WO2012072550A1

WO2012072550A1 - Method for the functional optimisation of the sequence of a program in an embedded system

Info

Publication number: WO2012072550A1
Application number: PCT/EP2011/071126
Authority: WO
Inventors: Ralph Mader; Rudolf Bauer; Maximilian Riepl
Original assignee: Continental Automotive Gmbh
Priority date: 2010-11-29
Filing date: 2011-11-28
Publication date: 2012-06-07
Also published as: DE102010062110A1

Abstract

The invention relates to a method for the functional optimisation of the sequence of a program in an embedded system. The program uses, for the sequence thereof, a stack of the embedded system, which is formed by a limited address range in a memory of the embedded system, and the size of which is defined by an address range having a lower address and an upper address. In the method according to the invention, a threshold value (SW) is defined for the stack, which is represented by an address of the stack, wherein said address is less than the upper address of the address range of the stack. Monitoring takes place to determine whether the program running on the embedded system using the stack reaches the address represented by the threshold value (SW). When the threshold value (SW) is reached, a predefined function is called up, by means of which the content of the stack is read out. It is determined from the content of the stack what part of the program has caused the predefined function to be called up. Finally, optimisation of the program and/or of the embedded system is carried out using the information about the part calling up.

Description

description

Method for optimizing the operation of a program in an embedded system

The invention relates to a method for optimizing the operation of a program in an embedded system, the program to its expiration uses a stack of the embedded system, which is formed by a delimited address range in a memory of the embedded system and its size by an address range with a lower (start) address and an upper (end) address is defined.

When running programs in an embedded system (so-called embedded software system), it is necessary to have the program's consumption of the stack under control during operation. An uncontrolled overflow of Sta ^¬ pelspeichers could have a reset of the embedded system with an unpredictable result wrongdoing. In order the behavior of the program to control at the end of which as regards the use of the stack, it is necessary to perform a reliable measurement of the consumption of the stack under real Be ^¬ conditions in the development of the program. It should be remembered that these real conditions are usually dynamic and

are non-deterministic, for example, due

unpredictable events that can occur at an unknown, arbitrary time and trigger an interrupt of the embedded system. Since interrupts significant impact on the real-time behavior of the program running on the embedded system program and thus can have the use of the stack must be considered when interrupts Ve ^¬ rifikation program flow.

Known methods usually perform static analyzes of the program. These are based on the compiler-generated assembler or ob ect code. Here, so-called call trees are created, which call trees of program parts (radio tions) of the program. In this case, the consumption of the stack of each program part is added up. A disadvantage of this approach is that dynamic effects, such as interrupts by external events, a long-term behavior or a task change of the operating system of the embedded system, can not be detected. It is only possible to carry out a worst-case analysis of the captured call trees. It is determined which consumption of the stack can occur maximum.

The dynamic load of the stack (so-called stack) can be determined by the so-called stack pattern method. However, there is no way to infer a conclusion about the part of the program that causes the stack memory usage. In the stack pattern method, the address space reserved for the stack memory is pre-initialized with a predetermined pattern (ie, a known date). The depth of penetration of the used Stapelbe ^¬ realm is determined in the predefined pattern then during the course of the program to be checked cyclically. It can be used to check how much of the stack was consumed. Again, there is the problem that it is not possible to determine which program part of the program has caused the consumption of the stack.

Since it is not possible to test the embedded system under real conditions with the known methods, it is not possible to accurately predict the behavior of the program when running in the embedded system with regard to the actual consumption of the stack. It can therefore come to the aforementioned, unpredictable misconduct of the embedded system.

It is therefore an object of the present invention to provide a method by which an improved function optimization of the flow of a program in an embedded system is made possible, so that in later real operation an overflow of a Stacked memory of the embedded system can be avoided with high security.

This object is achieved by a method according to the features of claim 1. Advantageous embodiments will be apparent from the dependent claims.

The invention provides a method for optimizing the operation of a program in an embedded system, wherein the program uses a stack of the embedded system, which is formed by a delimited address range in a memory of the embedded system and its size by an address range of a lower (Start) address and an upper (end) address is defined. In the method, a threshold value for the stack, is defined according to the invention, which is repre sented ^¬ by an address of the stack in which the address is smaller than the upper address of the address range of the stack. Next, whether running on the embedded system program reaches the repre ^¬ sented by the threshold address using the stack is monitored. Use of the stack means here that in the usual way starting at the lower address data is written to the stack, with data placed only on top of the stack and can be read from there again. Upon reaching the threshold, a predefined function is called which reads out the contents of the stack. The predefined function is eg a so-called trap, ie an error function with predefined functionality. Then it is determined from the contents of the stack which part of the program (program part) has triggered the call of the predefined function. Such a program part is also known as a function of the program. Based on the information about the calling part then an optimization of the program and / or the embedded system is performed.

