JP2011186999A - Device for supporting conversion of floating point representation program to fixed point representation and support program thereof - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a device for supporting the conversion of a floating point representation program to fixed point representation and a support program thereof for reducing the labor of setting attribute values to numerical elements to be converted to fixed point representation in a floating point representation program. <P>SOLUTION: The device for supporting the conversion of a floating point representation program to fixed point representation provides a user interface for: extracting numerical elements used in the floating point representation program (extracting numerical element information); extracting a relationship between the numerical elements used in the floating point representation program and one or a plurality of the other numerical elements which are connected by a formula (connection relationship information); generating attribute value relationship information representing a relationship between attribute values of the numerical elements represented by the numerical element information regarding attribute items required for fixed point representation of the numerical elements and attribute values of the other numerical elements, on the basis of a relationship between the numerical elements and one of a plurality of the other numerical elements which are connected by the formula; and setting an attribute value of each of the attribute items of the numerical elements on the basis of the attribute value relationship information. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention provides a user support for changing a floating-point expression program describing arithmetic processing using numerical elements such as variables of floating-point expression into a fixed-point expression program describing arithmetic processing using numerical elements of fixed-point expression. The present invention relates to a fixed-point expression support apparatus and a support program for a floating-point expression program that performs processing.

  In general, a program describing arithmetic processing executed by a computer (CPU) includes a variable used for the arithmetic processing, a floating point expression program in which numerical elements such as constants are expressed in floating point, and a variable used for the arithmetic processing. There are fixed-point expression programs in which numerical elements such as constants are expressed in fixed-point. Since numerical values (variables, constants, etc.) of floating-point expressions are represented by their mantissa and exponent parts, the range of numerical values that can be expressed is wider than that of fixed-point expressions of the same bit size. The accuracy is determined by the number of digits in the mantissa part. On the other hand, for numerical elements (variables, constants, etc.) of fixed-point representation, attribute item values (attribute values) required for the fixed-point representation, for example, resolution (LSB), size (bit size), and offset The value is defined, the range of numerical values that can be expressed by the attribute value is determined, and the accuracy is determined in particular by the resolution value.

  In general, arithmetic processing to be executed by a computer is described by a floating-point expression program. However, there are cases where the arithmetic processing is described in a fixed-point expression program because of restrictions on the performance of the computer (CPU). For example, an ECU (electronic control unit) mounted on a vehicle performs arithmetic processing by expressing a signal representing a physical quantity (voltage, speed, time, etc.) as a numerical value of several elements (variables and constants), but the arithmetic processing is fixed. It may be described by a decimal point expression program. In such a case, a “control designer” who is familiar with the control target of the vehicle engine or the like creates a floating-point expression program that describes arithmetic processing related to the control, and then stores the floating-point expression program in a computer such as an ECU. “Software designers” who are familiar with computer programs are rewritten into fixed-point representation programs in consideration of computer performance.

  There has also been proposed a program conversion apparatus for converting a floating-point expression program into a fixed-point expression program after the floating-point expression program is created (see Patent Document 1). In this program conversion apparatus, first, a profile program is created by modifying the floating-point expression program so that changes in the value of the target variable (several elements) are output as a history when the program is executed. Then, the profile program (modified floating-point expression program) is executed, and the variable is converted into the fixed-point expression based on the value range of the target variable output at that time. The expression program is changed.

JP 2008-33729 A

  In the above-described conventional program conversion apparatus, it is difficult to create a profile program that suits the purpose by modifying various floating-point expression programs, and the range of variables that can be changed under various conditions In this case, the profile program must be executed under many conditions (parameters), and the amount of processing becomes enormous. Therefore, the above-described program conversion apparatus has a problem in versatility.

  In addition, when a “software designer” who is familiar with a floating-point representation program rewrites it into a fixed-point representation program in consideration of the performance of the computer, the numerical elements (variables and constants) used in the floating-point representation program It is necessary to define the value of each attribute item (resolution (LSB), size (bit size), offset, etc.)) necessary for fixed-point representation, that is, the attribute value. There is a high possibility that the “software designer” will extract a numerical element that should be expressed as a fixed-point number from a floating-point expression program, and the attribute value of each attribute item is set to the extracted numerical element. It takes time and effort to set.

  The present invention has been made in view of such circumstances, and the fixed-point of a floating-point expression program that can reduce the effort for setting an attribute value for a numerical element to be expressed as a fixed-point in the floating-point expression program. An expression support device and a support program are provided.

