JPS63292333A - Data processor - Google Patents

Abstract

PURPOSE:To efficiently perform the arithmetic processing by storing the list length, as well of newly appearing table format list data in a main storage device as one item of list data. CONSTITUTION:Table format data where list data and the list structure are expressed with vectors using positions of respective elements is stored in a main storage device 1. In the case of processing of list data, list data is transferred to a structure data register 2 in a structure data arithmetic unit 5, and simultaneously, respective elements of list data are transferred to element registers 7 of each auxiliary arithmetic unit 10 independently of one another by a transfer device 6. When the structure data register 2 is subjected to operation, corresponding element registers 7 are subjected to required operation simultaneously, and the list data length is calculated by a list data length calculating means 4 in accordance with the list structure due to operation results to update a list data length area 22 of list data in the structure data register 2. Thus, the list data operation is efficiently performed.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of Industrial Application The present invention relates to a data processing device mainly intended for use in the field of artificial intelligence.

BACKGROUND OF THE INVENTION In recent years, the field of artificial intelligence has been actively researched as one of computer applications. In this field, it is necessary to process structured data, and therefore LISP, a language that can handle structured data, is widely used. but.

Since it is inefficient to execute the LISP language on a general-purpose computer, special-purpose machines with various improvements have been developed.

These dedicated machines were improved mainly by approaching the language aspect, and the typical improvements made are shown below.

(1) Primitive functions such as CAR and CDR are executed at the microprogram level.

(2) TAG for handling GeneLink data type
data format.

(3) Provide a hardware control stack to speed up stack processing.

(References rL I SP? Schiff J Information Processing Vol. 1.2
3N181) 9752-772) However, although improvements have been made to the execution system of the language as described above, the representation of structure data inside a computer basically uses pointers to represent the order of elements and the method of connection. (hereinafter referred to as the list).

In this method, all list operations involve sequentially following the pointer, so it is essentially inefficient for performing the following operations that frequently appear in application programs in the field of artificial intelligence.

*Pattern matching It is inefficient because it involves list decomposition.

Book Access to arbitrary elements, decomposition of lists.

This results in a list of replacements for specific elements, which is inefficient.

In order to solve the above problems, proposals have been made to change the method of expressing list data.

In general, a binary tree list starts from a starting node, branches left and right sequentially, and each branch ends at a leaf node. There are two types of 0-leaf nodes: atom nodes and NIL nodes. ′! Nodes other than J's nodes are list nodes indicating that the branch is continuing. This list node is used to indirectly indicate the position of the leaf node.

In the pointer representation, this structural representation is expressed as is by cells in which all nodes are connected by addresses. As a result, access to each element always requires repeated indirect references to the address.

However, if you can directly determine the position of a leaf node, there is no need to have list node information. Therefore, in a table that sequentially arranges leaf position information and leaf information,
A one-dimensional vector representation in which numbers are sequentially assigned in the length direction and items are sequentially assigned in the depth direction has been proposed as a method for expressing the node position of the 0th leaf that can represent equivalent list data. Therefore, the list data is expressed as a set of data that includes a vector indicating leaf position information and information about the leaf itself. With this representation method, there is no need to limit the interpretation of a list to a binary tree list, as is the case with pointer representations.

Figure 5 shows an example of the expression of list data, which shows the graphical expression (1) and tabular expression (2) of list data, which becomes (A (B (C) ') D) when expressed in 8 formulas. In the diagrammatic representation shown, circles represent list nodes, and squares represent leaf nodes. Further, the number string added above each node indicates the node position expressed according to the above-described method. The tabular representation (2) is a table representation (2) in which this leaf part is extracted and is made up of a table in which the ADDRESS section contains the node position vector and the VALUE section contains the leaf elements. By adopting such an expression format, it is possible to process list data in parallel for each element without passing through pointers, and the above-mentioned operations can be performed at high speed.

An example of a conventional data processing device based on this expression will be described below with reference to the drawings.

