JP2010097328A - Loop optimization system, loop optimization method, and program for loop optimization - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a loop optimization method in model inspection in a loop. <P>SOLUTION: This loop optimization system for performing processing to verify a source code by using a model inspection technology is provided with: a model inspection control part 10 for automatically generating a model from a source code, and for verifying the source code based on model inspection; and an arbitrary value loop control part 30 for moving the substitution formula of variables including an arbitrary value existing in the loop in the model to the outside of the loop, and for shortening the verification time, wherein a verification time for inspection duplicated due to repetition among model inspection in the loop is shortened. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a loop optimization system, and more particularly to a loop optimization system in model checking in a loop.

  Conventionally, several methods are known as methods for verifying a source code described in a programming language such as C language by a model checking method which is one method of formal verification.

  In general, when a model is automatically built from source code and verified by model checking, the number of states of the generated model becomes enormous and the amount of calculation becomes very large, so various optimizations are made during the modeling process. Need to do. In the process of generating an intermediate model called a control flow graph (CFG) that represents the control structure and the flow of processing, multiple optimizations are performed, but in the analysis of the loop control structure, "loop invariant calculation removal" etc. Only the optimization processing by the compiler optimization technique was performed.

  Here, “loop invariant calculation elimination” means that the loop invariant calculation that is not affected by the loop control variable is detected and the code movement process moves the calculation process of the variable in the loop to the outside of the loop. This reduces the amount of calculation processing and reduces the verification time.

  In model checking, if there is a calculation process that includes an arbitrary value in a loop and that calculation process is not subject to compiler optimization, the model check will take an arbitrary value for the number of iterations in the loop. Since exhaustive verification is performed for all of the obtained values, the amount of calculation becomes very large, and as a result, there is a problem that the verification time becomes long.

  Here, “arbitrary value” refers to a variable that cannot be calculated from a given condition in a model checking process, or a function that does not define the entity of the process. Means an element that expresses it specially.

  In model checking, if an arbitrary value is referenced in an expression, all states of possible values may exist, and all states are checked in a collapsed manner (exhaustively). For example, if the arbitrary value is an 8-bit signed integer type, any value of “−128, −127,..., 0,... 127” (−128 to 127) may be taken. In all cases, inspection is performed.

As a related technique, Japanese Patent Publication No. 2007-528059 (Patent Document 1) describes a system and method for software modeling, abstraction, and analysis.
This related technology is a technology for performing verification by converting source code into a model that can be model-checked in source code verification. However, in this related technique, even if there is a calculation process of a variable including an arbitrary value in the loop, no special process is performed and the model check is performed as it is. For this reason, verification related to an arbitrary value is repeatedly performed in the loop, and verification time for inspection of overlapping portions is required.

Special table 2007-528059 gazette

  In the process of verifying the source code using the model checking technique, the calculation process related to the arbitrary value is deleted before the model checking is performed, and the process of substituting the arbitrary value outside the loop is added.

  The loop optimization system according to the present invention automatically generates a model from source code, performs model check control on the source code by model checking, and a variable including an arbitrary value existing in a loop in the model. And an arbitrary value loop control unit for reducing the verification time.

  The loop optimization method of the present invention automatically generates a model from source code, performs verification by model checking on the source code, and any model existing in the loop in the model during the process of generating the model. Moving the substitution expression of the variable including the value out of the loop to shorten the verification time.

  The program of the present invention includes a step of automatically generating a model from source code, performing verification by model checking on the source code, and an arbitrary value existing in a loop in the model in the process of generating the model. This is a program for causing a computer to execute a step of reducing a verification time by moving a variable substitution expression out of a loop.

  Shorten the verification time for repeated inspections of model checks in the loop.

  Embodiments of the present invention will be described below with reference to the accompanying drawings.

  Referring to FIG. 1, the loop optimization system of the present invention includes a model checking control unit 10, a storage unit 20, and an arbitrary value loop control unit 30.

