CN106201874B - The MHP analysis method and device of concurrent program - Google Patents
The MHP analysis method and device of concurrent program Download PDFInfo
- Publication number
- CN106201874B CN106201874B CN201610527181.2A CN201610527181A CN106201874B CN 106201874 B CN106201874 B CN 106201874B CN 201610527181 A CN201610527181 A CN 201610527181A CN 106201874 B CN106201874 B CN 106201874B
- Authority
- CN
- China
- Prior art keywords
- thread
- node
- mhp
- analyzed
- flow graph
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/36—Prevention of errors by analysis, debugging or testing of software
- G06F11/3604—Analysis of software for verifying properties of programs
- G06F11/3608—Analysis of software for verifying properties of programs using formal methods, e.g. model checking, abstract interpretation
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Software Systems (AREA)
- Computer Hardware Design (AREA)
- Quality & Reliability (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Debugging And Monitoring (AREA)
Abstract
本发明实施例提供一种并行程序的MHP分析方法和装置，一种并行程序的MHP分析方法包括：从待分析并行程序的主线程开始，每次选择一个线程，模拟待分析并行程序的执行，其中，当正在模拟执行的线程无法继续执行时，跳转至任一可继续执行的线程模拟执行，直至待分析并行程序的所有线程模拟执行完毕，其中，待分析并行程序包括至少两个线程；根据对待分析并行程序进行模拟执行的结果，构建与待分析并行程序对应的多线程控制流图；对多线程控制流图进行处理，得到保守正确的多线程控制流图，保守正确的多线程控制流图中的各节点的可能并发MHP结果不存在漏报；对保守正确的多线程控制流图进行MHP分析，得到待分析并行程序的MHP分析结果。
An embodiment of the present invention provides an MHP analysis method and device for a parallel program. The MHP analysis method for a parallel program includes: starting from the main thread of the parallel program to be analyzed, selecting one thread at a time, simulating the execution of the parallel program to be analyzed, Wherein, when the thread being simulated and executed cannot continue to execute, jump to any thread that can continue to execute to simulate and execute until all threads of the parallel program to be analyzed are simulated and executed, wherein the parallel program to be analyzed includes at least two threads; According to the results of the simulated execution of the parallel program to be analyzed, construct a multi-threaded control flow graph corresponding to the parallel program to be analyzed; process the multi-threaded control flow graph to obtain a conservative and correct multi-threaded control flow graph, and conservative and correct multi-threaded control There is no false negative in the possible concurrent MHP results of each node in the flow graph; MHP analysis is performed on the conservative and correct multi-threaded control flow graph to obtain the MHP analysis result of the parallel program to be analyzed.
Description
Technical field
The present embodiments relate to field of computer technology more particularly to the MHP analysis methods and dress of a kind of concurrent program
It sets.
Background technique
With the development of computer technology, computer develops to multicore era by monokaryon, is developed to by single machine processing big
The processing of scale cluster.The development of computer equally promotes the development of computer software, even to this day, had up to hundreds of
Computer programming language be continuously created, such as universal C/C++, Java, Python etc..Various big software programs, example
Such as operating system, database, large software game, are exactly the code of rows up to a million easily, these big software programs may
Need team's experience that can just developed for many years.So, how more effectively software program to be debugged, and carried out
The problems such as program correctness verifying and optimization, becomes as the research hotspot in computer software exploitation.
With the development of multiple programming, the analysis to concurrent program is the hot issue of software program analysis again.And stroke
Sequence analysis wants complicated many relative to serial program analysis.The analysis of serial program be mainly based upon control stream, pointer analysis with
And program control flow and data-flow analysis on the fundamental analysis such as definite value reference, what key was serial program is in the nature any
It is carved at one and only a sentence can execute.And for concurrent program, any one moment, there may be a plurality of
Different threads or the sentence of process are performed simultaneously, and whether then how may concurrently execute between determining program sentence just becomes simultaneously
One research direction of line program analysis.
The basic reason of Parallel Program Analysis complexity is that existing simultaneously multiple and different threads or the parallel of process statements holds
Row, this makes different threads or the sentence of process on executing sequence there are uncertain relationship, and such uncertainty relation
It then may cause the serious error of program.Then a basic problem in Parallel Program Analysis is exactly possible concurrent (May
Happen in Parallel, MHP) analysis, the result of this analysis is exactly all possibility parallel subqueries an of sentence
Set, or judge any 2 sentences whether there may be parallel execution relationships.The supplementary set of this problem, which gives, to be certainly existed
The sentence set of precedence relationship in execution.After having MHP analysis, so that it may in conjunction with the method for serial analysis, do pointer analysis, number
According to flow point analysis etc..
