JP5589919B2 - Design support device, design support program - Google Patents

Description

  The present invention relates to a design support apparatus that executes arithmetic processing based on a model.

  Conventionally, for example, in development of a program for an engine ECU of a vehicle, a developer does not directly describe a source code, but describes a function of a target program in the form of a “model” that is easier to create and has better visibility. There is a case. A developer uses a workstation, personal computer, or the like in which a program development environment corresponding to this model is installed, performs simulation such as arithmetic processing defined in the model, and automatically generates source code from the model. .

  Program development for automatically generating source code based on a model in this way is called model-based development, and Matlab (registered trademark) is known as a program development environment for performing model-based development. In Matlab (registered trademark), a developer describes a model using Simulink (registered trademark) that provides an additional function of the program development environment.

  In Simulink (registered trademark), a model is described as a combination of blocks, which are functional units that perform various operations and data reading from a file, and connections that indicate input / output of data between blocks. That is, a model is an aggregate that expresses input / output of data in engine control, various arithmetic processes, and the like by connecting a plurality of blocks by connection.

  Patent Document 1 describes a source code generation apparatus used for such model-based development. According to the code generation device described in Patent Literature 1, when generating a source code corresponding to a specific type from one model in which functions of a plurality of types (for example, destination and model of the device) are described, Function descriptions for unsupported types are removed from the model. Thereby, it is possible to prevent generation of unnecessary source code, and it is possible to generate source code with a small ROM capacity.

JP 2004-227500 A

  By the way, in programming, as variable types that can represent numbers with decimal numbers, fixed-point numbers that represent a range of numbers with a preset precision and a wider range of numbers than fixed-point numbers can be represented. The floating-point number type is known. In general, it is possible to improve the accuracy of calculation by using a floating-point number type variable rather than a fixed-point number type variable. However, on the other hand, the ROM and RAM capacities required for operating the program, There is a problem that the processing load of the CPU increases.

  On the other hand, in the development of a program for an embedded system such as an in-vehicle device, a cost requirement is severe and severe restrictions are imposed on ROM capacity and the like. For this reason, conventionally, in model-based development for embedded systems, as a first step, a floating-point number model that represents a calculation process using a floating-point number type variable is created, and the calculation result obtained from the model by simulation is obtained. Validated. At the next stage, changing the type of all or some of the variables in the floating-point number model from the floating-point number type to the fixed-point number type reduced the ROM capacity and the like in exchange for a decrease in calculation accuracy. A fixed-point number model was created, and source code was automatically generated from the fixed-point number model.

  Here, in the case of a fixed-point number type variable, the fluctuation range of a numerical value represented by the variable when the value of the variable increases or decreases by 1 (for example, a numerical value represented by 0x000A if the variable is a fixed-point number type variable having a length of 2 bytes). And the absolute value of the difference between the numerical values represented by 0x0009 (hereinafter referred to as weight or 1LSB) is fixed. For this reason, the developer must set the weight when changing the variable type from the floating-point number type to the fixed-point number type. If the weight is not appropriate, the calculation accuracy will exceed the allowable range. There is a risk of getting worse.

  For this reason, it is necessary to verify how much the calculation accuracy is reduced in the fixed-point number model, and it is necessary to perform a simulation of each model and compare the calculation results. And when the deterioration of the calculation accuracy of the fixed-point number model is found, the calculation results obtained in each model are compared one by one at each stage of the calculation process in order to identify the cause of the deterioration. That was a very time-consuming task.

  The present invention has been made in view of the above problems, and design support that can easily find a location that causes a difference in calculation results obtained from two models representing the same calculation processing by different descriptions. An object is to provide a device or the like.

  The design support apparatus according to claim 1, which has been made in view of the above problems, uses a model representing an arithmetic process including a plurality of arithmetic steps as a reference model, and sets a predetermined input value to the reference model. A reference model execution means for executing the calculation processing and obtaining calculation results in each calculation step, and a model representing the same calculation processing as the reference model, wherein at least one calculation step is different from the reference model. The model described is a comparison model, a comparison model execution unit that sets the same input value as the reference model execution unit, executes the calculation process based on the comparison model, and obtains a calculation result in each calculation step; Is provided. Further, for the same calculation step, comparison means for comparing the calculation result obtained by the reference model execution means and the calculation result obtained by the comparison model execution means, and display means for displaying based on the comparison result by the comparison means And comprising.

