the overload determination unit 18c of the first embodiment a function for specifying the relationship between the current consumption Iss and the data processing load is derived in advance through experiments or the like, and when a packet is input based on the current consumption Iss, It is determined whether or not the data processing state can be overloaded.
- the overload determination unit 18c determines whether or not an overload state occurs when a packet is received based on the consumption current Iss acquired by the consumption current detection unit 18a. Specifically, it is determined whether or not the current consumption Iss is greater than or equal to the acceptance threshold value I1, that is, whether or not Iss ⁇ I1, and if it is greater than or equal to the acceptance threshold value I1, it is determined that an overload state occurs. If the overload determination unit 18c determines that the overload state is detected, the acceptance control unit 18 of the first embodiment stops receiving packets in the communication unit 12, and the packets transferred to the terminals T1 to T4 are ignored.
- the retransmission control unit 19 includes a consumption current detection unit 19a configured similarly to the consumption current detection unit 17a, a retransmission threshold storage unit 19b, a retransmission information determination unit 19c, and a retransmission determination unit 19d. And, when it is determined that the state is a high load state, control is performed to stop packet retransmission.
- the retransmission threshold storage unit 19b stores a threshold I2 for determining whether or not to retransmit a packet based on current consumption.
- the threshold value I2 of the first embodiment the consumption current Iss when the load in the chip multiprocessor 16 is high is measured and set in advance by experiments or the like. That is, the threshold I2 for determining whether or not the load on the mobile phone 11 as a node is a high load is stored.
- the retransmission information determination unit 19c determines whether the packet is information to be retransmitted. Whether or not the packet is retransmitted is determined by referring to, for example, an ID that identifies data included in the header of the packet or, as described in Patent Document 4, a start node ID and a message ID. It is also possible to discriminate from the above, and various conventionally known configurations can be adopted, and thus detailed description thereof is omitted.
- the retransmission determination unit 19d determines whether to perform retransmission based on the current consumption Iss.
- the retransmission determination unit 19d includes the consumption current Iss acquired by the consumption current detection unit 19a, the threshold I2 stored in the retransmission threshold storage unit 19b, the determination result by the retransmission information determination unit 19c, If the output packet is retransmission information and the total current consumption Iss is larger than the threshold value I2, it is determined that the own node is in a high load state, and packet retransmission is stopped. , Suppress.
- the ultra-low power consumption networking system S of the first embodiment having the above configuration, an ad hoc network N is adopted, and an infrastructure that consumes power constantly at a large scale, such as a server or an access point, increases power consumption. Compared to a network that requires a structure, a server or the like is not required for wireless communication between the terminals T1 to T4. Therefore, power consumption can be reduced.
- the ultra-low power consumption networking system S of the first embodiment has a data driven processor core 21 and receives packets compared to a processor that always waits for reception of packets and consumes power. In this case, processing is performed according to the received packet, so that power consumption is reduced.
- the acceptance control means 18 or the like rejects the acceptance of the packet. Therefore, the processing in the processor core 21 is avoided from being overloaded, the processing is prevented from being delayed, and the packet is stored as compared with the conventional configuration in which the packet waiting for processing is received and stored. It is possible to reduce the power required for storage, and to save power. Further, in the networking system S of the first embodiment, since the acceptance of the packet is rejected before the overload state occurs, the load margin in the processor core 21 can be reduced to the limit, the overhead can be minimized, and the efficiency can be improved. can do. Therefore, as in the technique described in Patent Document 4, it is possible to increase the amount of information processing at the same power consumption as compared with a technique that requires a margin to prevent overload. Power consumption can be reduced.
- the networking system S of the first embodiment when the same information is output, if the own node is in a high load state based on the current consumption by the retransmission control unit 19 or the like, the output is It is suppressed. Therefore, inefficient processing of retransmitting overlapping information to other nodes many times can be suppressed, and traffic on the network N can be reduced. Therefore, with simple flooding, it was difficult to get information to all nodes while using high power due to information collisions, etc., but unnecessary redundant retransmissions were reduced and traffic was reduced. In the first embodiment, information can be delivered to as many nodes as possible.
- suppression of retransmission is controlled based on the current consumption Iss. That is, as described in Patent Document 3, in a data driven processor realized by a self-synchronous elastic pipeline, throughput (processing speed, processing amount per unit time, load) and consumption in the processor core 21 There is a strong correlation with current, and specifically, it is in a proportional relationship. This is because in a data driven processor implemented with a self-synchronous elastic pipeline, only the pipeline stage in which valid data exists is driven. Therefore, if the current consumption is measured outside the data driven processor, the load inside the data driven processor can be observed. That is, it has observability.
- the drive voltage Vdd is changed based on the current consumption, that is, the load, and the power is consumed according to the load, so that useless power consumption can be reduced.
- a special circuit such as PLL (Phase-Locked Loop) is not required to realize a self-synchronous elastic pipeline, and a self-synchronous elastic pipeline can be realized only with standard logic gates.
- the delay in the circuit can be changed at the same ratio with respect to the change of the drive voltage Vdd. Therefore, the operation speed of the circuit changes only by changing the drive voltage Vdd, and as a result, the throughput can be controlled.
- the networking system S of the first embodiment in the case of a unary operation in the pipeline structure, data is transmitted from the second branching unit 39 to the second merging unit 33, and so-called bypass is formed. ing. Therefore, compared with the conventional pipeline structure via the ignition control unit 32, the number of pipeline stages can be reduced by the amount of the ignition control unit 32. That is, in the unary operation, since there is one input variable, it is not necessary to wait in the firing control unit 32 until the second input variable is input unlike the multinomial operation, and can be transmitted to the instruction fetch unit 34. The stage of the ignition control unit 32 can be omitted.
