lol.
All flameduck is saying is that you keep a copy of the land in main memory as a pixmap, and then if you need to change it you change the pixels in the pixmap with like WritePixel/ReadPixel, and then you also have access to the shape of the land using ReadPixel so that you can do your collision detection and so on. Then when you need to see the the changes you would transfer a rectangular portion of the pixmap to the graphics card, either by turning it into an image (where you have a grid of images making up the whole screen, say 64x64 size each, or smaller), or by directly transferring a rectangle that's changed (bounding box) to an existing texture/image using for example glTexSubImage2D() (opengl), or using DrawPixmap or similar to draw to backbuffer and grab into an image. You can grab to image fairly quickly, faster than to a pixmap.
But you said you didn't want to drop the land due to needing to keep tunnels and so on. But it would be an idea to put a duplicate copy of your landscape in a pixmap, as well as having it in an image (for drawing), so that you can do collision detection in main memory using the cpu.
Are you going to have it so that you can blow up some of the land, which leaves behind the `cave` texture?
All flameduck is saying is that you keep a copy of the land in main memory as a pixmap, and then if you need to change it you change the pixels in the pixmap with like WritePixel/ReadPixel, and then you also have access to the shape of the land using ReadPixel so that you can do your collision detection and so on. Then when you need to see the the changes you would transfer a rectangular portion of the pixmap to the graphics card, either by turning it into an image (where you have a grid of images making up the whole screen, say 64x64 size each, or smaller), or by directly transferring a rectangle that's changed (bounding box) to an existing texture/image using for example glTexSubImage2D() (opengl), or using DrawPixmap or similar to draw to backbuffer and grab into an image. You can grab to image fairly quickly, faster than to a pixmap.
But you said you didn't want to drop the land due to needing to keep tunnels and so on. But it would be an idea to put a duplicate copy of your landscape in a pixmap, as well as having it in an image (for drawing), so that you can do collision detection in main memory using the cpu.
Are you going to have it so that you can blow up some of the land, which leaves behind the `cave` texture?