C ++ Memory Management

The great Bill Gates once said:

640 K ought to be enough for everybody-Bill Gates 1981

ProgramDevelopers often write memory management programs, and are often worried. If you don't want to touch the mines, the only solution is to discover all the hidden mines and exclude them. The content of this article is much deeper than that of general textbooks. Readers must carefully read this article to truly understand memory management.

1. Memory Allocation Method

There are three memory allocation methods:

(1) distribution from the static storage area. The program has been allocated when it is compiled, and the program exists throughout the entire runtime. For example, global variables and static variables.

(2) create a stack. When a function is executed, the storage units of local variables in the function can be created on the stack. When the function is executed, these storage units are automatically released. Stack memory allocation computation is built into the processor's instruction set, which is highly efficient, but the memory capacity allocated is limited.

(3) allocate from the stack, also known as dynamic memory allocation. When the program runs, it uses malloc or new to apply for any amount of memory. The programmer is responsible for releasing the memory with free or delete. The lifetime of the dynamic memory is determined by us. It is very flexible to use, but the problem is also the most.

2. Common memory errors and Countermeasures

Memory Errors are very troublesome. The compiler cannot automatically detect these errors, which can be captured only when the program is running. Most of these errors do not have obvious symptoms, but they are often invisible and increase the difficulty of error correction. Sometimes the user finds you angrily, but the program has not encountered any problems. When you leave, the error occurs again. Common memory errors and their countermeasures are as follows:

* If the memory allocation is unsuccessful, it is used.

New programmers often make this mistake because they do not realize that memory allocation will fail. A common solution is to check whether the pointer is null before using the memory. If the pointer P is a function parameter, use assert (P! = NULL)

Check. If you use malloc or new to apply for memory, you should use if (P = NULL) or if (P! = NULL.

* Although the memory allocation is successful, it is referenced before initialization.

There are two main causes for this mistake: first, there is no idea of initialization; second, the default initial values of the memory are all zero, resulting in incorrect reference values (such as arrays ). There is no uniform standard for the default initial values of the memory. Although sometimes it is zero, we prefer to trust it without any trust. Therefore, no matter which method is used to create an array, do not forget to assign the initial value. Even the zero value cannot be omitted, so do not bother.

* The memory allocation is successful and initialized, but the operation is beyond the memory boundary.

For example, when an array is used, the subscript "more than 1" or "less than 1" is often performed. Especially in for loop statements, the number of loops is easy to make a mistake, resulting in array operations out of bounds.

* Forgot to release the memory, causing memory leakage.

A function containing such errors loses a piece of memory every time it is called. At the beginning, the system had sufficient memory and you could not see the error. Once a program suddenly died, the system prompts: memory is exhausted.

Dynamic Memory application and release must be paired. The usage of malloc and free in the program must be the same, otherwise there must be an error (the same applies to new/delete ).

* The memory is released but it is used again.

There are three scenarios:

(1) The object calling relationship in the program is too complex, so it is difficult to figure out whether an object has released the memory. At this time, we should re-design the data structure to fundamentally solve the chaos of Object Management.

(2) The Return Statement of the function is incorrect. Be sure not to return the "Pointer" or "Reference" pointing to "stack memory" because the function body is automatically destroyed when it ends.

(3) After the memory is released using free or delete, the pointer is not set to null. As a result, a "wild pointer" is generated ".

[Rule 1] after applying for memory with malloc or new, check whether the pointer value is null immediately. Prevents the use of memory with NULL pointer values.

Rule 2: Do not forget to assign initial values to arrays and dynamic memory. Avoid using uninitialized memory as the right value.

Rule 3: avoid overrunning the subscript of an array or pointer. Be careful when "more than 1" or "less than 1" is performed.

[Rule 4] dynamic memory application and release must be paired to prevent memory leakage.

[Rule 5] after the memory is released with free or delete, the pointer is immediately set to null to avoid "wild pointer ".

3. Comparison of pointers and Arrays

In C ++/C Programs, pointers and arrays can be replaced with each other in many places, which leads to the illusion that the two are equivalent.