The procedure described makes it possible to reduce the consumption of the stack in an embedded system taking into account assessment of all real-world conditions, such as unpredictable events, long-term behavior, etc., and, if necessary, to optimize the program and / or the embedded system.

According to an expedient embodiment, a Call Tree is determined (a so-called. Call tree of Pro ^¬ program parts of the program) from the content of the stack, wherein the part is determined from the Call Tree, which calls the predefined function of the embedded system. With knowledge of that part of the program which executes during execution of the program to call the predefined function, it is possible to detect problems in the program sequence and this to besei ^¬ term.

Appropriately, from the contents of the stack the value of the current stack pointer and the address of the called function set ^¬ (Program Counter) and, optionally, a time stamp of the function call, is determined. If this information is determined in a test environment ^¬ temporarily stored and evaluated later, the dynamic behavior of the program can be found in the embedded system regarding the use of the stack. On the basis of this ^information , weak points in the program or in the embedded system can be identified and remedied. From this data, both the consumption of the stack as a function of time and the call tree of those program parts can be rekon ^¬ strued, which have led to this observed consumption of the stack.

The call of the predefined function can also be seen as an indication that the stack space "limited" by the threshold value is not dimensioned large enough In a further, expedient refinement, after the call to the predefined function and the determination of the contents of the stack memory, the threshold value with a new address of the threshold moving closer to the upper address of the stack, and then re-running the program In the embedded system checks whether by a program part when reaching the new threshold again the default function is called. Based on this iterative approach, it can be determined which (maximum) area of the stack is actually used by the program running in the embedded system. A possible optimization may then be to increase the stack by changing the upper or lower address. An increase in the address space of the stack may become necessary if the threshold value assumes the upper address of the original stack area and the specified function is called upon a re-run of the program in the embedded system. A call to the predetermined function would indicate that the size of the stack for the expiry of the pre ^¬ given program is not sufficient. If, on the other hand, at a threshold value defined for the first time which is smaller than the upper address of the stack area, it is found that the predetermined function is not called when the program is running, the address space of the stack could also be reduced. The freed address space can then be used in the embedded system for other storage purposes. An alternative possibility of optimization is to make a change in the interleaving depth of a number of interdependent (program) parts, wherein the change in the interleaving depth of interdependent program parts is preferably realized by a compiler. In a further embodiment, the optimization comprises a change of the process in one or more parts of the program.

The invention is explained in more detail below with reference to an exporting ^¬ approximately embodiment in the drawing. Show it:

Fig. 1 is a schematic representation of a flow chart of the method according to the invention, and Figure 2 is a schematic representation of the consumption of a stack in an embedded system over time. The inventive method for optimizing the operation of a program in an embedded system is based on obtaining a reliable indication of the consumption of a stack of the embedded system under real conditions. In such an embedded system, the stack is often formed by a delineated address space in a memory of the embedded system. The size of the stack is initially defined by an address range having a lower (start) address and an upper (end) address. In a manner known to those skilled in the art, elements or data can only be placed on top of the stack from above, ie beginning with the start address, and read only from there. This means that items are stacked on top of each other and taken off the stack in reverse order. This principle is called last-in-first-out principle.

When carrying out a measurement of the consumption of the stack (so-called stack consumption), dynamic and non-deterministic events can be taken into account with the method according to the invention. The method is based on the definition of thresholds for the stack memory, wherein a threshold value is represented by an address of the stack whose address is smaller than the upper address of the address range of the stack. Fig. 1 shows the basic sequence of the invention

Process. S marks the start of the process. In a method step S1, a threshold value SW for the stack is initially defined. Preferably, the

Threshold selected to be near the top (end) address of the stack. At the end of the program in the embedded system is monitored whether the program reaches the repre ^¬ sented by the threshold address using the stack. In this case (step S2) becomes a predefined function of the embedded system called (step S3). The predefined function is an ^error function with predefined functionality and is called a trap.

