Basically what I'm trying to achieve here is having global variable with array of pointers to struct which size isn't known at compile time — in my example below it's my_struct **tab. In the final version I want to call a JNI method which will initialize my array of pointers and I want to keep them for some other methods usage.
Unfortunately I'm not a C programmer and I really struggle with this problem. Below I show what I tried to do; obviously, it's not working. Any constructive feedback would be really helpful.
(Sorry for missunderstanding with includes it's supposed to be a C code )
#include <stdio.h>
#include <stdlib.h>
typedef struct {
int tag;
} my_struct;
my_struct **tab;
void * get_pointer_to_struct() {
my_struct * s;
/* allocate memory */
if ((s = (my_struct *) malloc(sizeof (my_struct))) == NULL) {
return NULL;
}
return s;
}
void free_structures(int j) {
for (int a; a < j; a++) {
my_struct *s;
s = (my_struct *) tab[a];
/* free memory */
free(s);
tab[a] = NULL;
}
}
void init_pointers_array(int j) {
my_struct * temp_arr[j];
for (int i = 0; i < j; i++) {
temp_arr[i] = (my_struct *) get_pointer_to_struct();
temp_arr[i]->tag = i;
}
tab = temp_arr;
}
int main() {
//initialization
init_pointers_array(10);
//usage
for (int a = 0; a < 10; a++) {
if (tab[a]) {
my_struct * str_tmp = tab[a];
printf("Integer that you have entered is %d\n", str_tmp->tag);
}
}
//free mem
free_structures(10);
return 0;
}
This code is so unreadable I'm surprised anyone bothered to read it. Follow these guidelines and all your problems will be solved:
Use std::vector (or similar array class) instead of raw arrays
Don't use dynamic allocation if you don't need it, but if you do use new instead of malloc
Whenever you use dynamic allocation, do so within a class that owns the object and follows the RAII principle
Don't use global variables
my_struct * temp_arr[j];
then
tab = temp_arr;
is wrong. (Not only the placement of the * qualifier is horrible and there are superfluous casts that severely decrease code readibility, but) temp_array is a local auto array, so it will be deallocated when the function returns. Doing anything with its address afterwards results in undefined behavior. You may want to malloc()ate some memory for the struct instead (the casts are only there in order the code to be usable in C++. In C, it's strongly discouraged to make redundant typecasts):
my_struct **tab;
tab = (my_struct **)malloc(sizeof(tab[0]) * number_of_structs);
int i;
for (i = 0; i < number_of_structs; i++) {
tab[i] = (my_struct *)malloc(sizeof(tab[0][0]));
}
And for freeing it:
int i;
for (i = 0; i < number_of_structs; i++) {
free(tab[i]);
}
free(tab);
Several points:
get_pointer_to_struct can simply return the result of malloc. And change it's signature to avoid the extra casts to my_struct*.
temp_array is created on the stack, so it no longer exists when init_pointers_array exits. malloc it as well. This is your BIGGEST problem.
This is in general. You do alot of superfluous work, which bloats your code something terrible. Try to avoid redundant casts and unneeded variables.
Related
I need to:
a) define a data type (of my own choice) called mytype.
b) and have a function getmemory(int n), reserve memory for n values of type mytype and return a pointer. I have implemented this as shown below:
typedef int mytype;
// b)
mytype *getmemory(int n)
{
mytype *p;
//allocate space for n values of mytype.
p = (mytype *)malloc(n * sizeof(mytype));
return p;
}
Now I want to change the code to reserve and return memory for n pointers to mytype.
I'm not sure how to do that, I have searched a lot on the internet (also on StackOverflow), but I'm still not sure how to solve the task best. Below I have two possible solutions and would like to know which one is the right one.
Solutions:
1. Solution:
mytype **getmemoryForNPointer(int n_pointers, int n_mytype)
{
mytype **p = NULL;
p = (mytype **) malloc(n_pointers * sizeof(mytype *));
return p;
}
2. Solution
mytype **getmemoryForNPointer(int n_mytype, int n_pointers)
{
mytype **p = NULL;
//allocate n pointers to mytype first:
p = (mytype **) malloc(n_pointers * sizeof(mytype*));
//I'm not sure the for loop belongs in there.
for (int i = 0; i < n_pointers; i++)
{
// For each such pointer, allocate space for n mytype-values.
p[i] = (mytype *)malloc(n_mytype * sizeof(mytype));
}
return p;
}
is the following code equivalent to the second code?
mytype **getmemoryForNPointer(int n_mytype, int n_pointers)
{
mytype *p[n_pointers];
for (int i = 0; i < n_pointers; i++)
{
p[i] = (mytype *)malloc(n_mytype* sizeof(mytype));
}
return p;
}
Both your 'solutions' are correct, after a fashion (but see here: Do I cast the result of malloc?).
