JP2579185B2 - Data flow program execution control method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Application Field] [Prior Art] FIG. 6 is a block diagram showing an example of a data driven type information processing apparatus. FIG. 7 shows a field configuration example of a data packet in the information processing apparatus. FIG. 8 shows a part of the field configuration of the stored contents in the program storage means.

The destination field of the data packet shown in FIG. 7 stores destination information, the instruction field stores instruction information, and the data 1 field or the data 2 field stores operand data. 6 includes a program storage unit 1, a data pair generation unit 2, and an arithmetic processing unit 3. The program storage means 1 stores a data flow program shown in FIG.
As shown in FIG. 8, the program storage means 1 stores the next destination information, the next instruction information, and the copy presence / absence information by address designation based on the destination information of the input data packet in FIG. It reads and stores the destination information and the instruction information in the destination field and the instruction field of the input data packet and outputs them. The data pair generation unit 2 waits for a data packet output from the program storage unit 1. That is, two different data packets having the same destination information are detected,
The operand data of one of the two data packets having the same destination information, for example, the contents of the data 1 field in FIG. 7 are stored in the data 2 field of the other data packet and output. The arithmetic processing means 3 performs a predetermined arithmetic processing on the data packet output from the data pair generating means 2, stores the result in the data 1 field of the data packet, and outputs the result to the program storage means 1.

The program storage means 1 and the data pair generation means 2 are connected by data transmission paths 4 and 5, and the data pair generation means 2 and the arithmetic processing means 3 are connected by a data transmission path 6. The arithmetic processing means 3 and the program storage means 1 are connected by a data transmission path 7.

The data packet is stored in the program storage means 1 → data pair generation means 2 → operation processing means 3 → program storage means 1 →.
The calculation process proceeds based on the data flow program stored in the program storage means 1.

In the data flow program to be executed, when data copying is required as shown in FIG. 9, that is, when a certain operation result, for example, the output of the addition instruction in FIG. In such a case as referred to by the multiplication and subtraction instructions in FIG. As shown in FIG. 8, copy presence / absence information is stored in the program storage means 1. First, the contents of the addressed portion are read based on the destination information of the input data packet. At this time, the copy presence / absence information is "No"
If so, the data packet in which the contents of the destination field and the instruction field are updated is output, and the process ends. On the other hand, if the copy presence / absence information is “present”, a data packet in which the contents of the destination field and the instruction field are updated is output, and the destination information, the instruction information and the copy presence / absence stored subsequently are stored. The information is read. The subsequently read copy presence / absence information is "No"
If so, the same result data as the input data packet is stored in the data 1 field, and the data packet storing the currently read information in the destination field and the instruction field is output, and the process ends. If the subsequently read copy presence / absence information is “present”, the copy process is further continued.

FIG. 10 shows an example of a data flow program including a conventional selection structure. In FIG. 10, N1 to N7 are nodes. D1 to D7 are node numbers of each node, that is, destination information. op1, op2, TG, FG, and op5 to op7 are operation codes of each node, that is, instruction information. However, op
Reference numeral 2 denotes a condition determination instruction, which outputs a condition determination result according to input data as “true” / “false”. TG outputs the left input data as it is when the right input data is "true",
It has a function of outputting nothing when "false", that is, absorbing the input data pair. Conversely, the FG has a function of outputting the left input data as it is when the right input data is "false" and outputting nothing when the right input data is "true", that is, has a function of absorbing the input data pair. Accordingly, FIG. 10 shows a structure in which the node N5 is selectively executed when the output of the node N2 is "true", and the nodes N6 and N7 are selectively executed when the output is "false".

FIG. 11 shows the contents stored in the program storage means 1 corresponding to the data flow program shown in FIG.