Before carrying out MHP analysis to program, need first to construct controlling stream graph (Control Flow corresponding with program
Graph, CFG), then MHP analysis is carried out based on CGF.And for concurrent program, how to construct correct multithreading CFG figure
It is the key that guarantee MHP analysis correctness.But the method for building multithreading GFG is too simple at present, generally only investigates creation
(creat) primitive and the analysis of locking (lock)/unlock (unlock) primitive, and ignore and (Happens-Before) formerly occurs
The importance of analysis causes analysis result to have very big wrong report or fail to report when so that carrying out MHP analysis according to CFG.
Summary of the invention
The embodiment of the present invention provides the MHP analysis method and device of a kind of concurrent program, for improve to concurrent program into
The accuracy of row MHP analysis.
First aspect provides a kind of MHP analysis method of concurrent program, comprising:
Since the main thread of concurrent program to be analyzed, a thread is selected every time, simulates the concurrent program to be analyzed
Execution, wherein when the thread for simulating execution can not continue to execute, jump to it is any continue to execute thread simulation
It executes, until all threads simulation of the concurrent program to be analyzed is finished, wherein the concurrent program to be analyzed includes
At least two threads；
According to the concurrent program to be analyzed carry out simulation execution as a result, building with the concurrent program pair to be analyzed
The Multi-thread control flow graph answered；
The Multi-thread control flow graph is handled, obtains guarding correct Multi-thread control flow graph, it is described to guard just
There is no fail to report the concurrent MHP result of the possibility of each node in true Multi-thread control flow graph；
Correct Multi-thread control flow graph progress MHP analysis is guarded to described, obtains the MHP of the concurrent program to be analyzed
Analyze result.
MHP analysis method provided in an embodiment of the present invention, it is corresponding with concurrent program to be analyzed multi-thread program-controlled by constructing
Then flow graph processed is handled the Multi-thread control flow graph, obtain guarding correct Multi-thread control flow graph, conservative just to this
True Multi-thread control flow graph carries out MHP analysis, so that the MHP for obtaining concurrent program to be analyzed is analyzed as a result, due to multithreading
CFG be it is conservative correctly, therefore it is higher to the building accuracy of multithreading CFG, MHP analysis is carried out according to multithreading CFG
The result wrong report of acquisition is less also there is no failing to report, and improves the accuracy rate that MHP analysis is carried out to concurrent program.
It is described that the Multi-thread control flow graph is handled in a kind of possible implementation of first aspect, it obtains
Guard correct Multi-thread control flow graph, comprising:
If addition primitive node in the Multi-thread control flow graph and multiple thread match, delete it is all with it is described plus
Enter the side of primitive node connection；
If waiting primitive node and multiple triggering primitive node matchings in the Multi-thread control flow graph are deleted all
The side being connect with the waiting primitive node；
Using obtained Multi-thread control flow graph as guarding correct Multi-thread control flow graph.
Due to when carrying out MHP analysis, judging that it is correct matched excessively complexity with which thread actually that primitive, which is added, because
This can delete all sides connecting with addition primitive herein, can make in this way when carrying out MHP analysis, it is believed that after primitive is added
The sentence in face all may be concurrent with addition primitive, this is a kind of wrong report, but is not failed to report.Similarly, primitive and triggering primitive are waited
Match, if therefore the waiting primitive node in multi-thread program-controlled CFG and multiple triggering primitive node matchings, delete and wait original
The side of language node connection.By above-mentioned processing, that is, it can guarantee that obtained multithreading CFG is conservative correct.
It is described to guard correct Multi-thread control flow graph progress to described in a kind of possible implementation of first aspect
MHP analysis obtains the MHP analysis result of the concurrent program to be analyzed, comprising:
Calculate the non-concurrent set and productive set for guarding each node in correct Multi-thread control flow graph, wherein
The non-concurrent set expression of node A can not be concurrent with node A node set, the productive set of node A indicates that node A is produced
The set of raw node；
Correct Multi-thread control flow graph is guarded according to the iterative calculation of the non-concurrent set and productive set of each node is described
In each node MHP analysis as a result, until the MHP analysis result convergence, obtain the concurrent program to be analyzed MHP analysis
As a result.
It is described since the main thread in a kind of possible implementation of first aspect, a thread is selected every time,
Simulate the execution of the concurrent program to be analyzed, wherein when the thread for simulating execution can not continue to execute, jump to and appoint
The one thread simulation that can be continued to execute executes, until all threads simulation of the concurrent program to be analyzed is finished, comprising:
Step 1 selects a thread, the execution of the thread of analog selection, when described since the main thread every time
When the thread of selection can not continue to execute, locks the selected thread and record the current position letter of the selected thread
Breath；
Step 2 jumps to any first thread being not locked out and can continue to execute, simulates holding for the first thread
Row, when the first thread can not continue to execute, locks the first thread and records the current position of the first thread
Information, until all threads that is not locked out and can continue to execute are modeled execution；
Step 3 releases all locked threads, and repeats step 2, until the concurrent program to be analyzed
All thread simulations are finished.
The method simulation provided according to embodiments of the present invention executes the multithreading CFG of obtained result detection, can be true
The relationship in concurrent program to be analyzed between each thread is embodied, to obtain accurately and reliably multithreading CFG.
