JPH02270086A - Data flow type information processor - Google Patents

Abstract

PURPOSE:To reduce the width of a data line connecting respective processing parts by linking, generating only irreducible minimum data, which are required for processing in each processing part, and sending the data to a transmission line with a data packet passing through the respective processing parts. CONSTITUTION:A program storage part 1 and an identification tag separation part 21 are linked by a transmission line 31 and an arithmetic processing part 3 and the identification tag separation part 21 are linked by a transmission line 32. The identification tag separation part 21 and a pair data detection part 2 are linked by two transmission lines 33 and 34 and the pair data detection part 2 and an identification tag applying part 22 are linked by a transmission line 35. Further, the identification tag applying part 22 and program storage part 1 are linked by a transmission line 36 and the identification tag applying part 22 and arithmetic processing part 3 are linked by a transmission line 37. Namely, data separated by a branching part are linked according to identification information, which are applied to the data, and the irreducible minimum data required for each processing part are transmitted. Thus, the width of the data line connecting the respective processing parts can be reduced.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention relates to an information processing device, and in particular, to an information processing device that executes a data flow program that includes an instruction to perform numerical operation processing or logical operation processing on two pieces of data. The present invention relates to a data flow type information processing device.

[Prior Art] In a conventional Neumann type computer, various instructions are stored in a program memory in advance as a program, and the instructions are sequentially read out by sequentially specifying addresses in the program memory by a program counter, and the instructions are executed. be done.

On the other hand, a data flow type information processing device is a type of non-Neumann type computer that does not have the concept of sequential execution of instructions using a program counter.

Such a data flow type information processing device employs an architecture based on parallel processing of instructions. In a data flow type information processing device, instructions can be executed as soon as the data to be operated on is available, and multiple instructions are simultaneously driven by the data, so programs are executed in parallel according to the natural flow of data. be done. As a result, the time required for calculations is considered to be significantly reduced.

FIG. 3 is a block diagram showing an example of the configuration of a conventional data flow type information processing device. Further, FIG. 4 is a diagram showing an example of the field configuration of a data packet processed by the information processing device.

The data packet 60 shown in FIG. 4 includes a destination field 41, a command field 42, and a data 1 field 43.
, and a data 2 field 44. Destination field 41 stores destination information, command field 42 stores command information, and data 1 field 43 or data 2 field 44 stores operand data.

The information processing apparatus shown in FIG.
, a paired data detection section 2, and an arithmetic processing section 3. The program storage unit 1 stores a data flow program 70 shown in FIG.

The program storage unit 1 stores input data packets 60
By specifying the address based on the destination information of the data flow program 70, as shown in FIG.
1. Read command information 72 and copy, -presence/absence information 73, store destination information 71 and command information 72 in destination field 41 and command field 42 of data packet 60, respectively, and output data packet 60.

The data pair detection section 2 waits for data packets 60 to be output from the program storage section 1.

That is, two different data packets 60 having the same destination information are detected, and the operand data of one of the data packets 60, for example, the contents of the data 1 field 43 in FIG.
It is stored in the data 2 field 44 of the other data packet, and the other data packet 60 is output. The arithmetic processing unit 3 performs arithmetic processing on the data packet 60 output from the paired data detection unit 2 based on the instruction information, stores the result in the data 1 field 43 of the data packet, and stores the result in the data packet 60. ° is output to the program storage section 1.

Note that the program storage section 1 and the paired data detection section 2 are coupled by a data transmission path 4. Data transmission path 4 is 2
It branches into two input ports and is coupled to two input ports of the paired data detection section 2. The data packet 60 output from the program storage unit 1 is selectively input to one of the paired data detection units 2, depending on whether the operand data is right operand data or left operand data in arithmetic processing. The port is man-powered. Furthermore, the paired data detection section 2 and the arithmetic processing section 3 are coupled through a transmission path 5, and the arithmetic processing section 3 and the program storage section 1 are coupled through a transmission path 6.