An array is either created in a static storage area (such as a global array) or on a stack. The array name corresponds to (rather than pointing to) a piece of memory, and its address and capacity remain unchanged during the lifetime, only the content of the array can be changed.

A pointer can point to any type of memory block at any time, and its feature is "variable". Therefore, we often use pointers to operate dynamic memory. Pointers are far more flexible than arrays, but they are more dangerous.

The following uses a string as an example to compare the features of pointers and arrays.

3.1 modify content

In the example 3-1, the size of character array a is 6 characters, and its content is hello. The content of A can be changed, for example, a [0] = 'x '. The pointer P points to the constant string "world" (in the static storage area with the content of World). The content of the constant string cannot be modified. In terms of syntax, the compiler does not think that the statement P [0] = 'X' is inappropriate, but this statement attempts to modify the content of the constant string and causes a running error.

Char A [] = "hello "; A [0] = 'X '; Cout <A <Endl; Char * P = "world"; // note that P points to a constant string P [0] = 'X'; // the compiler cannot find this error. Cout <p <Endl;

Example 3.1 modify the array and pointer content

3.2 content replication and Comparison

The array name cannot be directly copied or compared. In Example 7-3-2, if you want to copy the content of array a to array B, you cannot use statement B = A. Otherwise, a compilation error is generated. Use the standard library function strcpy for replication. Similarly, if the content of B and A is the same, it cannot be determined by if (B = A). The standard library function strcmp should be used for comparison.

Statement P = A does not copy the content of A, but assigns the address of A to P. To copy the content of A, use the library function malloc as P to apply for a memory with a capacity of strlen (A) + 1 characters, and then use strcpy to copy strings. Similarly, the statement if (P = A) compares not the content but the address, and should be compared using the database function strcmp.

// Array... Char A [] = "hello "; Char B [10]; Strcpy (B, A); // B = A cannot be used; If (strcmp (B, A) = 0) // If (B = A) cannot be used) ... // Pointer... Int Len = strlen (); Char * P = (char *) malloc (sizeof (char) * (LEN + 1 )); Strcpy (P, A); // do not use P =; If (strcmp (P, A) = 0) // do not use if (P =) ...

Example 3.2 copying and comparing the array and pointer content

3.3 computing memory capacity

The sizeof operator can be used to calculate the array capacity (number of bytes ). In Example 7-3-3 (a), the value of sizeof (a) is 12 (don't forget ''). The pointer P points to A, but the value of sizeof (P) is 4. This is because sizeof (p) obtains the number of bytes of a pointer variable, which is equivalent to sizeof (char *) rather than the memory capacity referred to by P. C ++/C language cannot know the memory capacity referred to by the pointer unless you remember it when applying for memory.

Note: When an array is passed as a function parameter, the array will automatically degrade to a pointer of the same type. In Example 7-3-3 (B), sizeof (a) is always equal to sizeof (char *) regardless of the size of array *).

Char A [] = "Hello World "; Char * P =; Cout <sizeof (a) <Endl; // 12 bytes Cout <sizeof (p) <Endl; // 4 bytes

Example 3.3 (a) calculates the memory capacity of arrays and pointers

Void func (char a [1, 100]) { Cout <sizeof (a) <Endl; // 4 bytes instead of 100 bytes }

Example 3.3 (B) the array degrades to a pointer

4. How does the pointer Parameter Pass the memory?

If the function parameter is a pointer, do not expect this pointer to apply for dynamic memory. In Example 7-4-1, the getmemory (STR, 200) Statement of the test function does not enable STR to obtain the expected memory. STR is still null. Why?