FIG. 6 is a diagram showing the configuration of a conventional data processing device.

In FIG. 6, 1 is the main memory, 61 is a register, 6
2 is an arithmetic means, and the main arithmetic unit 63 is a general term for the plurality of registers 61 and the arithmetic means 62.

6 is a transfer device. 7 is an element register, 8 is a sub-storage device, and 9 is an element calculation means. The sub-processing unit IO is a general term for a plurality of element registers 7, sub-storage devices 8, and element calculation means 9, and a plurality of sub-processing units 1O exist.

Further, FIG. 7 shows the basic form of conventional list data. In FIG. 7, 21 is a tag area indicating that the data type is list data, and 23 is a pointer area to actual list structure data expressed in table format.

The operation of the data processing apparatus configured as described above will be described below. The list data is stored in the main storage device 1 in the form shown in FIG. 7, and if processing is required,
It is loaded from the main memory device 1 into the register 61 of the main operation bag M63. Each element of the list data is transferred element by element to the sub-processing unit IO by the transfer device 7 and processed in parallel.

Arithmetic instructions to be executed are stored in the main memory 1.

The arithmetic means 62 and the element arithmetic means 9 execute the arithmetic instructions transferred from the main memory l. For example, the function APPEND, which is a basic function of the LISP language and is often used,
Let us consider list operations such as (combining lists) and the function REVEPSE (reversing lists).

Bad APPEND'(ABC)' (DB(F
))). ? ! number register 61
List data related to (A13C) is stored in one register #1, and list data related to (DB(F)) is stored in another register #2. List (ABC), (D
For the elements of B (F)), a plurality of sub-processing units 10
Element A is stored in a certain element register #11 of a plurality of element registers 7 of a certain sub-processing unit #1, and D is stored in another element register outside 12.
1 is stored in B, E is stored in element register #22, 0 is stored in element register #31 of sub-processing unit #3, and F is stored in element register #32. To PPI! In the tabular representation, the NO function goes beyond the second argument! 1st in IADDRESS department
This is executed by merging the table format obtained by adding the list data length of the argument and the table format of the first argument. That is, ^PPIIN by the calculation means 62 and the element calculation means 9
The list data of the result of the D operation is stored in register #l, and element registers #ll to #13 of the plurality of element registers 7 of sub-processing device #l, which has a plurality of sub-processing devices 10, are the same as before execution, Element 1 is stored in element register #14, E is stored in element register #15, and F is stored in element register #16.

Next, consider the case where the command (REVER3E' (ABC)) is executed. List data regarding (ABC) is stored in one register #1 of the plurality of registers 61. The elements of the list (ABC) are stored as element A in element register #11, which has a plurality of element registers 7, of sub-processing unit #l, which has a plurality of sub-processing units 10, and thereafter in the element register of sub-processing unit #2. B is stored in #21, and C is stored in element register #31 of sub-processing unit #3. The REVER5E function starts with ADDR[! S
The S99 part is executed by the calculation means 62 and the element calculation means 9 as an operation of taking the two's complement number and adding the argument (list data length+1).

In order to execute the functions APPEND and REVEPSE, it is necessary to calculate the length of the argument list data.

Problems to be Solved by the Invention However, with the above configuration, for example, LISP
APPIIZND and REVER3, which are the basic functions of the language
When executing an instruction such as E, it is necessary to always calculate the argument list data length, which poses a problem in that a bottleneck occurs in terms of processing efficiency.

The present invention solves the above-mentioned problems, and solves the problems described above.
The present invention provides a data processing device with high processing efficiency in arithmetic processing such as ER3B.