A conventional execution control method of the selection structure will be described with reference to FIG. 10 and FIG. First, it is assumed that a data packet corresponding to the operation result of the node N1 is output from the operation processing unit 3 and input to the program storage unit 1 through the data transmission path 7. In the program storage unit 1, the storage contents shown in FIG. 11 are read out by the address designation based on the destination information D1 of the input data packet, and the destination information D3, the instruction information TG, and the copy presence / absence information “Yes” Is obtained. In this case, the program storage unit 1 outputs the next data packet having the destination information and the instruction information, and sets the next address of the address specified by D1 because the copy presence / absence information is “present”. By specifying, copy information, that is, the next destination information D4, the next instruction information FG and the next copy presence / absence information “absent” are read, and a copy having the next destination information and the next instruction information is read. Generates and outputs a data packet corresponding to. The data packet output from the program storage unit 1 is input to the data pair generation unit 2. Similarly, in the data packet having "true" / "false" as data according to the condition determination result of the node N2, the destination information and the instruction information are updated in the program storage means 1, and the data packet having the destination information D3 and the instruction information TG is updated. After a data packet having destination information D4 and instruction information FG is generated, the data packet is input to the data pair generation means 2. The data pair generation means 2 generates data pairs for the nodes N3 and N4, and inputs the data pairs to the arithmetic processing means 3, respectively. If the state determination result of the node N2 is “true”, the input data packet corresponding to the node N3, that is, the TG, passes through the arithmetic processing unit 3 as it is in the arithmetic processing unit 3,
The input data packet corresponding to the node N4, that is, the FG, is absorbed by the arithmetic processing means 3, and nothing is output from the arithmetic processing means 3. Therefore, the node N5 corresponding to the destination of the node N3 is continuously executed, but the input data does not arrive at the nodes N6 and N7 corresponding to the destination of the node N4, so that the node N5 is not executed. That is, the node N5 and the node
Selective execution for N6 and node N7 is realized.

Conversely, if the condition determination result of the node N2 is “false”, the input data packet corresponding to the node N3 is
And the input data packet corresponding to the node N4 passes through the arithmetic processing means 3 as it is, thereby selectively selecting the node N5 and the nodes N6 and N7. Execution has been realized.

As described above, the conventional selective execution of the data flow program is realized by the data processing unit 3 absorbing the data packet.

[Problems to be Solved by the Invention] In the above-described conventional execution control system of a selection structure, as shown in FIG. 10, execution of copying of "true" / "false" data representing selected data and selection conditions. And TG and FG
Is required to execute two nodes with the operation code of, and the execution efficiency of the data flow program is reduced by half compared to the case where the selection structure is realized by one node. Furthermore, in order to solve such a problem, it is desirable to reduce restrictions on the packet format as much as possible.

[Means for Solving the Problems] An execution control method for a data flow program according to the present invention is a method for controlling execution of a TG on a data flow program including a selection structure for selecting an instruction group to be executed according to a predetermined condition.
And the selection function realized by the two nodes having the operation code of FG is represented by one next node, and the destination information of the data packet for the next node input to the arithmetic processing means is inputted to the predetermined node. By operating on the basis of the above condition, the destination of the data packet is selectively designated. According to the prescribed conditions,
When executing a data flow program including a selection structure in which an instruction group to be executed on selected data is selected, an area of program storage means accessed by a data packet including condition data, and including selected data In the area of the program storage means accessed by the data packet, the same destination information and the same instruction information, which is to perform a predetermined operation on the destination information in accordance with the contents of the condition data, are stored. The instruction information includes relative position information between two destinations selected based on the condition data. In the arithmetic processing means,
If the instruction information of the input data packet is instruction information corresponding to the above-described selection structure, the relative position information is added to the destination information of the data packet or the destination information of the data packet according to the content of the condition data. Is selected, and the destination of the data packet is specified.

[Operation] By adopting the above-described execution control method, it is not necessary to execute the copying of the "true" / "false" data representing the selected data and the selection condition, which is necessary for realizing the conventional selection structure. Furthermore, a function equivalent to a function realized by two nodes using TG and FG as operation codes is realized by one node. For this reason, the amount of originally unnecessary data flowing on the information processing device can be reduced, the number of executed instructions can be reduced, and the execution efficiency of the data flow program can be increased. Further, by reducing the number of executed instructions, the storage capacity of the program storage unit can be effectively used. Also, since the command information only includes one destination information and the relative position information between the two selected destinations, it does not include the position information of the destination itself.
An area for storing information for designating two destinations is small.

EXAMPLES Hereinafter, the present invention will be described in detail based on examples.

The block diagram of the information processing apparatus according to one embodiment of the present invention is the same as the block diagram of the conventional example shown in FIG. The field configuration of the data packet and the field configuration of the storage contents in the program storage means 1 in this embodiment are the same as those in the conventional example shown in FIGS. 7 and 8, respectively.

FIG. 1 shows a conversion rule of a data flow program including a selection structure in this embodiment. In FIG. 1, n is the number of copies of the output of the node having the TG as the operation code in the conventional example, that is, a number representing the number of destinations of the data packet output from the node. In FIG. 1, a node having an operation code of SWn (n = 1, 2, 3,...) Has a logically complex function of two nodes having operation codes of TG and FG. That is, when the right input data is “true”, the node having the operation code SWn outputs the left input data as it is to the left output and outputs nothing to the right output, and conversely, the right input data is “false”. "", The function has the function of outputting the left input data to the right output as it is and outputting nothing to the left output. The operation code SWn is individually set according to the number of copies of the left output.