Void getmemory (char * P, int num) { P = (char *) malloc (sizeof (char) * num ); } Void test (void) { Char * STR = NULL; Getmemory (STR, 100); // STR is still null Strcpy (STR, "hello"); // running error }

Example 4.1 try to apply for dynamic memory with pointer Parameters

The fault lies in the getmemory function. The compiler always needs to make a temporary copy for each parameter of the function. The copy of the pointer parameter P is _ p, and the compiler makes _ p = P. If the program in the function body modifies the content of _ p, the content of parameter P is modified accordingly. This is why pointers can be used as output parameters. In this example, _ P applied for a new memory, but changed the memory address indicated by _ p, but P was not changed at all. Therefore, the getmemory function cannot output anything. In fact, each execution of getmemory will leak a piece of memory, because the memory is not released with free.

If you have to use the pointer parameter to request memory, you should use "pointer to Pointer" instead. See example 4.2.

Void getmemory2 (char ** P, int num) { * P = (char *) malloc (sizeof (char) * num ); } Void Test2 (void) { Char * STR = NULL; Getmemory2 (& STR, 100); // note that the parameter is & STR, not Str Strcpy (STR, "hello "); Cout <STR <Endl; Free (STR ); }

Example 4.2 apply for dynamic memory with a pointer pointing to the pointer

Because the concept of "pointer to Pointer" is not easy to understand, we can use function return values to transmit dynamic memory. This method is simpler. See example 4.3.

Char * getmemory3 (INT num) { Char * P = (char *) malloc (sizeof (char) * num ); Return P; } Void test3 (void) { Char * STR = NULL; STR = getmemory3 (100 ); Strcpy (STR, "hello "); Cout <STR <Endl; Free (STR ); }

Example 4.3 use function return values to transmit dynamic memory

Although it is easy to use the function return value to pass dynamic memory, some people often use the return statement wrong. It is emphasized that the return statement should not be used to return the pointer pointing to the "stack memory" because the function exists in it automatically disappears at the end, as shown in example 4.4.

Char * getstring (void) { Char P [] = "Hello World "; Return P; // the compiler will give a warning } Void test4 (void) { Char * STR = NULL; STR = getstring (); // STR content is junk Cout <STR <Endl; }

Example 4.4 Return Statement returns a pointer to "stack memory"

Use the debugger to track test4 step by step. After executing the STR = getstring statement, STR is no longer a null pointer, but the STR content is not "Hello World" but garbage.
What if I rewrite example 4.4 to example 4.5?

Char * getstring2 (void) { Char * P = "Hello World "; Return P; } Void test5 (void) { Char * STR = NULL; STR = getstring2 (); Cout <STR <Endl; }

Example 4.5 Return Statement returns a constant string

Although the function test5 runs without errors, the design concept of the function getstring2 is incorrect. Because "Hello World" in getstring2 is a constant string located in the static storage zone, it remains unchanged during the lifetime of the program. No matter when getstring2 is called, it returns the same read-only memory block.

5. Eliminate "wild pointer"

The "wild pointer" is not a null pointer, but a pointer to the "junk" memory. Generally, null pointers are not incorrectly used, because if statements are easy to judge. However, the "wild Pointer" is very dangerous, and the IF statement does not work for it. There are two main causes of "wild pointer:

(1) pointer variables are not initialized. When a pointer variable is created, it does not automatically become a null pointer. Its default value is random, which means it is random. Therefore, the pointer variable should be initialized at the same time when it is created, either set the pointer to null or set it to direct to the legal memory. For example

Char * P = NULL; Char * STR = (char *) malloc (100 );

(2) After the pointer P is free or deleted, It is not set to null, which makes people mistakenly think P is a valid pointer.

(3) pointer operations go beyond the scope of variables. This situation is hard to prevent. The example program is as follows:

Class { Public: Void func (void) {cout <"func of Class A" <Endl ;} }; Void test (void) { A * P; { A; P = & A; // note the life cycle of } P-> func (); // P is a "wild pointer" }

When the function test executes the statement p-> func (), object A has disappeared, and P points to a, so P becomes a "wild pointer ". But the strange thing is that I did not encounter any errors when running this program, which may be related to the compiler.

6. Why New/delete?

Malloc and free are standard library functions in C ++/C, and new/delete are operators in C ++. They can be used to apply for dynamic memory and release memory.

