Efficiency of iterators and alternatives? [migrated]
- by user48037
I have the following code for my game tiles:
std::vector<GameObject_Tile*>::iterator it;
for(int y = 0; y < GAME_TILES_Y; y++)
{
for(int x = 0; x < GAME_TILES_X; x++)
{
for (it = gameTiles[x][y].tiles.begin() ; it != gameTiles[x][y].tiles.end(); ++it)
{}}}
tiles is:
struct Game_Tile
{
// More specific object types will be added here eventually
vector<GameObject_Tile*> tiles;
};
My problem is that if I change the vector to just be a single GameObject_Tile* instead and remove the iterator line in the loop I go from about 200fps to 450fps.
Some context:
The vector/pointer only contains one object in both scenarios. I will eventually need to store multiple, but for testing I just set it to a single pointer.
The loop goes through 2,300 objects each frame and draws them. I would like to point out that if I remove the Draw (not seen int he example) method, I gain about 30 frames in both scenarios, the issue is the iteration.
So I am wondering why having this as a vector being looped through by an iterator (to get at a single object) is costing me over 200 frames when compared to it being a single pointer?
The 200+ frames faster code is:
std::vector<GameObject_Tile*>::iterator it;
for(int y = 0; y < GAME_TILES_Y; y++)
{
for(int x = 0; x < GAME_TILES_X; x++)
{
//gameTiles[x][y].tiles is used as a pointer here instead of using *it
}}
tiles is:
struct Game_Tile
{
// More specific object types will be added here eventually
GameObject_Tile* tiles;
};