  Note that the comparison model may be created by changing the type of a variable in the reference model, for example, by adding procedures to all or some of the calculation steps in the reference model, A part of the procedure may be deleted.

  As already mentioned, in model-based development for embedded systems such as in-vehicle devices, a model with high calculation accuracy is created by using a floating-point type variable, and then a floating-point type variable is By changing to the fixed-point number type, there was a case where a model with a reduced ROM capacity or the like was created in exchange for a decrease in calculation accuracy.

  In addition to this, in order to create a program corresponding to a low-spec product, there may be a case in which a model that has been created is intentionally reduced by changing the created model. On the other hand, in order to improve the program, it may be assumed that a model having a high calculation accuracy is newly created by changing a created model.

  In such a case, it is necessary to verify the calculation accuracy of the created model and the newly created model, and to suppress the difference in calculation accuracy between these models within a certain range.

  Furthermore, although details will be described later, when source code is automatically generated from a model, a description such as guard processing when the input / output data value of each calculation step exceeds the limit value may be newly added. In such a case, the calculation result obtained from the model may be different from the calculation result obtained from the source code. Therefore, it is necessary to newly create a model that performs the same processing as the source code, and to verify the calculation accuracy of the newly created model and the model related to the automatic generation of the source code.

  On the other hand, according to the design support apparatus of the first aspect, for each calculation step, it is easy to grasp how far the calculation result obtained by the reference model and the calculation result obtained by the comparison model differ. be able to. For this reason, for example, when the calculation result of the reference model and the calculation result of the comparison model are different from each other beyond an allowable range, the calculation step causing the difference can be easily identified.

  Therefore, in model-based development, when creating a new model that performs the same calculation processing as a model by changing a part of the model already created, the calculation results obtained from these models differ. It is possible to easily find the location that is causing the failure.

  In addition, as described in claim 2, the design support apparatus receives the input value of the same calculation step when the calculation process is executed by the reference model execution unit and when the calculation process is executed by the comparison model execution unit. An adjusting means for adjusting so as to have the same value may be further provided.

  By doing so, the input values of the reference model and the comparison model are the same when each calculation step is executed, so that the difference in the calculation results obtained from these models can be grasped more accurately.

  In addition, as described in claim 3, the calculation processing is executed in a predetermined order, and the calculation result in the calculation step is an input value of the calculation step to be executed later. If the calculation step is described in a manner in which the calculation result is obtained with higher accuracy in the reference model than in the comparison model, the adjustment unit performs the calculation by the comparison model execution unit. You may adjust so that the input value of the calculation step at the time of execution of a process may become the input value of the same calculation step obtained by execution of the calculation process by the reference model execution means.

  By doing so, in the comparative model, errors generated in the previously executed calculation step can be invalidated, and more accurate calculation results can be obtained in the calculation step executed later.

In addition, there may be a case where a calculation step causing a difference in calculation accuracy can be predicted between the reference model and the comparison model.
Therefore, the design support apparatus according to claim 4 uses a model representing a calculation process including a plurality of calculation steps as a reference model, sets a predetermined input value, and performs any calculation based on the reference model. A reference model execution means for executing steps and obtaining calculation results, and a model representing calculation processing similar to the reference model, wherein at least one calculation step is described in a manner different from the reference model And a comparison model execution unit that executes the calculation step based on the comparison model using the same input value as that of the reference model execution unit and obtains a calculation result with respect to the calculation step executed by the reference model execution unit. Further, for the same calculation step, comparison means for comparing the calculation result obtained by the reference model execution means and the calculation result obtained by the comparison model execution means, and display means for displaying based on the comparison result by the comparison means And comprising.

  By doing this, it is possible to compare the calculation results by focusing on the calculation step that is predicted to be the cause of the difference in calculation accuracy, and the cause of the difference in the calculation results of each model in a shorter period of time. It is possible to find out where it is.

Further, the design support apparatus may display the comparison result for the calculation step in the following manner.
That is, as described in claim 5, the display means may display the reference model or the comparison model in a state in which the location corresponding to the calculation step is associated with the comparison result for the calculation step. .