In a kind of possible implementation of first aspect, the execution of the thread of the analog selection, comprising:
Using the execution of the thread of depth-first search analytic approach analog selection；
The execution of the simulation first thread, comprising:
The execution of the first thread is simulated using depth-first search analytic approach.
It is described to record the current location information of the selected thread in a kind of possible implementation of first aspect,
Include:
The mark of the selected thread and the mark of current sentence are recorded in scheduling information table；
It is described to record the current location information of the first thread, comprising:
The mark of the first thread and the mark of current sentence are recorded in the scheduling information table.
Second aspect provides a kind of MHP analytical equipment of concurrent program, comprising:
Execution module is simulated, for selecting a thread every time since the main thread of concurrent program to be analyzed, simulates institute
State the execution of concurrent program to be analyzed, wherein when the thread for simulating execution can not continue to execute, jump to it is any can be after
The continuous thread executed, which is simulated, to be executed, until all threads simulation of the concurrent program to be analyzed is finished, wherein it is described to
Analyzing concurrent program includes at least two threads；
Controlling stream graph construct module, for according to the concurrent program to be analyzed carry out simulation execution as a result, building
Multi-thread control flow graph corresponding with the concurrent program to be analyzed；
Controlling stream graph processing module obtains conservative correctly multi-thread for handling the Multi-thread control flow graph
Process control flow graph, there is no fail to report the concurrent MHP result of possibility for guarding each node in correct Multi-thread control flow graph；
MHP analysis module, for guarding correct Multi-thread control flow graph to described and carry out MHP analysis, obtain it is described to
The MHP for analyzing concurrent program analyzes result.
In a kind of possible implementation of second aspect, the controlling stream graph processing module, if being specifically used for described more
Addition primitive node in thread controlling stream graph and multiple thread match are then deleted and all are added what primitive node was connect with described
Side；If waiting primitive node and multiple triggering primitive node matchings in the Multi-thread control flow graph, delete all and institute
State the side for waiting the connection of primitive node；Using obtained Multi-thread control flow graph as guarding correct Multi-thread control flow graph.
In a kind of possible implementation of second aspect, the MHP analysis module is specifically used for calculating described guard just
The non-concurrent set and productive set of each node in true Multi-thread control flow graph, wherein the non-concurrent set expression of node A without
The set of method and node A concurrent node, the productive set of node A indicate the set of node caused by node A；According to each section
The non-concurrent set and productive set of point iterate to calculate the MHP analysis for guarding each node in correct Multi-thread control flow graph
As a result, until MHP analysis result convergence, obtains the MHP analysis result of the concurrent program to be analyzed.
In a kind of possible implementation of second aspect, the simulation execution module is specifically used for executing following steps:
Step 1 selects a thread, the execution of the thread of analog selection, when the selected line since the main thread every time
When Cheng Wufa is continued to execute, locks the selected thread and record the current location information of the selected thread；Step 2,
Any first thread being not locked out and can continue to execute is jumped to, the execution of the first thread is simulated, when the First Line
When Cheng Wufa is continued to execute, lock the first thread and record the current location information of the first thread, until it is all not
Thread that is locked and can continuing to execute is modeled execution；Step 3 releases all locked threads, and repeats step
Rapid two, until all threads simulation of the concurrent program to be analyzed is finished.
In a kind of possible implementation of second aspect, the simulation execution module is specifically used for using depth-first
The execution of the thread of searching analysis method analog selection；The execution of the first thread is simulated using depth-first search analytic approach.
In a kind of possible implementation of second aspect, the simulation execution module is specifically used in scheduling information table
The mark of the middle mark for recording the selected thread and current sentence；The First Line is recorded in the scheduling information table
The mark of journey and the mark of current sentence.
MHP analysis method provided in an embodiment of the present invention and device are corresponding with concurrent program to be analyzed more by constructing
Then thread controlling stream graph is handled the Multi-thread control flow graph, obtain guarding correct Multi-thread control flow graph, to this
Guard correct Multi-thread control flow graph and carry out MHP analysis, thus obtain the MHP analysis of concurrent program to be analyzed as a result, due to
Multithreading CFG be it is conservative correctly, therefore it is higher to the building accuracy of multithreading CFG, carried out according to multithreading CFG
The result wrong report that MHP analysis obtains is less also there is no failing to report, and improves the accuracy rate that MHP analysis is carried out to concurrent program.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is this hair
Bright some embodiments for those of ordinary skill in the art without any creative labor, can be with
Other attached drawings are obtained according to these attached drawings.
Fig. 1 is the flow chart of the MHP analysis method embodiment one of concurrent program provided in an embodiment of the present invention；
Fig. 2 is the flow chart of the MHP analysis method embodiment two of concurrent program provided in an embodiment of the present invention；
Fig. 3 is the flow chart of the MHP analysis method embodiment three of concurrent program provided in an embodiment of the present invention；
Fig. 4 is the structural schematic diagram of the MHP analytical equipment embodiment one of concurrent program provided in an embodiment of the present invention.