  A floating-point representation program supporting apparatus for a floating-point representation program according to the present invention describes a floating-point representation program that describes a computation process that uses a number element of a floating-point representation and a computation process that uses a number element of a fixed-point representation. A floating-point representation program fixed-point representation support apparatus configured by a computer system that performs user support processing when changing to a fixed-point representation program, wherein the floating-point representation program is analyzed to analyze the floating-point representation program Number element extraction means for extracting number elements used in the expression program, number element information storage means for storing number element information representing the extracted number elements, and analysis of the description of the floating-point expression program And an expression with one or more other numerical elements used in the floating-point expression program. A relation extracting means for extracting a relation, a relation information storing means for storing joint relation information representing the extracted relation, and the number element information for each attribute item necessary for fixed-point expression of the number element. Attribute value relationship information representing the relationship between the attribute value of each number element and the attribute value of another number element, and one or more other number elements of the number elements represented by the connection relationship information Attribute value relationship generating means for generating based on the relationship connected by the formula: and means for providing a user interface for setting an attribute value for each attribute item of each of the number elements based on the attribute value relationship information; It becomes the composition which has.

  With such a configuration, based on attribute value relation information that represents the relationship between the attribute value of each attribute element required for fixed-point expression of each numerical element extracted from the floating-point expression program and the attribute value of another numerical element. Therefore, a user interface for setting an attribute value for each attribute item of each numerical element is provided, and the user can specify the attribute value of each numerical element included in the floating-point expression program by the provided user interface. Therefore, the program can be set according to the relationship with the attribute values of other numerical elements without directly analyzing the floating-point expression program and setting it individually.

  The program is a description representing processing by a computer and can be directly or indirectly translated into a language executable by the computer. Further, the attribute item is necessary for expressing a number element in a fixed point, and may include, for example, bit size, resolution, numerical range offset, presence / absence of a code, and the like.

  In the fixed-point expression support device for a floating-point expression program according to the present invention, the means for providing the user interface includes, in addition to the attribute value of each numerical element for each attribute item, based on the attribute value relation information. The display unit can be configured to display the relationship with the attribute values of several elements.

  With such a configuration, the relation between the attribute value of each numerical element extracted from the floating-point expression program and the attribute value of another numerical element is displayed on the display unit. There is no need to individually determine the attribute value for each attribute item of the element from the floating-point expression program, and the attribute value of the number element can be easily determined from the relationship with the attribute value of another number element whose attribute value is known. Can be set to

  Further, in the fixed-point expression support device for a floating-point expression program according to the present invention, the relationship extraction means indicates a relationship in which the number element and one or more other number elements are the right side and the left side of the assignment expression. It can be configured so as to be extracted as a relationship connected by.

  With such a configuration, for the number elements that are in the relationship between the right side and the left side of the assignment formula, the attribute value of the other number element (for example, the left side) of the other (eg, the left side) (for example, the left side) It is possible to generate attribute value relationship information that represents a relationship with the attribute values of one or more numerical elements on the right side).

  Further, in the fixed-point expression support device for a floating-point expression program according to the present invention, the relationship extracting means indicates a relationship in which the number element and one or more other number elements are the right side and the left side of the conditional expression. It can be configured so as to be extracted as a relationship connected by.

  With such a configuration, with respect to the number elements that are in the relationship between the right side and the left side of the conditional expression, the other (for example, the left side) attribute value of the number element (for example, the left side) (for example, the left side) It is possible to generate attribute value relationship information that represents a relationship with the attribute values of one or more numerical elements on the right side).

  Furthermore, in the fixed-point expression support device for a floating-point expression program according to the present invention, the number element extraction means may be configured to extract a variable used in the floating-point expression program as a number element. it can.

  With such a configuration, variables are extracted as numerical elements from the floating-point expression program, so that the user can directly set the attribute value of each attribute item necessary for the fixed-point expression of the variable included in the floating-point expression program. The user interface provided can be set according to the relationship with the attribute values of other numerical elements without analyzing and individually setting the floating point expression program.

  Further, in the fixed-point expression support device for a floating-point expression program according to the present invention, the number element extraction means may be configured to extract a constant used in the floating-point expression program as a number element. it can.

  With this configuration, constants are extracted as numerical elements from the floating-point representation program, so the user can directly enter the attribute values of each attribute item necessary for fixed-point representation of the constants included in the floating-point representation program. The user interface provided can be set according to the relationship with the attribute values of other numerical elements without analyzing and individually setting the decimal point expression program.