- the power switch is provided for each pipeline stage 51.
- the present invention is not limited to this configuration, and a plurality of pipeline stages may be provided as a lump, and a power switch may be provided for each lump. It is also possible to control by providing a power switch for each processor cluster (a cluster of processors).
- the autonomous distributed communication network by wireless communication is exemplified as the network.
- the present invention is not limited to this configuration, and an autonomous distributed communication network between terminals connected by wire can be used. is there.
- an autonomous distributed communication network is desirable from the viewpoint of low power consumption, but a communication network other than the autonomous distributed communication network is also possible.
- the drive voltage Vdd is changed.
- the present invention is not limited to this configuration.
- the drive voltage Vdd is constant, the effect of lowering the power consumption compared to the prior art is obtained by rejecting the acceptance of packets during overload and bypassing during unary operations, so the drive voltage Vdd is fixed. It is also possible.
- H05 In the above-described embodiment, it is desirable to execute the rejection of packet acceptance and the suppression of retransmission. However, because of the effect of reducing the power consumption by controlling the drive voltage Vdd and bypassing the unary operation, the acceptance rejection is performed. It is also possible to adopt a configuration that does not execute the above.
- the above-described invention of the present application can be expected to realize a communication infrastructure of an M2M (machine-to-machine) network when it is necessary to guarantee data communication regardless of human intervention. Therefore, ad hoc on mobile phones and smartphones in emergency situations where the communication infrastructure cannot function, such as natural disasters such as earthquakes, volcanic eruptions, tornadoes, large-scale incidents and accidents such as aircraft accidents and power plant accidents, etc. It can be used as an emergency communication means by building a network, and it can be expected to improve the lifesaving rate and reduce disasters.
- M2M machine-to-machine
- the M2M network that guarantees data communication can demonstrate its true value in places where it is difficult for people to enter. For example, replace batteries in high-rise building outer walls, cold places such as underwater, seabed, and polar areas, hot places such as furnaces of power plants and steelworks, and places with high humidity that are difficult to enter. For devices that are difficult to feed by wire or power and have large power consumption, long-term use is restricted.
- operation can be expected even in places where it is difficult for people to enter, for example, relatively small capacity self-sufficiency such as solar power generation or small capacity power supply by space power transmission.
- highly reliable data communication can be realized by avoiding an overload condition that causes congestion.
- environmental information temperature and humidity, user audio-visual information, position information
- a sensor network capable of communication and broadcasting can be constructed while avoiding an overload state and with low power consumption.