FIG. 2 shows an example of a data flow program including a selection structure in this embodiment. However, for the sake of simplicity, a case where the output of a node having an operation code of TG is not copied in the conventional example, that is, a case where n = 1 will be described. FIG. 2 shows an example in which the conversion rule shown in FIG. 1 is applied to a data flow program including a selection structure in the conventional example shown in FIG. 10 with n = 1.

FIG. 3 shows an example of the storage contents of the program storage means 1 corresponding to the data flow program shown in FIG. In FIG. 3, the address specified by the destination information D8 is SW
Information regarding the destination of the left output of the node N8 having an operation code of 1 is stored. Further, after the address specified by D8, that is, after the address specified by D8 + 1, information on the right output destination of the node N8 having SW1 as the operation code is stored.

Next, an execution control method of the selection structure in this embodiment will be described. First, it is assumed that a data packet corresponding to the operation result of the node N1 is output from the operation processing unit 3 and input to the program storage unit 1 through the data transmission path 7. In the program storage means 1, by specifying the address based on the destination information D1 of the input data packet, the storage contents shown in FIG. 3 are read out, and the destination information D8, the instruction information SW1, and the copy presence / absence information "None""
Is obtained. In this case, since the copy presence / absence information is "absence", the program storage means 1 outputs only the next data packet having the destination information and the instruction information.
The data packet output from the program storage unit 1 is input to the data pair generation unit 2. Similarly, the node
"True" / "False" as data according to the condition judgment result of N2
After the destination information and the instruction information are updated to D8 and SW1, respectively, in the program storage unit 1, the data packet having is stored in the data pair generation unit 2. The data pair generation means 2 generates a data pair for the node N8 having SW1 as an operation code, and inputs the data pair to the arithmetic processing means 3. If the condition determination result of the node N2 is “true”, the arithmetic processing means 3
That is, the left input data packet corresponding to SW1 is output from the arithmetic processing means 3 as it is, and is again input to the program storage means 1. Since the input data packet to the program storage means 1 has the destination information D8, the next destination information D5 and the instruction information op5 are designated in the program storage means 1 by addressing based on the destination information D8.
And the presence / absence of copy information “absence” is read out, and a data packet having the next destination information D5 and instruction information op5 is output. This means that the left input data packet of the node N8 is selectively output to the left output of the node N8 on the data flow program. Conversely, if the condition determination result of the node N2 is “false”, the arithmetic processing means 3 adds 1 to the destination information of the input data packet corresponding to the node N8, that is, SW1, and updates the destination information to D8 + 1. It is output from the arithmetic processing means 3 and input to the program storage means 1 again. Since the input data packet to the program storage means 1 has the destination information D8 + 1, the destination data D8 + 1 is stored in the program storage means 1.
The next destination information D6, instruction information op6 and copy presence / absence information "present" are read out by the address designation based on
A data packet having the next-order destination information D6 and instruction information op6 is output. Furthermore, if the above copy presence / absence information is “Yes”
Therefore, a copy process is subsequently performed, and a data packet having destination information D7 and instruction information op7 is output. This means that the left input data packet of the node N8 is selectively output to the right output of the node N8 on the data flow program.

Next, a case where the output of a node having an operation code of TG in the conventional example is copied will be described.

FIG. 4 shows another example of the data flow program including the selection structure in this embodiment. In FIG. 4, the output of a node having an operation code of TG in the conventional example, that is, the left output of the node representing the selection instruction in this embodiment is copied, and the number of copies is n = 2. Therefore, the operation code of the node representing the selection instruction is SW2.

FIG. 5 shows an example of contents stored in the program storage means 1 corresponding to the data flow program shown in FIG. Since the destination of the left output of the node N9 is copied, the address specified by the destination information D9 and the next address (D9 + 1) in FIG. Is stored. Further, after the next address (D9 + 2), information on the destination of the right output of the node N9 having the operation code of SW2 is stored.

In the operation until the data pair generation means 2 generates a data pair for the node N9 having SW2 as the operation code and inputs the data pair to the arithmetic processing means 3, the output of the node having SW2 as the operation code is not copied. Same as in the case. If the condition determination result of the node N2 is “true”, the arithmetic processing means 3 causes the node N9, that is, the left input data corresponding to SW2 to The packet is output from the arithmetic processing means 3 as it is, and is again input to the program storage means 1. Since the input data packet to the program storage means 1 has the destination information D9, the address designation based on the destination information D9 in the program storage means 1 causes the
Input data packets corresponding to N5 and node N5 'are generated. Conversely, if the condition determination result of the node N2 is “false”, the arithmetic processing means 3 causes the node N9, ie, SW
The destination information of the input data packet corresponding to 2 is updated to D9 + 2 by adding 2, and then output from the arithmetic processing means 3 and input again to the program storage means 1. The input data packet to the program storage means 1 is the destination information
D9 + 2, the program storage means 1 specifies the address based on the destination information D9 + 2,
Input data packets for nodes N6 and N7 are generated.

