passing vector in multi threaded application

[MJDB]

Hello there

I am having difficulties figuring out how to pass a vector from a thread to the thread function. Its an int vector and this is how I call the thread function for it:

hThread = (HANDLE) _beginthreadex(NULL, 0, iterative_mergesort, (void *)&Vector1, 0, 0);

and the thread function is as follow:

unsigned __stdcall iterative_mergesort(void *v2) {
		

	int increment;
	int left;
	int left_end;
	int right;
	int right_end;
	int current;
	int i;

	Vector tmp(v2.size());
	int size = v2.size();

	increment = 1;
         .
         .
         .

But it does not work.

I really appreciate any help in advance

[Zhuge]

MJDB wrote:

But it does not work.

That is so incredibly vague it isn't even funny.

I think you probably want to cast v2 to a Vector*, among other things.

[MJDB]

I have tried casting it as well...

hThread = (HANDLE) _beginthreadex(NULL, 0, iterative_mergesort, (void *)&Vector1, 0, 0);

unsigned __stdcall iterative_mergesort(void (vector<int>* v2)) {

but still...I am a beginner in multi threading. please bare with me.

thanks alot

[Moooce]

I think Zhuge is refering to the cast being here with your original code:
Vector tmp(v2.size());

[MJDB]

I am using VC++. It tells me that: "v2 is undefined".

Is it possible that the problem exist when passing the vector from thread to iterative_mergesort() function?... since I am sending it as void?

thanks in advance

[Moooce]

I'm not at a machimne with a compiler on to test this, but I don't think it's the passing to the function that is the problem, it's just how you're casting it in the sorting function, there's a capital V in your (and my copy+paste) vector tmp=that shouldn't be there.

Try simply setting size to the value obtaind from the cast void pointer without using a tmp, I think this would work?
int size = ((vector<int>*)v2)->size();

[MJDB]

I declared Vector as vector<int>:

typedef vector<int> Vector;	

Then I just used in the function:

Vector Vector1(tl2.size());

And from that function, I am sending it to a thread:

hThread = (HANDLE) _beginthreadex(NULL, 0, iterative_mergesort, (void *)&Vector1, 0, 0);

This is the whole iterative_mergesort function:

// implementing the mergesort iteratively (bottom up mergesort)
unsigned __stdcall iterative_mergesort(void* v2) {
		


	int increment;
	int left;
	int left_end;
	int right;
	int right_end;
	int current;
	int i;


	int size = ((vector<int>*)v2)->size();
	// int size = v2.size();
	vector<int> tmp(size);
	

	increment = 1;


	while (increment < size) {

		// I don't need to go over 1 for increment...so I can stop the last pass..
		if(increment == 2){ 
			return;}

		left = 0;
		right = increment;
		left_end = right - 1;
		right_end = (left_end + increment < size) ? left_end + increment : size - 1;
		current = 0;

		 while (current < size) {
			  while (left <= left_end && right <= right_end) {
					if (v2[right] < v2[left])
						  tmp[current] = v2[right++];
					else
						  tmp[current] = v2[left++];
						  current++;
				}
      
				while (right <= right_end)
						tmp[current++] = v2[right++];

				while (left <= left_end)
						tmp[current++] = v2[left++];
      
		 left = right;
		 right += increment;
		 left_end = (right - 1 < size) ? right - 1: size - 1;
		 right_end = (left_end + increment < size) ? left_end + increment : size - 1;
		}
    
		 increment *= 2;


		for (i = 0; i < size; i++)
			v2[i] = tmp[i];


		
		another_function(v2);

	 }
	
	tmp.empty();
}

you are right Moooce. The problem is not with the parameter passing because the casting works. Does it make a difference if I cast it as ((vector<int>*)v2) instead of ((Vector*)v2)?


Do I need to cast all my v2 in the iterative_mergesort() as well?

I really appreciate your help

regards

[Moooce]

I'm afraid I can't compile all that code in my head so you'll have to bare with me if I get something wrong here ;-)

Also I'm no STL guru so maybe somebody else could point you in a more efficient direction, but as far as I know, like you say, you'll need to cast each v2.

The other way may be to wrap your vector in a struct and pass that as you would a normal inbuilt type without having to cast each time, if that fits your needs better?

e.g.

struct mydata
{
	vector<int> Vector1;
	int somethingelse;
};

and in your sort function:

	mydata *tmp = (mydata*)v2;
	int size=tmp->Vector1.size();  // << here onwards you can use the vector members accessed from tmp. 

and in your calling function (probably main())

mydata myvector;
hThread = (HANDLE) _beginthreadex(NULL, 0, iterative_mergesort, (void *)&myvector, 0, 0);

But like I said, there may be better ways, If anyone knows, there will be at least 2 people who learn something from it.

[coder777]

MJDB wrote:

Then I just used in the function:


Vector Vector1(tl2.size());

'Vector1' is a local variable? Then your program will crash. except you're waiting for the end of the thread but then why bother using a thread.

MJDB wrote:

Does it make a difference if I cast it as ((vector<int>*)v2) instead of ((Vector*)v2)?

No.

MJDB wrote:

Do I need to cast all my v2 in the iterative_mergesort() as well?

Yes. Plus derefence it in order to use []. example: (*((Vector*)v2))[right]

or you use v3 instead like Vector &v3 = *((Vector*)v2);