For non-Internal data objects, maloc/free alone cannot meet the requirements of dynamic objects. The constructor must be automatically executed when the object is created, and the Destructor must be automatically executed before the object is extinct. Since malloc/free is a library function rather than an operator and is not controlled by the compiler, it is impossible to impose the tasks of executing constructor and destructor on malloc/free.

Therefore, the C ++ language requires a new operator that can complete dynamic memory allocation and initialization, and a delete operator that can clean up and release memory. Note that new/delete is not a database function. First, let's take a look at how malloc/free and new/delete implement dynamic memory management of objects. See example 6.

Class OBJ { Public: OBJ (void) {cout <"initialization" <Endl ;} ~ OBJ (void) {cout <"Destroy" <Endl ;} Void initialize (void) {cout <"initialization" <Endl ;} Void destroy (void) {cout <"Destroy" <Endl ;} }; Void usemallocfree (void) { OBJ * A = (OBJ *) malloc (sizeof (OBJ); // apply for dynamic memory A-> initialize (); // Initialization //... A-> destroy (); // clear the job Free (a); // releases the memory. } Void usenewdelete (void) { OBJ * A = new OBJ; // apply for dynamic memory and initialize //... Delete A; // clear and release the memory }

Example 6 How to Use malloc/free and new/Delete to manage the dynamic memory of Objects

The class OBJ function initialize simulates the constructor function, and the function destroy simulates the destructor function. In usemallocfree, because malloc/free cannot execute constructor and destructor, you must call the member functions initialize and destroy to complete initialization and clearing. The usenewdelete function is much simpler.

Therefore, we should not attempt to use malloc/free to manage the memory of dynamic objects. We should use new/Delete. Because the internal data type "object" does not have a process of construction and analysis, malloc/free and new/delete are equivalent to them.

Since the new/delete function completely covers malloc/free, why does C ++ not eliminate malloc/free? This is because C ++ programs often call C functions, and C Programs can only use malloc/free to manage dynamic memory.

If you use free to release the "New Dynamic Object", this object may cause program errors because it cannot execute the destructor. If you use Delete to release the "dynamic memory applied by malloc", theoretically, the program will not go wrong, but the program is poorly readable. Therefore, new/delete must be paired, and the same applies to malloc/free.

7. What should I do if the memory is exhausted?

If a large enough memory block cannot be found when applying for dynamic memory, malloc and new will return a null pointer, declaring that the memory application failed. There are usually three ways to handle the "memory depletion" problem.

(1) judge whether the pointer is null. If yes, use the return statement to terminate the function immediately. For example:

Void func (void) { A * A = new; If (A = NULL) { Return; } ... }

(2) judge whether the pointer is null. If yes, use exit (1) to terminate the entire program. For example:

Void func (void) { A * A = new; If (A = NULL) { Cout <"Memory Exhausted" <Endl; Exit (1 ); } ... }

(3) set exception handling functions for new and malloc. For example, in Visual C ++, you can use the _ set_new_hander function to set your own exception handling function for new, or enable malloc to use the same exception handling function as new. For more information, see the C ++ user manual.

The above (1) (2) method is the most common. If a function needs to apply for dynamic memory in multiple places, the method (1) is insufficient (it is troublesome to release the memory) and should be handled in the way (2.

A lot of people cannot bear to use exit (1). They asked, "Can the operating system solve the problem by itself without writing error handling programs ?"

No. In case of a "memory depletion" event, generally applications are no longer saved. If you do not use exit (1) to kill the program, it may kill the operating system. The truth is: if you do not kill a gangster, the gangster will commit more crimes before he dies.

There is a very important phenomenon to tell you. For 32-bit applications, no matter how malloc and new are used, it is almost impossible to cause "memory depletion ". In Windows 98, I wrote a test program using Visual C ++. See example 7. This program will run endlessly and will not be terminated at all. Because the 32-bit operating system supports "virtual storage" and the memory is used up, the hard disk space is automatically replaced. I only heard the sound of the hard drive. Window 98 was so tired that it didn't respond to the keyboard or mouse.