- the global energy crisis is expected to come due to the economic development of emerging countries in the future, but the realization of the ultra-low power consumption technology of the present invention is expected to help overcome the energy crisis. Is done. Also, in countries and regions where emergency contact means (disaster prevention radio equipment) such as depopulated areas and emerging countries are not sufficiently developed, mobile phones and routers with ad hoc modes can be deployed, so that emergency communication It can also be expected to construct a network capable of preferentially communicating safety information.
- emergency contact means diisaster prevention radio equipment
- 17d Voltage setting means
- 19 Output control means, 21 ... Data driven processor, 31 ... the first junction, 31-41, 51 ... pipeline stage, 32 ... ignition control unit, 33 ... second merging section, 34.
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Power Sources (AREA)
- Telephone Function (AREA)
Abstract
Description
このようなデータ駆動原理を使用した技術として、下記の特許文献1,2記載の技術が従来公知である。
さらに、特許文献4記載の技術では、自ノード内の処理の負荷の状況に応じて、メッセージの再送信を制御することで、ネットワーク全体において流れる情報量(トラフィック)を抑制し、ネットワークが過負荷の状態を抑制している。しかしながら、特許文献4記載の技術では、負荷を把握するために送信待ちパケット数を基準として用いており、パケットの種類や長さに応じて負荷が変動する場合に、一定の安全率(マージン)を考慮して再送信を待機させる必要があるため、再送信が過剰に抑えられた場合に、単位情報量当たりの消費電力に換算すると、消費電力の低減効果が限られる問題があった。
また、特許文献1~4に記載の従来の技術では、プロセッサに対して送信されたパケットを受信、格納しておき、処理待ちの状態とするキューイングを行っており、キューイングに要する電力も必要になる。
処理されるデータが分割されたパケットに対して命令実行に必要なパケットが全て受信されたか否かを判別することにより発火条件が成立したか否かを判別する発火制御部と、前記発火条件が成立した場合に前記発火制御部から送信されたパケットに応じた処理を実行するデータ処理部と、を少なくとも有するデータ駆動型プロセッサと、
前記データ駆動型プロセッサに電力を供給する電源回路と、
前記データ駆動型プロセッサにおける消費電流に基づいて、前記データ駆動型プロセッサにおけるデータ処理の負荷が、データ処理が滞る過負荷状態になり得ると判別される場合に、端末に対するパケットの入力を拒否する過負荷回避手段と、
を備えたことを特徴とする。
前記データ駆動型プロセッサにおける消費電流に基づいて、前記データ駆動型プロセッサにおけるデータ処理の負荷が予め設定された高負荷状態よりも高いと判別される場合に、再送パケットの出力を停止する出力制御手段と、
を備えたことを特徴とする。
前記データ駆動型プロセッサにおける消費電流に基づいて、前記データ駆動型プロセッサに印加する電圧を、前記データ駆動型プロセッサにおけるデータ処理の負荷に応じた電圧に、設定する電圧設定手段と、
前記電源回路を制御して、前記データ駆動型プロセッサに印加する電圧を、前記電圧設定手段で設定された電圧に制御する電源制御手段と、
を備えたことを特徴とする。
前記データ駆動型プロセッサに入力されるパケットと前記データ処理部で処理されたデータが合流する第1の合流部と、
前記第1の合流部から送信されたパケットまたはデータを受信する前記発火制御部と、
前記発火制御部から送信されたデータと前記データ処理部で処理されたデータが合流する第2の合流部と、
前記第2の合流部から送信されたデータに基づいて命令をフェッチする命令フェッチ部と、
前記命令フェッチ部から送信された命令の解析を行う命令デコード部と、
前記命令デコード部で解析された命令の処理を行う前記データ処理部と、
前記データ処理部で処理されたデータのメモリへの書き込みとメモリからの読み出しを行うメモリアクセス部と、
前記メモリアクセス部で書き込まれたデータに基づいて、前記データが単項演算であり且つ再度前記データ処理部で処理が必要な場合に、前記第2の合流部にデータを送信すると共に、前記データが多項演算である場合および前記データがデータ処理部での処理が必要ない場合に下流側の第1の分流部に送信する第2の分流部と、
前記第2の分流部から送信されたデータに基づいて、前記データが前記データ処理部で再度データ処理が必要なデータである場合に、前記第1の合流部にデータを送信すると共に、前記データが前記データ処理部でデータ処理が必要ない場合に、前記データ駆動型プロセッサから前記データを出力させる前記第1の分流部と、
を有するパイプラインステージ、
を備えたことを特徴とする。
処理されるデータが分割されたパケットに対して命令実行に必要なパケットが全て受信されたか否かを判別することにより発火条件が成立したか否かを判別する発火制御部と、前記発火条件が成立した場合に前記発火制御部から送信されたパケットに応じた処理を実行するデータ処理部と、を少なくとも有するデータ駆動型プロセッサと、
前記データ駆動型プロセッサに電力を供給する電源回路と、
前記データ駆動型プロセッサにおける消費電流に基づいて、前記データ駆動型プロセッサに印加する電圧を、前記データ駆動型プロセッサにおけるデータ処理の負荷に応じた電圧に、設定する電圧設定手段と、
前記電源回路を制御して、前記データ駆動型プロセッサに印加する電圧を、前記電圧設定手段で設定された電圧に制御する電源制御手段と、
を備えたことを特徴とする。
処理されるデータが分割されたパケットが全て受信されたか否かを判別することにより発火条件が成立したか否かを判別する発火制御部と、前記発火条件が成立した場合に前記発火制御部から送信されたデータに応じた処理を実行するデータ処理部と、を少なくとも有するデータ駆動型プロセッサと、
前記データ駆動型プロセッサに入力されるパケットと前記データ処理部で処理されたデータが合流する第1の合流部と、前記第1の合流部から送信されたパケットまたはデータを受信する前記発火制御部と、前記発火制御部から送信されたデータと前記データ処理部で処理されたデータが合流する第2の合流部と、前記第2の合流部から送信されたデータに基づいて命令をフェッチする命令フェッチ部と、前記命令フェッチ部から送信された命令の解析を行う命令デコード部と、前記命令デコード部で解析された命令の処理を行う前記データ処理部と、前記データ処理部で処理されたデータのメモリへの書き込みとメモリからの読み出しを行うメモリアクセス部と、前記メモリアクセス部で書き込まれたデータに基づいて、前記データが単項演算であり且つ再度前記データ処理部で処理が必要な場合に、前記第2の合流部にデータを送信すると共に、前記データが多項演算である場合および前記データがデータ処理部での処理が必要ない場合に下流側の第1の分流部に送信する第2の分流部と、前記第2の分流部から送信されたデータに基づいて、前記データが前記データ処理部で再度データ処理が必要なデータである場合に、前記第1の合流部にデータを送信すると共に、前記データが前記データ処理部でデータ処理が必要ない場合に、前記データ駆動型プロセッサから前記データを出力させる前記第1の分流部と、を有するパイプラインステージと、