The data packet 60 is stored in the program storage unit 1, the paired data detection unit 2, the arithmetic processing unit 3, and the program storage unit 1.
By continuing to rotate in order, arithmetic processing based on the data flow program 70 stored in the program storage unit 1 progresses.

In the data flow program 70 to be executed, the sixth
Consider the case where data copying is required as shown in the figure. That is, when the result of an operation, such as the output of the addition instruction in FIG. 6, is referenced by two or more other operations, such as the multiplication and subtraction instructions in FIG. . As shown in FIG.
Copy presence/absence information 73 is stored in .

First, the contents of the portion addressed based on the destination information of the input data packet 60 are read from the data flow program 70. At this time, if the copy existence/non-copy information 73 is "absent", the data packet 6 whose contents of the destination field 41 and the command field 42 have been updated
o is output and the process ends. On the other hand, if the copy existence/non-copy information 73 is “Yes”, the data packet 6 in which the contents of the destination field 41 and the command field 42 have been updated
0 is output, and subsequently the stored destination information 71, command information 72, and copy presence/absence information 73 are read out. Copy presence/absence information 73 subsequently read out
If "None", the same data as the input data packet 60 is stored in the data 1 field 43, and the currently read destination information 71 and command information 72 are stored in the destination field 41 and command field 42, respectively. The stored data packet 60 is output and the process ends. The copy presence/absence information 73 read out subsequently indicates “Yes”.
If so, the same copying process is continued.

In the above information processing apparatus, once the copying process is performed, the transmission path 6 for input to the program storage section 1 and the transmission path 4 for output from the program storage section 1 are transmitted.
The flow rate ratio of the data packet 60 with the data packet 60 is always 1:2. Here, the flow rate ratio of data packets 60 refers to the number of data packets 60 that pass through those transmission lines per unit time.
is the ratio of the numbers of FIG. 7 shows the flow rate of data packets 60 on each transmission path when the flow rate of data packets 60 on transmission path 4 is set to 1. The data pair detection unit 2 outputs one data packet 60 in response to the input of two data packets 60. Therefore, even if it is assumed that the transmission line 4 operates at maximum performance, only half the flow rate of the maximum performance is guaranteed from the transmission line 5 onwards.

In this way, when the copy process is performed in the program storage unit 1, the transmission lines 5 and 6 are operated at the maximum performance of 2.
It only works at half the performance.

For this reason, the arithmetic processing unit 3 also has half of its potential.
As a result, the performance of the information processing device when executing a program is reduced.

FIG. 8 is a block diagram showing another example of the configuration of a conventional data flow type information processing device.

The information processing device shown in FIG.
1 and 12 are provided. A data packet 60 output from the arithmetic processing section 3 is inputted to either one of the program storage sections 11 and 12 via the allocation section 13.

The arithmetic processing section 3 and the allocation section 13 are coupled by a transmission path 6. The allocation section 13 is coupled to the program storage sections 11 and 12 via transmission lines 16 and 17, respectively. Further, the program storage units 11 and 12 are individually coupled to the paired data detection unit 2 via transmission lines 14 and 15, respectively. The functions of each part of the information processing apparatus shown in FIG. 8 are exactly the same as the functions of each part of the information processing apparatus shown in FIG.

The allocation unit 13 allocates the data packets 60 output from the arithmetic processing unit 3 to either of the two program storage units 11 and 12 according to the order of arrival.

FIG. 9 shows the flow rate of data packets 60 on each transmission path, assuming that the flow rate of data packets 60 on transmission paths 14 and 15 is 1, respectively. In the information processing apparatus shown in FIG. 8, the input of the data packet 60 to the program storage units 11 and 12 is distributed by the allocation unit 13, so that the data packet 60 is input to the program storage units 11 and 12 at a rate of 1/2 of the maximum performance. A data packet 60 is input. When the copy process is performed in the program storage units 11 and 12, the flow rate of data packets 60 output from the program storage units 11 and 12 to the transmission lines 14 and 15 is changed to This is twice the flow rate of data packets 60 input to 12.

That is, in the transmission lines 14 and 15 for inputting the data packet 60 to the data detection section 2, the flow rate of maximum performance can be ensured.