Means for Solving the Problems In order to solve the above problems, the data processing device of the present invention includes a main storage device that stores list data as tabular data expressed in vectors using the position of each element; A structured data calculation device includes a plurality of structured data registers that store list data, structured data calculation means that performs calculations on the list data, list data length calculation means that calculates the list length for the list data, and a table-format It is equipped with a plurality of element registers for storing each element of list data, an element calculation means for performing calculations on the elements, and a sub-storage device, and each element of the list data in tabular form is stored under the control of the structure data calculation device. The apparatus includes a plurality of sub-processing units that perform parallel processing, and a transfer unit that sequentially allocates elements of the tabular data stored in the main memory to each of the sub-processing units.

Operation With the above-described configuration, the present invention calculates the list length of newly appearing list data in table format, and also calculates the list length after the calculation is performed on the list data on the structured data register that is the target of the list calculation. By calculating the list length and storing the list length in the main memory as one item of list data, APPEND and REVE in the LISP language can be easily performed.
List calculation processing such as R3E can be performed efficiently.

Embodiment Hereinafter, a data processing apparatus according to an embodiment of the present invention will be described with reference to the drawings. FIG. 1 shows the configuration of a data processing device in an embodiment of the present invention.

In FIG. 1, 1 is a main memory, 2 is a structured data register, 3 is a structured data calculation means, 4 is a list data length calculation means, and 5 is a structured data calculation device, which operates with a plurality of structured data registers 2 and structured data. This is a general term for means 3. Reference numeral 6 denotes a transfer device, 7 an element register, 8 a sub-storage device, 9 an element operation means, and 1O a sub-operation device, which is a general term for the element register 7, the sub-storage device 8 and the element operation means 9.

FIG. 2 shows the basic form of list data handled. 21 is a tag area, 22 is a list data length M area, 2
3 is a pointer area. The tag area 21 indicates that the data type is list data, the list data length area 22 indicates the list length of the list data, and the pointer area 23 stores a pointer to the actual list structure data expressed in table format. do.

The operation of the data processing apparatus configured as described above will be described below with reference to FIGS. 1 and 2.

Processing of the list data is performed by the structured data calculation device 5 and the sub-processing device 1G. One structure data register 2 of the structure data calculation device 5 corresponds to one of the element registers 7 in each sub-processing device 10. That is, list data is stored in the structural data register 2, and each element of the list data is stored in a plurality of corresponding element registers 7, representing one list data as a whole.

The main storage device l stores list data and tabular data in which the list structure is expressed as a vector using the position of each element. When list data is processed, the list data is first transferred to the structured data register 2 in the structured data calculation device 5, and at the same time each element of the list data is transferred to the structured data register 2 in the structured data calculation device 5. Each sub-operation by 6! JfIO element register 7
are transferred separately. and structure data register 2
When the operation is performed on the corresponding element register 7, the necessary operation is simultaneously performed on the corresponding element register 7, and the list data length calculation means 4 calculates the list data length from the list structure generated from the operation result, and the structure data The list data length area 22 of the list data in the register 2 is updated.

Figure 3 shows the LISP function APPEND in the above embodiment.
Take up (APPEND' (ABC) ' (DE (F
) List data related to (ABC) in one structure data register #l of multiple structure data registers 2, which indicates the operation when executing ) ) ), is stored in another structure data register #2 (DB (F)). List data related to is stored. List (ABC).

For the elements of (DB (F)), element register #11 with multiple element registers 7 of sub-processing unit #l with multiple sub-processing units 10, and element register #1 with element A
D is stored in element register #2 of sub-processing unit #2, B is stored in element register #22 of sub-processing unit #2, and E1 is stored in element register #22 of sub-processing unit #3.
0 is stored in element register #31 and F is stored in element register #32. The function APPEND uses the structure data register # that stores the list data of the first argument.
The list data length obtained from the list data length area 22 of l is the first ADD in the list data expression of the second argument.
This is done by adding the RESS part and then merging each representation. That is, as a result of the APPEND operation by the structure data calculation means 3 and the element calculation means 9, the list data having the updated list data length area 22 is stored in the structure data register #1, and the list data having the updated list data length area 22 is stored in the structure data register #1, and the result is stored in the structure data register #1.
Element registers #11 to #13 of the plurality of element registers 7 of sub-processing unit #1 with O are the same as before execution, the element register is stored in element register #14, and the following element registers #1
E and F are stored in element register #15 and element register #16, respectively. If this expression is expressed using equation 8, it becomes (ABCDE (F)), which means that the function APPEND has been executed.

