Question

I need to render a 1027 * 768 bitmap to the client window (same size) and I don’t have more than 10-15 ms to complete this task. I am using bufferedGraphics allocated from a bufferedGraphicsContect object and still notice huge performance issues.

I am using unsafe code to perform my copy operations and found unbelievable results. I know the Graphics / BufferedGraphics objects should have some sort of a drawing surface in memory. I was wondering if someone could point me in the right direction on how to write to this surface using Marshal or some other unsafe low level method.

I am in the process of porting an older c# graphics application. I know c# is not designed for heavy graphics and that there are better tools than GDI+ available, unfortunately I don’t have those luxuries.

This is what I have come up with so far… Any insight what so ever is greatly apperciated.

byte[] _argbs = null;
static readonly Bitmap _bmUnderlay = Properties.Resources.bg;
static Bitmap _bmpRender = new Bitmap(1024, 768, System.Drawing.Imaging.PixelFormat.Format24bppRgb); 
int bmpHeight = Properties.Resources.bg.Height;
int bmpWidth = Properties.Resources.bg.Width;
static BufferedGraphicsContext _bgc = new BufferedGraphicsContext();

internal unsafe void FillBackBuffer(Point cameraPos)
{
        // lock up the parts of the original image to read (parts of it)
        System.Drawing.Imaging.BitmapData bmd = _bmUnderlay.LockBits(
            new Rectangle(cameraPos.X, cameraPos.Y, 1024, 768),
            System.Drawing.Imaging.ImageLockMode.ReadOnly, System.Drawing.Imaging.PixelFormat.Format24bppRgb);
        // get the address of the first line.
        IntPtr ptr = bmd.Scan0;

        //if (_argbs == null || _argbs.Length != bmd.Stride * bmd.Height)
        //    _argbs = new byte[bmd.Stride * bmd.Height];
        if (_argbs == null || _argbs.Length != 1024 * 3 * 768)
            _argbs = new byte[1024 * 3 * 768];

        // copy data out to a buffer
        Marshal.Copy(ptr, _argbs, 0, 1024 * 3 * 768);

        _bmUnderlay.UnlockBits(bmd);

        // lock the new image to write to (all of it)
        System.Drawing.Imaging.BitmapData bmdNew = _bmpRender.LockBits(
            new Rectangle(0, 0, 1024, 768),
            System.Drawing.Imaging.ImageLockMode.ReadWrite, System.Drawing.Imaging.PixelFormat.Format24bppRgb);
        // copy data to new bitmap
        Marshal.Copy(_argbs, 0, bmdNew.Scan0, 1024 * 3 * 768);
        _bmpRender.UnlockBits(bmdNew);
}

private unsafe void _btnGo_Click(object sender, EventArgs e)
{
    // less than 2 ms to complete!!!!!!!!
    FillBackBuffer(new Point());

    using (BufferedGraphics bg = _bgc.Allocate(CreateGraphics(), ClientRectangle))
    {
        System.Diagnostics.Stopwatch sw = new System.Diagnostics.Stopwatch();
        sw.Start();
        /////
        ///
        // This method takes over 17 ms to complete
        bg.Graphics.DrawImageUnscaled(_bmpRender, new Point());
        //
        ///
        /////
        sw.Start();
        this.Text = sw.Elapsed.TotalMilliseconds.ToString();

        bg.Render();
    }
}

EDIT:

Forgot to mention I’m looking for a low level alternative to Graphics.DrawImage(), preferrably writing to Graphics surface memory, using pointers? Thanks again

Answer 1

Pay attention to the pixel format of the bitmap. On standard 32bpp video hardware, Format32bppPArgb renders ten times faster than any of the other ones. Because the pixels don’t need any translation. The 24bpp format you use now needs to be widened to 32bpp and that doesn’t come for free. Don’t skip the P of PArgb and don’t forget to set the alpha value to 255 in your code.

Using BufferedGraphics is fishy btw. You should always use the one you get for free in the OnPaint method. And you probably don’t need one at all since you’re blitting this fast. That’s an automatic 2x speed-up.