  By doing so, it becomes easier to grasp the comparison result for the calculation step, and it is possible to more easily find a location that causes a difference between the calculation results obtained from the reference model and the comparison model.

  Further, as described in claim 6, the comparison means calculates a difference between the calculation result obtained by the reference model execution means and the calculation result obtained by the comparison model execution means for the calculation step, Information regarding the difference may be used as a comparison result.

By doing so, it is possible to accurately grasp how much the calculation result is different in each calculation step between the reference model and the comparison model.
Further, as described in claim 7, in the reference model, the calculation step is described in a mode in which the calculation result can be obtained with higher accuracy than in the comparison model, and the comparison means includes a difference between the calculation steps. May be determined within a predetermined allowable range, and the determination result may be used as a comparison result.

  By doing so, it is possible to easily identify the calculation step whose calculation result is different from the allowable range, and more easily, the calculation result obtained from the reference model and the comparison model is the cause of the difference. It becomes possible to find the location.

Note that the allowable range of the difference between the calculation results in the calculation step may be set as follows.
That is, as described in claim 8, the allowable range used for the determination of the difference by the comparison means is set based on the maximum error that is assumed to be included in the input value of the calculation step related to the difference. May be.

  By doing this, the allowable range of the difference between the calculation results in the calculation step can be set appropriately, and the location causing the difference between the calculation results obtained from the reference model and the comparison model can be found easily and reliably. It becomes possible.

  In claim 9, a design support program for operating a computer as the design support apparatus described in claim 1 is described. By using such a design support program, in model-based development, when a new model that performs the same arithmetic processing as the model is created by changing a part of the created model, these It is possible to easily find a location that causes a difference in calculation results obtained from the model.

  Further, claim 10 describes a design support program for operating a computer as the design support apparatus according to claim 4. By using such a design support program, it is possible to compare the calculation results with the calculation steps predicted to be the cause of the difference in calculation accuracy, and the calculation results of each model can be compared in a shorter period of time. It becomes possible to find the location causing the difference.

It is a block diagram which shows the structure of PC which operate | moves as a model verification apparatus. It is explanatory drawing which shows an example of a model. It is a work flow figure showing a flow of model base development. It is a flowchart which shows a verification work flow. It is a flowchart about a comparison process. It is explanatory drawing about the display screen which shows the result of having compared the calculation result obtained by each block.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. The embodiment of the present invention is not limited to the following embodiment, and can take various forms as long as they belong to the technical scope of the present invention.

[Description of configuration]
(1) Configuration of Model Verification Device FIG. 1 is a block diagram showing the configuration of a known personal computer (PC) 1 that operates as a model verification device in the present embodiment. As shown in FIG. 1, the PC 1 includes a display 10, an HDD (hard disk drive) 20, a CPU (arithmetic unit) 30, a ROM 40, a RAM 50, an input device 60, and the like.

The display 10 displays the video signal received from the CPU 30 as a video to the user.
The input device 60 includes a keyboard, a mouse, and the like, and outputs a signal corresponding to the operation to the CPU 30 when operated by the user.

  The RAM 50 is a readable / writable volatile memory, the ROM 40 is a read-only nonvolatile memory, and the HDD 20 is a readable / writable nonvolatile memory. In the ROM 40 and the HDD 20, programs that are read and executed by the CPU 30 are stored in advance. The HDD 20 stores a model to be described later.

  When the CPU 30 executes a program stored in the ROM 40 or the HDD 20, the RAM 50 is used as a storage area for temporarily storing the program and a storage area for temporarily storing work data. .

  When the CPU 30 is activated by turning on the power of the PC 1, the CPU 30 reads and executes a predetermined boot program from the ROM 40, and reads and executes an operating system (OS) and other programs defined in the boot program from the HDD 20. The startup process is performed. After the startup process, until the power is turned off, the CPU 30 executes various programs recorded in the HDD 20 as processes on the OS based on a signal from the input device 60, a schedule predetermined by the OS, and the like. . In the startup process and process, the CPU 30 receives a signal input from the input device 60 as necessary, outputs a video signal to the display 10, and reads / writes data to / from the RAM 50 and the HDD 20. Control.

