JP7352107B2 - Arithmetic error warning device and arithmetic error warning program - Google Patents

Description

The present invention relates to an arithmetic error warning device and an arithmetic error warning program.

In calculations performed on a computer, the formats of the numbers on the left and right sides of the operator must match. Here, the numerical format is, for example, a unit. For example, an operation that adds length m (meter) and mm (millimetre) is an error.

Another example is LSB resolution information representing resolution below the decimal point. For example, the decimal number "221" becomes "11011101" in binary, and if the lower 2 bits are defined as below the decimal point, "110111.01" = 55.25 can be expressed. "Define the lower 2 bits as below the decimal point" is long, so in the case of "define the lower 2 bits as below the decimal point", the lowest bit is 0.25, so write it as "221 [LSB]" By multiplying by "0.25[1/LSB]", the decimal number can be obtained as 221[LSB]*0.25[1/LSB]=55.25. Since "0.25" appears to be a unit, it can be said that it represents resolution.

Numerical values handled in "embedded" development etc. are physical quantities and have units, so for example, in the case of mV, the LSB is expressed as 0.25 [mV/LSB]. Note that in the above example, the resolution is a power of 2, but since the target value to be handled is a physical quantity, values other than powers of 2 are also possible, such as 0.77 [mV/LSB]. It is. If the LSB of a 16-bit signed integer is 0.001 [m/LSB], a length range of -32.768 m to 32.767 m can be expressed. Also, when using LSB, a bias value called offset is used, and if the LSB of a 16-bit signed integer is 0.01 [°C/LSB] and the offset is -273.15°C, -273 It becomes possible to express temperatures in the range of .15°C to 382.20°C.

As mentioned above, when performing calculations on a computer, errors may occur when changing the format so that the formats of the numbers on the left and right sides of the operator in the calculation match, and if left as is, the embedded device may This means that you will not be able to make it work. However, currently, defects in such source programs are checked by humans, and there is a possibility that they may be overlooked. In addition, human checking requires a huge amount of time and cost, which is still insufficient.

Patent Document 1 discloses a compiler device that changes data types. The device of this cited document 1 concerns data types. Further, although Patent Document 2 describes that type conversion is performed, it is not a type conversion in the case of an operation. Further, Patent Document 3 discloses a method for deleting unnecessary type conversion operations. However, the calculation does not detect defects that lead to errors, and it is not possible to make the source program appropriate.

Japanese Patent Application Publication No. 2003-150386 Japanese Patent Application Publication No. 2011-85967 Japanese Patent Application Publication No. 5-324344

The present invention has been made in view of the current state of technology regarding calculation errors as described above, and its purpose is to detect calculation errors that can issue an appropriate warning when there is a possibility that a calculation error may occur in a source program. An object of the present invention is to provide a warning device and a calculation error warning program.

The calculation error warning device according to the present embodiment includes a check-required operation detection means for detecting a check-required operation that changes the expression format of a numerical value by an operation included in the source code of a source program; a format conversion/inverse conversion processing means that performs a format conversion process to change the expression format , and further performs a format inverse conversion process to restore the format conversion process result changed by the format conversion process; and the format conversion/inverse conversion process. Error detection means that calculates the difference between the returned value obtained by the processing means and the original value before the format conversion process as an error, and compares the error with a preset error threshold and outputs a calculation error warning. The invention is characterized by comprising an error warning means.

FIG. 1 is a block diagram of an arithmetic error warning device according to a first embodiment of the present invention. FIG. 3 is a diagram showing an example of an error threshold used in the calculation error warning device according to the first embodiment of the present invention. 1 is a flowchart showing the operation of the arithmetic error warning device according to the first embodiment of the present invention. FIG. 3 is a block diagram of an arithmetic error warning device according to a second embodiment of the present invention. FIG. 7 is a diagram illustrating an example of an error rate threshold used in the calculation error warning device according to the second embodiment of the present invention. 7 is a flowchart showing the operation of the arithmetic error warning device according to the second embodiment of the present invention.

DESCRIPTION OF THE PREFERRED EMBODIMENTS A computation error warning device and a computation error warning program according to embodiments of the present invention will be described below with reference to the accompanying drawings. In each figure, the same components are given the same reference numerals and redundant explanations will be omitted.

FIG. 1 shows an arithmetic error warning device according to a first embodiment of the present invention. The calculation error warning device according to this embodiment includes a database 200, a CPU (central processing unit) 100, an input unit 150 such as a keyboard for inputting information, and a display unit 300 as an output unit for outputting information. ing. That is, the calculation error warning device can be configured by a computer device such as a workstation, a server, or a personal computer.