Specific embodiment
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention
In attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is
A part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art
All other embodiment obtained without creative efforts, shall fall within the protection scope of the present invention.
During being compiled to concurrent program, when constructing the CFG of concurrent program, need to match the wound in each thread
Build (creat)/addition (join) primitive, triggering (signal)/wait (wait) primitive, locking (lock)/unlock (unlock)
Primitive.Wherein creat/join primitive, the matching between signal/wait primitive constitute Happens-Before problem, because
The sentence that these primitive act on it is shown in that there are specific partial ordering relation, this partial ordering relation is inevitable during program really executes
The execution order relation of satisfaction.So if analysis program cannot catch this partial order feature well, analysis result is necessarily adjoint
A large amount of error analysis (such as the sentence that centainly cannot concurrently execute be judged as may be concurrent).
Fig. 1 is the flow chart of the MHP analysis method embodiment one of concurrent program provided in an embodiment of the present invention, such as Fig. 1 institute
Show, MHP analysis method provided in this embodiment includes:
Step S101 selects a thread since the main thread of concurrent program to be analyzed every time, simulates to be analyzed parallel
The execution of program, wherein when the thread for simulating execution can not continue to execute, jump to any thread continued to execute
Simulation executes, until all threads simulation of concurrent program to be analyzed is finished, wherein concurrent program to be analyzed includes at least
Two threads.
Specifically, MHP analysis method provided in this embodiment is applied to during being compiled to concurrent program.For
One concurrent program to be analyzed, due to parallel feature, including multiple threads, one of thread is main thread, should be to
Analysis concurrent program is executed since main thread.
In the present embodiment, first since the main thread of concurrent program to be analyzed, a thread is selected every time, to each
Thread is analyzed, the execution of the thread of analog selection.There is a principle for the simulation of per thread, exactly when simulating
When the thread of execution can not continue to execute, then jump to any one thread simulation execution that can be continued to execute.It is above-mentioned to continue
The thread of execution can meet the condition that can be continued to execute during certain threads are performed and therefore repeat the above steps, can
It is all simulated with to be analysed to all threads of concurrent program and is finished.
Step S102, according to concurrent program to be analyzed carry out simulation execution as a result, building with concurrent program to be analyzed
Corresponding Multi-thread control flow graph.
Specifically, after all threads to concurrent program to be analyzed are all simulated execution, you can learn that each thread
Between relationship, i.e., whether there is creat/join, the matching of signal/wait, lock/unlock primitive are closed between each thread
System.According to the matching relationship between each thread, multithreading GFG corresponding with concurrent program to be analyzed can be constructed.
Step S103 handles Multi-thread control flow graph, obtains guarding correct Multi-thread control flow graph, guards just
There is no fail to report the MHP result of each node in true Multi-thread control flow graph.
Specifically, multithreading CFG corresponding with concurrent program to be analyzed embody in concurrent program to be analyzed thread it
Between logical relation, carrying out analysis to multithreading CFG can be obtained analysis result to concurrent program to be analyzed.Due to right
When program carries out MHP analysis, in fact it could happen that two kinds of mistakes, one kind to fail to report, i.e., will likely concurrent sentence be judged as can not be simultaneously
Hair, another kind are wrong report, that is, would be impossible to concurrent sentence and be judged as concurrent.When carrying out static analysis to program, it is ensured that point
The case where there is no fail to report in analysis result then claims to guard when the result correct.
Step S104 carries out MHP analysis to correct Multi-thread control flow graph is guarded, obtains concurrent program to be analyzed
MHP analyzes result.
Specifically, after obtaining guarding correct multithreading CFG, MHP analysis can be carried out to it, obtains analysis result.
The analysis result is to reflect the MHP analysis result of concurrent program to be analyzed.Since multithreading CFG is to guard correctly,
It is higher to the building accuracy of multithreading CFG, it is less that the result wrong report that MHP analysis obtains is carried out according to multithreading CFG
There is no failing to report, the accuracy rate that MHP analysis is carried out to concurrent program is improved.
MHP analysis method provided in this embodiment, by constructing Multi-thread control stream corresponding with concurrent program to be analyzed
Figure, then handles the Multi-thread control flow graph, obtains guarding correct Multi-thread control flow graph, conservative correct to this
Multi-thread control flow graph carries out MHP analysis, so that the MHP for obtaining concurrent program to be analyzed is analyzed as a result, since multithreading CFG is
It is conservative correct therefore higher to the building accuracy of multithreading CFG, carry out what MHP analysis obtained according to multithreading CFG
As a result it reports less be also not present by mistake to fail to report, improves the accuracy rate for carrying out MHP analysis to concurrent program.