  Furthermore, in the fixed-point expression support device for a floating-point expression program according to the present invention, the attribute value relationship information generating unit is configured to display an internal state in the arithmetic processing described by the floating-point expression program among the extracted number elements. Information representing the relationship between the attribute value of the number element to be represented and the attribute value of another number element may be included in the attribute value relationship information.

  With such a configuration, the attribute value information includes information indicating the relationship between the attribute value of the number element representing the internal state in the arithmetic processing described by the floating point expression program and the attribute value of the other number element. The user must directly set the attribute value of each attribute item necessary for the fixed-point representation of the numerical elements representing the internal state in the arithmetic processing described by the floating-point program by directly analyzing the floating-point representation program. Instead, the provided user interface allows the user interface to be set according to the relationship with the attribute values of several other elements.

  In the fixed-point expression support device for a floating-point expression program according to the present invention, the attribute relationship information generating means includes a constant attribute value and other numbers in the floating-point expression program among the extracted number elements. Information representing a relationship with an attribute value of an element can be configured to be included in the attribute value relationship information.

  With such a configuration, the attribute value relationship information includes information indicating the relationship between the attribute values of constants in the floating point representation program and the attribute values of other numerical elements. The user interface does not directly set the attribute value of each attribute item required for fixed-point representation of constants by directly analyzing the floating-point representation program. Can be set according to the relationship.

  The support program according to the present invention is a user who changes a floating-point expression program describing an arithmetic process using a number element of a floating-point expression into a fixed-point expression program describing an arithmetic process using a number element of a fixed-point expression. A support program for causing a computer system to execute support processing, wherein the computer analyzes the description of the floating-point expression program and extracts a number element used in the floating-point expression program Storing numerical element information representing the extracted numerical elements in a numerical element storage means; analyzing the description of the floating-point expression program; and each numerical element used in the floating-point expression program A relationship extracting step of extracting a relationship with one or more other numerical elements, and the extracted relationship A step of storing the relationship information to be represented in the relationship storage means, and the attribute value of another number element other than the attribute value of each number element represented by the number element information for each attribute item required for fixed-point representation of the number element An attribute value relationship generating step for generating attribute value relationship information representing a relationship between the attribute value and the attribute value based on a relationship with one or more other number elements of the number elements represented by the relationship information; And a step of providing a user interface for setting an attribute value for each attribute item of each of the several elements based on the relationship information.

  In the support program according to the present invention, the step of providing the user interface is based on the attribute value relationship information, and the relationship between the attribute values of the number elements and the attribute values of other number elements for the attribute items. Can be displayed on the display unit.

  According to the fixed-point expression device and the support program for a floating-point expression program according to the present invention, other numbers of attribute values for each attribute item necessary for fixed-point expression of each numerical element extracted from the floating-point expression program A user interface is provided for setting an attribute value for each attribute item of each number element based on attribute value relationship information representing a relationship with the attribute value of the element, and the user can float by the provided user interface. Because the attribute value of each number element included in the decimal point representation program can be obtained according to the relationship with the attribute value of other number elements without directly analyzing the floating point representation program directly, Reduces the hassle of setting attribute values for numeric elements that should be represented in fixed-point representation in floating-point representation programs So that it is Rukoto.

It is a flowchart which shows an example of the preparation procedure of a fixed point expression program. It is a block diagram which shows the computer system which comprises the fixed point expression support apparatus of the floating point expression program which concerns on embodiment of this invention. It is a figure which shows an example of the floating point expression program described by C language. It is a flowchart which shows the procedure of the arithmetic processing described by the floating point expression program shown in FIG. It is a flowchart which shows the process sequence in the fixed point expression support apparatus of the floating point expression program which concerns on embodiment of this invention. It is a figure which shows an example of a variable list (number element information). It is a figure which shows an example of the formula list (joining relationship information) of an assignment formula. It is a figure which shows an example of the formula list | wrist (joining relationship information) of a conditional expression. It is a figure which shows an example of the attribute list | wrist of an empty state. It is a figure which shows the result list (the 1) in which the relationship (attribute value relationship information) of the attribute value was described. It is a figure which shows the result list (the 2) in which the relationship (attribute value relationship information) of the attribute value was described. It is a figure which shows a state (the 1) when an attribute value is set to each attribute item of the variable u of a result list (user interface). It is a figure which shows a state (the 2) when an attribute value is set to each attribute item of the variable u of a result list (user interface). It is a figure which shows a state (the 1) when an attribute value is set to each attribute item of the variable y of a result list (user interface). It is a figure which shows a state when an attribute value is set to each attribute item of the variable C of a result list (user interface). FIG. 4 is a diagram showing an example of a fixed-point expression program created by converting the floating-point expression program described in the C language shown in FIG. 3. It is a figure which shows an example of the rule for setting the attribute value about the resolution (attribute item) of a type | formula. It is a figure which shows an example of the rule for setting the attribute value about the size (attribute item) of a formula. It is a figure which shows an example of the rule for setting the attribute value about the offset (attribute item) of a type | formula. It is a figure which shows an example of the rule for setting the attribute value about the code | symbol (attribute item) of a formula.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

