JPH01243126A - Arithmetic and control unit - Google Patents

Abstract

PURPOSE:To execute a programming without being conscious of a data type and to dispense with the instruction of a kind according to the data type by generating a control code to indicate the discrimination and conversion methods of the data type even in an operation mutually between different data types, and executing the operation. CONSTITUTION:A data type conversion control code generating part 115 inputs the data type code of the instruction sent from a data type part 113a of an instruction register 113 with type and the data type code of an operand sent from a data type part 114a of an operand register 114, and the part 115 generates and outputs a data conversion control code to an instruction code part 111c and the data type code of an arithmetic result to an instruction stack 109. Namely, despite the data types are different, the data types are judged, a prescribed control code and the code to indicate the data type of the arithmetic result are generated according to a conversion rule, and an arithmetic processing in the specific data type is executed. Thus, without being conscious of the data type, the programming can be executed, and an arithmetic processing speed can be improved.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to an arithmetic control device that directly executes an intermediate language of a program written in a functional format.

(Prior Art) A conventional arithmetic and control device of this type is configured as shown in FIG. 4, for example. In other words, in the figure, the arithmetic control unit includes a syntax analysis unit 100' that expands the intermediate language of a program into a format executable by the arithmetic unit 200', and performs operations using instructions (macro instructions) in the expanded executable format. It is composed of a calculation section 200'.

Of these, the syntax analysis unit 100' includes a program memory 101 in which a machine language intermediate language program is stored, a program counter 102 that specifies the address of the program, and a program memory 102 in which instructions read from the program memory 101 are stored. instruction register 103 and arithmetic unit 20
A macro instruction control unit 104 that controls expansion into instructions in a format that can be directly executed by the 0'IC, an instruction stack 109' that controls the execution order of instructions, and an instruction stack pointer 110 that serves as a pointer thereto. , a macro instruction register 111' in which executable instructions are stored;
It consists of a macro logic bus 112. The macro instruction register 111' has a data section 111b where data is stored and an instruction code section 1 where instruction codes are stored.
11C.

On the other hand, the calculation section 200' includes a microprogram control section 201 connected to the instruction code section 111C, an address conversion section 202 connected to the data section 111b for converting a logical address into a physical address, and a data memory. 203, an arithmetic operation unit (ALU) 204, a data stack 205, a data stack pointer 206 as a pointer thereof, and a data bus 207 that exchanges data between the arithmetic operation unit 204 and the data memory 203.
It consists of a micrologic bus 208.

Here, the instruction stack 109' and instruction stack pointer 110, as well as the data stack 205 and data stack pointer 206, are all L I F 0 (las
t-in first-out) stack.

Next, this operation will be explained. First, the syntax analysis unit 100'
The operation of is explained according to the flowchart in Fig. 5.
First, the macro instruction control unit 104 sets the initial value of the program address in the program counter 102 via the macro logic bus 112 in step Sl).
According to this address, the one-word instruction taken out from the program memory 101 is stored in the instruction register 103 (S2).

Next, the contents of the instruction register 103 are analyzed (S3).
, if this is an instruction, it is stored in the instruction code section 111C of the macro instruction register 111' and the instruction stack 109' (541), the program counter 102 is incremented (+1) (S5), and the syntax of the next word is analyzed. Move to. Furthermore, when the contents of the instruction register 103 are operands, the contents of the instruction register 103 are transferred to the data section 111b.
After reading out the first instruction of the instruction stack 109' and storing it in the instruction code section 111C (543), the CPU 200' starts arithmetic processing (544). Thereafter, the program counter 102 is incremented (S5).

Note that for processing the next instruction, as soon as the arithmetic processing of the arithmetic unit 200' is completed, the instruction is taken out from the program memory 101, and the above-mentioned processing is repeatedly executed for the instructions.

In the arithmetic unit 200', the microprogram control unit 201 first reads the instruction code from the instruction code section 111c of the macro instruction register 111' and decodes it. If it is a read command from the memory 203, data is read from the data memory 203 via the address conversion unit 202 and written to the data stack 205 via the data bus 207 and the arithmetic operation unit 204.