It, can be to step in order to guarantee that the multithreading CFG obtained after handling multithreading CFG is to guard correctly
The multithreading CFG that S102 is obtained is handled as follows: if addition primitive node and multiple threads in Multi-thread control flow graph
Match, then deletes all sides connecting with addition primitive node；If waiting primitive node and multiple touchings in Multi-thread control flow graph
Primitive node matching is sent out, then deletes the side connecting with the waiting primitive node；Using obtained Multi-thread control flow graph as guarantor
Keep correct Multi-thread control flow graph.
It specifically, is for matching with creation primitive since primitive is added, but multiple creation primitive may be using one
A identical mark, therefore addition primitive may match with multiple threads, but only one of them thread is that primitive is added to answer
It is correct matched.Due to when carrying out MHP analysis, judging that it is correct matched excessively multiple with which thread actually that primitive, which is added,
It is miscellaneous, therefore all sides connecting with addition primitive can be deleted herein, it can make in this way when carrying out MHP analysis, it is believed that be added
The subsequent sentence of primitive all may be concurrent with addition primitive, this is a kind of wrong report, but is not failed to report.Similarly, primitive and touching are waited
Hair primitive matches, if therefore the waiting primitive node in multi-thread program-controlled CFG and multiple triggering primitive node matchings, delete and
Wait the side of primitive node connection.By above-mentioned processing, that is, it can guarantee that obtained multithreading CFG is conservative correct.
Further, in the embodiment shown in fig. 1, it in step S104, is carried out to correct Multi-thread control flow graph is guarded
The step of MHP is analyzed, and obtains the MHP analysis result of concurrent program to be analyzed, can specifically include: calculating conservative correctly multi-thread
The non-concurrent set and productive set of each node in process control flow graph, wherein the non-concurrent set expression of node A can not be with node
The set of A concurrent node, the productive set of node A indicate the set of node caused by node A；According to each node it is non-simultaneously
Hair set and productive set iterative calculation guard the MHP analysis of each node in correct Multi-thread control flow graph as a result, until MHP
Result convergence is analyzed, the MHP analysis result of concurrent program to be analyzed is obtained.
Specifically, when to correct multithreading CFG progress MHP analysis is guarded, it is necessary first to each in multithreading CFG
Node is analyzed, and determines the non-concurrent set and productive set of each node.The non-concurrent set expression of one of node A without
The set of method and node A concurrent node, that is, certainly can not be concurrent with node A node set.The generation collection of node A
Close the set for indicating node caused by node A.Then a node is selected every time, is saved according to its input node set, output
Point set, non-concurrent set and productive set calculate the MHP result of the node.Wherein, the input node knot of each node
It closes, the union of output node set and productive set, then subtracts non-concurrent set, as the MHP of the node analyzes result.It will be right
Each calculated MHP analysis result of node is iterated if having circulation along the execution sequence transmitting downwards of multithreading CFG
It calculates, finally gathered the MHP of each node points after obtaining, set obtains the MHP analysis result of concurrent program to be analyzed.
Fig. 2 is the flow chart of the MHP analysis method embodiment two of concurrent program provided in an embodiment of the present invention, such as Fig. 2 institute
Show, MHP analysis method provided in this embodiment includes:
Step S201 selects a thread, the execution of the thread of analog selection, when selection since main thread every time
When thread can not continue to execute, locks the thread of selection and record the current location information of thread of selection.
Specifically, MHP analysis method provided in this embodiment is that the specific of step S101 in embodiment illustrated in fig. 1 executes step
Suddenly.Firstly, selecting a thread every time since main thread, simulating the execution of selected thread, until the thread of the selection
It can not continue to execute, that is, the thread of each selection is required as much as possible to simulate execution, until can not continue to execute.
Then the thread of the selection is locked, and records the current location of the thread of the selection.Before the thread UNLOCKS to the selection, no
It can continue the thread that simulation executes the selection.That is, can holding using the thread of depth-first search analytic approach analog selection
Row.The mark of the thread of selection and the mark of current sentence can be recorded in scheduling information table, to record the line of selection
The location information of Cheng Dangqian.
Step S202 jumps to any first thread being not locked out and can continue to execute, simulates the execution of first thread,
When first thread can not continue to execute, locks first thread and record the current location information of first thread, until not locked
Thread that is fixed and can continuing to execute is modeled execution.
Specifically, after having locked the thread of above-mentioned selection, any one can be jumped to and be not locked out and can continue to hold
Capable first thread, and simulate the execution of first thread.Ability simulation maximum to the greatest extent is also required to the simulation of first thread to execute,
Until first thread can not continue to execute.Then it locks first thread and records the current position of first thread.Repeat selection not
First thread that is locked and can continuing to execute, until all threads that is not locked out and can continue to execute are modeled and hold
Row.That is, can be using the execution of depth-first search analytic approach simulation first thread.Can be recorded in scheduling information table
The mark of one thread and the mark of current sentence, thus the location information that the thread for recording selection is current.
Step S203 releases locked thread, and repeats step S202, until concurrent program to be analyzed is all
Thread simulation is finished.