を備えたことを特徴とする。
請求項2に記載の発明によれば、高負荷の場合にもパケットを再送する場合に比べて、ネットワーク全体のトラフィックを抑制でき、効率的な処理で消費電力を削減することができる。
請求項4、6に記載の発明によれば、第2の合流部および第2の分流部を有しない構成に比べて、単項演算の場合に処理時間を短縮することができ、処理時間を従来と同等に維持または向上しつつ印加電圧を低減して、ネットワーキングシステムにおける消費電力を削減することもできる。
なお、以下の図面を使用した説明において、理解の容易のために説明に必要な部材以外の図示は適宜省略されている。
図1において、本発明の実施例1の超低消費電力化ネットワーキングシステムSは、複数の端末T1~T4を有する。実施例1の超低消費電力ネットワーキングシステムSでは、端末T1~T4の無線通信により形成される自律分散型通信網、いわゆる、アドホックネットワーク(Ad hoc Network)Nが採用されており、データの一例としての音声データや動画データ(例えば、災害現場のライブ中継映像等)がストリーミング配信可能である。また、実施例1の各端末T1~T4は、アドホックネットワーク通信機能を有する情報処理装置(いわゆるPC:パーソナルコンピュータ、ノート型PCまたはデスクトップ型PC)や、携帯型情報端末(いわゆるPDA:Personal Digital Assistant)、携帯型音楽再生装置、携帯電話、あるいは、通信機能を有する携帯型ゲーム機等により構成することが可能である。
図2において、実施例1の端末T1~T4の一例であって、過負荷を回避する超低消費電力化データ駆動ネットワーキング処理装置の一例としての携帯電話11は、他の端末T1~T4との間でデータの送受信を行う通信手段12と、前記通信手段12と接続されてデータが分割されたパケットの送受信を行うオフローディングインターフェース13と、前記オフローディングインターフェース13と電気的に接続されて携帯電話11に内蔵されたカメラや表示画面11aの制御や、入力キー11bからの入力の処理等を行うアプリケーションプロセッサ14と、前記オフローディングインターフェース13と電気的に接続されて通信処理等のデータ処理を行うチップマルチプロセッサ16と、前記チップマルチプロセッサ16に対する電源供給を制御する電源制御手段17と、を有する。そして、実施例1の携帯電話11では、前記通信手段12に対して、過負荷回避手段の一例としての受入制御手段18と、出力制御手段の一例としての再送信制御手段19から制御信号が入力される。
なお、実施例1では、前記チップマルチプロセッサ16は、超低消費電力化データ駆動型チップマルチプロセッサ、ULP-DDCMP:Ultra Low Power - Data Drive Chip Multi Processorにより構成されている。また、実施例1のチップマルチプロセッサ16では、通信規約の一例としてのUDP/IP(User Datagram Protocol / Internet Protocol)の処理を行っているが、これに限定されず、従来公知の通信規約(例えば、TCP/IP:Transmission Control Protocol / Internet Protocol)等を使用することも可能である。
図3において、実施例1のチップマルチプロセッサ16は、データ駆動型プロセッサにより構成されたプロセッサコア21と、各プロセッサコア21どうしを接続して情報の送受信を行うコア間を接続する手段の一例としてのコア間接続ネットワーク22(Interconnection Network)とを有する。実施例1のコア間接続ネットワーク22は、各プロセッサコア21に対して送受信されるパケットを、パケットの送信先に転送するトークンルータにより構成されている。
図4に示す実施例1のプロセッサコア21のアーキテクチャの説明図において、実施例1のプロセッサコア21は、いわゆる循環型のパイプライン構造を有している。
実施例1の循環型のパイプライン構造では、コア間接続ネットワーク22から出力されたパケットは、第1の合流部(M:Merge)31に入力される。
第1の合流部31に入力されたパケットは、発火制御部(FC:Firing Control)32に送信され、発火制御部32は、命令実行に必要なパケットが全て受信されたか否かを判別することにより、発火条件が成立したか否かを判別する。
第2の合流部33に入力されたパケットは、命令フェッチ部(IF:Instruction Fetch)34に入力され、入力されたパケットの命令をフェッチ、すなわち、命令を読み出す。
命令フェッチ部34で処理されたパケットは、命令デコード部(ID:Instruction Decode)36に入力されて、命令の解析、制御信号の生成が行われる。
命令デコード部36で処理されたパケットは、データ処理部(演算部、EX:EXecution)37において命令が処理される。
データ処理部37で処理されたパケットは、メモリアクセス部(MA:Memory Access/WB:Write Back)38において、メモリ(レジスタ)への書き込みと読み出しが行われる。
すなわち、実施例1のプロセッサコア21では、従来のデータ駆動型のプロセッサに設けられていない第2の合流部33および第2の分流部39が設けられており、単項演算に関しては、発火制御部32を介さずに、命令フェッチ部34における処理が実行される。
図5において、実施例1の各プロセッサコア21は、自己同期式エラスティックパイプラインによるパイプライン構造を有している。実施例1のプロセッサコア21は、前記発火制御部32等の機能ブロック上(アーキテクチャ上)のパイプラインステージ31~41に対応する複数のパイプラインステージ51を有する。
電源供給線56には、検流計58が接続されており、図示しないADC(アナログデジタルコンバータ)を介して、電源供給線56における消費電流Issが電源制御手段17の消費電流検出手段17aにより検出される。なお、実施例1では、各プロセッサコア21のそれぞれの消費電流Issが、電源制御手段17で検出される。
電源制御手段17の電圧設定テーブル記憶手段17cは、消費電流Issに基づいて、プロセッサコア21に印加される駆動電圧Vddの設定を行う際に参照する電圧設定テーブルが記憶されている。前記電圧設定テーブルは、予め実験等により、消費電流IssおよびQosに対して、最適な駆動電圧Vddが導出、選択されており、消費電流Issと駆動電圧Vddとの関係がテーブルとして記憶されている。図6において、消費電流Issが0より大きく、予め設定された閾値α1以下である場合には、「省電力優先」の場合には、駆動電圧Vddとして、V1が設定され、「速度優先」の場合には、V1′(>V1)が設定されている。同様にして、消費電流Issがα1~α2である場合には、「省電力優先」の場合に、V2(>V1′が設定され、「速度優先」の場合には、V2′(>V2)が設定されている。
なお、実施例1の電源制御手段17は、電圧設定手段17dで駆動電圧Vddが設定された場合、現在の駆動電圧から設定された目標の駆動電圧Vddに向けて、フィードバック制御の一例としてのPID制御(Proportional Integral Differential)により、目標の駆動電圧値Vddに集束するように電源回路57を制御する。
