Creating Compressed Textures

[C++]

To describe a compressed texture surface in the DDSURFACEDESC2 structure when creating the surface, you must include the following steps.

Specify the DDSCAPS_TEXTURE flag in the dwFlags member of the DDSCAPS structure, just as you would for any texture.
Set the dwFourCC member of the DDPIXELFORMAT structure to one of the DXT codes described later.
Include DDPF_FOURCC in the dwFlags member of DDPIXELFORMAT. Do not set the DDPF_RGB flag.
Specify a width and height that are a multiple of 4 pixels.

There are two ways to load image data into a compressed texture surface:

Create a regular RGB or ARGB surface and load a normal bitmap into it, then use IDirectDrawSurface7::Blt or IDirectDrawSurface7::BltFast to blit from the uncompressed surface to the compressed surface. DirectDraw does the compression for you.
Load the compressed data from a file and copy it directly into the surface memory. (See Accessing Surface Memory Directly.) You can create and convert compressed texture (DDS) files using the DirectX Texture Tool (Dxtex.exe) supplied with the SDK. You can also create your own DDS files and either copy the data from compressed surfaces or else use your own routines to convert regular bitmap data to one of the compressed formats.

Note When you call IDirectDrawSurface7::Lock or IDirectDrawSurface7::GetSurfaceDesc on a compressed surface, the DDSD_LINEARSIZE flag is set in the dwFlags member of the DDSURFACEDESC structure, and the dwLinearSize member contains the number of bytes allocated to contain the compressed surface data. The dwLinearSize parameter resides in a union with the lPitch parameter, so these parameters are mutually exclusive, as are the flags DDSD_LINEARSIZE and DDSD_PITCH.

[Visual Basic]

To describe a compressed texture surface in the DDSURFACEDESC2 type when creating the surface, you must include the following steps.

Specify the DDSCAPS_TEXTURE flag in the lCaps member of the DDSCAPS2 type, just as you would for any texture.
Set the lFourCC member of the DDPIXELFORMAT type to one of the DXT codes described later.
Include DDPF_FOURCC in the lFlags member of DDPIXELFORMAT. Do not set the DDPF_RGB flag.
Specify a width and height that are a multiple of 4 pixels.

There are two ways to load image data into a compressed texture surface:

Create a regular RGB or ARGB surface and load a normal bitmap into it, then use DirectDrawSurface7.Blt or DirectDrawSurface7.BltFast to blit from the uncompressed surface to the compressed surface. DirectDraw does the compression for you.
Load the compressed data from a file and copy it directly into the surface memory. (See Accessing Surface Memory Directly.) You can create and convert compressed texture (DDS) files using the DirectX Texture Tool (Dxtex.exe) supplied with the SDK. You can also create your own DDS files and either copy the data from compressed surfaces or else use your own routines to convert regular bitmap data to one of the compressed formats.

Note When you call DirectDrawSurface7.Lock or DirectDrawSurface7.GetSurfaceDesc on a compressed surface, the DDSD_LINEARSIZE flag is set in the lFlags member of the DDSURFACEDESC2 type, and the lLinearSize member contains the number of bytes allocated to contain the compressed surface data. The lLinearSize parameter resides in a union with the lPitch parameter, so these parameters are mutually exclusive, as are the flags DDSD_LINEARSIZE and DDSD_PITCH.

The advantage of this behavior is that an application can copy the contents of a compressed surface to a file without having to calculate for itself how much storage is required for a surface of a particular width and height in the specific format.

The following table shows the five types of compressed textures. For more information on how the data is stored (you need to know this only if you are writing your own compression routines) see Compressed Texture Formats.

FOURCC	Description	Alpha premultiplied?
DXT1	Opaque / one-bit alpha	n/a
DXT2	Explicit alpha	Yes
DXT3	Explicit alpha	No
DXT4	Interpolated alpha	Yes
DXT5	Interpolated alpha	No

Note When you blit from a non-premultiplied format to a premultiplied format, DirectDraw scales the colors based on the alpha values. Blitting from a premultiplied format to a non-premultiplied format is not supported. If you try to blit from a premultiplied-alpha source to a non-premultiplied-alpha destination, the method will return DDERR_INVALIDPARAMS. If you blit from a premultiplied-alpha source to a destination that has no alpha, the source color components, which have been scaled by alpha, will be copied as is.