  The fixed-point expression program is created, for example, according to the procedure shown in FIG. In FIG. 1, first, a “control designer” who is familiar with target control (for example, vehicle engine fuel injection control, vehicle travel control, etc.) creates a floating-point program according to the contents of arithmetic processing related to control (S1). ). Then, the created floating-point expression program is tested, and the “control designer” confirms whether or not the processing (algorithm) proceeds as intended (S2). If it does not proceed as intended (NO in S2), the “control designer” modifies the created floating-point expression program until it proceeds as intended (YES in S2) (S1). ). If the processing (algorithm) proceeds as intended by the “control designer” according to the created and modified floating-point expression program (YES in S2), the floating-point expression program is completed, and “control designer” Is finished.

  Next, a “software designer” who is familiar with the performance of a computer system (for example, an ECU mounted on a vehicle) that is actually used on behalf of the “control designer” has the constraints (memory capacity, calculation amount) of the computer system. ) And the like are converted into a fixed-point representation program (S3, S4). Specifically, the variables and constants (number elements) used in the floating-point program are extracted from the floating-point representation program, and the attribute items necessary for the fixed-point representation for each of the extracted variables and constants Attribute values (for example, bit size, resolution, numerical range offset, presence / absence of code, etc.) are set. Then, a fixed-point expression program describing an arithmetic process having the same contents as the arithmetic process described in the above-described floating-point expression program is created with the variables and constants in which attribute values of the attribute items are set (defined). (S3).

  Then, the “software designer” causes the actual computer system to execute arithmetic processing according to the created fixed-point expression program, and confirms the execution result (S4). The result of the process according to the fixed-point expression program differs from the result of the process according to the floating-point expression program due to numerical error and overflow (becomes a value that exceeds the range that can be expressed by the fixed point). If this is not permitted as the operation of the computer system, the fixed-point expression program is corrected by resetting the attribute values of the variable and constant attribute items. This operation is repeated, and when the operation result in the computer system of processing according to the fixed-point expression program becomes acceptable, the fixed-point expression program is completed.

  A fixed-point expression support device for a floating-point expression program according to an embodiment of the present invention (hereinafter referred to as a support device in the description of the embodiment) is a numerical element of variables and constants used in the above-described floating-point expression program. This is used to support the user's work when setting the attribute value of each attribute item necessary for the fixed-point expression (S3 in FIG. 1).

  A computer system constituting the support apparatus according to the embodiment of the present invention is configured as shown in FIG.

  In FIG. 2, the computer system (supporting device) has a processing unit (CPU) 101 that executes various processes according to a program. The processing unit 101 is connected to a memory 102, an external storage device 103 such as a hard disk device, an operation unit 105 such as a mouse and a keyboard, and a monitor 106 (display unit) via a bus 107. The processing unit 101 reads, from the external storage device 103, a program related to support processing for fixed-point expression of a floating-point expression program to the memory 103 in accordance with an operation signal based on an operation performed by the user on the operation unit 105, and in accordance with the program Execute the process. Support information generated by executing the processing is displayed on the monitor 106 (user interface).

  A floating-point expression program to be converted into a fixed-point expression is stored in the external storage device 103 in advance. This floating point expression program is written in C language as shown in FIG. 3, for example. The floating-point expression program shown in FIG. 3 describes a calculation process for correcting the input u from the sensor to obtain the output y, and the processing procedure can also be expressed by the flowchart shown in FIG. This calculation process will be briefly described.

Since the input u from the sensor includes a lot of noise, the input u is subjected to low-pass filter processing. In particular,
y = C · u + (1−C) · y
y: previous output value C: filter coefficient (0.0 ≦ C ≦ 1.0)
Accordingly, the filter output y is calculated.
Furthermore, the circuit of the sensor itself and the signal line connecting the sensor to the CPU are at risk of physical failure, and the CPU needs to detect the failure and perform appropriate processing. Therefore, if the input u sampled from the sensor is continuous (Nf 1 time) and out of the specified range, the value (true) is set to the variable fx, assuming that the sensor is faulty (disconnected, shorted), and sampled input. If the change amount (| u1-u |) of u is continuous (Nf 2 times) and becomes larger than a predetermined value (Udelta), the value (true) is set in the variable fx as a sensor failure (sensor destruction). . When the variable fx is set to a value (ture) (in the case of a sensor failure), the filter output y is forcibly set to a safe fixed value (Ysafe).