Moreover, the flow rate with maximum performance can be ensured also in the transmission path 5 for the data packet 60 output from the data pair detection section 2, that is, the transmission path 5 for the data packet 60 inputted to the arithmetic processing section 3.

Therefore, in the conventional information processing device shown in FIG.
The performance of the arithmetic processing unit 3 is maximized, and the effect of parallel processing is theoretically twice that of the conventional information processing device shown in FIG.

[Problem to be Solved by the Invention] In the conventional data flow type information processing device, in order to improve the processing efficiency of the data flow type information processing device shown in the above-mentioned FIG. In addition, it has become possible to create a data flow type information processing device with dual program storage units.

However, in the data flow type information processing device shown in FIG. 8, the program storage section 1 is duplicated, so the total storage capacity of the program storage section 1 is smaller than that of the device shown in FIG. It will be doubled. Therefore, there is a problem in that the information processing apparatus itself becomes large-scale.

Furthermore, based on the execution order of the data flow program 70 stored in the program storage unit 1, the data packets 60 that circulate among the processing units in the device in order are always processed in the form shown in FIG. 4 above. Therefore, the width of the data lines (paths) connecting each processing section was wasted. In other words, the data flow program 70 originally from the program storage unit 1
, the data packet 60 pointing to the next processing instruction is read.
Only the data in the destination field 41 in the data packet 60 was necessary, and the data in the command field 42, data 1 field 43, and data 2 field 44 in the data packet 60 were unnecessary. In addition, the arithmetic processing unit 3 also requires the data of the instruction field 42, data 1 field 43, and data 2 field 44 in the data packet 60, and the data of the destination field 41 in the data packet 60.
data was unnecessary. Therefore, there is a problem in that a large amount of physical waste is generated when configuring the device.

Therefore, an object of the present invention is to reduce the expansion of the physical scale of the device itself by configuring data packets that circulate between each processing section in the device with the minimum amount of data necessary for each processing section. The object of the present invention is to provide a data flow type information processing device that can perform the following tasks.

[Means for Solving the Problems] A data flow type information processing device according to the present invention is a data flow type information processing device that performs information processing based on data including a portion necessary for calculation and a portion necessary for reading a program. A program storage means for storing a data flow program, an arithmetic processing means for performing arithmetic processing on data according to the data flow program, information read from the program storage means, and a result computed by the arithmetic processing means. merging means for combining; paired data detection means for converting the data combined by the merging means into a form that can be processed; and output from the paired data detecting means into a portion necessary for calculation and a portion necessary for program reading. and branching means for separating into two parts.

[Operation] In the information processing device according to the present invention, there is a branching unit in the device that adds identification information to new data and separates the data, and a branching unit that separates data by adding identification information to the data, and converts the data separated by the branching unit into the identification information attached to the data. Since a merging section is provided for connecting each processing section and the minimum amount of data required by each processing section is sent, the width of the data line connecting each processing section can be reduced.

[Example] Hereinafter, one embodiment of the present invention will be described with reference to the drawings.

FIG. 1 is a diagram showing the configuration of a data flow type information processing apparatus according to an embodiment of the present invention.

In FIG. 1, the processing contents of the program storage section 1, paired data detection section 2, and arithmetic processing section 3 are similar to those of the conventional information processing apparatus. In this embodiment, an identification tag separating section 21 is located at the entrance of the paired data detecting section 2, and an identification tag attaching section 22 is located at the exit of the paired data detecting section 2.

In this embodiment, the output of the program storage section 1 and the output of the arithmetic processing section 3 are connected to two inputs of the identification tag separation section 21, and the two outputs of the identification tag separation section 21 are connected to the output of the paired data detection section 2. Connected to two inputs. Further, the output of the paired data detection section 2 is connected to the input of the identification tag adding section 22, and the two outputs of the identification tag adding section 22 are connected to the respective inputs of the program storage section 1 and the arithmetic processing section 3.