The difference is that the first allocates the array of pointers but then does nothing with those pointers (i.e. they will all contain undefined and unusable addresses). You could improve this version by using calloc in place of malloc, which would initialize all those pointers to NULL (so that each can then be tested, elsewhere in your code, before trying to dereference it).
The second solution also allocates memory for each block of data and assigns the addresses of those blocks to the pointers. In that sense, the second is more complete, but not necessarily more correct.
But, don't forget to free the data when you're done:
void freememoryForNPointer(mytype** ptr, int n_pointers)
{
for (int i = 0; i < n_pointers; i++) free(ptr[i]);
free(ptr);
}
this is how i declare this struct
typedef struct cache{
int vaild;
char* tag;
char* data;
}cache;
this is part of my main which called this function
struct cache **cacheA = createCache(Setnum,(int)pow(2,blocksize),cachesize);
struct cache **cacheB = createCache(Setnum,(int)pow(2,blocksize),cachesize);
and now this is my called function
struct cache ** createCache(int numset, int blocksize, int cachesize){
int numcache = (int)((cachesize/blocksize)*numset);
struct cache out[numset][numcache];
int i,j;
for (i=0; i < numset; i++){
for (j=0; j < numcache; j++){
out[i][j].tag = "0";
out[i][j].vaild = 0;
out[i][j].data ="0";
}
}
return out;
}
and when i try to compile this, it tells me that
return from incompatible pointer type
function returns address of local variable
(which points to the line "return out;")
I have no idea whats wrong with my code, i mean the type of the function return is the same as how i declear "out", so what causes this problem?
You create struct cache out[numset][numcache];
within the function prototyped as: struct cache ** createCache(...).
Then attempt to return out.
It is because struct cache [][] is typed differently than struct cache ** that you are getting the return errors.
Other comments:
1) If you truly do want a pointer to pointer to struct, then malloc or calloc will need to be used at some point to allocate memory.
2) the char * members of the struct also need to be assigned memory before assigning values. For illustration below, they are changed to char []
3) assigning values to strings does not work by using = assignment operator. Use a string function such as strcpy, sprintf, etc.
4) you've named the struct with the same symbol as that of the new type you have created, i.e. cache. In this application, the name cache is not necessary. Also, purely for style, I show the new type in CAPS. This is not necessary, but just a style I use to make the new type more recognizable in code.
In consideration of the comments above, the struct could be changed to the following:
typedef struct { /// don't need name here when it in this application
int vaild;
//char *tag;
char tag[20];//for illustration, to avoid additional dynamic allocation of memory
//char* data;
char data[80];
}CACHE;//capitalization is style only, not a necessity here.
Note, there is no name, but the new type CACHE was created. Now, you can create the function createCache:
CACHE ** createCache(int ncache, int nset)//note for simplicity of freeing this
//object later, simplify number of arguments
{
CACHE **out;
out = calloc(ncache, sizeof(CACHE *));//create array of pointers to CACHE
if(!out) return NULL;
int i;
for (i=0; i < nset; i++)
{
out[i] = calloc(nset, sizeof(CACHE));//create space for each instance
//of CACHE pointed to by array pointers
}
return out;
}
Anytime memory is created on the heap, it needs to be freed. This method will free the CACHE object memory:
void freeCashe(CACHE **a, int nset)
{
int i;
for(i=0; i<nset; i++)
{
if(a[i])free(a[i]);
}
if(a)free(a);
}
Calling these functions as shown below, will create an array of pointers, each pointing to an instance of CACHE where you can use them as intended, then free all of the memory when finished:
int main(void)
{
int cachesize = 20;
int blocksize = 20;
int numset = 10;
//move the calculation out of creation function
//to simplify freeing object later.