  Examples of the loop optimization system of the present invention include PCs (personal computers), thin client terminals / servers, workstations, mainframes, supercomputers and other computers, mobile phones, PDAs (Personal Digital Assistants), car navigation systems (car navigation systems) ), Portable terminals such as portable game machines or electronic devices having a similar wireless communication function. However, actually, it is not limited to these examples.

  Examples of the model checking control unit 10 and the arbitrary value loop control unit 30 include a processing device such as a CPU (Central Processing Unit) and a microprocessor, or a semiconductor integrated circuit (Integrated Circuit: IC) having a similar function. Can be considered. The model checking control unit 10 and the arbitrary value loop control unit 30 may be programs for causing a computer to execute each function. However, actually, it is not limited to these examples.

  Examples of the storage unit 20 include a semiconductor storage device such as a memory, an external storage device (storage) such as a hard disk, or a storage medium (media). The storage unit 20 may be a relay device or a peripheral device. Further, the storage unit 20 may be shared by a plurality of model checking control units 10 and arbitrary value loop control units 30. However, actually, it is not limited to these examples.

  Each of the model checking control unit 10, the storage unit 20, and the arbitrary value loop control unit 30 may be a physically independent computer. For example, a network system in which the model checking control unit 10 is a terminal and the storage unit 20 or the arbitrary value loop control unit 30 is a server can be considered. However, actually, it is not limited to these examples.

  The model check control unit 10 includes a CFG model creation unit 11, a check model creation unit 12, a model check execution unit 13, and a check result output unit 14.

  The storage unit 20 includes a pre-conversion CFG model storage unit 21, a post-conversion CFG model storage unit 22, and an inspection model storage unit 23.

  The arbitrary value loop control unit 30 includes a loop detection unit 31, a data dependence graph creation unit 32, an out-loop use variable mark unit 33, a subgraph detection unit 34, and a CFG model conversion unit 35.

  The CFG model creation unit 11 reads the inspection specification 1 and the inspection target code 2, performs compiler optimization processing, creates a CFG model, and outputs the CFG model to the pre-conversion CFG model storage unit 21 of the storage unit 20. The inspection specification 1 is information defining a model inspection method. The inspection target code 2 is a source code to be inspected.

  The inspection model creation unit 12 reads a CFG model from the converted CFG model storage unit 21 of the storage unit 20, creates an inspection model based on a symbolic model inspection method using computation tree logic (CTL), The data is output to the inspection model storage unit 23 of the storage unit 20.

  The model checking execution unit 13 reads the checking model from the checking model storage unit 23 of the storage unit 20 and executes checking based on the symbol model checking method using the above-described calculation tree logic (CTL).

  The inspection result output unit 14 outputs the result of the inspection performed by the symbol model inspection execution unit 12 to the inspection result 3.

  The pre-conversion CFG model storage unit 21 acquires and holds the pre-conversion CFG model from the CFG model creation unit 11, and provides the pre-conversion CFG model to the loop detection unit 31.

  The post-conversion CFG model storage unit 22 acquires and stores the post-conversion CFG model from the CFG model conversion unit 35, and provides the post-conversion CFG model to the inspection model creation unit 12.

  The inspection model storage unit 23 acquires and holds an inspection model from the inspection model creation unit 12 and provides the inspection model to the model inspection execution unit 13.

  The loop detection unit 31 reads the CFG model held in the pre-conversion CFG model storage unit 21 of the storage unit 20 and detects loop processing in the CFG model.

  The data dependence graph creation unit 32 creates a data dependence graph of values appearing in the loop detected by the loop detection unit 31 based on the data dependence graph creation method.

  The out-of-loop use variable mark unit 33 marks each variable in the data dependence graph created by the data dependence graph creation unit 32 if the variable (each variable in the data dependence graph) is used after the loop. wear. That is, the out-of-loop use variable mark unit 33 marks variables used after the loop among the variables in the data dependence graph.

  The subgraph detection unit 34 detects a subgraph that includes an arbitrary value as an element and includes only one variable marked by the out-of-loop use variable mark unit 33 from the data dependency graph generated by the data dependency graph generation unit 32. To do.