(2) Model Verification Tool Next, an outline of the model verification tool of the present embodiment installed in the PC 1 will be described. The model verification tool is one of applications stored in the HDD 20 and executed as a process on the OS, and the PC 1 operates as a model verification apparatus for performing model-based development by operating the model verification tool. . The model verification tool is activated by the user's operation of the input device 60 according to the specifications of the OS, and then performs model simulation and the like based on the operation of the user's input device 60.

  The model verification device can be used for developing programs for embedded systems such as in-vehicle devices, but is not limited to this, and can be used for developing programs for all purposes.

[About model-based development]
Next, model-based development will be described.
A model is an expression form of a program that is created based on a model language specification that is defined for the purpose of making description easier than source code and improving human readability. As a model, for example, there is a Simulink model created by Simulink (registered trademark). When a Simulink (registered trademark) model is input to a Real Time Workshop (registered trademark: hereinafter referred to as RTW) operating on Matlab (registered trademark), the RTW is a source corresponding to the Simulink (registered trademark) model. Generate code.

  The Simulink (registered trademark) model is created as an aggregate composed of combinations of functional units called blocks. A combination refers to an input / output connection between functional units. As a specific example of the block, for example, there is an addition block for adding two input numerical data and outputting the result. In the model, the type of a variable in which input / output data for the block is stored, the data length, the weight of a fixed-point number type variable (details will be described later), and the like are defined.

  An aggregate of a plurality of blocks can also be defined as a kind of block called a subsystem. This subsystem is equivalent to a function in programming, and it is allowed to create a block having a nested structure in which the block is included in the block. In the source code generated based on the model having the subsystem, Processing in the system is described as a subroutine.

  Each block in the Simulink (registered trademark) model can have attribute information. The attribute information includes, for example, the name of the block, the range definition of input / output data, and the like.

  Note that FIG. 2 shows, as an example, a model 100 including A calculation blocks 101 to C calculation blocks 103 and the like (it goes without saying that these blocks may be configured as subsystems). I). In the model 100, the calculation result (output data) obtained in each block is input data of other blocks. That is, the connection connected to each block indicates the input / output of numerical data, the connection with the arrow pointing to the block is the input data to the block, and the arrow starting from the block is attached The connection is the output data from the block. Each connection is defined with a variable type or the like in which numerical data corresponding to the connection is stored (not shown).

  Needless to say, in the model 100, the block to which the input data to the model is input is executed first, and thereafter, the input data is executed in order from the block from which all the input data is obtained (in other words, the connection The blocks are executed in the order from the block located on the upstream side based on the direction of the arrow to the block on the downstream side. To be specific, the blocks are executed in the order of the A calculation block 101 to the C calculation block 103).

  In addition, as shown in FIG. 2, in Simulink (registered trademark), the calculation procedure of the block in the model can be described by the source code 110, thereby performing the same calculation processing as the source code. (Registered trademark) model can be created.

  The model in the present embodiment is also composed of blocks as in the Simulink (registered trademark) model, and description of blocks as a subsystem and block nesting structure are allowed, each block has attribute information, and calculation by source code The procedure can be described. In addition, it is possible to arbitrarily define the type of variable in which input / output data for the block or subsystem is stored.

  Next, the flow of model-based development in this embodiment will be described using the work flow 200 shown in FIG. In this embodiment, model creation and source code automatic generation are performed through the following steps using Simulink (registered trademark).

  In other words, in programming, as variable types that can represent numbers with decimal numbers, a fixed-point number type that represents a certain range of numbers with preset precision and a wider range of numbers than fixed-point number types can be represented The floating-point number type is known. In general, it is possible to improve the accuracy of calculation by using a floating-point number type variable rather than a fixed-point number type variable. However, on the other hand, the ROM and RAM capacities required for operating the program, There is a problem that the processing load of the CPU increases.

  Therefore, in the present embodiment, as a first stage, a floating-point number model 220 representing an arithmetic process based on the control specification 210 is created using a floating-point number type variable. Then, at the next stage, by changing the type of all or some of the variables in the floating-point model 220 from the floating-point number type to the fixed-point number type, the ROM capacity and the like can be reduced in exchange for a decrease in calculation accuracy. A fixed-point number model 230 is created.