Further, if the instruction is a write instruction using the contents of the data section 111b as an address, contrary to the above, from the data stack 205 to the arithmetic operation section 204 and the data bus 2.
The data is written to the address of the data memory 203 indicated by the address conversion unit 202 via the address 07.

The arithmetic operation unit 204 also performs arithmetic and logical operations on data within the data stack 205.

Next, according to a specific calculation example, the macro instruction control unit 104
The functions of the arithmetic operation unit 204 and the arithmetic operation unit 204 will be explained focusing on the operation of the LIFO stack for instructions and data respectively.

Now, consider the following equation (1) as an example of a functional type operation, and expand this equation (1) as shown in equation (2) using Borland notation.

f=A÷BX(C+D) ・・・・・・・・・
・(1) Then, in the program memory 01 shown in FIG. 4, it is assumed that the formula (2) is expanded into machine language and stored in the order ① to [phase] shown in FIG. In addition, this order■
.about.[phase] is also the processing order by the syntax analysis unit 100', and hereinafter each process itself will also be represented by the number .about.[phase].

Hereinafter, for each of the above processing ■~[phase], the instruction stack 109'
The operation of the data stack 205 and the data stack 205 will be explained based on FIG.

First, since the processes ■, ■, and ■ are instruction code reading, these instructions are stored in the instruction code section ]11c as described above.
The instructions are stored in the instruction stack 109', and are simultaneously pushed down and stored in the instruction stack 109'. Also.

Processes (1), (2), and (2) are operands, so after evaluating these operands, they are finally pushed down and stored in the data stack 205 as described above. here,
``Evaluate'' means that if the operand is a constant, it is treated as the value as is, and if it is an address, the contents indicated by the address are read and treated as the value.

Furthermore, in process ■, the operand is read following process ■, and after evaluating the operand, data stack 2 is
The first data of 05 (last stacked) and
The arithmetic operation section 204 performs an operation using the instruction at the head of the instruction stack 109' (same as the same), and the operation result is stored in the data stack 205 again. In addition, the processing in Figure 7 ■
The content f1 of the data stack 205 in FIG. 2 shows f +, =C+D, which is the intermediate progress of the calculation.

Next, processing ■, ■, [phase] indicates the end of the processing of the corresponding instruction, and after popping up the instruction at the head of the instruction stack 109', the operation of the instruction at the head is performed on the data at the head of the data stack 205. and the data stored next, and the result of the calculation is transferred to the data stack 2.
After popping up 05, store it again. In addition, the seventh
The content f2 of the data stack 205 in process ■ in the figure shows the intermediate progress of the calculation, f2=B .

By the above-mentioned second operation, the arithmetic processing of equation (1) is completed.

(Problem to be Solved by the Invention) Here, when the types of data A, B, C, and D in equation (1) are different, and when performing arithmetic processing using these, the following points need to be kept in mind. be. Specifically, when A, B, and D are given as different data types such as integer values and C as a floating point value, expression (1) is converted into an instruction to convert the data type and a specific data type. must be programmed by replacing it with the target instruction.

That is, as shown in FIG. 8(a), calculate f from data C and D by calculating C+D, and then use this fl and data B.
When calculating f2 by calculating BXfl from , and calculating A÷f2 from this f2 and data A to obtain the solution f, if the data types are different in this way, as shown in Figure 8 ( As shown in (b) or (c), it becomes necessary to insert a conversion instruction to match the data type to either one. In addition,
In FIGS. 8(a) to (d), data A-D and calculation results f1.1f2. f is the state stored in the data memory 203 or the data stack 205.

First, in Figure 8 (b), in order to unify the data types to floating point values, an instruction "I/F" is inserted to convert data A, B, and D to floating point values, and then the floating point values are converted to floating point values. It shows the process of replacing the target operation. In this figure, "H, rr b II, and u d n are data A and B after conversion to floating point values, respectively.

D, II 10F II, II X F II,
If÷p II indicates addition, 9 multiplication, and division by floating point operations, respectively. Further, data a, b, and d are stored in the data memory 203 or the data stack 205.