  The CFG model conversion unit 35 deletes the CFG model element corresponding to the calculation processing of the variable included in the subgraph detected by the subgraph detection unit 34 from the CFG model, and the subgraph detection unit 34 detects immediately after the loop. A CFG model element corresponding to a calculation process for substituting an arbitrary value is added to the CFG model for the variable marked by the out-of-loop use variable mark unit 33 that is included in only one subgraph.

  Next, the operation of the loop optimization system of the present invention will be described in detail with reference to the flowchart of FIG.

(1) Step S101
First, the CFG model creation unit 11 performs compiler optimization processing on the inspection target code 2, creates a CFG model based on the inspection specification 1, and stores it in the pre-conversion CFG model storage unit 21.

(2) Step S102
Next, the loop detection unit 31 confirms whether there is a loop to be processed in the CFG model stored in the pre-conversion CFG model storage unit 21. Here, the loop detection unit 31 detects a loop to be processed in the CFG model stored in the pre-conversion CFG model storage unit 21. The loop to be processed is a loop in which the subsequent processing from step S103 to step S107 is not performed (not performed).

(3) Step S103
When the loop detection unit 31 detects a loop to be processed, the data dependency graph creation unit 32 creates a data dependency graph of values that appear in the loop detected by the loop detection unit 31.

(4) Step S104
Next, the out-of-loop use variable mark unit 33 marks each variable in the data dependence graph created by the data dependence graph creation unit 32 if the variable is used after the loop.

(5) Step S105
Next, the partial graph detection unit 34 includes an arbitrary value as an element from the data dependency graph generated by the data dependency graph generation unit 32 and includes only one variable marked by the out-of-loop use variable mark unit 33. Check if the graph exists. Here, the partial graph detection unit 34 includes an arbitrary value as an element from the data dependency graph generated by the data dependency graph generation unit 32 and includes only one variable marked by the out-of-loop use variable mark unit 33. Detect the graph.

(6) Step S106
The CFG model conversion unit 35 includes an arbitrary value as an element, and when there is a subgraph including only one variable marked by the out-of-loop use variable mark unit 33, that is, the subgraph detection unit 34 determines that the corresponding subgraph is If detected, the CFG model element corresponding to the calculation processing of the variable included in the subgraph detected by the subgraph detection unit 34 is deleted from the CFG model.

(7) Step S107
Furthermore, immediately after the loop, the CFG model conversion unit 35 assigns an arbitrary value to the variable marked by the out-of-loop use variable mark unit 33 that is included in only one subgraph detected by the subgraph detection unit 34. A CFG model element corresponding to the calculation process to be substituted is added to the CFG model.

(8) Step S108
After performing the above processing (from step S102 to step S107) until there is no loop to be processed, the inspection model creation unit 12 stores the CFG model converted by the series of processing, and stores the converted CFG model storage unit 22. An inspection model is created from the CFG model and stored in the inspection model storage unit 23. The inspection model creation process may be performed by a conventional technique.

(9) Step S109
The model checking execution unit 13 reads the checking model stored in the checking model storage unit 23 and performs model checking.

(10) Step S110
Finally, the inspection result output unit 14 outputs the result to the inspection result 3.

  The operation of the arbitrary value loop control unit 30 will be specifically described below.

  First, an example in which a CFG model is converted will be described with reference to FIGS. 3A and 3B. Here, FIG. 3A shows a procedure before conversion of the CFG model. FIG. 3B shows the procedure after conversion of the CFG model.

  The loop detection unit 31 refers to the CFG model before conversion, and detects a loop from “if (i <10)” to “t = f ()” as shown in FIG. 3A.

  The data dependence graph creation unit 32 creates a data dependence graph of values appearing in the loop.

  The out-of-loop use variable mark unit 33 marks each variable in the data dependence graph created by the data dependence graph creation unit 32 if the variable is used after the loop. In this example, as shown in FIG. 4, it is assumed that “x” and “y” are used after the loop, and the “*” mark is given.