int numcache = (int)((cachesize/blocksize)*numset);
CACHE **a = createCache(numcache, numset);
/// use a, then free a
freeCashe(a, numset);
return 0;
}
Your function needs to allocate the memory on the heap rather than the stack. You will need to allocate space on the heap for your array of pointers, and for what they point too.
struct cache ** createCache(int numset, int blocksize, int cachesize){
cache ** out;
int numcache = (int)((cachesize/blocksize)*numset);
size_t headerSize = sizeof(*out)*numset;
size_t bodySize = sizeof(**out)*numcache;
out = malloc(headerSize + (bodySize*numset));
if (out == NULL) {
/* Should probably output some message about being
* insufficient memory here. */
return NULL;
}
int i,j;
for (i=0; i < numset; i++){
/* need to assign our point */
out[i] = (cache*)(((char*)out)+(headerSize+bodySize*i));
for (j=0; j < numcache; j++){
out[i][j].tag = "0";
out[i][j].vaild = 0;
out[i][j].data ="0";
}
}
return out;
}
/* importantly, you want a way to free your allocated memory */
void destroyCache(cache ** ptr) {
free(ptr);
}
PS: You don't have to typedef your struct if you reference it with the struct keyword.
You are wanting a pointer pointer type to be returned, but in order to do that you need to dynamically allocate it. Local stack allocations (i.e. struct cache[x][y]) won't work. You will either get an error or your program will crash when attempting to use the 2D array.
The solution is to either pre-allocate space and pass it in to the function or allocate in the function itself.
Allocation In Function Example:
struct cache ** createCache(int numset, int blocksize, int cachesize){
int numcache = (int)((cachesize/blocksize)*numset);
struct cache **out = malloc(sizeof(struct cache *) * numset); // This line changed.
int i,j;
for (i=0; i < numset; i++){
out[i] = malloc(sizeof(struct cache) * numcache); // This line added.
for (j=0; j < numcache; j++){
out[i][j].tag = malloc(sizeof(char)); // This line added.
out[i][j].data = malloc(sizeof(char)); // This line added.
strcpy(out[i][j].tag, "0");
out[i][j].vaild = 0;
strcpy(out[i][j].data, "0");
}
}
return out;
}
I know how to create an array of structs but with a predefined size. However is there a way to create a dynamic array of structs such that the array could get bigger?
For example:
typedef struct
{
char *str;
} words;
main()
{
words x[100]; // I do not want to use this, I want to dynamic increase the size of the array as data comes in.
}
Is this possible?
I've researched this: words* array = (words*)malloc(sizeof(words) * 100);
I want to get rid of the 100 and store the data as it comes in. Thus if 76 fields of data comes in, I want to store 76 and not 100. I'm assuming that I don't know how much data is coming into my program. In the struct I defined above I could create the first "index" as:
words* array = (words*)malloc(sizeof(words));
However I want to dynamically add elements to the array after. I hope I described the problem area clearly enough. The major challenge is to dynamically add a second field, at least that is the challenge for the moment.
I've made a little progress however:
typedef struct {
char *str;
} words;
// Allocate first string.
words x = (words) malloc(sizeof(words));
x[0].str = "john";
// Allocate second string.
x=(words*) realloc(x, sizeof(words));
x[1].FirstName = "bob";
// printf second string.
printf("%s", x[1].str); --> This is working, it's printing out bob.
free(x); // Free up memory.
printf("%s", x[1].str); --> Not working since its still printing out BOB even though I freed up memory. What is wrong?
I did some error checking and this is what I found. If after I free up memory for x I add the following:
x=NULL;
then if I try to print x I get an error which is what I want. So is it that the free function is not working, at least on my compiler? I'm using DevC??
Thanks, I understand now due to:
FirstName is a pointer to an array of char which is not being allocated by the malloc, only the pointer is being allocated and after you call free, it doesn't erase the memory, it just marks it as available on the heap to be over written later. – MattSmith
Update
I'm trying to modularize and put the creation of my array of structs in a function but nothing seems to work. I'm trying something very simple and I don't know what else to do. It's along the same lines as before, just another function, loaddata that is loading the data and outside the method I need to do some printing. How can I make it work? My code is as follows:
# include <stdio.h>
# include <stdlib.h>
# include <string.h>
# include <ctype.h>
typedef struct
{
char *str1;
char *str2;
} words;
void LoadData(words *, int *);
main()
{
words *x;
int num;
LoadData(&x, &num);
printf("%s %s", x[0].str1, x[0].str2);
printf("%s %s", x[1].str1, x[1].str2);
getch();
}//
void LoadData(words *x, int * num)
{
x = (words*) malloc(sizeof(words));
x[0].str1 = "johnnie\0";
x[0].str2 = "krapson\0";
x = (words*) realloc(x, sizeof(words)*2);
x[1].str1 = "bob\0";
x[1].str2 = "marley\0";
*num=*num+1;
}//
This simple test code is crashing and I have no idea why. Where is the bug?