図2において、実施例1の受入制御手段18は、前記消費電流検出手段17aと同様に構成された消費電流検出手段18aと、受入閾値記憶手段18bと、過負荷判別手段18cとを有する。
受入閾値記憶手段18bは、オフローディングインターフェース13からパケットの入力を受け入れるか、拒否するかを判別するための閾値I1を記憶する。実施例1の閾値I1は、各プロセッサコア21における負荷が過負荷である場合の消費電流Issが、実験等により予め測定され、設定されている。
前記再送信制御手段19は、前記消費電流検出手段17aと同様に構成された消費電流検出手段19aと、再送信閾値記憶手段19bと、再送信情報判別手段19cと、再送信判別手段19dとを有し、高負荷状態と判別される場合に、パケットの再送信を停止する制御を行う。
再送信閾値記憶手段19bは、消費電流に基づきパケットを再送信するかしないかを判別するための閾値I2を記憶する。実施例1の閾値I2は、チップマルチプロセッサ16における負荷が高負荷である場合の消費電流Issが、実験等により予め測定され、設定されている。すなわち、ノードとしての携帯電話11における負荷が高負荷であるか否かを判別するための閾値I2が記憶されている。
前記構成を備えた実施例1の超低消費電力化ネットワーキングシステムSでは、アドホックネットワークNが採用されており、サーバーやアクセスポイント等の大規模で常時電力を消費して、消費電力が多くなるインフラストラクチャーが必要なネットワークに比べて、端末T1~T4間の無線通信でサーバー等が必要ない。したがって、低消費電力化することができる。また、実施例1の超低消費電力化ネットワーキングシステムSでは、データ駆動型のプロセッサコア21を有しており、常時パケットの受信を待機して電力を消費するプロセッサに比べて、パケットを受信した場合に受信したパケットに応じて処理を行うため、電力消費が低減されている。
したがって、従来の構成に対して、同一の駆動電圧Vddにおいて、処理時間を短縮することができるため、処理時間を従来と同様にすれば、駆動電圧Vddを低減することができる。したがって、実施例1のネットワーキングシステムでは、処理速度を従来と同等の速度に維持しつつ駆動電圧Vddを低減することもでき、電力消費を低減することができる。
以上、本発明の実施例を詳述したが、本発明は、前記実施例に限定されるものではなく、特許請求の範囲に記載された本発明の要旨の範囲内で、種々の変更を行うことが可能である。本発明の変更例(H01)~(H07)を下記に例示する。
(H01)前記実施例において例示したように、プロセッサコア21やコア間接続ネットワーク22において、パワースイッチを有する自己同期式エラスティックパイプラインを採用することを例示したが、これに限定されず、アプリケーションプロセッサ14やオフローディングインターフェース13も、プロセッサコア21等と同様に、自己同期式エラスティックパイプラインを採用することが望ましい。なお、アプリケーションプロセッサ14として、プロセッサコア21と同様のデータ駆動型プロセッサを使用した場合には、コア間接続ネットワーク22で接続可能となり、オフローディングインターフェース13を省略することも可能である。
(H03)前記実施例において、ネットワークとして、無線通信による自律分散型通信網を例示したが、この構成に限定されず、有線で接続された端末間の自律分散型通信網とすることも可能である。さらに、低消費電力化の観点から自律分散型通信網であることが望ましいが、自律分散型通信網以外の通信網とすることも可能である。
(H05)前記実施例において、パケットの受入拒否や再送信の抑制を実行することが望ましいが、駆動電圧Vddの制御や単項演算時のバイパスにより、低消費電力化の効果があるため、受入拒否等を実行しない構成とすることも可能である。
(H07)前記実施例において、プロセッサコア21を複数有するチップマルチプロセッサの構成を例示したが、これに限定されず、プロセッサコア21の単体のみを有するプロセッサの構成とすることも可能である。
したがって、地震や火山噴火、竜巻等の自然災害や航空機事故や発電所の事故等の大規模な事件・事故等が発生した場合等の、通信インフラストラクチャーが機能できない緊急時に携帯電話やスマートフォンのアドホックネットワークを構築して、緊急時の通信手段として、利用可能であり、これによる救命率の向上や減災が期待できる。
また、過疎地や新興国のような緊急時の連絡手段(防災無線設備)等の整備が十分でない国や地域において、携帯電話やアドホックモード付きのルータを展開することで、緊急時に救助連絡や安否情報を優先的に疎通させることが可能なネットワークを構築することも期待できる。
17…電源制御手段、
18…過負荷制御手段、
19…出力制御手段、
21…データ駆動型プロセッサ、
31…第1の合流部、
31~41,51…パイプラインステージ、
32…発火制御部、
33…第2の合流部、
34…命令フェッチ部、
36…命令デコード部、
37…データ処理部、
38…メモリアクセス部、
39…第2の分流部、
41…第1の分流部、
57…電源回路、
Iss…消費電流、
S…ネットワーキングシステム、
Vdd…電圧。
Claims (6)
- 処理されるデータが分割されたパケットに対して命令実行に必要なパケットが全て受信されたか否かを判別することにより発火条件が成立したか否かを判別する発火制御部と、前記発火条件が成立した場合に前記発火制御部から送信されたパケットに応じた処理を実行するデータ処理部と、を少なくとも有するデータ駆動型プロセッサと、
前記データ駆動型プロセッサに電力を供給する電源回路と、
前記データ駆動型プロセッサにおける消費電流に基づいて、前記データ駆動型プロセッサにおけるデータ処理の負荷が、データ処理が滞る過負荷状態になり得ると判別される場合に、端末に対するパケットの入力を拒否する過負荷回避手段と、
を備えたことを特徴とする過負荷を回避する超低消費電力化データ駆動ネットワーキング処理装置。 - 前記データ駆動型プロセッサにおける消費電流に基づいて、前記データ駆動型プロセッサにおけるデータ処理の負荷が予め設定された高負荷状態よりも高いと判別される場合に、再送パケットの出力を停止する出力制御手段と、
を備えたことを特徴とする請求項1に記載の過負荷を回避する超低消費電力化データ駆動ネットワーキング処理装置。 - 前記データ駆動型プロセッサにおける消費電流に基づいて、前記データ駆動型プロセッサに印加する電圧を、前記データ駆動型プロセッサにおけるデータ処理の負荷に応じた電圧に、設定する電圧設定手段と、
前記電源回路を制御して、前記データ駆動型プロセッサに印加する電圧を、前記電圧設定手段で設定された電圧に制御する電源制御手段と、
を備えたことを特徴とする請求項1または2に記載の過負荷を回避する超低消費電力化データ駆動ネットワーキング処理装置。 - 前記データ駆動型プロセッサに入力されるパケットと前記データ処理部で処理されたデータが合流する第1の合流部と、
前記第1の合流部から送信されたパケットまたはデータを受信する前記発火制御部と、
前記発火制御部から送信されたデータと前記データ処理部で処理されたデータが合流する第2の合流部と、
前記第2の合流部から送信されたデータに基づいて命令をフェッチする命令フェッチ部と、
前記命令フェッチ部から送信された命令の解析を行う命令デコード部と、
前記命令デコード部で解析された命令の処理を行う前記データ処理部と、
前記データ処理部で処理されたデータのメモリへの書き込みとメモリからの読み出しを行うメモリアクセス部と、
前記メモリアクセス部で書き込まれたデータに基づいて、前記データが単項演算であり且つ再度前記データ処理部で処理が必要な場合に、前記第2の合流部にデータを送信すると共に、前記データが多項演算である場合および前記データがデータ処理部での処理が必要ない場合に下流側の第1の分流部に送信する第2の分流部と、
