Question

I’m a novice in Open GL.

Is there ever a need to do texture compression at runtime?

Surely, the way it works is a big texture file is compressed at build time. At runtime, you expand portions of the compressed texture file, as needed, to apply to a surface.

Are there any (credible) circumstances where you have expanded texture data, and you need to compress it at runtime?

Thanks!

Answer 1

Are you talking about compressed image formats (like JPEG or even a zip file containing an image) or compressed texture formats (like DXT1, etc)? When you have a compressed texture (such as DXT) you don’t have to decompress it at runtime, the graphics card can do it on the fly as it samples the texture.

For games, where you can precompile all your assets ahead of time, it’s generally a good idea to apply something like DXT compression at (asset) build time so you get all the benefits of texture compression (faster load time, less memory bandwidth usage, etc) without the cost of actually performing the compression at runtime. That said, in any circumstance where you wanted to render with compressed textures, but you don’t have access to images you’ll be using ahead of time (maybe you let the user pick image files from their machine or something) you would have no choice but to do the compression at runtime.

EDIT:

The way you would do DXT compression at runtime would be to call glTexImage2D, specifying the actual format of the source image you have (GL_RGBA, etc) for the ‘format’ parameter and a compressed format for the ‘internal format’ parameter, such as GL_COMPRESSED_RGBA_S3TC_DXT1_EXT for DXT1, assuming your card supports the gl_ext_texture_compression_s3tc extension.

If you have pre-compressed texture data then you can load it directly with glCompressedTexImage2D.