  Hereinafter, the operation of the support apparatus will be described by taking as an example a case where a floating-point expression program (see FIGS. 3 and 4) describing such arithmetic processing is expressed as a fixed-point expression.

  In accordance with a predetermined operation of the operation unit 105 by the user (“software designer”), the processing unit 101 stores, for example, a floating-point expression program written in C language from the external storage device 103 on the memory 102 as shown in FIG. And the support process is executed according to the procedure shown in FIG.

The processing unit 101 analyzes the lexical structure of the floating-point expression program expanded in the memory 102, and uses variables and constants (number elements: hereinafter simply referred to as variables for convenience) used in the floating-point expression program. A variable list is created by extracting and listing the information (number element information) (S11). The variable list is stored in the memory 102 (several element storage means). The processing unit 101 extracts information (definition information) defined for the variable in the floating-point expression program together with the variable, and the definition information is included in the variable list so as to correspond to the variable. Yes. For example, the following variables are extracted from the floating-point expression program described in the C language shown in FIG.
u
y
nf1
nf2
fx
C
Umin
Umax
u1
Udelta
Nf1
Nf2
Ysafe

  Along with these variables, “type” and “type”, which are definition information defined for each variable, are extracted. From the definition information of the variable (variable name), “type”, and “type”, for example, FIG. A variable list as shown in FIG. The “type” is definition information about the type of variable (number element) such as input, output, internal information, and constant. The internal information refers to an internal state such as a count value and a set value held in the arithmetic processing described by the floating point expression program. "Type" is a variable (number element) type defined in the C language, and is a floating point type (float), integer value type (int), boolean value type (bool), double precision floating point type (Double), character type (char), etc.

Next, the processing unit 101 analyzes the lexical structure of the floating-point expression program expanded in the memory 102 and connects it with an expression with one or more other variables of each variable used in the floating-point program. Are extracted (S12). Specifically, the relationship between the variable and one or more other variables is the right and left sides of the assignment expression for updating the value of the variable, and the variable and the other one or more variables are conditional branches. The relationship between the right side and the left side of the conditional expression related to is extracted, and an expression list in which information representing the relationship (binding relation information) is listed is generated. The expression list is stored in the memory 102 (relation information storage means). For example, the following relationship is extracted from the floating-point expression program written in C language shown in FIG. 3 (see FIG. 4).
Substitution formula nf1 = nf1 + 1
fx = true
nf 1 = 0
nf2 = nf2 + 1
nf2 = 0
u1 = u
y = C * u + (1-C) * y
y = Ysafe

Conditional expression u <umin
nf1> Nf1
Umax <u
u <u1-Udelta
u1 + Udelta <u
nf2> Nf2
fx = true

  Then, an expression list as shown in FIG. 7 is created from the relationship related to the substitution expression, and an expression list as shown in FIG. 8 is generated from the relation related to the conditional expression.

  When the processing unit 101 creates the variable list (see FIG. 6) and the expression list (see FIGS. 7 and 8), it selects a variable from the variable list (S13), and the variable is not included in the attribute list described later. And whether or not the “type” of the variable is a floating-point type (float) or a double-precision floating-point type (double) is determined (S14). When the variable selected from the variable list is not yet included in the attribute list and the “type” is a floating-point type (float) or a double-precision floating-point type (double) (YES in S14), The processing unit 101 adds the variable selected from the variable list to the attribute list read from the external storage device 103 and developed in the memory 102 (S15).

This attribute list is a list of attribute values of each attribute item (bit size, resolution, numerical range offset, sign presence / absence) required for fixed-point representation of each variable. 9 is configured. From the variable list shown in FIG. 6, the next variable whose “type” is a floating-point type (float) or a double-precision floating-point type (double) is added to the attribute list (see FIG. 9).
u
y
C
Umin
Umax
u1
Udelta
Ysafe

  When the processing unit 101 finishes the processing (S14, S15) for all variables in the variable list (see FIG. 6) (YES in S16), the floating-point type included in the floating-point expression program is stored in the memory 102. An attribute list shown in FIG. 9 is formed in which all variables of (float) and double-precision floating-point type (double) are listed. Next, the processing unit 101 selects a variable from the attribute list (S17), and refers to the variable list to determine whether the “type” of the selected variable is a constant or an internal state (S18). ). When the “type” of the selected variable is a constant or an internal state (YES in S18), the processing unit 101 determines the attribute of the other variable of the attribute value of each attribute item necessary for the fixed-point representation of the selected variable. The relationship with the value is set according to the following rule from the relationship between the variables shown in the formula list (see FIGS. 7 and 8) (S19).