On the other hand, Fig. 8 (c) shows an instruction "F/I"' to convert data C into an integer value in order to unify the data type to an integer value.
, and then replaces it with an operation that targets integer values. In this figure, LL , , II is the process of converting data C into an integer value, and LL + II , II
X# and LL÷'' indicate addition, squaring, and division by integer operations, respectively.

Furthermore, Figure 8 (d) shows that instead of processing data in one data type, each data has its own value and data type, and the data type is evaluated during calculations. It similarly outputs the value and data type as the result of the operation, stacks these, and moves on to the next operation. In this figure, data A, B, C, D and the calculation result f□.

LL I II and LL F+'' attached to f, , f are integer values, respectively.

Indicates the data type of floating point values. By this method, the contents of the data stack 205 at the time of performing the process (2) in FIG. 7 are as shown in FIG.
will be stored respectively.

Here, as shown in Figures 8 (b) and (c) above, when the data is only a value and the instruction has a type of operation that targets only a specific data type, the following problems arise. . In other words, in this case, the programmer must be aware of the data type when programming by writing data type conversion instructions and selecting arithmetic instructions according to the data type. Programming becomes extremely complicated because it is necessary to prepare as many as needed.

Further, the arithmetic control unit also needs to process instructions for a plurality of data types, which increases the hardware and firmware processing of the instruction syntax analysis section and the arithmetic section.

Furthermore, as shown in FIG. 8(d), when data has a value and a data type, and the data type is evaluated during calculation to perform calculation processing, as shown in FIG. 9, during calculation, After extracting the data type from the data stack 205 and determining whether or not it is necessary to convert it, the actual operation is performed, and the operation result f1
．． Since f, . . . also have a value and a data type, there is an inconvenience that it takes time to recognize the data type when evaluating the calculation result, which takes a lot of time to process the data. At the same time, a large memory capacity of the data stack 205 was required to store the data types.

The present invention was proposed to solve the above-mentioned problems, and its purpose is to enable programming to be performed in operations between different data types without being aware of the data types, and to enable programmers and operators to perform operations on different data types. In addition to reducing the burden on the hardware of the control device, it also increases the memory efficiency of the data stack and reduces the time required to evaluate the calculation results by allowing the calculation results to contain only the values and not the data type. An object of the present invention is to provide an arithmetic control device that enables a significant improvement in arithmetic processing speed.

(Means for Solving the Problems) In order to achieve the above object, the present invention provides an arithmetic control unit that directly executes an intermediate language of a program written in a functional form, in which a syntax analysis unit of the intermediate language is configured to perform the calculations of the program. Determines the data type of an operand that constitutes an instruction, and generates a control code that indicates a method of converting this operand to the data type to be processed, and a data type code that indicates the data type of the result of an operation using the operand. storage means such as an instruction stack for controlling the operation order and stacking data typed instructions consisting of the data type code and the instruction code of the operation instruction; It is characterized by being executed as

(Operation) According to the present invention, a control code indicating a data type conversion method and a code indicating a data type of an operation result are generated based on the data type of the operand and the data type of the typed instruction.

Among these, data type conversion is performed as necessary by the control code, and an operation is executed based on the instruction code of the typed instruction.

Further, the code indicating the data type of the operation result is sent to the instruction stack and stacked together with the instruction code as a new data typed instruction, so that the reception is carried over when the next instruction is executed.

(Example) An embodiment of the present invention will be described below with reference to the drawings.

First, FIG. 1 shows the configuration of an arithmetic control device according to the present invention. In the figure, 100 is a syntax analysis unit that develops the intermediate language of a program into a format executable by the arithmetic unit 200, and 200 is a macro Arithmetic unit 20 that performs arithmetic operations based on instructions
It is 0.

Here, the calculation unit 200, as shown in FIG. 4 above,
Microprogram control section 201 and address conversion section 2
02, a data memory 203, an arithmetic operation unit 204,
A data stack 205, a data stack pointer 206 that together with the data stack 205 constitute a LIFO stack, an arithmetic operation unit 204, and a data memory 203.
It consists of a data bus 207 and a micrologic bus 208 for exchanging data between them.