The CPU 100 is equipped with a check-required operation detection means 110, a format conversion/inverse conversion processing means 120, an error detection means 130, and an error warning means 140 as means realized by a program. The check-required operation detection means 110 analyzes the source code of a source program input from the input unit 150 and arrives at the CPU 100, and changes the expression format of a numerical value by the operation included in the source code of the source program. It detects calculations. The expression format may include, for example, LSB resolution information representing resolution below the decimal point. The source program may be stored in the database 200 from a storage medium or network (not shown).

In other words, a check-required operation in which the expression format of a numerical value is changed by an operation included in the source code refers to a case where the expression format of a numerical value is changed by, for example, LSB conversion and LSB inverse conversion.

The format conversion/inverse conversion processing means 120 performs a format conversion process to change the expression format of a numerical value by the above-mentioned check-required operation, and further performs a format inverse conversion process to restore the format conversion process result changed by this expression format process. This is what we do. The error detection means 130 determines as an error the difference between the return value obtained by the format conversion/inverse conversion processing means 120 and the original value before the format conversion processing. The error warning means 140 compares the above error with a preset error threshold and outputs a calculation error warning.

The preset error threshold is given for each operation included in the source code of the source program, and is stored in the error threshold file 210 of the database 200. That is, as shown in FIG. 2, an error threshold value is set numerically for each address where an operation is described in the source program and for each operator of the operation. Of course, in the case of numerical values for which no error occurs due to format conversion, no error threshold is set.

The error threshold can be composed of a first error threshold for alerting and a second error threshold larger than the first error threshold. The error warning means 140 outputs a warning message when the error exceeds the first error threshold and is between the second error threshold and indicates that the error exceeds the second error threshold. In some cases, a warning message can be output. Here, an example of a warning message can be ``There is no problem with this format conversion, but please be careful when calculating this number several times.'', and a warning message can be `` For example, there is a large error in this format conversion, which will result in a calculation error, so we would like to warn you.''. Further, the error warning means 140 may output a message notifying appropriate conversion such as "There is no problem with this format conversion" when the error is less than or equal to the first error threshold. These warning messages are stored in the warning message file 220 of the database 200.

The arithmetic error warning device configured as described above issues an arithmetic error warning as shown in the flowchart of FIG. For example, the source code of a source program that is input from the input unit 150 and arrives at the CPU 100 is analyzed starting from the upper line, and check-required operations that change the representation format of numerical values by operations included in the source code of the source program are detected. (S11), and it is determined whether there is a corresponding calculation (S12).

When branching to YES in step S12, the process proceeds to step S13, where a format conversion process is performed to change the expression format of the numerical value based on the check-required operation detected in step S12, and further, a format conversion process changed by this expression format process is performed. A format inverse conversion process is performed to return the result to its original state (S13). In the process of step S13, if the numerical value is a physical quantity and the format of the numerical value is LSB conversion, in the format conversion process, logical value = (type conversion) (physical quantity x LSB value) (Formula 1 ) to obtain a logical value that is the result of format conversion processing. The above "x" may also become "÷". In addition, in the format inverse conversion process that returns the format conversion process result to the original format, a physical quantity (restored result) that is the format conversion result is obtained using the logical value obtained by the following (Equation 2). Physical quantity = (type conversion) (logical value ÷ LSB value)... (Formula 2)
In the above (Formula 2), the above "÷" may become "×".

Let us take as an example a case where, for example, an integer 32767 with a 16-bit sign is converted as a physical quantity into a logical value with an LSB value of 0.001 using the above (Formula 1), and then the LSB is inversely converted to return it to a physical quantity. When the physical quantity 32767 is converted into a logical value with an LSB value of 0.001, a logical value of 32.767 is calculated.
That is, physical quantity (32767) x LSB value (0.001) = logical value (32.767)
It is.

Assuming that the result of calculating this logical value is 32.7679, the formula for calculating the error is:
Error = (Logical value/LSB value) - (Physical quantity type) (Logical value/LSB value)... (Formula 3)
It is.

As described above, since the physical quantity is a 16-bit signed integer (physical quantity type), the error is the value below the decimal point that is rounded down when LSB conversion is performed from a logical value to a physical value.
In the above example, the value obtained by dividing the logical value by the LSB value (logical value/LSB value) is 32.7679, so the difference from the logical value (32.767) of 0.9 is an error.

The result of the format inversion process is obtained as described above, and the error is obtained from error=(logical value/LSB value)−(physical quantity type)(logical value/LSB value) (Formula 3) (S14). “(Physical quantity type) (logical value/LSB value)” in (Formula 3) is the result of calculating the physical quantity before the format inversion processing (result of the format inversion processing).

When the error is determined as described above, a message is determined by comparing the error with a preset error threshold, and the message is output from the display means 300 (outputting a calculation error warning (S15)). When the error threshold is composed of a first error threshold for alerting and a second error threshold larger than the first error threshold, the following processing is performed. In other words, if the above error exceeds the first error threshold and is between the second error threshold, the message "There is no problem with this format conversion, but be careful when calculating this number several times. If the above error exceeds the second error threshold above, a warning message will be output such as ``The error in this format conversion is large and will result in a calculation error.''. ” etc. can be operated to output a warning message. Further, the error warning means 140 may output a message notifying appropriate conversion when the error is less than or equal to the first error threshold.

FIG. 4 shows an arithmetic error warning device according to a second embodiment of the present invention. The apparatus of the second embodiment has the configuration of the apparatus of the first embodiment, and in addition, the CPU 100A is equipped with error rate calculation means 160. The error rate calculation means 160 determines that the error calculated by the error detection means 130 is the maximum value of the value range of the numerical values (here, physical values) handled in the calculation included in the source code of the source program. This method calculates the error rate, which indicates how much of the error rate there is. In this second embodiment, the error warning means 140 compares the error rate with a preset error rate threshold and outputs a calculation error warning.

The preset error rate threshold is given for each operation included in the source code of the source program, and is stored in the error rate threshold file 230 of the database 200. That is, as shown in FIG. 5, an error rate threshold value is set numerically for each address where an operation is described in the source program and for each operator of the operation. Of course, in the case of numerical values for which no error occurs due to format conversion, no error rate threshold is set.

The error threshold can be composed of a first error threshold for alerting and a second error threshold larger than the first error threshold. The error warning means 140 outputs a warning message when the error exceeds the first error threshold and is between the second error threshold and indicates that the error exceeds the second error threshold. In some cases, a warning message can be output. Here, an example of a warning message can be ``There is no problem with this format conversion, but please be careful when calculating this number several times.'', and a warning message can be `` For example, there is a large error in this format conversion, which will result in a calculation error, so we would like to warn you.''. Further, the error warning means 140 may output a message notifying appropriate conversion such as "There is no problem with this format conversion" when the error is less than or equal to the first error threshold. These warning messages are stored in warning message file 220A of database 200.

The computation error warning device according to the second embodiment configured as described above issues a computation error warning as shown in the flowchart of FIG. The flowchart in FIG. 6 showing the operation of the second embodiment is almost the same as the flowchart in FIG. 3 showing the operation in the first embodiment, and step S16 is inserted after step S14 in FIG. Step S15 has been changed to step S15A. Therefore, below, processing that is different from the first embodiment will be mainly explained.

That is, when an error is detected in step S14, next, the error determined in step S14 is within the value range of the numerical values (here, physical values) handled in the calculation included in the source code of the source program. An error rate indicating the percentage of the maximum value is calculated (S16).

For example, if the range of physical quantities (numeric values) handled in the above source program is -10 to 10 and the obtained error is 0.9, the error rate will be 9% (0.9/10). . In addition, if the range of physical quantities (numeric values) handled in the above source program is -100 to 100 and the obtained error is 0.9, the error rate is 0.9% (0.9/100) becomes. In this way, even if the error is the same, if the range of physical quantities (numeric values) handled in the source program is different, the error rate will be different. That is, the error rate represents the degree to which errors affect the overall calculations of the system, and when the error rate is small (large), it can be said that the degree of error imparted to the system is small (large). In this embodiment, the error rate is used from this viewpoint.

When the error rate is determined as described above, a message is determined by comparing the error rate and a preset error rate threshold, and the message is output from the display means 300 (a calculation error warning is output (S15A). )). When the error rate threshold is composed of a first error rate threshold for alerting and a second error rate threshold larger than the first error rate threshold, the following processing is performed. In other words, if the above error exceeds the first error rate threshold and is between the second error rate threshold, the message ``There is no problem with this format conversion, but if you calculate this number several times, If the above error exceeds the second error rate threshold, a warning message such as ``This format conversion has a large error and will cause a calculation error'' will be output. It can be operated to output a warning message such as ". Further, the error warning means 140 may output a message notifying appropriate conversion when the error is less than or equal to the first error rate threshold.

As described above, according to this embodiment, when a variable is LSB-converted into a numerical value such as a logical value by an operation in an operation and then returned to its original expression format by LSB conversion (inverse LSB conversion), an error occurs. For example, if a particularly large error occurs, a warning can be reliably given to prompt the program to be changed, and this can be expected to contribute to the creation of an appropriate program.

In the first embodiment described above, a warning message based on an error is output, and in the second embodiment, a warning message based on an error rate is output. A configuration may be adopted in which an operation selection command is given from the input unit 150 so as to cause the operation to be performed according to the function.

100, 100A CPU
110 Check operation detection means 120 Format conversion/inverse conversion processing means 130 Error detection means 140 Error warning means 150 Input section 200 Database 210 Error threshold file 220, 220A Warning message file 230 Error rate threshold file 300 Display means

Claims (10)

Check-required operation detection means for detecting a check-required operation that changes the representation format of a numerical value by an operation included in the source code of the source program;
Format conversion/inverse conversion processing means that performs a format conversion process to change the representation format of a numerical value by the check-required operation, and further performs a format inverse conversion process to restore the format conversion process result changed by the format conversion process; ,
error detection means for determining the difference between the return value obtained by the format conversion/inverse conversion processing means and the original value before the format conversion processing as an error;
An arithmetic error warning device comprising: error warning means for comparing the error with a preset error threshold and outputting an arithmetic error warning. The error threshold includes a first error threshold for alerting and a second error threshold larger than the first error threshold,
The error warning means outputs a warning message when the error exceeds the first error threshold and is between the second error threshold, and the error warning means outputs a warning message when the error exceeds the second error threshold. 2. The arithmetic error warning device according to claim 1, wherein the arithmetic error warning device outputs a warning message if the calculation error occurs. comprising error rate calculation means for calculating an error rate indicating what percentage of the error is with respect to the maximum value of the value range of the numerical value handled in the calculation,
2. The calculation error warning device according to claim 1, wherein the error warning means compares the error rate with a preset error rate threshold and outputs a calculation error warning. The error rate threshold includes a first error rate threshold for alerting and a second error rate threshold larger than the first error rate threshold,
The error warning means outputs a warning message when the error rate exceeds the first error rate threshold and is between the second error rate threshold, and when the error rate exceeds the second error rate threshold. 4. The arithmetic error warning device according to claim 3, wherein a warning message is output when the rate threshold value is exceeded. The check-required operation detection means detects an operation that is a change in the expression format using LSB resolution information representing the resolution below the decimal point, which is the format of the left and right numbers of the operator in the operation included in the source code of the source program. The arithmetic error warning device according to any one of claims 1 to 4, wherein the arithmetic error warning device detects. The computer in the arithmetic error warning device,
A check-required operation detection means for detecting a check-required operation that changes the representation format of a numerical value by an operation included in the source code of a source program;
Format conversion/inverse conversion processing means that performs a format conversion process to change the representation format of a numerical value by the check-required operation, and further performs a format inverse conversion process to restore the format conversion process result changed by the format conversion process;
error detection means for determining the difference between the return value obtained by the format conversion/inverse conversion processing means and the original value before the format conversion processing as an error;
A computation error warning program that functions as an error warning unit that compares the error with a preset error threshold and outputs a computation error warning. The error threshold includes a first error threshold for alerting and a second error threshold larger than the first error threshold,
The computer serves as the error warning means, and when the error exceeds the first error threshold and is between the second error threshold, outputs a warning message, and outputs a warning message when the error exceeds the second error threshold. 7. The calculation error warning program according to claim 6, wherein the calculation error warning program is configured to output a warning message when the calculation error exceeds . causing the computer to function as an error rate calculation means for calculating an error rate indicating what percentage of the error is with respect to the maximum value of the value range of the numerical value handled in the calculation;
7. The calculation error according to claim 6, wherein the computer is operated as the error warning means to compare the error rate with a preset error rate threshold and output a calculation error warning. Warning program. The error rate threshold includes a first error rate threshold for alerting and a second error rate threshold larger than the first error rate threshold,
The computer serves as the error warning means, and when the error rate exceeds the first error rate threshold and is between the second error rate threshold, outputs a warning message, 9. The calculation error warning program according to claim 8, wherein the program is configured to output a warning message when the error rate threshold of 2 is exceeded. Using the computer as the check-required operation detection means, changing the expression format using LSB resolution information representing the resolution below the decimal point, which is the format of the numerical values on the left and right sides of the operator in the operation included in the source code of the source program. The calculation error warning program according to any one of claims 6 to 9, characterized in that the program functions to detect a calculation that is.

Images