You've tagged this as C++ as well as C.
If you're using C++ things are a lot easier. The standard template library has a template called vector which allows you to dynamically build up a list of objects.
#include <stdio.h>
#include <vector>
typedef std::vector<char*> words;
int main(int argc, char** argv) {
words myWords;
myWords.push_back("Hello");
myWords.push_back("World");
words::iterator iter;
for (iter = myWords.begin(); iter != myWords.end(); ++iter) {
printf("%s ", *iter);
}
return 0;
}
If you're using C things are a lot harder, yes malloc, realloc and free are the tools to help you. You might want to consider using a linked list data structure instead. These are generally easier to grow but don't facilitate random access as easily.
#include <stdio.h>
#include <stdlib.h>
typedef struct s_words {
char* str;
struct s_words* next;
} words;
words* create_words(char* word) {
words* newWords = malloc(sizeof(words));
if (NULL != newWords){
newWords->str = word;
newWords->next = NULL;
}
return newWords;
}
void delete_words(words* oldWords) {
if (NULL != oldWords->next) {
delete_words(oldWords->next);
}
free(oldWords);
}
words* add_word(words* wordList, char* word) {
words* newWords = create_words(word);
if (NULL != newWords) {
newWords->next = wordList;
}
return newWords;
}
int main(int argc, char** argv) {
words* myWords = create_words("Hello");
myWords = add_word(myWords, "World");
words* iter;
for (iter = myWords; NULL != iter; iter = iter->next) {
printf("%s ", iter->str);
}
delete_words(myWords);
return 0;
}
Yikes, sorry for the worlds longest answer. So WRT to the "don't want to use a linked list comment":
#include <stdio.h>
#include <stdlib.h>
typedef struct {
char** words;
size_t nWords;
size_t size;
size_t block_size;
} word_list;
word_list* create_word_list(size_t block_size) {
word_list* pWordList = malloc(sizeof(word_list));
if (NULL != pWordList) {
pWordList->nWords = 0;
pWordList->size = block_size;
pWordList->block_size = block_size;
pWordList->words = malloc(sizeof(char*)*block_size);
if (NULL == pWordList->words) {
free(pWordList);
return NULL;
}
}
return pWordList;
}
void delete_word_list(word_list* pWordList) {
free(pWordList->words);
free(pWordList);
}
int add_word_to_word_list(word_list* pWordList, char* word) {
size_t nWords = pWordList->nWords;
if (nWords >= pWordList->size) {
size_t newSize = pWordList->size + pWordList->block_size;
void* newWords = realloc(pWordList->words, sizeof(char*)*newSize);
if (NULL == newWords) {
return 0;
} else {
pWordList->size = newSize;
pWordList->words = (char**)newWords;
}
}
pWordList->words[nWords] = word;
++pWordList->nWords;
return 1;
}
char** word_list_start(word_list* pWordList) {
return pWordList->words;
}
char** word_list_end(word_list* pWordList) {
return &pWordList->words[pWordList->nWords];
}
int main(int argc, char** argv) {
word_list* myWords = create_word_list(2);
add_word_to_word_list(myWords, "Hello");
add_word_to_word_list(myWords, "World");
add_word_to_word_list(myWords, "Goodbye");
char** iter;
for (iter = word_list_start(myWords); iter != word_list_end(myWords); ++iter) {
printf("%s ", *iter);
}
delete_word_list(myWords);
return 0;
}
If you want to dynamically allocate arrays, you can use malloc from stdlib.h.
If you want to allocate an array of 100 elements using your words struct, try the following:
words* array = (words*)malloc(sizeof(words) * 100);
The size of the memory that you want to allocate is passed into malloc and then it will return a pointer of type void (void*). In most cases you'll probably want to cast it to the pointer type you desire, which in this case is words*.