At this time, the program storage section 1 and the identification tag separation section 21 are coupled through a transmission path 31, and the arithmetic processing section 3 and the identification tag separation section 21 are connected through a transmission path 32. Further, the identification tag separation section 21 and the paired data detection section 2 are coupled by two transmission paths, a transmission path 33 and a transmission path 34, and the paired data detection section 2 and the identification tag attaching section 22 are connected to each other by a transmission path 33 and a transmission path 34, which are two transmission paths.
5. Further, the identification tag attaching section 22 and the program storage section 1 are coupled through a transmission line 36, and the identification tag attaching section 22 and the arithmetic processing section 3 are coupled through a transmission line 37.

FIG. 2 is a diagram showing the field structure of a data packet processed by the information processing apparatus of FIG. 1. Each data packet shown in FIG. 2 is generated by each processing section shown in FIG. 1. In FIG. 2, data packet 6
1 consists of an identification tag 50, an instruction field 42, a data 1 field 43, and a data 2 field 44.

The identification tag 50 serves as one of the marks when connecting two data packets. The data packet 62 is
It consists of an identification tag 50 and a destination field 41. Data packet 63 consists of identification tag 50 and result data field 45. The result data feed 45 stores the results of calculations performed by the calculation processing section 3. Data packet 64 consists of identification tag 50, destination field 41 and command field 42. The identification tag 50 serves as one of the marks when connecting two data packets, and the result data field 45 stores the calculation results performed by the calculation processing section 3. The other fields store the same data as before.

Next, the processing operation of the information processing apparatus shown in FIG. 1 will be described with reference to FIGS. 1, 2, 4, and 5.

First, a data packet 60 is sent from the program storage section 1 to the identification tag separation section 21 . In the data node 60, data to be subjected to arithmetic processing is stored in a data 1 field 43 or a data 2 field 44. At this time, since the identification tag 50 is not attached to the data kerat 60, the identification tag 50 is not separated by the identification tag separation section 21 and is sent to the data detection section 2 via the transmission line 33. The data detection unit 2 waits for packets whose destination fields 41 match,
When matching packets are collected, a data packet 60 is generated using the same process as in the conventional process.

The data packet 60 generated by the data detection unit 2 is
It is sent to the identification tag attaching section 22 via the transmission path 35.

At this time, the data packet 60 transmits the next instruction to the data flow program 7 stored in the program storage section 1.
Specify the specified address for reading from 0 in the destination field 41.
It is stored in. In addition, the content of the data operation is stored in the instruction field 42, and the data to be operated on is stored in the data 1 field 43 or the data 2 field 44 (in operations that require two data, they are used in pairs, and one data One is used in calculations).

The identification tag attaching unit 22 divides the data packet 60 into a data packet 61 and a data packet 62.

The identification tag 50 attached to the data packet 61 and the data packet 62 indicates that both packets are the same data packet 6.
This indicates that the pair is generated from 0. The data packet 62 generated by the identification tag adding unit 22 is sent to the program storage unit 1 via the transmission line 36 in order to read the next instruction from the data flow program 70 stored in the program storage unit 1. It will be done. Further, the data packet 61 is sent to the transmission path 3 in order to perform data arithmetic processing.
The data is sent to the arithmetic processing unit 3 via 7.

The program storage section 1 generates a data packet 64 based on the information of the data packet 62 sent from the identification tag attaching section 22. That is, destination information 71 of the data flow program 70 specified by the address stored in the destination field 41 of the data packet 62 is stored in the destination field 41 of the data packet 64.

Further, command information 72 of the data flow program 70 is stored in the command field 42 of the data packet 64. The identification tag 50 of the data packet 62 is attached to the identification tag 50 of the data packet 64.

The data packet 64 generated in the program storage section 1 is sent to the identification tag separation section 21 via the transmission path 31.

The arithmetic processing section 3 generates a data packet 63 based on the information of the data packet 61 sent from the identification tag attaching section 22. That is, based on the operation contents stored in the command field 42 of the data packet 61, using the data stored in the data 1 field 43 and the data packet 2 field 44 of the data packet 61,
Performs calculation processing. The calculation result is stored in the result data field 45 of the data packet 63. The identification tag 50 of the data packet 61 is attached to the identification tag 50 of the data packet 63. The data packet 63 generated by the arithmetic processing section 3 is sent to the identification tag separation section 21 via the transmission path 32.