  The subgraph detection unit 34 includes “f ()” having an arbitrary value in the data dependence graph as an element, and is used outside the loop as shown in the left side (“x” side) of FIG. 4. Find the subgraph containing only one variable “x” marked by.

  As shown in FIG. 3B, the CFG model conversion unit 35 deletes the CFG model element corresponding to the calculation processing of the variable included in the partial graph detected by the partial graph creation unit 34 from the CFG, and immediately after the loop, A CFG model element corresponding to a calculation process for substituting an arbitrary value for the variable “x” marked by the out-of-loop use variable mark unit 33 that is included in only one subgraph detected by the graph detection unit 34. Add to. The CFG model to which the CFG model element is added is the converted CFG model.

  Next, an example in which the CFG model is not converted will be described with reference to FIG.

  The loop detection unit 31 detects a loop from “if (i <10)” to “t = f ()”. At this time, there is a calculation process of “x = t + i” and “y = t”, and “t” has a dependency relationship with both “x” and “y”.

  The data dependence graph creation unit 32 creates a data dependence graph of values appearing in the loop. The out-of-loop use variable mark unit 33 marks each variable in the data dependence graph created by the data dependence graph creation unit 32 if the variable is used after the loop. In this example, as shown in FIG. 6, it is assumed that “x” and “y” are used after the loop, and the “*” mark is given.

  The subgraph detection unit 34 attempts to detect a subgraph that includes “f ()” having an arbitrary value as an element in the data dependence graph and includes only one variable marked by the outside loop use variable mark unit 33. However, since this subgraph includes two of “x” and “y”, the condition does not match. Therefore, the CFG model is not converted for this subgraph. That is, the subgraph detection unit 34 restricts the conversion of the CFG model for a subgraph including two or more variables marked by the out-of-loop use variable mark unit 33 in the data dependence graph. In this case, the partial graph detection unit 34 restricts the operation of the CFG model conversion unit 35. The CFG model conversion unit 35 outputs the converted CFG model to the converted CFG model storage unit 22 without changing the CFG model.

  If this determination is not made, “x = y + i” should be retained in the loop and after exiting the loop (after exiting the loop, “i” is the value used at the end of the loop. (In the example of FIG. 5, i = 9)) is not established. Therefore, when there are a plurality of marked variables in the subgraph, the CFG model cannot be converted.

  As described above, according to the present invention, in the process of verifying the source code using the model checking technique, the assignment expression of the variable including the arbitrary value existing in the loop is moved out of the loop in the process of generating the model. This shortens the verification time.

  The loop optimization system of the present invention includes an inspection specification, an inspection object code, an inspection result, a model inspection control unit, a storage unit, and an arbitrary value loop control unit. Here, the model check control unit may be configured by a conventional technique such as a compiler optimization process and a symbol model check method using calculation tree logic (CTL).

  The effect of the present invention is that the verification time of model checking in a loop can be shortened. The reason is that the arbitrary value loop control unit 30 moves the calculation process including the arbitrary value existing in the loop and independent of other calculation processes in the loop to the outside of the loop.

  For example, in the conventional model check, when an arbitrary value exists in a loop, the check is performed for all the states in every round of the arbitrary value. Therefore, in the case of the example before the conversion in FIG. 3, the case from the minimum value to the maximum value that can be represented by the variable t is generated and checked in the first loop check, and the first loop check first. From each of the generated cases, cases from the minimum value to the maximum value that can be expressed by the variable t are further generated and inspected. Accordingly, the number generated by the ninth time is the ninth power of the number from the minimum value to the maximum value.

  On the other hand, in the present invention, since the calculation process of an arbitrary value is moved out of the loop, it is only necessary to check the number of cases from the minimum value to the maximum value after the loop. In this case, the difference in the number affects the calculation amount, and the verification time is shortened.

  Since the arbitrary value means “all states of possible values”, the result of the calculation process for the arbitrary value is included in the state of the arbitrary value. Accordingly, if the inspection is performed when there is an arbitrary value, it corresponds to the inspection performed for all cases of a certain variable.