(1) Rule 1
The unary variable on the right side or the left side in the expression list has the same attribute value as the variable on the other side (may be multiple terms).
(2) Rule 2
When the variable is a term of the addition / subtraction expression, the variable has the same attribute value as the variable of the other term of the addition / subtraction expression.

  On the other hand, if the “type” of the variable selected from the attribute list is not a constant or an internal state (NO in S18), that is, if the “type” is input or output, the processing unit 101 performs processing for the variable. No particular action is taken.

  As a result of the above processing, each attribute item (bit size, resolution, offset, presence / absence of sign) necessary for the fixed-point expression of the variable Umin according to the rule 1 based on the relationship between the variable Umin and the variable u in the expression list shown in FIG. Is determined to be the same as the attribute value of the variable u. Similarly, regarding the variable Umax, it is determined that the attribute value of each attribute item necessary for the fixed-point expression is the same as the attribute value of the variable u from the relationship with the variable u of the expression list shown in FIG. Is done. Regarding the variable u1, it is determined that the attribute value of each attribute item necessary for the fixed-point expression is the same as the attribute value of the variable u according to the rule 1 from the relationship with the variable u in the expression list shown in FIG. . Regarding the variable Ysafe, it is determined that the attribute value of each attribute item necessary for the fixed-point expression is the same as the attribute value of the variable y according to the rule 1 from the relationship with the variable y in the expression list shown in FIG. . Therefore, as shown in FIG. 10, information indicating that each attribute item corresponding to each of the variables Umin, Umax, u1 in the attribute list has the same attribute value as the attribute value of the variable u (attribute value relation information) A reference expression (= u) is associated, and each attribute item corresponding to the variable Ysafe has a reference expression (attribute value relation information) indicating that the attribute value is the same attribute value as the attribute value of the variable y. = Y) is associated.

  For the variable Udelta, the attribute value of each attribute item necessary for its fixed-point expression is the attribute value of the variable u1 according to the rule 2 because it is included in the same addition / subtraction formula as the variable u1 on the right side in the formula list shown in FIG. It is determined that Therefore, as shown in FIG. 11, each attribute item corresponding to the variable Udelta in the attribute list has a reference expression (= attribute value relation information) indicating that the attribute value is the same attribute value as the attribute value of the variable u1. u1) are associated.

  When such processing is performed for all variables listed in the attribute list (YES in S20), the processing unit 101 stores the variables Umin, Umax, u1, Udelta, and Ysafe in the memory 102. A result list as shown in FIG. 11 is formed in which attribute relationship information (= u, = u1, = y) indicating that the attribute values of these variables are the same as other attribute values is associated with the attribute items. Is done.

  The variables u and y have the “type” of input and output, and are neither constants nor internal states. Therefore, information (attributes) indicating the relationship between the attribute values of other variables for each attribute item. (Related information) is not set. The attribute value of each attribute item necessary for the fixed-point expression of the variables u and y whose “type” is input and output is to be set by the user (“software designer”) as an external requirement. In addition, the variable C has a “type” that is a constant. In either of the rule 1 and the rule 2, the information of the expression list (see FIG. 7) indicates that the attribute value of another variable Since the relationship cannot be determined, information representing the relationship (attribute relationship information) is not set.

The processing unit 101 displays a result list as shown in FIG. 11 on the monitor 106 (display unit). The user looks at the result list (user interface) displayed on the monitor 106 and knows that no attribute value is set for each attribute item of the variables u, y, and C, and operates the operation unit 105 to operate the variable u. , Y, and C are input (set) for attribute values. For example, as shown in FIG. 12, for each attribute item (bit size, resolution, offset, sign) of the variable u in the result list displayed on the monitor 106, its attribute value (bit size: 8, resolution: 2 ^{− 4} , offset: 0, sign: none), the processing unit 101 sets the input attribute value to each attribute item of the variable u in the result list, and the same attribute value as the variable u Based on the attribute relation information (reference expression = u) indicating that the same as the attribute value of the variable u is set in each attribute item of the variables umin, Umax, u1. As described above, when the attribute value of each attribute item of the variable u1 is set according to the attribute relationship information (= u) with the variable u, the processing unit 101 further represents the attribute relationship indicating the same attribute value as the variable u1. Based on the information (reference formula = u1), the attribute value of each attribute item of the variable Udelta is set to the variable u1, that is, the same attribute value as the attribute value of the variable u, as shown in FIG.