Specifically, when thread that is no unlocked and can continuing to execute, need to release all locked threads simultaneously.
Then step S202 is repeated, until all threads that must analyze program are modeled and are finished.That is, at this
In embodiment, the principle that simulation execution is carried out to each thread is first, and the simulation that needs to do one's best executes the line currently selected
Journey, second, if still there is executable thread, locked thread cannot be executed.
The multithreading CFG that obtained result detection is executed according to method provided in this embodiment simulation, being capable of real embodiment
Relationship in concurrent program to be analyzed between each thread, to obtain accurately and reliably multithreading CFG.
Figures 1 and 2 show that the implementing procedure of the MHP analysis method of concurrent program provided by the invention, below with one
The MHP analysis method of Fig. 1 and concurrent program shown in Fig. 2 is further described in specific embodiment.
Fig. 3 is the flow chart of the MHP analysis method embodiment three of concurrent program provided in an embodiment of the present invention, such as Fig. 3 institute
Show, MHP analysis method provided in this embodiment includes:
Step S301 constructs a scheduler, using the execution of scheduler simulation dynamic routine, selects a thread every time
Carry out static analysis.
Specifically, firstly, constructing a scheduling information list, the information of the thread for saving concurrent program to be analyzed,
This information is mark (Identity, ID) and the analysis position of present analysis thread, for example, where function, place it is basic fast with
Which sentence of place.Each cross-thread is using random schedule or follows bad scheduling scheme, and using the depth-first of standard in thread
Search for (Depth First Search, DFS) analysis method.Then for a thread, in scheduling information list
The schedulable position of only one will be occupied, and its information can behave as the form of stack, function interbed when for expressing DFS analysis
The relationship that layer calls.Device to be compiled scans through source code, and carries out basic morphology, multithreading CFG in syntactic analysis and process
Building after, scheduler start from principal function to multithreading CFG scan, if a creation primitive is met, in scheduling information
The information of this new wound thread is established in list, and this new wound thread is analyzed in right times later.If adjusted
Degree device meets linear function calling, then is analyzed according to DFS, the location information pop down of calling function, then jump to and adjusted
Function is analyzed.If current thread cannot be analyzed again, for example meet an addition primitive etc., scheduler saves current
Analyze and the location informations such as the place function of primitive, place basic block, place sentence be added, then select another thread after
It is continuous to be analyzed.If current thread analysis finishes, the information of this thread can be cancelled in scheduling information list.
Step S302, scheduler analyze each thread, occur blocking or cannot divide again in currently analyzed thread
It is switched to other threads when analysis to be analyzed, after meeting certain conditions, scheduler can continue to analyze before those
It is blocked or unanalyzable thread.
Specifically, according to the meaning of each primitive, it may be defined that and primitive is added, triggering primitive, waits primitive that there is obstruction
Meaning.Namely: primitive, triggering primitive is added, primitive is waited to have the ability that analysis program is prevented to continue to analyze the energy of current thread
Power, until some condition meets.Analysis is carried out to each thread and needs to meet following two rule:
Rule one: analysis program should use up its best endeavors and do analysis, (because of quilt before it cannot do again any analysis
Obstruction or part Thread Analysis finish), analysis program cannot do any creation/addition primitive and the matching of triggering/waiting primitive
Work.Rule two: once creation/addition primitive and the matching work of triggering/waiting primitive are done in analysis program preparation, journey is analyzed
Sequence should handle current all creation/addition primitive and triggering/waiting primitive, then unlock again after blocked state continues analysis
Continuous program.
Step S303, circulation execute step S302, finish until institute's source code is analyzed.Then pass through two method of purifications
Then guarantee that constructed multithreading CFG is conservative correct.
First purification rule: if an addition primitive node (can include standard thread with multiple thread match
The case where (normalized thread)), just delete all sides connecting with the addition primitive node.Article 2 purifies rule:
As soon as if waiting primitive node can be deleted all with multiple triggering primitive node matchings (the case where including standard thread)
The side being connect with the waiting primitive node.Above-mentioned standard thread is defined as: the thread created in circulation；These threads are usual
There is identical subject thread function body, but in static analysis, due to that can not know the number of creation thread, usually according to list
A thread or 2 threads are analyzed, but to be labeled as standard thread, to indicate the difference with general thread.
Step S304 calculates the multithreading CFG after creation and carries out MHP calculating.
Specifically, define some symbols first: t indicates that a thread, N (t) indicate all nodes of this thread；N table
Show a node, (it is online that the node returned belongs to n institute to localpred (n) by the predecessor node immediately for returning to the local n
Journey), localsucc (n) returns to the descendant node immediately of the local n；Pred (n) returns to all predecessor nodes of n, and succ (n) is returned
Return all descendant nodes of n；All ancestor nodes of anc (n) return n；All descendant nodes of des (n) return n；path_
Des (n1, n2) returns to all paths that n2 is reached from n1, these paths obtain before all passing through to traversal；P is a paths, p ∈
path_des(n1,n2)；N (p) returns to all nodes on this paths p；Path_anc (n1, n2) is returned from n1 and is reached n2's
All paths, these paths obtain after all passing through to traversal.