By calling the predetermined function, the content of the stack is determined according to steps S4 and S5 to determine which program part of the program has triggered the call of the predefined function. Further, in step S5, a call tree (a

Call tree), whereby Festge particular this can be ^¬ represents which has called the predefined function on ^¬. In steps S7 and S8, information is thus extracted and stored which represent the previously highest, used address of the stack (step S7) and the call tree CT, which is temporarily stored in step S8 for further evaluation.

According to step S6, after the triggering of a trap, the threshold value SW is increased, but this is not increased above the upper (end) address of the stack. The increase of the threshold value SW ensures that the threshold value checked in step S2 is set to the now valid threshold value. This ends a loop of the procedure (end E). The flow of the program in the embedded system can now begin again. If the now valid threshold value SW reached in step S2, in turn, as a new call of the prede ^¬-defined function in step S3, etc. As a result, it can be determined whether the reserved for the execution of the program stack is sufficiently large. This is the case if the threshold is located below the upper (end) address, even after repeated increases, and no call is made to the predefined function. In this case, an optimization could be to reduce the top (end) address of the stack, eg to the address of the threshold or at least one address higher. If, on the other hand, the upper (end) address is reached by the threshold value, without the program being able to run to its end as intended, then either the size of the stack memory is to be increased, for example by increasing the upper (end) address. Likewise, the optimization could include changing the nesting depth of the number of interdependent program parts. This is preferably done automatically by a compiler. Another optimization is to make a change to the process on one or more parts of the program, preferably manually by the programmer of the program.

It is also possible to determine from the contents of the stack the value of the current stack pointer and the address of the called function (Program Counter), and optionally a timestamp of the function call. From these data, both the stack consumption as a function of time and a call tree of those functions can be determined, which led to this stack consumption.

From the data determined in the process, the stack consumption can be determined specifically for specific scenarios. For example, the maximum stack consumption of individual program parts or the entire program can be determined. Furthermore, the call tree can be determined at the time when the valid threshold value was reached by the program and the predefined function was executed.

Furthermore, a minimum, maximum and average stack consumption of a single operating system process, including the call tree, can be determined. This course is illustrated by way of example in FIG. 1, in which the stack usage SU is shown over the time t. Of the

Stack consumption of a first process ("process 1") is shown as a function of the call of its functions f1, f2, f3, f4 and as a function of interrupt events il, i2. Analogously, the stack consumption of a second process ( "Process 2"), which also runs in the embedded system. This results in a total stack consumption ("Total"), where From this, a maximum value (max) of the stack consumption, an With ^¬ average value (avg) and a minimum value (min) of the overall velvet stack usage "Total" are illustrated. As a result, the method of the invention takes into account real conditions, it is possible , to determine the stack usage of the Pro ^¬ program of the embedded system taking into account all existing in the real operating conditions, such as events, long-term behavior, etc., and to optimize the program or the embedded system if necessary.

Claims

claims

A method for optimizing the operation of a program in an embedded system, wherein the program to its expiration uses a stack of the embedded system, which is formed by a delimited address range in a memory of the embedded system and its size by an address range with a lower address and an upper address is defined in which

- a threshold value (SW) is defined for the stack memory, which is represented by an address of the stack, the address being smaller than the upper address of the address area of the stack;

monitoring whether the program running on the embedded system, using the stack, reaches the address represented by the threshold value (SW); upon reaching the threshold value (SW) a predefined function is called, by which the content of the stack is read out;

- It is determined from the contents of the stack, which part of the program has triggered the call of the predefined function; and

an optimization of the program and / or the embedded system is performed on the basis of the information about the calling part.

2. The method according to claim 1, characterized in that a call tree is determined from the contents of the stack, wherein from the call tree that part is determined which calls the predefined function.

3. The method of claim 1 or 2, characterized in that from the contents of the stack, the value of the current stack pointer and the address of the called function (Program Counter), and optionally a time stamp of Func ^¬ onsaufrufs be determined.

4. The method according to any one of the preceding claims, characterized in that after the call of the predefined function and the determination of the contents of the stack of the threshold value (SW) is increased, with a new address of the threshold moves closer to the upper address of the stack.

5. The method according to any one of the preceding claims, characterized in that the optimization comprises an enlargement of the stack by changing the upper or lower address.

6. The method according to any one of the preceding claims, characterized in that the optimization comprises a change of the nesting depth of a number of interdependent (program) parts by a compiler.

7. The method according to any one of the preceding claims, characterized in that the optimization comprises a change of the process in one or more parts of the program.