The sizeof keyword is used here to find out the size of the words struct, then that size is multiplied by the number of elements you want to allocate.
Once you are done, be sure to use free() to free up the heap memory you used in order to prevent memory leaks:
free(array);
If you want to change the size of the allocated array, you can try to use realloc as others have mentioned, but keep in mind that if you do many reallocs you may end up fragmenting the memory. If you want to dynamically resize the array in order to keep a low memory footprint for your program, it may be better to not do too many reallocs.
This looks like an academic exercise which unfortunately makes it harder since you can't use C++. Basically you have to manage some of the overhead for the allocation and keep track how much memory has been allocated if you need to resize it later. This is where the C++ standard library shines.
For your example, the following code allocates the memory and later resizes it:
// initial size
int count = 100;
words *testWords = (words*) malloc(count * sizeof(words));
// resize the array
count = 76;
testWords = (words*) realloc(testWords, count* sizeof(words));
Keep in mind, in your example you are just allocating a pointer to a char and you still need to allocate the string itself and more importantly to free it at the end. So this code allocates 100 pointers to char and then resizes it to 76, but does not allocate the strings themselves.
I have a suspicion that you actually want to allocate the number of characters in a string which is very similar to the above, but change word to char.
EDIT: Also keep in mind it makes a lot of sense to create functions to perform common tasks and enforce consistency so you don't copy code everywhere. For example, you might have a) allocate the struct, b) assign values to the struct, and c) free the struct. So you might have:
// Allocate a words struct
words* CreateWords(int size);
// Assign a value
void AssignWord(word* dest, char* str);
// Clear a words structs (and possibly internal storage)
void FreeWords(words* w);
EDIT: As far as resizing the structs, it is identical to resizing the char array. However the difference is if you make the struct array bigger, you should probably initialize the new array items to NULL. Likewise, if you make the struct array smaller, you need to cleanup before removing the items -- that is free items that have been allocated (and only the allocated items) before you resize the struct array. This is the primary reason I suggested creating helper functions to help manage this.
// Resize words (must know original and new size if shrinking
// if you need to free internal storage first)
void ResizeWords(words* w, size_t oldsize, size_t newsize);
In C++, use a vector. It's like an array but you can easily add and remove elements and it will take care of allocating and deallocating memory for you.
I know the title of the question says C, but you tagged your question with C and C++...
Another option for you is a linked list. You'll need to analyze how your program will use the data structure, if you don't need random access it could be faster than reallocating.
Your code in the last update should not compile, much less run. You're passing &x to LoadData. &x has the type of **words, but LoadData expects words* . Of course it crashes when you call realloc on a pointer that's pointing into stack.
The way to fix it is to change LoadData to accept words** . Thi sway, you can actually modify the pointer in main(). For example, realloc call would look like
*x = (words*) realloc(*x, sizeof(words)*2);
It's the same principlae as in "num" being int* rather than int.
Besides this, you need to really figure out how the strings in words ere stored. Assigning a const string to char * (as in str2 = "marley\0") is permitted, but it's rarely the right solution, even in C.
Another point: non need to have "marley\0" unless you really need two 0s at the end of string. Compiler adds 0 tho the end of every string literal.
For the test code: if you want to modify a pointer in a function, you should pass a "pointer to pointer" to the function. Corrected code is as follows:
#include <stdio.h>
#include <stdlib.h>
#include <ctype.h>
typedef struct
{
char *str1;
char *str2;
} words;
void LoadData(words**, int*);
main()
{
words **x;
int num;
LoadData(x, &num);
printf("%s %s\n", (*x[0]).str1, (*x[0]).str2);
printf("%s %s\n", (*x[1]).str1, (*x[1]).str2);
}
void LoadData(words **x, int *num)
{
*x = (words*) malloc(sizeof(words));
(*x[0]).str1 = "johnnie\0";
(*x[0]).str2 = "krapson\0";
*x = (words*) realloc(*x, sizeof(words) * 2);
(*x[1]).str1 = "bob\0";
(*x[1]).str2 = "marley\0";
*num = *num + 1;
}
Every coder need to simplify their code to make it easily understood....even for beginners.