  In the case of a fixed-point number type variable, the fluctuation range of the numerical value represented by the variable when the value of the variable is increased or decreased by 1 (for example, in the case of a fixed-point number type variable of 2 bytes length, the numerical value represented by 0x000A) The absolute value of the difference between the numerical values represented by 0x0009 (hereinafter referred to as weight or 1LSB) is fixed. For this reason, when creating the fixed-point number model 230, the weight of each variable of the fixed-point number type is set in consideration of the required calculation accuracy.

  In the next stage, a model verification operation 240 (details will be described later) for comparing the calculation accuracy of the floating point number model 220 and the fixed point number model 230 is performed. If it is determined that the reduction in the calculation accuracy is within the allowable range, the fixed-point number code 250 is automatically generated based on the fixed-point number model 230 using the RTW.

  Here, when automatically generating source code from a model, RTW performs a conversion that unifies the types and weights of these variables in calculation procedures using different types of variables and fixed-point number variables with different weights. Add a description of the process. In addition to this, for example, a description such as guard processing when input / output data to each block exceeds a value range defined by the attribute information is added. Therefore, different calculation results may be obtained from the fixed-point number model 230 and the fixed-point number code 250.

  Therefore, in the next stage, the calculation procedure corresponding to each block in the fixed-point number code 250 is described in the corresponding block in the fixed-point number model 230, thereby performing the same calculation processing as the fixed-point number code 250. A model 260 is created. Then, a code verification operation 270 (details will be described later) for comparing the calculation accuracy of the fixed-point number model 230 and the source code model 260 is performed, and the calculation processing in the source code model 260 is verified.

[About verification work]
Next, a model verification operation 240 for comparing the calculation accuracy of the floating-point number model 220 and the fixed-point number model 230 and a code verification operation 270 for comparing the calculation accuracy of the fixed-point number model 230 and the source code model 260 will be described. These operations are performed according to the verification work flow shown in FIG. 4 using the model verification apparatus.

  In the following description, in the model verification operation 240, the floating-point number model 220 is a reference model and the fixed-point number model 230 is a comparison model, and in the code verification operation 270, the fixed-point number model 230 is a reference model. The source code model 260 is used as a comparison model.

In S305, simulation of the comparison model is executed using the model verification apparatus.
In subsequent S310, the calculation result obtained from the comparison model is compared with the theoretical value, and it is determined whether or not the error included in the calculation result is within an allowable range. This allowable range is determined based on the control specification 210 and the like. If the determination is affirmative (S310: Yes), the process proceeds to S315. If the determination is negative (S310: No), the verification operation ends.

  In S315, using the model verification apparatus, a comparison process is performed in which the calculation results obtained in the respective blocks of the reference model and the comparison model are compared, and a block in which an error exceeding the allowable range is generated is identified.

  In subsequent S320, it is determined whether or not a block in which an error exceeding the allowable range is generated is determined in S315. When an affirmative determination is obtained (S320: Yes), the process proceeds to S325, and when a negative determination is obtained (S320: No), the process proceeds to S330.

In S325, the description of the block in which the error exceeding the allowable range is generated is reviewed, and the work of improving the calculation accuracy is performed.
Specifically, if the model verification work 240 is performed according to the verification work flow, for example, a mistake in setting the weight when changing the variable type to the fixed-point number type is verified. In such a case, it is conceivable to take measures such as reducing the weight of the fixed-point number type variable. Further, for example, it is conceivable to return a fixed-point number type variable to a floating-point number type.

  In addition, if the code verification operation 270 is performed according to the verification work flow, for example, the validity of the conversion processing and guard processing added at the time of automatic source code generation is examined, and if necessary It may be possible to change attribute information, variable type, or the like.

  On the other hand, in S330, it is determined that the calculation results of the comparison model are accumulated errors within the allowable range generated in each block, and the design of the entire calculation process is reviewed again.

Then, the verification work flow is repeatedly performed until the error included in the calculation result obtained from the comparison model is within the allowable range.
Next, the comparison process performed by the model verification apparatus will be described with reference to the flowchart shown in FIG.

  In S405, the model verification apparatus reads the reference model stored in the HDD 20 or the RAM 50, and in subsequent S410, reads the comparison model stored in the HDD 20 or the like to obtain the reference model (or comparison model). Is displayed on the display 10. Then, the same numerical value is set as input data for each model, and one block to be executed first (any block to which the input data is input) is selected from the blocks represented by each model. To do.