Next, the configuration of the syntax analysis section 100 will be described in detail. This syntax analysis unit 100 includes a program memory 101 in which an intermediate language program converted into machine language is stored in the same manner as described above, and a program counter 1 that specifies the address of this program.
02, an instruction register 103 in which instructions read from the program memory 101 are stored, a macro instruction control unit 104 that performs control to expand instructions into instructions in a format that can be directly executed by the arithmetic unit 200, and an instruction execution unit. An instruction stack 109 for controlling the order and this instruction stack 1
09 and an instruction stack pointer 110 that constitutes a LIFO stack.

Thus, the program memory 01 includes a typed instruction register 113 in which data-typed instructions are stored via the instruction bus 116, and an operand register 114 in which operands of instructions read from the program memory 101 are stored. is connected. Of these, the typed instruction register 113 consists of a data type section 113a in which a data type code indicating a data type is stored, an instruction code section 113b in which an instruction code indicating the type of instruction is stored, and an operand register. 114 consists of a data type section 114a in which a data type code indicating the data type of the operand is stored, and a data section 114b in which a logical address or a direct numerical value is stored. Typed instruction register 113 and operand register 114 via logic bus 112
has been added to.

Furthermore, 111 is a macro instruction register in which instructions in executable format are stored, and this register 111 stores the data type section 114a and data section 1 of the operand register 114.
14b, and the data type section 113a of the typed instruction register 113 via the data type conversion control code generation section 115, and the instruction code section 113b. It consists of an instruction code section 111c. Note that the data section 111b and instruction code section 111C are also connected to the macro instruction control section 104.

Here, the data type conversion control code generation unit 115 generates a data type code of the instruction sent from the data type section 113a of the typed instruction register 113 and a data type code of the operand sent from the data type section 114a of the operand register 114. is input, and a data type conversion control code for the instruction code section 111c and a data type code of the operation result for the instruction stack 109 are generated and output.

Note that the instruction stack 109 constitutes a wired OR with the program memory 101 and the typed instruction register 113, and the instruction stack 09 contains an instruction indicating the data type code of the operation result and the instruction type of the typed instruction register 113. It is configured so that it can be combined with code and stacked as data-typed instructions.

Also, the data type section 111a of the macro instruction register 111
The data section 111b is connected to the address conversion section 202 in the calculation section 200, and a logical address is converted into a physical address based on the data type code and data of the operand.

Next, with reference to FIGS. 2 and 3, an explanation will be given of the operation in the case of performing the arithmetic processing of equation (1) according to this embodiment.

First, FIGS. 2(a) and 2(b) respectively show the format of the intermediate language operands and formats 1 to 1 of the instructions stored in the program memory 101 in this embodiment. It is identified by the most significant bit 111 I+ or O''. Among these, the format of the operand is a data type indicating the data type, such as an integer value or a floating point value, as shown in Figure 2 (a). An operand consisting of a code and a logical address or direct numerical value, each with its own value.

It is sent to the macro instruction register 111 via the register 114.

In addition, as shown in FIG. 2 (b), the instruction format does not have a data type code (=0) and consists only of an instruction code indicating the instruction type, and the instruction stack 109 described above
By wire FOR with
sent to. Regarding the data type section 113a,
A control code for converting data types is generated by the data type conversion control code generation section 115, and this control code is sent to the instruction code section 111C. Note that FIG. 2(c) shows the format of the instruction read from the instruction stack 109, and this format has a data type code and an instruction code indicating the data type of the operation result.
Figure (d) shows the instruction code section 1 of the macro instruction register 111.
11c format, and consists of a data type conversion control code and an instruction code.

In addition, in FIGS. 2(a) to 2(d), the symbols attached below each format indicate the locations where these codes and the like are stored.

Now, to explain the control of the wired OR, the processes ① and ② in FIGS. 6 and 7 above.

Since step (2) is for reading an instruction code, the program memory 101 is selected and only the instruction code is sent to the instruction code section 113b of the typed instruction register 113.

Further, since the processes (1), (2), (2), and [phase] are operations, the instruction stack 109 is selected, and the data type code and instruction code are sent to the data type section 113a and instruction code section 113b. The above control is based on the contents loaded into the instruction register 103 by the macro instruction control unit 104 for each process.
Decide whether it is an order or not.