The above is the case where the left output of a node having a selection instruction as an operation code is not copied, and the case where the left output is copied and input to two nodes, that is, the number of copies is N
= 1 and N = 2, the same applies to the case where the left output of a node whose operation code is a selection instruction is copied and input to any of n nodes. When the right input data of the input data packet of the set node is “false”, the arithmetic processing means 3 updates the value of the destination information of the input data packet by adding n and outputs it from the arithmetic processing means 3 Implementation of the selection structure can be realized.

[Effects of the Invention] As described above in detail, if the selective structure execution control method according to the present invention is employed in the data driven type information processing apparatus, the number of copies and the number of executed instructions can be reduced as compared with the related art. Thus, the execution efficiency of the data flow program is improved. Further, by reducing the number of executed instructions, the storage capacity of the program storage unit can be effectively used. Also, by simply including the relative position information between the two destinations in the instruction information, it is possible to specify both of the two destinations in the selection structure, which is compared with the case where the two destination information are stored as they are. Therefore, the required storage area is small. Therefore, the packet format of the data packet can be the same as that of the conventional one, and there is no need to perform complicated processing to determine the destination.

[Brief description of the drawings]

FIG. 1 is a diagram showing a conversion rule of a selection structure in one embodiment of the present invention. FIG. 2 is a diagram showing a first example of a data flow program including a selection structure in the embodiment. FIG. 3 is a diagram showing the contents stored in a program storage means corresponding to the data flow program shown in FIG.
FIG. 4 is a diagram showing a second example of the data flow program including the selection structure in the embodiment. FIG. 5 is a diagram showing storage contents of a program storage means corresponding to the data flow program shown in FIG. FIG. 6 is a block diagram of a data driven type information processing apparatus according to a conventional example and an embodiment of the present invention. FIG. 7 is a diagram showing a field configuration of a data packet in the conventional example and the embodiment of the present invention. FIG. 8 is a diagram showing the field configuration of the storage contents of the program storage means in the conventional example and the embodiment of the present invention. FIG. 9 is a diagram showing an example of a data flow program including a copy process. FIG. 10 is a diagram showing an example of a data flow program including a selection structure in a conventional example. FIG. 11 is a diagram showing the storage contents of a program storage means corresponding to the data flow program shown in FIG. In the figure, 1 is a program storage means, 2 is a data pair generation means, 3 is an arithmetic processing means, 4 to 7 are data transmission paths, N1
~ N9, N5 'are nodes, D1 ~ D9, D5' are destination information, op1, op2,
op5 to op7, op5 ', TG, FG, SW1, and SW2 indicate operation codes.

Claims (1)

(57) [Claims] A data flow program comprising a plurality of destination information and a plurality of instruction information is stored, and a data packet comprising a destination field, an instruction field, a first data field, and a second data field is inputted, and the data packet is inputted. By specifying an address based on the contents of the destination field of the packet, the next destination information and the next instruction information of the data flow program are read, and those informations are respectively stored in the destination field and the instruction field of the data packet. A program storage unit for storing and outputting the data packet; and a data packet output from the program storage unit for queuing the data of the first data field of one of the two data packets having the same destination information. The other data packet Second
Data pair generating means for storing the other data packet stored in the data field, and inputting the data packet output from the data pair generating means, decoding the instruction information in the instruction field of the data packet, Information processing means for performing predetermined arithmetic processing on the two data of the first and second data fields, storing the result in the first data field of the data packet, and outputting the result; In the processing device, when executing a data flow program including a selection structure in which an instruction group to be executed on selected data is selected according to a predetermined condition, the program storage accessed by a data packet including the condition data The area of the means and before being accessed by the data packet containing the selected data In the area of the program storage means, the same destination information and the same instruction information, which is to perform a predetermined operation on the destination information in accordance with the content of the condition data, are stored, and the instruction information is The relative processing information including the relative position information between the two destinations selected based on the condition data; and if the instruction information of the input data packet is the instruction information, the content of the condition data The destination of the data packet is specified by selecting either the destination information of the data packet or the destination information obtained by adding the relative position information to the destination information of the data packet. , A data flow program execution control method.