So array of structures using dynamically is easy, if you understand the concepts.
// Dynamically sized array of structures
#include <stdio.h>
#include <stdlib.h>
struct book
{
char name[20];
int p;
}; //Declaring book structure
int main ()
{
int n, i;
struct book *b; // Initializing pointer to a structure
scanf ("%d\n", &n);
b = (struct book *) calloc (n, sizeof (struct book)); //Creating memory for array of structures dynamically
for (i = 0; i < n; i++)
{
scanf ("%s %d\n", (b + i)->name, &(b + i)->p); //Getting values for array of structures (no error check)
}
for (i = 0; i < n; i++)
{
printf ("%s %d\t", (b + i)->name, (b + i)->p); //Printing values in array of structures
}
scanf ("%d\n", &n); //Get array size to re-allocate
b = (struct book *) realloc (b, n * sizeof (struct book)); //change the size of an array using realloc function
printf ("\n");
for (i = 0; i < n; i++)
{
printf ("%s %d\t", (b + i)->name, (b + i)->p); //Printing values in array of structures
}
return 0;
}
If you want to grow the array dynamically, you should use malloc() to dynamically allocate some fixed amount of memory, and then use realloc() whenever you run out. A common technique is to use an exponential growth function such that you allocate some small fixed amount and then make the array grow by duplicating the allocated amount.
Some example code would be:
size = 64; i = 0;
x = malloc(sizeof(words)*size); /* enough space for 64 words */
while (read_words()) {
if (++i > size) {
size *= 2;
x = realloc(sizeof(words) * size);
}
}
/* done with x */
free(x);
Here is how I would do it in C++
size_t size = 500;
char* dynamicAllocatedString = new char[ size ];
Use same principal for any struct or c++ class.
How to return 1000 variables from a function in C?
This is an interview question asked which I was unable to answer.
I guess with the help of pointers we can do that. I am new to pointers and C can anyone give me solution to solve this problem either using pointers or different approach?
Pack them all in a structure and return the structure.
struct YourStructure
{
int a1;
int b2;
int z1000;
};
YouStructure doSomething();
If it's 1000 times the same type (e.g. int's):
void myfunc(int** out){
int i = 0;
*out = malloc(1000*sizeof(int));
for(i = 0; i < 1000; i++){
(*out)[i] = i;
}
}
This function allocates memory for 1000 integers (an array of integers) and fills the array.
The function would be called that way:
int* outArr = 0;
myfunc(&outArr);
The memory held by outArr must be freed after use:
free(outArr);
See it running on ideone: http://ideone.com/u8NX5
Alternate solution: instead of having myfunc allocate the memory for the integer array, let the caller do the work and pass the array size into the function:
void myfunc2(int* out, int len){
int i = 0;
for(i = 0; i < len; i++){
out[i] = i;
}
}
Then, it's called that way:
int* outArr = malloc(1000*sizeof(int));
myfunc2(outArr, 1000);
Again, the memory of outArr must be freed by the caller.
Third approach: static memory. Call myfunc2 with static memory:
int outArr[1000];
myfunc2(outArr, 1000);
In that case, no memory has to be allocated or freed.
Array Pointer approach:
int * output(int input)
{
int *temp=malloc(sizeof(int)*1000);
// do your work with 1000 integers
//...
//...
//...
//ok. finished work with these integers
return temp;
}
Struct pointer approach:
struct my_struct
{
int a;
int b;
double x;
...
//1000 different things here
struct another_struct;
}parameter;
my_struct * output(my_struct what_ever_input_is)
{
my_struct *temp=malloc(sizeof(my_struct));
//...
//...
return temp;
}
This is how you do it in C.
void func (Type* ptr);
/*
Function documentation.
Bla bla bla...
Parameters
ptr Points to a variable of 'Type' allocated by the caller.
It will contain the result of...
*/
If your intention wasn't to return anything through "ptr", you would have written
void func (const Type* ptr);
instead.
This question already has answers here:
Closed 11 years ago.
Possible Duplicate:
Simple C implementation to track memory malloc/free?