Furthermore, for each attribute item (bit size, resolution, offset, sign) of the variable y in the result list (see FIG. 13) displayed on the monitor 106, its attribute value (bit size: 8, resolution: 2 ⁻⁴ , When an input operation of offset: 0, sign: none is performed, the processing unit 101 sets the input attribute value in each attribute item of the variable y in the result list and has the same attribute value as the variable y. Based on the attribute relationship information indicating that (reference expression = y), the attribute value of each attribute item of the variable Ysafe is set to the same attribute value as the attribute value of the variable y as shown in FIG. Further, when the attribute value is input to each attribute item of the variable C in the result list (see FIG. 14) displayed on the monitor 106, the processing unit 101 causes each attribute of the variable C in the result list to be input. The input attribute value is set in the item, and the result list displayed on the monitor 106 is updated as shown in FIG.

  When the attribute values of the respective attribute items necessary for the fixed-point representation of all the variables listed in the attribute list (see FIG. 9) are set in this way, the user (“software designer”) With reference to the attribute value of each attribute item for each variable appearing in the result list (see FIG. 15) displayed on the monitor 106, an operation of creating a fixed-point expression program from the floating-point expression program is performed. As a result, the floating-point expression program described in the C language shown in FIG. 3 is converted into, for example, a fixed-point expression program shown in FIG. At this time, well-known application software (for example, a fixed-point expression program for creating a fixed-point expression program from a floating-point expression program using attribute value information of each attribute item necessary for fixed-point expression set for each variable) You can also use the Fixed Point Toolbox (MathWorks, USA).

  In the support apparatus as described above, other variables (u, Umax, Umax, u1, Udelta, Ysafe) of the attribute values necessary for the fixed-point representation of each variable extracted from the floating-point representation program. Since a result list (see FIG. 11) in which attribute value relationship information (reference expression = u, = y) representing a relationship with the attribute value of y) is set is displayed on the monitor 106 as a user interface, the user (“soft The designer ”) does not need to analyze the floating-point representation program for all the variables included in the floating-point representation program and individually determine the attribute values thereof, but some variables (u, y, C ), The attribute values of other variables can be obtained according to the attribute value relation information listed in the displayed result list (see FIG. 11).

  Further, in the support device, if the attribute values of the some variables (u, y, C) are input to the result list displayed on the monitor 106 (see FIGS. 12, 13, and 14), the support device For other variables (Umin, Umax, u1, Udelta, Ysafe), attribute values are automatically set based on the relationship with the attribute values of the input variables (attribute value relationship information) and displayed in the result list. Since it is displayed, the user can easily know the attribute value of each attribute item necessary for the fixed-point expression of the variable included in the floating-point expression program.

  As described above, according to the support device, it is possible to reduce time and effort for setting attribute values for variables to be expressed in fixed-point numbers in a floating-point expression program. Therefore, it becomes possible to create a fixed-point expression program more efficiently.

  In the above-described support device, the relationship between the attribute value of the variable and the attribute value of the other variable (attribute value relationship) is determined according to the rules 1 and 2, but by determining the attribute value of the entire expression You can also determine that relationship. The resolution attribute item is, for example, according to the relationship shown in FIG. 17, and the bit size attribute item is, for example, according to the relationship shown in FIG. Further, according to the relationship, for the attribute item with or without the sign, for example, according to the relationship shown in FIG. 20, the addition formula (a + b), the subtraction formula (ab), the multiplication formula (a × b), and the division formula ( The attribute value of a / b) can be determined.

  In the support apparatus, the result list (FIGS. 11 to 15) is displayed on the monitor 106 as a user interface. However, the variable list (FIG. 6) and the expression list (FIGS. 7 and 8) are displayed. It may be displayed on the monitor 106. Further, only the result list (FIG. 11) that displays the relationship (attribute value relationship information) of the attribute value of each variable (Umin, Umax, u1, Udelta, Ysafe) with the attribute value of another variable (u, y). It may be displayed. In this case, according to the attribute value relation information, the user may use the operation unit 105 to perform an operation for inputting attribute values for each variable.