  In subsequent S415, the model verification apparatus executes a calculation procedure corresponding to the selected block based on the reference model, and executes a calculation procedure corresponding to the block based on the comparison model, and the process proceeds to S420. .

  In S420, the model verification apparatus calculates a difference between the calculation result calculated based on the reference model and the calculation result calculated based on the comparison model as an error generated in the selected block in the comparison model, and the error is calculated. Determine whether it is within the allowable range.

Here, it is conceivable to determine whether or not the error is within the allowable range as follows.
That is, for example, in the comparative model, if a fixed-point number type variable is used in the selected block, the largest weight among the weights of these fixed-point number type variables is the absolute value of the error. It may be an allowable range. If the output data is stored in a fixed-point number type variable, the weight of the fixed-point number type variable may be set as the allowable range of the absolute value of the error.

  Also, for example, in the comparison model, if the input data to the selected block is stored in a fixed-point number type variable, from the lower limit value obtained by subtracting the weight from the input value indicated by the input data, A range up to an upper limit value obtained by adding the weight to the input value may be set as an assumed fluctuation range of the input value. The fluctuation range in which the calculation result of the block fluctuates when each input value fluctuates within the assumed fluctuation range is set as the allowable range of the calculation result including the error obtained in the block. A range may be specified. The weight of the variable storing the input data corresponds to the maximum error in the claims.

  As a specific example, the input data of the selected block is x, y, the calculation procedure of the block is x + y = z, the weight for x is 0.5, the weight for y is 0.1, and no error is included. Assume that the values of x and y are X and Y. At this time, x and y vary in the range of X−0.5 <x <X + 0.5 and Y−0.1 <y <Y + 0.1, and the calculation result including an error obtained in the block is X + Y−0.6 <z <X + Y + 0.6. For this reason, the allowable range of the error ΔZ of the block is −0.6 <ΔZ <0.6.

  Further, not limited to the above-described method, based on the control specification 210, an estimated range of errors included in input data in the selected block (for example, it is assumed that a maximum of ± 3% error is included). It may be set in advance and the assumed range may be an assumed fluctuation range of the input value indicated by the input data. The fluctuation range in which the calculation result of the block fluctuates when each input value fluctuates within the assumed fluctuation range is set as the allowable range of the calculation result including the error obtained in the block. A range may be specified. The maximum value and the minimum value in the assumed range of the error correspond to the maximum error in the claims.

  If the error is within the allowable range (S420: Yes), the model verification apparatus shifts the process to S425. If the error is not within the allowable range (S420: No), the model verification apparatus shifts the process to S430.

  In S425, the model verification apparatus performs highlighted display (warning display) of the selected block in the model currently displayed on the display 10, and shifts the processing to S430. In addition, not only when the error in the block exceeds the allowable range, a display indicating that an error has occurred may be performed.

  In S430, the model verification apparatus determines whether or not the selected block is the block to be executed last. If an affirmative determination is obtained (S430: Yes), the process ends. If a negative determination is obtained (S430: No), the process proceeds to S435.

  In S435, the model verification apparatus replaces the calculation result (output data) of the block being selected in the comparison model with the output data of the block being selected in the reference model, so that the input data to the block to be executed next and thereafter (In other words, the error generated in the selected block in the comparison model is canceled). Then, the process proceeds to S440.

In S440, the model verification apparatus selects a block to be executed next, and proceeds to S415.
Next, a display screen 500 showing a comparison result of each block in the comparison process displayed on the display 10 of the model verification apparatus will be described with reference to FIG.

  The display screen 500 is provided with a model display area 510 for displaying a reference model (or a comparison model). Each block (A calculation block 511 to C calculation block 513) constituting the reference model is provided in the model display area 510. Etc.) is displayed. These blocks are displayed in different modes depending on the error determination result in the comparison process.

  Note that the display screen 500 of FIG. 6 shows, as an example, a fixed-point number type variable used in a block (or a fixed-point number type in which output data from the block is stored) as an allowable range of absolute values of errors. The display mode of each block when the weight (1LSB) of (variable) is set is shown.