In the identification tag separation section 21, the data packet 64 sent from the program storage section 1 and the data packet 63 sent from the arithmetic processing section 3 are combined.

In the identification tag separation unit 21, the data packets 63 to be merged are
A data packet 60 is generated from the data packet 64 and the data packet 64 . That is, the data bucket soto 63 and the data packet 64 are generated into one packet using the identification tag 50 as a mark. The data packet 60 generated by the identification tag separation section 21 is sent to the paired data detection section 2 via the transmission path 33 or the transmission path 34. Thereafter, processing according to the data flow program 70 progresses through each device in turn.

Therefore, in this embodiment, a data packet that connects only the data required by each processing unit is generated and transmitted, so the data line through which the data packet passes is limited to the width necessary for the transmitted data packet. Can be reduced.

In the above-described embodiment, the identification tag separation unit 21 waits for data packets with matching identification tags 50. Therefore, the identification tag separation unit 21 uses an internal buffer or an external buffer (not shown) for waiting.
have.

[Effects of the Invention] As described above, according to the present invention, data packets passing between each processing unit in an information processing device are connected by connecting only the minimum amount of data required by each processing unit for processing. Since the data is generated and sent to the transmission path, the width of the data line connecting each processing section can be reduced, and furthermore, the area of the IC chip that is the constituent unit of each processing section can be reduced.

When the information processing device has a plurality of processing units such as a program storage unit, a data pair detection unit, and an arithmetic processing unit, the above effect is effective in reducing the scale of the device itself.

Further, even when the information processing apparatus is configured as a multiprocessor by combining a plurality of units, each unit being the unit shown in FIG. 1, the above effect is effective in reducing the scale of the apparatus itself.

[Brief explanation of drawings]

FIG. 1 is a diagram showing the configuration of a data flow type information processing apparatus according to an embodiment of the present invention. FIG. 2 is a diagram showing the field structure of a data packet processed by the information processing apparatus of FIG. 1. FIG. 3 is a diagram showing an example of the configuration of a conventional data flow type information processing device. Figure 4 shows
FIG. 2 is a diagram showing a field configuration of a data packet processed in a data flow type information processing device. Figure 5 shows
FIG. 3 is a diagram illustrating part of a data flow program stored in a program storage unit of the data flow type information processing device. FIG. 6 is a diagram showing part of a data flow program including copy processing. FIG. 7 is a diagram showing the flow rate of data packets in each transmission path of the information processing apparatus shown in FIG. FIG. 8 is a diagram showing another example of the configuration of a conventional data flow type information processing device. FIG. 9 is a diagram showing the flow rate of data packets in each transmission path of the information processing apparatus shown in FIG. 8. In the figure, 1 is a program storage unit, 2 is a paired data detection unit, and 3
21 is an arithmetic processing unit, 21 is an identification tag separating unit, 22 is an identification tag adding unit, 41 is a destination field, 42 is an instruction field, 43 is a data 1 field, 44 is a data 2 field, 45 is a result data field, 50 is an identification tag and 60.61.62.63.64 is a data packet. In each figure, the same reference numerals indicate the same or corresponding parts. Atsushi Figure 1 Figure 2

Claims (1)

[Scope of Claim] A data flow type information processing device that performs information processing based on data including a portion necessary for calculation and a portion necessary for reading the program, comprising: a program storage means for storing a data flow program; an arithmetic processing means for performing arithmetic processing on data according to a flow program; a merging means for combining information read from the program storage means and a result calculated by the arithmetic processing means; and a merging means for combining the data combined by the merging means. Data flow type information, comprising: a pair of data detecting means that converts the data into a form that can be processed by arithmetic operations; and a branching means that separates the output from the pair of data detecting means into a part necessary for the calculation and a part necessary for reading the program. Processing equipment.