The non-concurrent set and productive set of some key nodes are calculated according to the following formula, and wherein KILL (n) indicates section
The non-concurrent set of point n, GEN (n) indicate the productive set of node n.Signal_node indicates triggering primitive node, wait_
Node indicates that waiting primitive node, begin_node indicate that the ingress when opening of thread, join_node indicate that primitive is added
Node, the egress create_node at the end of exit_node indicates thread indicate creation primitive node, create_succ_
Node indicates the descendant node of creation primitive.
It calculates KILL (signal_node)
Wherein wait_node is the node to match with signal_node, and wait_node and begin_node belong to together
One thread.
It calculates GEN (signal_node)
GEN (signal_node) :=KILL (signal_node)
It calculates KILL (wait_node)
KILL (wait_node) :=n | n ∈ anc (signal_node)-des (signal_node) }
Signal_node is the node to match with wait_node.
It calculates KILL (begin_node)
KILL (begin_node) :=∪n∈signal_nodeKILL(n)
Begin_node and signal_node belongs to the same thread.
It calculates KILL (join_node)
KILL (join_node) :=ane (exit_node)
It calculates GEN (begin_node/create_succ_node)
Each create_node is there are two descendant node immediately, and one is begin_node, and one is create_succ_
node。
It calculates GEN (begin_node/create_succ_node)
Then the calculating of MHP is carried out according to the formula of following data-flow equations.
IN (n) :=∪p∈localpred(n)out(p)
POUT (n) :=OUT (n) ∪ IN (n) ∪ GEN (n)-KILL (n)
MHP (n)=: OUT (n)-LOCK (n)
Cycle calculations are carried out to entire concurrent program to be analyzed, so as to obtain the MHP analysis of concurrent program to be analyzed
As a result.
Fig. 4 is the structural schematic diagram of the MHP analytical equipment embodiment one of concurrent program provided in an embodiment of the present invention, is such as schemed
Shown in 4, the MHP analytical equipment of concurrent program provided in this embodiment includes:
Execution module 41 is simulated, for since the main thread of concurrent program to be analyzed, selecting a thread, simulation every time
The execution of the concurrent program to be analyzed, wherein when the thread for simulating execution can not continue to execute, jump to it is any can
The thread simulation continued to execute executes, until all threads simulation of the concurrent program to be analyzed is finished, wherein described
Concurrent program to be analyzed includes at least two threads.
Controlling stream graph construct module 42, for according to the concurrent program to be analyzed carry out simulation execution as a result, structure
Build Multi-thread control flow graph corresponding with the concurrent program to be analyzed.
Controlling stream graph processing module 43 obtains conservative correctly more for handling the Multi-thread control flow graph
Thread controlling stream graph, there is no leakages for the concurrent MHP result of possibility for guarding each node in correct Multi-thread control flow graph
Report.
MHP analysis module 44 obtains described for guarding correct Multi-thread control flow graph progress MHP analysis to described
The MHP of concurrent program to be analyzed analyzes result.
The MHP analytical equipment of concurrent program provided in this embodiment for realizing concurrent program shown in Fig. 1 the analysis side MHP
The process of method, it is similar that the realization principle and technical effect are similar, and details are not described herein again.
Further, in the embodiment shown in fig. 4, controlling stream graph processing module 43, if being specifically used for described multi-thread program-controlled
Addition primitive node and multiple thread match in flow graph processed then delete all sides connecting with the addition primitive node；If
Waiting primitive node and multiple triggering primitive node matchings in the Multi-thread control flow graph, then delete all and waiting
The side of primitive node connection；Using obtained Multi-thread control flow graph as guarding correct Multi-thread control flow graph.
Further, in the embodiment shown in fig. 4, MHP analysis module 44 is specifically used for calculating described conservative correctly more
The non-concurrent set and productive set of each node in thread controlling stream graph, wherein the non-concurrent set expression of node A can not be with section
The set of point A concurrent node, the productive set of node A indicate the set of node caused by node A；According to the non-of each node
Concurrency set and productive set iterate to calculate it is described guard each node in correct Multi-thread control flow graph MHP analysis as a result,
Until MHP analysis result convergence, obtains the MHP analysis result of the concurrent program to be analyzed.
Further, in the embodiment shown in fig. 4, execution module 41 is simulated, is specifically used for executing following steps: step
One, since the main thread, every time select a thread, the execution of the thread of analog selection, when the selected thread without
When method continues to execute, locks the selected thread and record the current location information of the selected thread；Step 2 jumps
To any first thread being not locked out and can continue to execute, simulate the execution of the first thread, when the first thread without
It when method continues to execute, locks the first thread and records the current location information of the first thread, do not locked until all
Thread that is fixed and can continuing to execute is modeled execution；Step 3 releases all locked threads, and repeats step
Two, until all threads simulation of the concurrent program to be analyzed is finished.