Figure 4 shows the LISP function REVER3 in the above example.
List data regarding (ABC) is stored in one structure data register #1 of the plurality of structure data registers 2, which indicates the operation when E is taken and (R[!VER3E' (AI3C)) is executed.

The elements of (ABC) are stored in the element register #11, which has the plurality of element registers 7, of the sub-processing device #1, which has the plurality of sub-processing devices 10, and are stored as elements in the element register #11, which has the plurality of element registers 7 of the sub-processing device #1, which has the plurality of sub-processing devices 10. 81 sub-processing unit #3 in register #21
C is stored in each element register #31.

Function REνI! R3E is an operation that takes the two's complement of the first ADDRESS part in the expression of the list data of the argument and further adds (list data length + 1) obtained from the list data length area 22 of structure data register #1 where list data is stored. This is executed by the structure data calculation means 3 and the element calculation means 9 as follows. When the obtained expression is expressed by Equation 8, it becomes (CB A), which means that the function REVER3E has been executed.

Effects of the Invention As described above, the present invention includes a main storage device that stores list data as tabular data expressed as a vector using the position of each element, a plurality of structured data registers that store list data, and list data. a structured data calculation device consisting of a structured data calculation means for calculating a list length for the list data, a list data length calculation means for calculating the list length for the list data, and a plurality of element registers for storing each element of the list data in a tabular format. and a plurality of sub-processing devices, which are provided with element calculation means and a sub-storage device for performing calculations on the elements, and process each element of the tabular list data in parallel under the control of the structured data calculation device; By providing a transfer device that sequentially allocates elements of the tabular data stored in the main memory to each of the sub-processing units, each element can be processed simultaneously and list data operations such as APPEND and REVEPSE in the LISP language can be performed. This can be done efficiently.

[Brief explanation of the drawing]

FIG. 1 is a configuration diagram of a data processing device in an embodiment of the present invention, FIG. 2 is an explanatory diagram of the basic form of list data in an embodiment of the present invention, and FIG. 3 is a diagram of the LISP function APPEND.
FIG. 4 is a diagram showing an example of the operation in the present invention.
Function REV[! R3! ! (7) An explanatory diagram of an example of the operation in the present invention, FIG. 5 is an explanatory diagram of an example of tabular representation of list data, FIG. 6 is a configuration diagram of a conventional data processing device, and FIG. 7 is a diagram of conventional data processing. FIG. 3 is an explanatory diagram of the basic form of list data in the device. 1... Main memory device, 2... Structure data register, 3... Structure data calculation means, 4...
... List data length calculation means, 5 ... Structure data calculation device, 6 ... Transfer device, 7 ...
...Element register, 8...Sub-storage device, 9...
...Element calculation means, 10...Sub calculation device,
21...Tag area, 22...List data length, 23...Pointer area, 61...
・・Register, 62・・・・Arithmetic unit, 63・Old・
・Main processing unit.

Claims (1)

[Claims] A main storage device that stores list data as tabular data expressed as a vector using the position of each element, a plurality of structured data registers that store list data, and structured data calculation means that performs calculations on list data. A structured data calculation device comprising a list data length calculation means for calculating a list length for list data, a plurality of element registers for storing each element of list data in a tabular format, and an element calculation means for performing calculations on the elements. a plurality of sub-processing devices each having a sub-storage device and processing each element of tabular list data in parallel under the control of the structured data processing device; What is claimed is: 1. A data processing device comprising: a transfer device that sequentially allocates elements of stored tabular data; and processing each element simultaneously when calculating list data.