前記第2の分流部から送信されたデータに基づいて、前記データが前記データ処理部で再度データ処理が必要なデータである場合に、前記第1の合流部にデータを送信すると共に、前記データが前記データ処理部でデータ処理が必要ない場合に、前記データ駆動型プロセッサから前記データを出力させる前記第1の分流部と、
を有するパイプラインステージ、
を備えたことを特徴とする請求項1ないし3のいずれかに記載の過負荷を回避する超低消費電力化データ駆動ネットワーキング処理装置。 - 処理されるデータが分割されたパケットに対して命令実行に必要なパケットが全て受信されたか否かを判別することにより発火条件が成立したか否かを判別する発火制御部と、前記発火条件が成立した場合に前記発火制御部から送信されたパケットに応じた処理を実行するデータ処理部と、を少なくとも有するデータ駆動型プロセッサと、
前記データ駆動型プロセッサに電力を供給する電源回路と、
前記データ駆動型プロセッサにおける消費電流に基づいて、前記データ駆動型プロセッサに印加する電圧を、前記データ駆動型プロセッサにおけるデータ処理の負荷に応じた電圧に、設定する電圧設定手段と、
前記電源回路を制御して、前記データ駆動型プロセッサに印加する電圧を、前記電圧設定手段で設定された電圧に制御する電源制御手段と、
を備えたことを特徴とする過負荷を回避する超低消費電力化データ駆動ネットワーキング処理装置。 - 処理されるデータが分割されたパケットが全て受信されたか否かを判別することにより発火条件が成立したか否かを判別する発火制御部と、前記発火条件が成立した場合に前記発火制御部から送信されたデータに応じた処理を実行するデータ処理部と、を少なくとも有するデータ駆動型プロセッサと、
前記データ駆動型プロセッサに入力されるパケットと前記データ処理部で処理されたデータが合流する第1の合流部と、前記第1の合流部から送信されたパケットまたはデータを受信する前記発火制御部と、前記発火制御部から送信されたデータと前記データ処理部で処理されたデータが合流する第2の合流部と、前記第2の合流部から送信されたデータに基づいて命令をフェッチする命令フェッチ部と、前記命令フェッチ部から送信された命令の解析を行う命令デコード部と、前記命令デコード部で解析された命令の処理を行う前記データ処理部と、前記データ処理部で処理されたデータのメモリへの書き込みとメモリからの読み出しを行うメモリアクセス部と、前記メモリアクセス部で書き込まれたデータに基づいて、前記データが単項演算であり且つ再度前記データ処理部で処理が必要な場合に、前記第2の合流部にデータを送信すると共に、前記データが多項演算である場合および前記データがデータ処理部での処理が必要ない場合に下流側の第1の分流部に送信する第2の分流部と、前記第2の分流部から送信されたデータに基づいて、前記データが前記データ処理部で再度データ処理が必要なデータである場合に、前記第1の合流部にデータを送信すると共に、前記データが前記データ処理部でデータ処理が必要ない場合に、前記データ駆動型プロセッサから前記データを出力させる前記第1の分流部と、を有するパイプラインステージと、
を備えたことを特徴とする過負荷を回避する超低消費電力化データ駆動ネットワーキング処理装置。
Priority Applications (4)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
KR1020147003808A KR101557596B1 (ko) | 2011-07-15 | 2012-07-09 | 과부하를 회피하는 초저소비 전력화 데이터 구동 네트워킹 처리장치 |
US14/232,951 US9363196B2 (en) | 2011-07-15 | 2012-07-09 | Ultra low power-data driven networking processing device for avoiding overload |
JP2013524594A JP5971731B2 (ja) | 2011-07-15 | 2012-07-09 | 過負荷を回避する超低消費電力化データ駆動ネットワーキング処理装置 |
EP12814871.5A EP2738683B1 (en) | 2011-07-15 | 2012-07-09 | Ultra low power-data driven networking processing device for avoiding overload |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2011-157238 | 2011-07-15 | ||
JP2011157238 | 2011-07-15 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2013011653A1 true WO2013011653A1 (ja) | 2013-01-24 |
Family
ID=47557855
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2012/004420 WO2013011653A1 (ja) | 2011-07-15 | 2012-07-09 | 過負荷を回避する超低消費電力化データ駆動ネットワーキング処理装置 |
Country Status (5)
Country | Link |
---|---|
US (1) | US9363196B2 (ja) |
EP (1) | EP2738683B1 (ja) |
JP (1) | JP5971731B2 (ja) |
KR (1) | KR101557596B1 (ja) |
WO (1) | WO2013011653A1 (ja) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105579996A (zh) * | 2013-10-10 | 2016-05-11 | 国际商业机器公司 | 硬件加速器的性能测量 |
EP2972912A4 (en) * | 2013-03-15 | 2016-11-09 | Soft Machines Inc | Method for implementing a line speed link structure |
JP2016218956A (ja) * | 2015-05-26 | 2016-12-22 | 国立大学法人 筑波大学 | データ駆動型処理装置 |
WO2018097242A1 (ja) | 2016-11-25 | 2018-05-31 | 国立大学法人筑波大学 | ネットワーキングシステム |
US10282170B2 (en) | 2013-03-15 | 2019-05-07 | Intel Corporation | Method for a stage optimized high speed adder |
US11003459B2 (en) | 2013-03-15 | 2021-05-11 | Intel Corporation | Method for implementing a line speed interconnect structure |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9817666B2 (en) | 2013-03-15 | 2017-11-14 | Intel Corporation | Method for a delayed branch implementation by using a front end track table |