  In the display screen 500, the C calculation block 513 in which the absolute value of the error of the output data exceeds the weight is highlighted in the most conspicuous manner. In the comparison processing of the present embodiment, highlighting is performed only for blocks that exceed the allowable range. However, on the display screen 500, highlighting is also performed for the B operation block 512 in which an error within the allowable range occurs. Has been made.

  An error display area 520 for displaying the error of each block is provided below the model display area 510. In this area, an error in each block and whether or not the error is within an allowable range are displayed. ing.

[effect]
In the verification work flow of the present embodiment, if it is found by comparison between the calculation result obtained from the comparison model and the theoretical value that the calculation result includes an error exceeding the allowable range, the model verification apparatus Comparison processing is performed. In the comparison process, the result of comparing the calculation results obtained from each model is displayed for each block in the reference model and the comparison model.

  For this reason, for each block, it can be easily grasped how much the calculation result obtained from the reference model and the calculation result obtained from the comparison model are different, which causes a calculation error in the comparison model. Can be easily identified.

[Other Embodiments]
(1) In S415 of the comparison process of the present embodiment, all the blocks in the reference model and the comparison model are executed in a predetermined order, and in S420, the operation results are compared for all the blocks. However, the design support apparatus may select a specific block based on an instruction from the developer and set the same value for each model as an input value of the selected block. Then, only the block selected using the set input value may be executed in S415 of the comparison process, and only the comparison result for the block may be displayed in subsequent S420 and S425. Even in such a case, it is possible to easily find a location that causes a difference in the calculation result of each model.

  (2) Further, in S435 of the comparison process of the present embodiment, a process for canceling the error is performed, but the process may not be performed. By doing so, it is possible to find a location that causes a difference in the calculation result of each model based on the error generated cumulatively in each block.

  (3) In this embodiment, a model with high calculation accuracy is used as a reference model, and a model with low calculation accuracy is used as a comparison model. However, the present invention is not limited to this. A model with high accuracy may be used as a comparative model. Even in such a case, the same effect can be obtained.

  (4) In the comparison processing of this embodiment, the calculation results are compared in units of one block, but the calculation results may be compared in units of a plurality of blocks. Even in such a case, it is possible to easily identify the location that causes the calculation error that occurs in the comparison model.

  (5) In the present embodiment, the case where model-based development is performed according to the work flow 200 has been described as an example, but the comparison process performed by the model verification apparatus has been described. Comparison processing can be performed using a model verification apparatus. That is, the same arithmetic processing is represented as in the floating-point model 220 and the fixed-point model 230, and the fixed-point model 230 and the source code model 260, but in the type of variable used and the corresponding block. If there are two models that differ in part of the procedure, comparison processing can be performed on these models. By doing so, it is possible to easily find a location that causes a difference in calculation accuracy in these models.

[Correspondence with Claims]
The correspondence between the terms used in the description of the above embodiment and the terms used in the description of the claims is shown.

The model verification device corresponds to a design support device, the model verification tool corresponds to a design support program, and the blocks correspond to calculation steps.
Further, S415 of the comparison process is a reference model execution means, a comparison model execution means, a reference model execution procedure, a comparison model execution procedure, S420 is a comparison means, a comparison procedure, S425 is a display means, a display procedure, and S435 is an adjustment means. It corresponds to.

  DESCRIPTION OF SYMBOLS 1 ... PC, 10 ... Display, 20 ... HDD, 30 ... CPU, 40 ... ROM, 50 ... RAM, 60 ... Input device, 100 ... Model, 110 ... Source code, 200 ... Work flow, 210 ... Control specification, 220 ... Floating point number model, 230 ... Fixed point number model, 240 ... Model verification work, 250 ... Fixed point number code, 260 ... Source code model, 270 ... Code verification work, 500 ... Display screen, 510 ... Model display area, 520 ... Error display area.