  In the above-described embodiment, the program written in the C language (floating point expression program and fixed point expression program) has been described. However, the present invention is not limited to such a program, and is directly applied to a computer executable language. It can be applied to programs written in other languages or in other formats as long as they can be translated either indirectly or indirectly.

  As described above, the fixed-point representation support device for a floating-point representation program according to the present invention reduces the effort for setting attribute values for the number elements to be fixed-point representation in the floating-point representation program. The floating-point representation program that describes arithmetic processing using numerical elements such as variables of floating-point representation is changed to a fixed-point representation program that describes arithmetic processing using numerical elements of fixed-point representation. This is useful as a fixed-point expression support apparatus and support program for a floating-point expression program that performs user support processing.

101 processing unit (CPU)
102 Memory 103 External storage device (HDD)
105 Operation unit 106 Monitor (display unit)
107 bus

Claims (10)

In a computer system that performs user support processing when changing a floating-point expression program that describes arithmetic processing using a number element of a floating-point expression to a fixed-point expression program that describes arithmetic processing using a number element of a fixed-point expression A fixed-point expression support device for a configured floating-point expression program,
Analyzing a description of the floating-point expression program, and extracting a number element used in the floating-point expression program;
Number element information storage means for storing number element information representing the extracted number elements;
Analyzing the description of the floating-point expression program, and extracting a relationship connected by an expression with one or more other numerical elements of each numerical element used in the floating-point expression program;
Relationship information storage means for storing connection relationship information representing the extracted relationship;
The attribute value relation information representing the relation between the attribute value of each number element represented by the number element information for each attribute item necessary for the fixed-point representation of the number element and the attribute value of another number element. Attribute value relationship generating means for generating based on a relationship connected by an expression with one or a plurality of other number elements represented by the relationship information;
A fixed-point support apparatus for a floating-point expression program, comprising: means for providing a user interface for setting an attribute value for each attribute item of each numerical element based on the attribute value relation information.   The means for providing the user interface causes a display unit to display a relationship between an attribute value of each number element and an attribute value of another number element for each attribute item based on the attribute value relationship information. A fixed-point expression support device for a floating-point expression program according to claim 1.   The floating point expression program according to claim 1, wherein the relation extraction unit extracts a relation in which a numerical element and one or a plurality of other numerical elements are a right side and a left side of an assignment expression as a relation connected by the expression. Fixed-point expression support device.   The floating point according to any one of claims 1 to 3, wherein the relation extracting unit extracts a relation in which a numerical element and one or more numerical elements are a right side and a left side of a conditional expression as a relation connected by the expression. Fixed-point expression support device for decimal point expression program.   5. The floating-point expression program fixed-point expression support apparatus according to claim 1, wherein the number element extraction unit extracts a variable used in the floating-point expression program as a number element.   6. The fixed-point expression device for a floating-point expression program according to claim 1, wherein the number element extraction unit extracts a constant used in the floating-point expression program as a number element.   The attribute value relationship information generating means is configured to determine a relationship between an attribute value of another number element of the attribute value of the number element representing an internal state in the arithmetic processing described by the floating point expression program among the extracted number elements. 7. The floating-point expression program fixed-point expression support apparatus according to claim 1, wherein information to be included is included in the attribute value relation information.   The attribute relation information generating means includes information representing a relation between a constant attribute value in the floating-point expression program and an attribute value of another number element among the extracted number elements in the attribute value relation information. Item 8. A fixed-point expression support device for a floating-point expression program according to any one of Items 1 to 7. Causes a computer system to execute user-supported processing when changing a floating-point representation program describing arithmetic processing using a number element of floating-point representation into a fixed-point representation program describing arithmetic processing using a number element of fixed-point representation A support program,
In the computer,
Analyzing the description of the floating-point expression program and extracting a number element used in the floating-point expression program; and
Storing number element information representing the extracted number elements in a number element storage means;
Analyzing the description of the floating-point expression program, and extracting a relationship with one or more other numerical elements of each numerical element used in the floating-point expression program;
Storing relationship information representing the extracted relationship in a relationship storage means;
Attribute value relationship information representing a relationship between an attribute value of each number element represented by the number element information for each attribute item necessary for fixed-point representation of the number element and an attribute value of another number element. An attribute value relationship generating step for generating based on a relationship with one or more other numerical elements of each of the numerical elements represented by information;
Providing a user interface for setting an attribute value for each attribute item of each of the number elements based on the attribute value relation information.   The step of providing the user interface causes a display unit to display a relationship between an attribute value of each numerical element and an attribute value of another numerical element for each attribute item based on the attribute value relationship information. 9. Support program according to 9.

Images