CN110391670A (zh) * | 2018-04-16 | 2019-10-29 | 广西师范大学 | 一种微电网系统灾变预测方法及装置 |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2004013602A (ja) | 2002-06-07 | 2004-01-15 | Handotai Rikougaku Kenkyu Center:Kk | データ駆動プロセッサのエミュレーションシステム |
JP2005108086A (ja) | 2003-10-01 | 2005-04-21 | Handotai Rikougaku Kenkyu Center:Kk | データ処理装置 |
JP2010020598A (ja) | 2008-07-11 | 2010-01-28 | Univ Of Tsukuba | ネットワークシステムおよびネットワークシステムにおける電源制御方法 |
JP2011030210A (ja) | 2009-06-26 | 2011-02-10 | Tokai Univ | 無線通信装置、無線ネットワークシステム及び通信処理方法 |
Family Cites Families (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP3853098B2 (ja) * | 1999-01-18 | 2006-12-06 | シャープ株式会社 | データ駆動型情報処理システム |
WO2003003753A2 (en) | 2001-05-14 | 2003-01-09 | Telefonaktiebolaget Lm Ericsson | Method for protecting against overload of a packet switching network node of a communication network |
US7120804B2 (en) * | 2002-12-23 | 2006-10-10 | Intel Corporation | Method and apparatus for reducing power consumption through dynamic control of supply voltage and body bias including maintaining a substantially constant operating frequency |
JP4149360B2 (ja) * | 2003-11-10 | 2008-09-10 | シャープ株式会社 | データ駆動型情報処理装置およびデータフロープログラムの実行制御方法 |
US7562234B2 (en) * | 2005-08-25 | 2009-07-14 | Apple Inc. | Methods and apparatuses for dynamic power control |
TW200723632A (en) * | 2005-12-15 | 2007-06-16 | Inventec Corp | Current overload status-informing system and the method |
JP4353990B2 (ja) * | 2007-05-18 | 2009-10-28 | 株式会社半導体理工学研究センター | マルチプロセッサ制御装置 |
US8261112B2 (en) * | 2008-12-08 | 2012-09-04 | International Business Machines Corporation | Optimizing power consumption by tracking how program runtime performance metrics respond to changes in operating frequency |
-
2012
- 2012-07-09 WO PCT/JP2012/004420 patent/WO2013011653A1/ja active Application Filing
- 2012-07-09 EP EP12814871.5A patent/EP2738683B1/en not_active Not-in-force
- 2012-07-09 JP JP2013524594A patent/JP5971731B2/ja active Active
- 2012-07-09 US US14/232,951 patent/US9363196B2/en not_active Expired - Fee Related
- 2012-07-09 KR KR1020147003808A patent/KR101557596B1/ko active IP Right Grant
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2004013602A (ja) | 2002-06-07 | 2004-01-15 | Handotai Rikougaku Kenkyu Center:Kk | データ駆動プロセッサのエミュレーションシステム |
JP2005108086A (ja) | 2003-10-01 | 2005-04-21 | Handotai Rikougaku Kenkyu Center:Kk | データ処理装置 |
JP2010020598A (ja) | 2008-07-11 | 2010-01-28 | Univ Of Tsukuba | ネットワークシステムおよびネットワークシステムにおける電源制御方法 |
JP2011030210A (ja) | 2009-06-26 | 2011-02-10 | Tokai Univ | 無線通信装置、無線ネットワークシステム及び通信処理方法 |
Non-Patent Citations (1)
Title |
---|
See also references of EP2738683A4 |
Cited By (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108255521A (zh) * | 2013-03-15 | 2018-07-06 | 英特尔公司 | 一种用于实现线路速度互连结构的方法 |
US11003459B2 (en) | 2013-03-15 | 2021-05-11 | Intel Corporation | Method for implementing a line speed interconnect structure |
CN108255521B (zh) * | 2013-03-15 | 2022-05-31 | 英特尔公司 | 一种用于实现线路速度互连结构的方法 |
US9740499B2 (en) | 2013-03-15 | 2017-08-22 | Intel Corporation | Method for implementing a line speed interconnect structure |
US10282170B2 (en) | 2013-03-15 | 2019-05-07 | Intel Corporation | Method for a stage optimized high speed adder |
US10303484B2 (en) | 2013-03-15 | 2019-05-28 | Intel Corporation | Method for implementing a line speed interconnect structure |
EP2972912A4 (en) * | 2013-03-15 | 2016-11-09 | Soft Machines Inc | Method for implementing a line speed link structure |
CN105579996B (zh) * | 2013-10-10 | 2018-01-05 | 国际商业机器公司 | 硬件加速器的性能测量的方法、系统和计算机可读介质 |
CN105579996A (zh) * | 2013-10-10 | 2016-05-11 | 国际商业机器公司 | 硬件加速器的性能测量 |
JP2016218956A (ja) * | 2015-05-26 | 2016-12-22 | 国立大学法人 筑波大学 | データ駆動型処理装置 |
JPWO2018097242A1 (ja) * | 2016-11-25 | 2019-11-21 | 国立大学法人 筑波大学 | ネットワーキングシステム |
WO2018097242A1 (ja) | 2016-11-25 | 2018-05-31 | 国立大学法人筑波大学 | ネットワーキングシステム |
JP6989959B2 (ja) | 2016-11-25 | 2022-01-12 | 国立大学法人 筑波大学 | ネットワーキングシステム |
US11233808B2 (en) | 2016-11-25 | 2022-01-25 | University Of Tsukuba | Networking system |
Also Published As
Publication number | Publication date |
---|---|
EP2738683A1 (en) | 2014-06-04 |
US20140219084A1 (en) | 2014-08-07 |
EP2738683B1 (en) | 2018-04-04 |
KR101557596B1 (ko) | 2015-10-05 |
KR20140057551A (ko) | 2014-05-13 |
JP5971731B2 (ja) | 2016-08-17 |
US9363196B2 (en) | 2016-06-07 |
JPWO2013011653A1 (ja) | 2015-02-23 |
EP2738683A4 (en) | 2015-09-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
JP5971731B2 (ja) | 過負荷を回避する超低消費電力化データ駆動ネットワーキング処理装置 | |
US9811374B1 (en) | Partial resume for operating system modules | |
Abazeed et al. | Routing protocols for wireless multimedia sensor network: a survey | |
WO2018201476A1 (zh) | 多播承载的管理方法和终端设备 | |
US12067540B2 (en) | Method and system for selectively using network coding for propagating transactions in a blockchain network | |
TW201643607A (zh) | 用於計算裝置內多個soc之操作狀態間協調之方法及系統 | |
US20130329618A1 (en) | Dynamic power mode switch in a wireless ad-hoc system | |
US8094731B2 (en) | System and method for transferring telemetry data between a ground station and a control center | |
WO2021146911A1 (zh) | 路由的建立方法及装置、电子设备以及计算机存储介质 | |
KR20090115479A (ko) | 무선 근거리 통신망에서 데이터 전송 방법 및 장치 및 이를위한 시스템 | |
US9891698B2 (en) | Audio processing during low-power operation | |
US10542489B2 (en) | Terminal wakeup method and apparatus | |
Lien et al. | A multi-hop walkie-talkie-like emergency communication system for catastrophic natural disasters | |
WO2016209351A1 (en) | Dynamic management of inactivity timer during inter-processor communication | |
JP5517875B2 (ja) | 無線通信システム、データ送信装置、データ無線受信装置、及び、無線通信方法 | |
US9722949B2 (en) | Reducing power utilization by transferring communication sessions from a main processor | |
US20090016313A1 (en) | Method and apparatus for switching a communication channel | |
JP5294449B2 (ja) | ネットワークシステムおよびネットワークシステムにおける電源制御方法 | |
Nishikawa et al. | Intermediate achievement of ultra-low-power data-driven networking system: ULP-DDNS | |
US20190174412A1 (en) | Wi-Fi Station Power Optimization Using Bluetooth/BLE | |
WO2017157116A1 (zh) | 一种流量拥塞控制方法、装置及存储介质 | |
TW201633765A (zh) | 無線終端、儲存應用程式的計算機可讀媒體及方法 | |
US11290516B1 (en) | Prioritized MSRP transmissions to reduce traffic interruptions | |
WO2023241559A1 (zh) | QoS参数分割方法、装置、通信设备、通信系统及可读存储介质 | |
CN103095562A (zh) | 云计算智能网关 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 12814871 Country of ref document: EP Kind code of ref document: A1 |
|
DPE1 | Request for preliminary examination filed after expiration of 19th month from priority date (pct application filed from 20040101) | ||
ENP | Entry into the national phase |
Ref document number: 2013524594 Country of ref document: JP Kind code of ref document: A |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
ENP | Entry into the national phase |
Ref document number: 20147003808 Country of ref document: KR Kind code of ref document: A |
|
WWE | Wipo information: entry into national phase |
Ref document number: 14232951 Country of ref document: US |