  INDUSTRIAL APPLICABILITY According to the present invention, the present invention can be applied to uses such as improving the processing time of a source code verification tool using model checking.

  As mentioned above, although embodiment of this invention was explained in full detail, actually it is not restricted to said embodiment, Even if there is a change of the range which does not deviate from the summary of this invention, it is contained in this invention.

FIG. 1 is a block diagram showing a configuration example of a loop optimization system of the present invention. FIG. 2 is a flowchart showing the operation of the loop optimization system of the present invention. FIG. 3A is a diagram illustrating a procedure before conversion when a CFG model is converted. FIG. 3B is a diagram illustrating a procedure after conversion when a CFG model is converted. FIG. 4 is a diagram illustrating an example of a data dependence graph when the CFG model is converted. FIG. 5 is a diagram illustrating a procedure when the CFG model is not converted. FIG. 6 is a diagram illustrating an example of a data dependence graph when the CFG model is not converted.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 ... Model inspection control part 11 ... CFG model preparation part 12 ... Inspection model preparation part 13 ... Model inspection execution part 14 ... Inspection result output part 20 ... Storage part 21 ... CFG model storage part 22 before conversion CFG model storage part after conversion DESCRIPTION OF SYMBOLS 23 ... Examination model memory | storage part 30 ... Arbitrary value loop control part 31 ... Loop detection part 32 ... Data dependence graph preparation part 33 ... Out-of-loop utilization variable mark part 34 ... Subgraph detection part 35 ... CFG model conversion part

Claims (14)