Further, in the embodiment shown in fig. 4, execution module 41 is simulated, is specifically used for using depth-first search point
The execution of the thread of analysis method analog selection；The execution of the first thread is simulated using depth-first search analytic approach.
Further, in the embodiment shown in fig. 4, execution module 41 is simulated, specifically for recording in scheduling information table
The mark of the selected thread and the mark of current sentence；The mark of the first thread is recorded in the scheduling information table
Know the mark with current sentence.
Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention., rather than its limitations；To the greatest extent
Pipe present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that: its according to
So be possible to modify the technical solutions described in the foregoing embodiments, or to some or all of the technical features into
Row equivalent replacement；And these are modified or replaceed, various embodiments of the present invention technology that it does not separate the essence of the corresponding technical solution
The range of scheme.
Claims (12)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610527181.2A CN106201874B (en) | 2016-07-06 | 2016-07-06 | The MHP analysis method and device of concurrent program |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610527181.2A CN106201874B (en) | 2016-07-06 | 2016-07-06 | The MHP analysis method and device of concurrent program |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106201874A CN106201874A (en) | 2016-12-07 |
| CN106201874B true CN106201874B (en) | 2018-12-28 |
Family
ID=57465874
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610527181.2A Active CN106201874B (en) | 2016-07-06 | 2016-07-06 | The MHP analysis method and device of concurrent program |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106201874B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP0406602B1 (en) * | 1989-06-29 | 1995-08-16 | International Business Machines Corporation | Method and apparatus for debugging parallel programs by serialization |
| CN101937365A (en) * | 2009-06-30 | 2011-01-05 | 国际商业机器公司 | Deadlock detection method of parallel programs and system |
| CN102073588A (en) * | 2010-12-28 | 2011-05-25 | 北京邮电大学 | Code static analysis based multithread deadlock detection method and system |
| CN104536898A (en) * | 2015-01-19 | 2015-04-22 | 浙江大学 | C-program parallel region detecting method |
-
2016
- 2016-07-06 CN CN201610527181.2A patent/CN106201874B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP0406602B1 (en) * | 1989-06-29 | 1995-08-16 | International Business Machines Corporation | Method and apparatus for debugging parallel programs by serialization |
| CN101937365A (en) * | 2009-06-30 | 2011-01-05 | 国际商业机器公司 | Deadlock detection method of parallel programs and system |
| CN102073588A (en) * | 2010-12-28 | 2011-05-25 | 北京邮电大学 | Code static analysis based multithread deadlock detection method and system |
| CN104536898A (en) * | 2015-01-19 | 2015-04-22 | 浙江大学 | C-program parallel region detecting method |
Non-Patent Citations (1)
| Title |
|---|
| 高可扩展性的MHP分析算法;印乐;《软件学报》;20131130;第2289-2299页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106201874A (en) | 2016-12-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Fellner et al. | Model-based, mutation-driven test-case generation via heuristic-guided branching search | |
| US8365152B2 (en) | Path-sensitive analysis through infeasible-path detection and syntactic language refinement | |
| CN104536883B (en) | A kind of static defect detection method and its system | |
| Bensalem et al. | Confirmation of deadlock potentials detected by runtime analysis | |
| Santone et al. | Abstract reduction in directed model checking ccs processes | |
| Remenska et al. | Using model checking to analyze the system behavior of the LHC production grid | |
| Baluda et al. | Bidirectional symbolic analysis for effective branch testing | |
| Slaby et al. | Compact symbolic execution | |
| CN104267936A (en) | Semantic tree based asynchronous dynamic push-down network reachability analysis method | |
| CN106201874B (en) | The MHP analysis method and device of concurrent program | |
| Chechik et al. | Property-based methods for collaborative model development | |
| Wang et al. | Symbolic predictive analysis for concurrent programs | |
| Ito et al. | A method for detecting bad smells and ITS application to software engineering education | |
| Wang et al. | Interactive inconsistency fixing in feature modeling | |
| Binalialhag et al. | Static slicing of Use Case Maps requirements models | |
| Novella et al. | Automatic test set generation for event-driven systems in the absence of specifications combining testing with model inference | |
| Remenska | Bringing model checking closer to practical software engineering | |
| Cheng et al. | Algorithms for synthesizing priorities in component-based systems | |
| Mavridou et al. | Automated Formalization of Probabilistic Requirements from Structured Natural Language | |
| Chen et al. | On interleaving space exploration of multi-threaded programs | |
| Xiao et al. | An On-the-fly Synthesis Framework for LTL over Finite Traces | |
| Altaie et al. | Automated test suite generation tool based on GWO algorithm | |
| Xu et al. | A log-linear probabilistic model for prioritizing extract method refactorings | |
| Marchant et al. | Delivering Rules-Based Workflows for Science | |
| Zhang et al. | Regression test generation approach based on tree-structured analysis |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |