JPS6252663A - Method and device for processing information - Google Patents

Abstract

PURPOSE:To automonically perform a specialized processing jog and to enable a job which is highly made in parallel by giving a function broadcasting information on the processed result and a function selecting information regarding an own purpose to each element device. CONSTITUTION:When an information T1 showing that an input element 12-i is activated and a train R2 passes through a point S1 in a time T1 is transmitted to a transmission line 5 through a transmission control part 11-i, it is broadcasted to all elements. The information T1 is selectively received by an arithmetic element 32-i and a memory element 42-i. According to the order of a queue, the arithmetic element 32-i broadcasts said information, while an output element 12-i selectively receives the information, and controls a switching unit A1 through an actuator 13-i. In accordance with the information T1, the memory element 42-i broadcasts the scheduled passing time T<0>1 of the train R2 at the point S1, and the arithmetic element 32-i selectively receives the information, calculates a delay time and broadcasts an information DELTAT1.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Application of the Invention] The present invention relates to an information processing method and apparatus, and particularly to an information processing method and apparatus that are fully equipped with a high degree of parallelism and autonomous decentralization. The present invention is applicable to control system monitoring, abnormality response, image processing, etc., which require parallel execution of various processes that are difficult to formulate as a whole.
Particularly suitable.

[Background of the invention]

The capabilities required for information processing, such as control systems and image processing, are the ability to perform highly diverse processing that is difficult to formulate or structure as a whole, online or in a data-driven manner, on a large amount of various types of data. , to process in parallel. To achieve this, it is desirable to ensure a sufficient degree of freedom in processing procedures through thorough parallelism and autonomous activity in the novels of each word storage element and calculation element.

In a conventional AI type processing device, overall management by a host computer is required in order to allocate each processing element device and storage area.

An AI-type data flow processing device has also been proposed (Takahashi, Nishinomiya ° ``Architecture of a data flow processor array computer for ultra-high-speed scientific and technical calculations'' IEICE Transactions 84/1. Vol. J 67-D.

A1, pages 62 to 69), but in this device as well, the processing element device (PE) receives distribution of programs and initial data from the host processor. Additionally, inside each PE, there are multiple general-purpose arithmetic units (Function Units).
In order to achieve parallel operation of the data 70 of these units, an instruction memory for storing operation nodes specifying the operation contents, an operand memory for waiting operand data, an operation section, and an external A communication control unit and a link memory for storing all link nodes indicating distribution destinations of calculation results are loosely connected in a pipeline via a queue. Therefore, there is room to further improve parallelism and autonomous decentralization.

[Purpose of the invention]

The purpose of the present invention is to provide information processing with a high degree of parallelism and autonomous decentralization, and with a large degree of freedom in processing procedures, suitable for solving problems that require large-scale parallel processing or problems in which processing procedures are difficult to clarify. An object of the present invention is to provide a method and apparatus for.

[Summary of the invention]

A feature of the present invention is that each part of human output, calculation, and memory is
Input source and output light in the input/output section, calculation contents in the calculation section, type and source of target data, whether temporary or fixed storage in the storage section, and information that should be or is being stored. The idea is to subdivide the system into different elemental devices (such as type, source, supply destination, etc.), make each elemental device independent, and have these elemental devices perform parallel operations. These elemental devices are interconnected via transmission paths, autonomously broadcast information or select information relevant to their respective purposes from the broadcast information according to their respective purposes, and furthermore, depending on the situation. depending on,
Broadcast calculation results, storage information, etc.

That is, according to the present invention, each element device has at least one of a function of broadcasting processing result information and a function of selecting information related to its own purpose, in addition to an information processing function according to its own purpose. By having this, specialized processing can be performed autonomously, and as a result, highly parallel processing is possible, and even processing with unspecified steps can be performed.

[Embodiments of the invention]

FIG. 1 shows a control system (command and control system) to which the present invention is applied.

The control target 1 includes a sensor or actuator 13-c.
(c=1.2...., f) are connected to each sensor or actuator 13-C, and each sensor or actuator 13-C is connected to a control input element or control output element 12-c (c=1.2....・・・
・, f) are connected, and each control input element or control output element 12-c is connected to the transmission control unit 1l-
It is connected to the information transmission path 5 via c (C=1, 2..., f)'. The operator 2 uses a control terminal device equipped with a display section 3 and an input section (for example, a keyboard 4) to know the status of the system and give necessary information or commands. -1 and the transmission control unit 2 through the information transmission path 5, and the input unit 4 is connected to the man-machine input element 22-2 and the transmission control unit 2.
It is connected to the information transmission line 5 via 1-2. These input/output elements 12-c, 22-1. 22-2 performs processing such as signal format conversion and data editing.

Arithmetic element 32-d (d=1. 2......, g
) are each assigned a specific and limited purpose (designation of calculation content, target data, etc.), and the transmission control unit 3
It is connected to the information transmission path 5 via 1-d (a=1. 2. . . . , g). Storage element 42-e+(
e=1. 2. ..., h) are each assigned a specific and limited purpose (designation of data to be stored, designation of fixed data and programs for control, etc.), and are used for transmission control. Part 4l-e (e=1°2,...
. . . are connected to the information transmission path 5 via h[-. The designation of these purposes and the provision of information to be held in advance can be performed by the operator 2 in advance via the input unit 4, for example. Transmission control unit 11-C.

21-1. 21-2. 31-d, 41-e and the information transmission line 5 are a known autonomous decentralized transmission system (Special Publication No. 58
-47111, JP-A-56-111353, etc.).

An overview of the information processing in the system shown in FIG. 1 will first be described with reference to an example of train V-pass and delay situation monitoring in a train operation management system. FIG. 2 is a track diagram showing the train operation status. The symbols R+1 and R2 each represent a train, and 8+ and S, ya, and V- represent a train, respectively.
sAt and AI+1 each represent a switch at a branch point requiring course control.

Train detection device SI and SIya! are respectively sensor 3-
1 and 13-i+1, switch machines A1 and A++
1 is the actuator 13 = +2゜13-i+3
Suppose it is connected to.

Now, when an event occurs that train R2 passes point 8+ at time Tl, input element 12-1 is activated and information T+ (St, R/2) representing the event is sent to transmission control unit 11-i. It is sent to the transmission line 5 via f. The sent information is sent to each transmission control unit 11-C.

21-1. 21-2. 31-d and 41-ef, and then broadcast to all elements. Transmission line 5
The information flowing over takes the form of packets, and each packet is
As shown in FIG. 3, it consists of a header and information content. The header includes a source code and an information content code, and the information content code is a group of codes representing the gist of the information content. For example, in the case of the above information T+ (81, R2),
This indicates that the information is the actual passing time of train R2 at point SI.

Broadcast information Tl (S t , R2)
consists of a calculation element 32-1 to which the purpose of delay time calculation at point Ss is assigned, a storage element 42-1 that holds data on the planned passage time of train R2 at point SI, and a queue process for turning machine A1. are selectively received by the calculation elements 32=+2 to which the purpose of is assigned. Computation element 3
2=-+-z is the output element 12=+ which broadcasts this information according to the order of the queue and is responsible for the switch Al.
2 selectively receives this information from the calculation element 32-i+2 and transmits it to the switch A+ via the actuator 13-1+ze.
i control and guide train R2 to a predetermined course. The storage element 42-1 still has information Tl (St.

Rz), the planned passage time T+0(St, R2)'fr of train R2 at point S1 is broadcast. The calculation element 32-1 performs this Tlo(Sl).

R2)' is selectively received and, knowing that the necessary operands are available, calculates the delay time [Tlo-T1=ΔTt)e and broadcasts the information ΔTr(81, R2)'ff. The storage element 42-i+2 for storing the delay time at the point SL of the train R2 selectively receives and stores this delay time information. Thereafter, when the operator requests monitoring of the delay time of train R2 at point S1 and instructs the input unit 4 to that effect, this monitor request information is sent to the operator's man-machine input element 22-2 into a packet with an appropriate header. will be edited and broadcast. The storage element 42-i+2 selectively receives this monitor request information and stores the information ΔT+ stored therein.
(si, R2)k broadcast.

The man-machine output element 22-1 selectively receives this information, performs necessary processing such as signal conversion, and then displays the display unit 3.
send to

Next, when an event occurs that the train R1 passes the point S++tk at time T2, the input element 12-11 is activated and the information Tz (S+++, Rt) is broadcast, and this information is transmitted to the point S++tk. The calculation element 32-i+1 for delay time calculation corresponding to SI+1 and the point S
The data is selectively received by the storage element 42-i+1 for planned passage time data of the train R1 at L+1 and the queue processing calculation element 32-i+3 for the switch A1. The calculation element 32-1+3 broadcasts this information according to the order of the queue, and the output element 12-i+3 selectively receives this information, and transmits it to the switch A, +++e via the actuator 13-1-+-31. The train R+'Th is controlled and guided to a predetermined course. On the other hand, the storage element 42-1+t stores information T2
According to (S +++, Rt ), the planned passing time T2° (St. +, Rt ) of trains R and I at point S +++
) to broadcast. The arithmetic element 32-i+1, which has received all selections, knows that the operands are complete, calculates the delay time CT2°-T2=ΔT2], and broadcasts the information ΔT2(St, +, Rt)k. The storage element 42-i+3 for storing the delay time at the point S1 or 1 of the train R1 selectively receives and stores this information. The rest is the same as last time.

Since the processing related to the occurring event at point S1 at time T1 and the processing related to the occurring event at point S+++ at time T2 are performed independently and in parallel,
There is no problem even if TI and T2 are at the same time. Maximum,
A number of arithmetic operations equal to the number of arithmetic elements are simultaneously executed in parallel within a time difference comparable to the transmission delay time. Also,
Since there is no need to strictly position the processing by individual elements within the overall processing, it is possible to perform even diverse processing that is difficult to formulate. The same process regarding the same event may be performed by multiple elements at the same time, and the results may be mutually checked.

FIG. 4 shows an example of the calculation element 32-d. The header decoder 11 sends the information contents of the received information bucket to the route changeover switch 102, decodes the header,
The result is sent to the opening speed determining section 111. State control unit 11
2 has a built-in indicator that displays the state of the element, and sends a control signal to each part in all the elements according to the state and the received signal. The opening speed determination unit 111 examines the output of the header decoder 101 and the status display information from the status control unit 112, determines the degree of association between the received information and the element, and returns the result to the status control unit 112.

The first operand registers 1 and 03 and the second operand register 104 are V registers for holding the first operand and the second operand, respectively, and the outputs of both registers are calculated by the calculation unit 105. The output of the calculation unit 105 is sent to the header encoder 107 via the path changeover switch 106 or returned to the first operand V register 103 through the path changeover switch 102. The header encoder 107 adds a header to the output of the path changeover switch 106 and sends it to the transmission control section 31-d. The number of operand V registers may be increased or decreased as necessary.

A storage device may also be added for receiving from the storage element a program for controlling the processing steps.

Next, the operation of this calculation element will be explained assuming that it is the calculation element 32-1 for calculating the delay time of the point SL. A flowchart of this operation is shown in FIG. In this case, the possible states of the arithmetic element are a first state in which it does not contain any operands, a second state in which it contains only the first operand T10, and a second state in which it contains only the second operand TI. There are three states and a fourth state where both operands are present and delay time calculation is being executed. Assume that the calculation element is in the first state as an initial state. Input element 12-
The information T+ (Sl, R
Upon receiving the message 2), the header decoder 101 decodes the header.

The open speed determination unit 111 receives the header decoding output from the header decoder 101 and the status signal from the status control unit 112, and determines whether the source of the packet is the input element 12-1 and the content is at the train point sI. Actual passing time,
Since the element is in the first state, the information content is related to the element, specifically, it is determined that it is the second operand to be processed, and the result is sent to the state controller. Pass it to 112. Since the received information is the second operand, the state control unit 112 selects the route changeover switch 1.
02 and transfers this information to the second operand V register 1.
Store in 04. Is the state indicator in the third state? Updated as shown. As a result of the opening speed determination, the other calculation elements learn that the received information is not relevant to themselves, and discard it.

Then, upon receiving the packet of information Ti' (S t , R2) broadcast from the storage element 42-1, the header decoder 111 decodes the header,
The opening speed determination unit 112 determines that the packet source is the storage element 4.
In 2-1, the information content is the planned passage time of the train at the point SI, and the element is in the third state, so the information content is related to the element, and specifically, It is determined that this is the first operand to be processed, and the result is passed to the state control unit 112. The state control unit 112 switches the path changeover switch 102 in response to reception of the first operand, and stores the received information in the first operand V register 103. The status indicator is the fourth
Updated to indicate status.

When the fourth state is reached, the state control unit 112 activates the calculation unit 105, and the calculation unit 105 executes a predetermined calculation, that is, subtraction of the second operand from the first operand, and the difference output is Changeover switch 106, 102'!
i- and stored in the first operand register. Upon receiving the computation end signal from the computation unit 105, the state control unit 112 switches the path changeover switch 106 and instructs the header encoder 107 to select the element 32-'-ii.
The source code and information ΔT+ (81, R2)
Then, the state control device sends the contents of the first operation second register 103 (the difference which is the calculation result) to the header encoder through the calculation unit 105 and the path changeover switch 106. 107. The calculation result with the header attached here is sent to the transmission control unit 3l.
-d (in this example, 3l-i) is broadcast via i. The status indicator is updated to indicate the first status.

FIG. 6 shows an example of the storage element 42-e. Header decoder 301, header encoder 307, opening speed determination unit 3
11 and the state control section 312 are basically the same as the corresponding components in FIG. Information to be written is written into a one-word storage area 303 through a write gate 302, and its contents are read out through a read gate 304.

The operation will be described assuming that this storage element is an element 42-i+2 for storing delay time at point S1 of train R2. A flowchart of this operation is shown in FIG. In this case, the states of the storage element include a writable standby state (first state) and a holding state (second state), and operations that can be performed include writing delay time information in the first state. , simple read in the second state, and erase after read. The simple read operation is in response to a monitor request from the operator, and the read-and-delete operation is performed when data is recorded (for example, transferred to an external storage device) after the target train has arrived.

Assume that this storage element is in the first state. Computing element 32-
Delay time information ΔT1 (
SL, R2) is received and its header is decoded by the header decoder 301, the opening speed determining unit 311 determines that the information content is the delay time at the point SL of the train R2 and the state of the element is the first one. Since it is in the 1 state, it is determined that it is related to the element in question and, specifically, information that should be written there, and the result is passed to the state control unit 312. The state control unit 312 controls the write gate 302 to write the received information content to the storage area 30.
3 to update all status indicators to the second status.

Thereafter, when the header decoder 301 detects the delay monitor request information for the train R2 at the point S1 from the man-machine input element 22-2, the opening speed discrimination circuit 3
11 is based on the output of the header decoder 301 and the state indicator output indicating the second state, and this information is related to the element in question, specifically, a monitor request for stored information (delay time of train R2 at point S+). , and notifies the state control unit 312 of this fact.

Therefore, the state control unit 312 controls the read gate 304 to read the contents of the storage area 303, adds an appropriate header with the header encoder 307, and then broadcasts the contents. For read operations performed in response to monitor requests, no state updates occur.

Next, when information indicating the end of train R2's operation is broadcast from the appropriate input element 12-r, the header decoder 301 determines that the received information indicates the end of train R2's operation and that the element is in the second state. Based on this fact, the state control unit 31 determines that the information is not related to the element, specifically, that it is necessary to record the memory information.
Tell 2. In response, the state control unit 312 controls the read gate 304'i, sends the contents of the storage area 303 to the header encoder 307, adds an appropriate header and broadcasts, and at the same time clears the storage area 303. The status indicator is updated to status 1. Thereafter, this storage element is either inactive until the next run of train R2, or used for other purposes as specified by the operator. In the latter case, the second purpose may be set in advance in the storage element.

In the case of an input element, a part related to information reception may be omitted, and conversely, in the case of an output element, a part related to information transmission may be omitted. In the case of these elements, format conversion of information (A/D conversion, code conversion, format conversion) is performed instead of the calculation mechanism in the calculation element or the storage mechanism in the storage element.

(editing, etc.)).

〔Effect of the invention〕

According to the present invention, each information processing element specialized for each purpose can operate independently and autonomously, and as a result, there are cases where a high degree of parallelization is required or where it is difficult to specify processing procedures. Even if effective information processing? can be carried out.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing an example of an information processing device according to the present invention, FIG. 2 is a train track diagram as an example of a control target, and FIG. 3 is a format diagram of an example of an information packet used in the device shown in FIG. , FIG. 4 is a block diagram of an example of a calculation element, FIG. 5 is a flowchart of an example of processing in the calculation element of FIG. 3 is a flowchart of an example of processing in a storage element in the figure. 5... Transmission line, 1l-c (c=1 to f), 21-
1°21-2. 31-d (d=1~g L 41
-e (e = 1~h)--Transmission control device, 12-c
(c=1 to f), 22-1.22-2... input/output element, 32-d (d=1 to g) ・= calculation element, 42-
e (e = 1~h)--Storage element, 101. 3
01-Header decoder, 103, 104... Operand V register, 105... Arithmetic unit, 107, 307...
- Header encoder, 111, 311... Opening speed verification unit, 112, 312... State control unit, 302... Write gate, 303... Storage area, 304... Read gate.

Claims (1)

[Scope of Claims] 1. Information processing comprising an input/output unit, an arithmetic unit, and a storage unit each consisting of a plurality of information processing elements specialized for each purpose, and a transmission line interconnecting these information processing elements. In the device, a certain information processing element broadcasts information via the transmission path, and another information processing element determines the relevance of the content of the broadcasted information to its own purpose and accepts the information. and a step of executing information processing corresponding to the purpose in the information processing element that has received the information, and broadcasting the processing result via the transmission path as necessary for the purpose. information processing method. 2. In Claim 1, the step when information is accepted in an element of the calculation unit is to save or accept the received information according to the relationship between the content of the accepted information and the state of the element. An information processing method comprising the step of performing processing corresponding to the purpose between the information obtained and the information already stored, and broadcasting the result via the transmission path. 3. In claim 1 or 2, the step of accepting information in the information processing element of the storage unit is to process the received information according to the relationship between the content of the received information and the state of the information processing element. An information processing method comprising the step of storing or reading stored information and broadcasting it via the transmission path. 4. The input/output section, calculation section, and storage section each consist of a plurality of information processing elements specialized for each purpose, and these information processing elements are interconnected to transmit the information sent by each element to all other information processing elements. a transmission network for broadcasting to the network, and each of the information processing elements includes an information processing means for performing processing corresponding to its purpose;
Information processing that is equipped with at least one of means for transmitting processing result information to the transmission network and means for selectively accepting information on the transmission network based on the relevance of the content and its own purpose, depending on the purpose. Device. 5. In claim 4, the elements of the arithmetic unit are:
determining means for determining the correlation between the content of information on the transmission network, its own purpose, and its own state; operand holding means for holding the accepted information as an operand; The operand holding means includes a calculation means for performing an operation, a sending means for sending the result of the calculation by the calculation means to the transmission network, and a means for displaying the state of the element, and the operand holding means responds to the output of the determining means. An information processing device comprising a calculation means and a state control means for controlling a sending means. 6. In claim 4 or 5, the element of the storage unit includes a determination means for determining the correlation between the content of information on the transmission network, its own purpose, and its own state, a storage means, and an receiving unit. It includes a writing means for writing information into the storage means, a sending means for reading out the contents of the storage means and sending it to the transmission network, and a means for displaying the state of the element and responding to the output of the determining means. and a state control means for controlling the writing means and the sending means.