I need to know how much memory I have used till now in a C program and here is the pseudo code
#include <stdio.h>
int usedMemory =0;
void *MyMalloc(int size){
usedMemory = usedMemory +size ;
return malloc(size);
}
void MyFree(void *pointer){
/*****************what should i write here????*************/
}
int main(int argc, char *argv[])
{
char *temp1= (char *)MyMalloc(100);
char *temp2= (char *)MyMalloc(100);
/*......other operations.........*/
MyFree(temp1);
MyFree(temp2);
return 0;
}
Can anyone tell me what to write in the MyFree method(which decrements the amount of memory freed from usedMemory.
You could allocate few extra bytes more than asked, and store the size in the extra bytes, so that you could know the size later on, in MyFree function, with little calculation as:
unsigned long int usedMemory = 0;
void *MyMalloc(int size)
{
char *buffer = (char *) malloc(size + sizeof(int)); //allocate sizeof(int) extra bytes
if ( buffer == NULL)
return NULL; // no memory!
usedMemory += size ;
int *sizeBox = (int*)buffer;
*sizeBox = size; //store the size in first sizeof(int) bytes!
return buffer + sizeof(int); //return buffer after sizeof(int) bytes!
}
void MyFree(void *pointer)
{
if (pointer == NULL)
return; //no free
char *buffer = (char*)pointer - sizeof(int); //get the start of the buffer
int *sizeBox = (int*)buffer;
usedMemory -= *sizeBox;
free(buffer);
}
In C++, you could keep a global std::map<void*, std::size_t> around to track the size of each allocated block; your own allocator function would register the size when allocating, and the deallocation function would remove the entry. (Update: Or do as the linked question suggests and allocate a bit more memory and save the size there.)
The more fundamental problem is that this will probably only be of very limited use in a typical C++ program: Allocations there are done predominantly in two ways: 1) through explicit new expressions, which call ::operator new(), which in turn (usually) calls malloc(), and 2) through std::allocator<T>::allocate(), which on many platforms is implemented in terms of ::operator new().
The problem is that you don't have control over the specifics of your platform. You can replace the global operator-new to use your own MyMalloc(), but the default std::allocator might use malloc() directly and thus not be affected by that.
A cleaner approach for debugging purposes is to use an external tool like valgrind to track heap usage. For permanent internal use, tracking the allocation sizes is going to cause a significant performance hit, too.
You could allocate memory and store its size in the allocated block (error checking omitted for brevity):
unsigned int totalAlloc = 0;
void *MyAlloc(unsigned int size)
{
void *p;
totalAlloc += size;
p = malloc(size + sizeof(int));
*(int *) p = size;
return (void *)(((int *) p) + 1)
}
void MyFree(void *ptr)
{
ptr = (void *)(((int *) ptr) -1 );
totalAlloc -= * (int *) ptr;
free(ptr);
}
This code actually reserves more memory than requested in order to store the block's size in the (usually) first four bytes. This information can then be retrieved later on when you free the memory.
You need to manage a list of all malloc() you have done with pointer + size. Then you can search for the size in that list, and decrement it in free().
Check for example in that example how they are doing:
http://developers.sun.com/solaris/articles/lib_interposers_code.html#malloc_interposer.c
You might have other possibilities to track memory, like:
Valgrind with massif tool for tracking memory usage over time. You can even generate png output graphics
Interposed libraries. You can found some libraries that you can use by LD_PRELOAD=thelib.so ./yourprogram, and they will output some statistics like jemalloc
(A side note, please accept some answers to your question !)
you could try something like this... i'd strongly recommend to use this only for debugging purpose!
#define MAXMEMBLOCKS 10000
typedef struct {
int size;
void* ptr;
} memblock;
typedef struct {
int totalSize;
int current;
memblock memblocks[MAXMEMBLOCKS];
} currentMemory;
currentMemory mem;
void *MyMalloc(int size) {
if (mem.current < MAXMEMBLOCKS) {
mem.current += size;
mem.memblocks[mem.current].size = size;
mem.memblocks[mem.current].ptr = malloc(size);
return mem.memblocks[mem.current++].ptr;
} else {
// you needed more memblocks than estimated
return NULL;
}
};
int MyFree(void *pointer) {
int i;
for (i = 0; i < mem.current; i++) {
if (mem.memblocks[i].ptr == pointer) {
mem.totalSize -= mem.memblocks[i].size;
free(mem.memblocks[i].ptr);
mem.current--;
return 0;
}
}
// you tried to free a block wich hasn't been allocated through MyMalloc()
return -1;
}