I can conclude that for 32-bit or more applications, the "memory used up" error handler is useless. Now, UNIX and Windows programmers are happy with this: the error handler does not work, and I will not write it, saving a lot of trouble.

I don't want to mislead readers. I must emphasize that without error handling, the quality of the program will be poor, and never be compromised.

Void main (void) { Float * P = NULL; While (true) { P = new float [1000000]; Cout <"Eat memory" <Endl; If (P = NULL) Exit (1 ); } }

Example 7 try to exhaust the operating system memory

8. Usage of malloc/free

The following is a prototype of the malloc function:

Void * malloc (size_t size );

Use malloc to apply for an integer-type memory with a length. The program is as follows:

Int * P = (int *) malloc (sizeof (INT) * length );

We should focus on two elements: "type conversion" and "sizeof ".

* The type returned by malloc is void *. Therefore, you must explicitly convert the type when calling malloc and convert void * to the required pointer type.

* The malloc function does not recognize the type of memory to be applied for. It only cares about the total number of bytes in the memory. We usually cannot remember the exact number of bytes for int, float, and other data types. For example, the int variable is 2 bytes in a 16-bit system and 4 bytes in a 32-bit system, and the float variable is 4 bytes in a 16-bit system, it is also 4 bytes in 32 bits. It is best to use the following program for a test:

Cout <sizeof (char) <Endl; Cout <sizeof (INT) <Endl; Cout <sizeof (unsigned INT) <Endl; Cout <sizeof (long) <Endl; Cout <sizeof (unsigned long) <Endl; Cout <sizeof (float) <Endl; Cout <sizeof (double) <Endl; Cout <sizeof (void *) <Endl;

Using the sizeof operator in malloc's "()" is a good style, but be careful when we sometimes get dizzy and write P = malloc (sizeof (p )) such a program.

* The prototype of function free is as follows:

Void free (void * memblock );

Why isn't the free function as complicated as the malloc function? This is because the pointer p type and the memory capacity it refers to are known in advance, and the statement free (p) can correctly release the memory. If P is a null pointer, no matter how many times the free P operation will fail. If P is not a null pointer, the free operation on P will cause a program running error.

9. Usage of new/delete

The new operator is much easier to use than the malloc function, for example:

Int * P1 = (int *) malloc (sizeof (INT) * length ); Int * P2 = new int [length];

This is because new has built-in sizeof, type conversion, and type security check functions. For non-Internal data objects, new completes initialization while creating dynamic objects. If an object has multiple constructors, the new statement can also have multiple forms. For example

Class OBJ { Public: OBJ (void); // a constructor without Parameters OBJ (int x); // constructor with a parameter ... } Void test (void) { OBJ * A = new OBJ; OBJ * B = new OBJ (1); // The initial value is 1. ... Delete; Delete B; }

If you use new to create an object array, you can only use the non-parameter constructor of the object. For example

OBJ * objects = new OBJ [100]; // create 100 Dynamic Objects

Cannot be written

OBJ * objects = new OBJ [100] (1); // create 100 dynamic objects and assign initial value 1

When releasing an object Array Using Delete, do not lose the symbol '[]'. For example

Delete [] objects; // correct usage Delete objects; // incorrect usage

The latter is equivalent to delete objects [0], and 99 Other objects are missing.

10. Some experiences

I know many well-developed C ++/C programmers. Few people can pat their chests and say they are familiar with pointer and memory management (including myself ). When I first learned the C language, I was especially afraid of pointers, which led to the development of the first application software (about 10 thousand lines of CCode), Instead of using a pointer, it is too stupid to use arrays to replace the pointer. It was not a solution to avoid pointers. Later I changed the software and reduced the amount of code to half of the original one.

My lessons are:

(1) The more you are afraid of pointers, the more you need to use pointers. If pointer is not used correctly, it is definitely not a qualified programmer.

(2) You must develop the habit of gradually tracing programs using the debugger. Only in this way can you discover the essence of the problem.