Claims (10)

A model representing a calculation process composed of a plurality of calculation steps is set as a reference model, a predetermined input value is set, the calculation process is executed based on the reference model, and calculation results in the calculation steps are obtained. A reference model execution means;
A model representing the arithmetic processing similar to the reference model, in which at least one of the calculation steps is described in a manner different from the reference model, and the same input as the reference model execution means Comparison model execution means for setting the value and executing the calculation processing based on the comparison model to obtain the calculation result in each of the calculation steps;
Comparing means for comparing the calculation result obtained by the reference model execution means with the calculation result obtained by the comparison model execution means for the same calculation step;
Display means for displaying based on a comparison result by the comparison means;
A design support apparatus comprising: The design support apparatus according to claim 1,
An adjusting unit configured to adjust an input value of the same calculation step to be the same value when the calculation process is performed by the reference model execution unit and when the calculation process is performed by the comparison model execution unit; Further preparation,
Design support device characterized by The design support apparatus according to claim 2,
The calculation process is configured such that each calculation step is executed in a predetermined order, and the calculation result in the calculation step is an input value of the calculation step to be executed later,
In the reference model, the calculation step is described in a manner in which the calculation result is obtained with higher accuracy than the comparison model,
The adjustment means is configured such that the input value of the calculation step at the time of execution of the calculation process by the comparison model execution means is the same as the input value of the calculation step obtained by the execution of the calculation process by the reference model execution means. Adjusting to be
Design support device characterized by A model representing a calculation process composed of a plurality of calculation steps is used as a reference model, a predetermined input value is set, one of the calculation steps is executed based on the reference model, and a reference model is executed to obtain a calculation result. Means,
A model representing the same arithmetic processing as the reference model, wherein at least one of the calculation steps is a model described in a different form from the reference model, and is executed by the reference model execution means With respect to the calculation step, a comparison model execution unit that executes the calculation step based on the comparison model using the same input value as the reference model execution unit, and obtains a calculation result;
Comparing means for comparing the calculation result obtained by the reference model execution means with the calculation result obtained by the comparison model execution means for the same calculation step;
Display means for displaying based on a comparison result by the comparison means;
A design support apparatus comprising: The design support apparatus according to any one of claims 1 to 4,
The display means displays the reference model or the comparison model in a state in which the location corresponding to the calculation step is associated with the comparison result for the calculation step;
Design support device characterized by The design support apparatus according to any one of claims 1 to 5,
The comparison means calculates a difference between the calculation result obtained by the reference model execution means and the calculation result obtained by the comparison model execution means for the calculation step, and information about the difference is calculated as the comparison result. To do,
Design support device characterized by The design support apparatus according to claim 6,
In the reference model, the calculation step is described in a manner in which the calculation result is obtained with higher accuracy than the comparison model,
The comparing means determines whether the difference with respect to the calculation step is within a predetermined allowable range, and sets the determination result as the comparison result;
Design support device characterized by The design support apparatus according to claim 7,
The allowable range used for the determination of the difference by the comparison means is set based on a maximum error assumed to be included in the input value of the calculation step related to the difference;
Design support device characterized by A model representing a calculation process composed of a plurality of calculation steps is set as a reference model, a predetermined input value is set, the calculation process is executed based on the reference model, and calculation results in the calculation steps are obtained. Reference model execution procedure,
A model representing the same arithmetic processing as the reference model, in which at least one of the calculation steps is described in a mode different from the reference model, and the same input as the reference model execution procedure A comparison model execution procedure for setting the value and executing the calculation process based on the comparison model to obtain the calculation result in each of the calculation steps;
For the same calculation step, a comparison procedure for comparing the calculation result obtained in the reference model execution procedure with the calculation result obtained in the comparison model execution procedure;
A display procedure for displaying based on the comparison result obtained in the comparison procedure;
A design support program characterized by causing a computer to execute. A model representing a calculation process composed of a plurality of calculation steps is used as a reference model, a predetermined input value is set, one of the calculation steps is executed based on the reference model, and a reference model is executed to obtain a calculation result. Procedure and
A model representing the calculation process similar to the reference model, wherein at least one of the calculation steps is a model described in a different form from the reference model, and is executed in the reference model execution procedure. With respect to the calculation step, a comparison model execution procedure for executing the calculation step based on the comparison model using the same input values as the reference model execution procedure and obtaining a calculation result;
For the same calculation step, a comparison procedure for comparing the calculation result obtained in the reference model execution procedure with the calculation result obtained in the comparison model execution procedure ;
A display procedure for displaying based on the comparison result obtained in the comparison procedure;
A design support program characterized by causing a computer to execute.

Images