Thread: Slow drawing code

I'm using the WinAPI directly for drawing a grid, and extrapolating a 32x32 bitmap image to fit inside of that grid. (Basically, it's a zoom function with a grid-overlay.) Anyways, I've been looking for alternative ways to do this that would be faster but I can't quite find any, so I'm here asking for help!

The code takes around 3-4 seconds to draw the "image." I would like it to be instant. I'm not sure how I can optimize the code, so, any suggestions/pointers in the right direction would be _wonderful!_ Thanks in advanced if anyone can help me:

(btw: I've tested the code with just drawing the grid and just drawing the bitmap, it appears to take roughly 3-4 seconds regardless of weather I render only the grid, only the bitmap, or both. It doesn't make any sense to me, seems like it should take longer to draw both together if it was the drawing code at fault, so I'm wondering if it's either my loops or something else, but I'm not sure how to optimize any of it :x)

EDIT: Oh, also CSize is a class that basically stores two ints, int x and int y. getX() returns that x, and getY() returns that y. (It's really basic, the prototype for the functions is pretty much simply void getX() { return x; } void getY() { return y; }) I doubt that's the slow-down...but maybe?

Code:

bool CCharEdit::renderGrid(HDC hdc)
{
	// Stuff we need
	RECT gridRect;
	CSize gridRectSize;

	HDC hdcMem;
	HBITMAP hBmOld;
	BITMAP bm;

	// Create a Black color
	HBRUSH hBr = CreateSolidBrush(RGB(0,0,0));

	// find the size of each grid "pixel"
	gridRectSize.setSize((m_gridZoom*m_gridSize.getX()), (m_gridZoom*m_gridSize.getY()));

	// Set the upper-bounds for the grid
	gridRect.top = 10;
	gridRect.left = 10;

	// Find the lower-bounds for the grid
	gridRect.bottom = gridRect.top + gridRectSize.getY();
	gridRect.right = gridRect.left + gridRectSize.getX();

	// Initialize the bitmap
	hdcMem = CreateCompatibleDC(hdc);

	hBmOld = (HBITMAP)SelectObject(hdcMem, m_hBmp);
	GetObject(m_hBmp, sizeof(bm), &bm);

	// Draw the grid
	for ( int i = 0; i < m_Size.getX(); ++i )			// the rows
	{
		for ( int j = 0; j < m_Size.getY(); ++j )		// the columns
		{
			// Draw the grid-square
			MoveToEx(hdc, gridRect.left, gridRect.top, NULL);
			LineTo(hdc, gridRect.left, gridRect.bottom);
			LineTo(hdc, gridRect.right, gridRect.bottom);
			LineTo(hdc, gridRect.right, gridRect.top);
			LineTo(hdc, gridRect.left, gridRect.top);

			// Render the portion of the model
			// that corresponds to this portion
			// of the grid
			for ( int k = 1; k < m_gridZoom; ++k )
			{
				for ( int l = 1; l < m_gridZoom; ++l )
				{
					BitBlt(hdc, gridRect.left+k, gridRect.top+l, 1, 
						1, hdcMem, i, j, SRCCOPY);
				}
			}

			// Translate on to the next grid-square
			gridRect.top = gridRect.bottom;
			gridRect.bottom = gridRect.top + gridRectSize.getY();
		}
		// Translate on to the next row
		gridRect.left = gridRect.right;
		gridRect.right = gridRect.left + gridRectSize.getX();

		// Set the grid column back to the top
		gridRect.top = 10;
		gridRect.bottom = gridRect.top + gridRectSize.getY();
	}

	// Clean Up
	DeleteObject(hBr);

	SelectObject(hdcMem, hBmOld);
	DeleteDC(hdcMem);
	DeleteObject(hBmOld);

	// return true for success
	return true;
}

Small benefit can perhaps be had from copying this out of the loop condition:

Code:

xSize = m_Size.getX();
ySize = m_Size.getY();
	for ( int i = 0; i < xSize; ++i )			// the rows
	{
		for ( int j = 0; j < ySize; ++j )		// the columns
		{

Blitting a single pixel at a time is definitely going to be HIGHLY time consuming. Perhapa StretchBlt may help here.

--
Mats

Ohh that's a nifty function - thanks matsp! I'll go implement that right now.

edit:

That helped _a lot_ !! It's instant, it flashes a lot but it's instant! I just need to double buffer then I think to get rid of the flashing and it should be good to go, hooray XD Thanks again matsp!

Use StretchBlt or some other appropriate GDI blit function.

Only draw from rect.top to rect.bottom in rect.left to rect.right order. The moment you go beyond rect.bottom, reset your Y to rect.top, and increment your map_offset. There are fewer pixels down the screen than across the screen so drawing top to bottom left to right is faster than left to right, top to bottom.

Get rid of the inner loop and use a linear offset inside of the outer loop to access your world tile map data. In this way you can completely remove the multiply from the equation and your loop boils down to a few adds.

But beyond that GDI is slow. It was meant for robustness not speed. Use OpenGL, Direct3D or some other graphics API for doing graphics in Windows. GDI is not worth the trouble.