inserting into vectors c++

Question

I need to push 66,000 vectors (the number of vectors is not fixed, it can be 90,000 vectors also. For the sake of brevity I am showing the below code with the example of 66,000 vectors) of type vector into the following vector:

vector<int> vec;

The size of each of the 66,000 vectors is 9,000 elements. I am using the following for doing the same:

vec.reserve(66000*9000);
for(int j=0;j<66000;j++)
    for(int i=0;i<9000;i++) //9000 elements in vec1[i] per vector is not fixed
        vec.push_back(vec1[i]); //i am pushing i as an example

Is there someway by which I may increase the efficiency of this code?

I need to concatenate too many vectors, so a solution for the same is potentially different from concatenating two vectors. Also I can not use multi-threading as mentioned in the previous question

if the size is fixed, why not use an array / `std::array` instead? — Nikola Dimitroff, Jul 13 '15 at 16:22
@NikolaDimitroff Because it is big. Of course one could dynamically allocate an `std::array`, but what would that help? — Baum mit Augen, Jul 13 '15 at 16:23
possible duplicate of [Best way to concatenate two vectors?](http://stackoverflow.com/questions/3177241/best-way-to-concatenate-two-vectors) — NathanOliver, Jul 13 '15 at 16:26

score 6 · Accepted Answer · edited Jul 13 '15 at 16:28

6

Try the following

std::vector<int> vec;
vec.reserve( 66000*other_vec.size() );

for ( size_t i = 0; i < 66000; i++ ) 
{
    vec.insert( vec.end(), other_vec.begin(), other_vec.end() );
}

edited Jul 13 '15 at 16:28

Galimov Albert

7,269
1
24
50

answered Jul 13 '15 at 16:26

Vlad from Moscow

301,070
26
186
335

is insert faster then push_back? If yes, what is the reason for the same – Steg Verner Jul 13 '15 at 16:27
@Steg Verner I think that insert is faster because at least general calculations are made only once. – Vlad from Moscow Jul 13 '15 at 16:30
@StegVerner interval insertion is known to be faster. Check Scott Meyers "Effective STL" book. Good reading, though – Galimov Albert Jul 13 '15 at 16:40

Satish Chalasani · Answer 2 · 2015-07-13T16:47:01.543

1

You may use resize() instead of reserve(). Remove push_back when resize() is used. Memory is allocated with resize() and initializes. reserve just allocates but doesn't initialize.

edited Jul 13 '15 at 16:47

answered Jul 13 '15 at 16:22

Satish Chalasani

167
7

1

He must remove `push_back` then. Also i think this will give too small benefit. – Galimov Albert Jul 13 '15 at 16:23
1

@PSIAlt _"Also i think this will give too small benefit."_ Not with many small reallocations. – πάντα ῥεῖ Jul 13 '15 at 16:25
1

@πάνταῥεῖ What reallocation? He already reserves enough memory. – Baum mit Augen Jul 13 '15 at 16:25
I do not see how this is supposed to help. – Baum mit Augen Jul 13 '15 at 16:26
I have to down vote as using `resize()` will break using `push_back()` – NathanOliver Jul 13 '15 at 16:28

FooTheBar · Answer 3 · 2015-07-13T16:46:38.660

0

You can go to the underlying data directly via data():

http://en.cppreference.com/w/cpp/container/vector/data

Then you can copy the data directly (e.g. just memcpy).

std::vector<int> foo;
for (int i=0; i<10;++i){
    foo.push_back(i);
}

std::vector<int> bar;

bar.resize(10*10);

int pos = 0;
int size = foo.size()*sizeof(int);
for (int i=0; i<10; ++i){
    memcpy(bar.data()+pos/sizeof(int),foo.data(),size);
    pos += size;
}


for (size_t i=0; i<bar.size(); ++i){
    cout << i << " " << bar[i] << endl;
}

edited Jul 13 '15 at 16:46

answered Jul 13 '15 at 16:29

FooTheBar

818
1
9
20

The STL uses memcpy internally anyway when it is safe to do so. See: http://maintainablecode.logdown.com/posts/159916-memcpy-memmove-and-memset-are-deprecated. This is way too complicated/error prone and doesn't actually buy you anything. – Mark Jul 13 '15 at 20:17

inserting into vectors c++

3 Answers3