Next, a method for generating a data type conversion control code will be described. When converting the data type, the data type of the operation and the result of the operation can be uniquely determined by matching the data word length or the width that the data can represent (data range), whichever is larger. For example, if the data is of three types: a one-word long integer value, a two-word long integer value, and a floating point value, the conversion rule shall follow the larger data type according to the size relationship in the following equation (3). .

One word long integer value〈Two word long integer value〈Floating point value...
(3) Focusing on this point, data type conversion control code generation section]15
As shown in Table 1 below, inputs the data type code of the operand and the data type code of the typed instruction, and generates a control code indicating data type conversion and a code indicating the type of the operation result.

(Hereinafter, blank space) In this Table 1-, 10'' indicates the state value stored in the program memory 101 before having the data type.
L D I +1 is an operand code that indicates a two-word long integer, "SI"' is an operand code that indicates a one-word long integer, II F p ++ is an operand code that indicates a floating point value, and rr D ++ is a two-word operand code. LL S II is a type code that indicates long data, tt F ++ is a type code that indicates floating point data, and II D C+1 is a type code that indicates two word length operations without data type conversion. 11P C++ is the code that performs floating point operations without data type conversion, 11P C++ is the code that performs floating point operations without data type conversion, and "SDC" is the code that performs the operand data type. "FDC" is the code that converts to the type and performs the two-word length operation.
The code that converts to the type of ``DS'' and performs floating-point operations is
C” is the LL SI+ of the data on the data stack 205
"FSC'" converts the data type "S" on the data stack 205 to the "IIF" type and performs floating point operations. "DFC" and "SFC" respectively indicate codes that convert the data type of the operand to the L(F11 type) and perform floating point operations.

Next, when performing the arithmetic processing shown in equation (1) in this embodiment, the operations will be described in detail, particularly regarding processes (1), (2), and (2).

First, in process (3), when the operand related to data C (floating point value) is read, the data type section 114a of the operand register 114 and the macro instruction register 1 are read.
From the data type part 111a of No. 11, the data type code "'F
P” is output, and the data section 114b of the operand register 114 and the data section 11 of the macro instruction register 111
A logical address or a direct numerical value is output from 1b.

At this time, the data type section 113 of the typed instruction register 113
A and the instruction code section 113b store the instruction "(10)" of process (2), and as shown in FIG. 2 (b), this instruction does not have a data type code (="O"), It has only the instruction code.

Therefore, the data type conversion control code generation unit 115 has the following information:
Operand data type code from the data type section 114a
FP'' and the data type code LL OI+ of the typed instruction from the data type section 113a are input, and the control code ``FC'''' shown in the data type conversion method and the code indicating the data type of the operation result according to Table 1 above are input. "F'" is output. Among these, the control code "FC" is the macro instruction register 11.
The code 11 F I+ is sent to the instruction code section 111c of instruction stack 1 and indicates the data type of the operation result.
09 and composes the next data typed instruction together with the instruction code.

Next, when the operand is read by the process (2) concerning data D (integer value), the data type code "DI" of the operand is output from the data type sections 114a and 111a, and the logical address or direct numerical value is written to the data section 11.
4b and 1llb, the data type code II D I I+ is input to the data type conversion control code generation section 115, and a code indicating the data type of the operation result is sent from the instruction stack 109 via the data type section 113a. “FI+ is input.

Therefore, according to Table 1, data type conversion control code generator 1
From 15 onwards, there is a control code “DF” that indicates the data type conversion method.
C'' and a code II F +1 indicating the type of the operation result are output, and these are sent to the instruction code section 111c and the instruction stack 109, respectively, in the same manner as described above.

Here, the control code "DFC" is a code that converts the operand to floating point data and performs floating point operation.
Code llF'' is a code for obtaining the calculation result as floating point data.

In other words, even though the data C and D in processing ■ and ■ have different data types, the data type conversion control code generation unit 115 determines these data types and converts them into the equation (3) above. A predetermined control code and a code indicating the data type of the calculation result are generated according to the conversion rule, and calculation processing with the specific data type is executed.