A model check control unit that automatically generates a model from source code and performs verification by model check on the source code;
A loop optimization system comprising: an arbitrary value loop control unit that moves an assignment expression of a variable including an arbitrary value existing in a loop in the model to the outside of the loop, and reduces a verification time. The loop optimization system of claim 1,
The model checking control unit
A CFG model creation unit that reads the source code and the inspection specification, performs a compiler optimization process, creates a CFG (control flow graph) model, and provides it to the arbitrary value loop control unit;
A test model creation unit that reads a CFG model provided from the arbitrary value loop control unit and creates a test model based on a symbolic model test method using computation tree logic (CTL);
A model checking execution unit that reads the checking model and executes checking based on a symbolic model checking method using the calculation tree logic (CTL);
A loop optimization system comprising: an inspection result output unit that outputs a result of the executed inspection. The loop optimization system according to claim 2, comprising:
The arbitrary value loop control unit
A loop detection unit that reads a CFG model provided from the model checking control unit and detects a loop in the CFG model;
A data dependence graph creating unit for creating a data dependence graph of values appearing in the detected loop;
Of each variable in the data dependence graph, an out-loop use variable mark part that marks a variable used after the detected loop;
A subgraph detection unit for detecting a subgraph including the arbitrary value as an element and including the marked variable from the data dependence graph;
The CFG model element corresponding to the calculation processing of the variable included in the detected subgraph is deleted from the CFG model, and the marked variable included in the detected subgraph is immediately after the detected loop. In contrast, a loop optimization system comprising: a CFG model conversion unit that adds the deleted CFG model element to the CFG model and provides the CFG model element to the model check control unit. The loop optimization system according to claim 3,
A pre-conversion CFG model storage unit for obtaining and holding a pre-conversion CFG model from the CFG model creation unit, and providing the pre-conversion CFG model to the loop detection unit;
A converted CFG model storage unit for acquiring and holding the converted CFG model from the CFG model conversion unit and providing the converted CFG model to the inspection model creation unit;
A loop optimization system further comprising: an inspection model storage unit for acquiring and holding the inspection model from the inspection model creation unit and providing the inspection model to the model inspection execution unit. The loop optimization system according to claim 3 or 4,
The subgraph detection unit includes an assignment expression of a variable including the arbitrary value as an element from the data dependence graph, and detects a subgraph including only one of the marked variables.
The CFG model conversion unit deletes the CFG model element corresponding to the substitution expression of the variable including the arbitrary value from the CFG model, and includes only one in the detected subgraph immediately after the detected loop. A loop optimization system that adds the deleted CFG model element to the CFG model for the marked variable to be provided to the model checking control unit. The loop optimization system of claim 5,
The subgraph detection unit restricts the conversion of the CFG model for a subgraph including two or more of the marked variables.
The CFG model conversion unit provides the CFG model to the model check control unit. Automatically generating a model from source code and performing verification by model checking on the source code;
A step of reducing a verification time by moving an assignment expression of a variable including an arbitrary value existing in a loop in the model in a process of generating the model to reduce a verification time. The loop optimization method according to claim 7, comprising:
Reading the source code and inspection specification, performing compiler optimization processing, and creating a CFG (control flow graph) model;
Reading the CFG model and detecting a loop in the CFG model;
Creating a data dependence graph of values appearing in the detected loop;
Marking each variable in the data dependency graph that is used after the detected loop;
Detecting a subgraph including the arbitrary value as an element and including the marked variable from the data dependence graph;
The CFG model element corresponding to the calculation processing of the variable included in the detected subgraph is deleted from the CFG model, and the marked variable included in the detected subgraph is immediately after the detected loop. In contrast, the deleted CFG model element is added to the CFG model, the CFG model is converted, and the converted CFG model is read, and a symbol using computation tree logic (CTL) is used. Creating an inspection model based on a model checking method;
Reading the check model and performing a check based on a symbolic model check method using the computational tree logic (CTL);
And outputting the result of the executed inspection. A loop optimization method. The loop optimization method according to claim 8, comprising:
Detecting, from the data dependence graph, a subgraph including an assignment expression of a variable including the arbitrary value as an element and including only one of the marked variables;
The CFG model element corresponding to the substitution expression of the variable including the arbitrary value is deleted from the CFG model, and immediately after the detected loop, only one of the marked variables included in the detected subgraph is included. In contrast, the method further comprises: adding the deleted CFG model element to the CFG model and transforming the CFG model. The loop optimization method according to claim 9, comprising:
For a subgraph including two or more of the marked variables, limiting the conversion of the CFG model;
Reading the CFG model and creating a check model based on a symbolic model check method using the computational tree logic (CTL). A loop optimization method. Automatically generating a model from source code and performing verification by model checking on the source code;
A program for causing a computer to execute a step of reducing a verification time by moving an assignment expression of a variable including an arbitrary value existing in a loop in the model in the process of generating the model. The program according to claim 11,
Reading the source code and inspection specification, performing compiler optimization processing, and creating a CFG (control flow graph) model;
Reading the CFG model and detecting a loop in the CFG model;
Creating a data dependence graph of values appearing in the detected loop;
Marking each variable in the data dependency graph that is used after the detected loop;
Detecting a subgraph including the arbitrary value as an element and including the marked variable from the data dependence graph;
The CFG model element corresponding to the calculation processing of the variable included in the detected subgraph is deleted from the CFG model, and the marked variable included in the detected subgraph is immediately after the detected loop. In contrast, the deleted CFG model element is added to the CFG model, the CFG model is converted, and the converted CFG model is read, and a symbol using computation tree logic (CTL) is used. Creating an inspection model based on a model checking method;
Reading the check model and performing a check based on a symbolic model check method using the computational tree logic (CTL);
A program for causing a computer to further execute the step of outputting the result of the executed inspection. A program according to claim 12,
Detecting, from the data dependence graph, a subgraph including an assignment expression of a variable including the arbitrary value as an element and including only one of the marked variables;
The CFG model element corresponding to the substitution expression of the variable including the arbitrary value is deleted from the CFG model, and immediately after the detected loop, only one of the marked variables included in the detected subgraph is included. On the other hand, a program for causing the computer to further add the deleted CFG model element to the CFG model and convert the CFG model. The program according to claim 13,
For a subgraph including two or more of the marked variables, limiting the conversion of the CFG model;
A program for causing the computer to further execute a step of reading the CFG model and creating a check model based on a symbol model check method using the calculation tree logic (CTL).

Images