On the other hand, in the calculation section 200, the address conversion section 202 converts the previous data C into a physical address based on the contents of the data type section 111a and the data section 111b.
Read from data memory 203 and data stack 20
Using the data stored in the instruction code section 1 and the data read out using the same operation for the data D,
The typed instruction code from 11c is decoded by the microprogram control unit 201, the arithmetic operation unit 204 performs an operation, and the result (f,=C+D) is stored in the data stack 205.

Although no data type is attached to this operation result, the data type of the operation result is determined by the instruction stack 109 as described above.
will be managed by.

Note that when the above calculation is executed by the calculation unit 200, =24
- While the syntax analysis unit 100 is
14, the typed instruction register 113, etc., syntax analysis of the next instruction is being performed in parallel.

FIG. 3 shows how the data type of the operand is inherited as a typed instruction as described above. Y2. Y3 indicates the flow of data that generates the data type conversion control code by comparing and determining data types, the codes i, ii, and ni indicate the order in which the data types are inherited, and [F] indicates the generated data type. Each indicates that it is a data-typed instruction.

In this embodiment, the macro instruction register 111 may be omitted and the processing may be performed by the typed instruction register 113 and the operand register 114. In addition, these typed instruction register 113 and operand register 1
14 is provided to perform syntax analysis processing and arithmetic processing in parallel, and can be omitted if such parallel processing is not necessary.

(Effects of the Invention) As detailed above, according to the present invention, a code indicating a data type is added to the operands constituting the program, while the instruction code has a structure of an instruction word without a specific data type, Even when performing operations between different data types, we have created a control code that indicates the data type discrimination and conversion method to perform the operation, so programming can be performed without being aware of data types. Since there is no need to provide different types of instructions depending on the type, the burden on programming and hardware can be reduced, and the program memory capacity can be reduced.

Furthermore, a code indicating the data type of the operation result is generated, this code and the instruction code of the operation are combined to form a data typed instruction, and this data typed instruction is stacked to be used as the instruction for the next operation data. Since the data type can be executed again as follows, the data type will be inherited to the next instruction.
Since the data stack that stores the calculation results only needs to store values excluding the data type, it is possible to reduce the memory capacity of the stack.

Therefore, when evaluating a calculation result, processing for recognizing the data type can be omitted, and the processing speed of the calculation can be improved.

[Brief explanation of the drawing]

Figures 1 to 3 show one embodiment of the present invention. Figure 1 is a block diagram, Figure 2 #(a) to #(d) are explanatory diagrams of each format, and Figure 3 is a calculation diagram. An explanatory diagram that abstractly expresses the flow of data in an example of processing, Fig. 4 is a block diagram showing a conventional example, Fig. 5 is a flowchart showing the operation of the syntax analysis section, and Fig. 6 shows the contents of the program memory. Explanatory diagram, Figure 7 is an explanatory diagram showing the contents of the instruction stack and data stack for each process, Figure 8 (a) to (d)
9 is an explanatory diagram abstractly expressing the flow of data in an example of arithmetic processing, and FIG. 9 is an explanatory diagram of the contents of the data stack. 10〇-Syntax analysis section 113...Typed instruction register 109...Instruction stack 111...Macro instruction register 111a...Data type section 111b...Data section 113a...Data type section 113b... Instruction code section 114... Operand register -14a...
Data type section 114b...Data section 115...Data type conversion control code generation section 200...
・The calculation part GρG has the same ρ (a◎ (R-7)

Claims (1)

[Scope of Claims] An arithmetic control device that includes a syntax analysis unit that analyzes the syntax of an intermediate language of a program written in a functional format and generates instructions that can be executed by an arithmetic unit, and is capable of directly executing the intermediate language. In the syntax analysis unit, a control code indicating a method for determining the data type of an operand constituting an operation instruction of the program and converting this operand into a data type to be processed, and a control code indicating a method for converting the operand into a data type to be processed, and a calculation result using the operand. means for generating a data type code indicating the data type of the data type; and storage means for controlling the operation order and stacking data typed instructions consisting of the data type code and the instruction code of the operation instruction; An arithmetic control device characterized in that an attached instruction is executed as an instruction for the next operation.