Texture assets

IRadientTextureAsset::GetDesc() returns an immutable RadientTextureAssetDesc with the texture's width, height, pixel format, and mip-level count.

const RadientTextureAssetDesc& Desc = Texture->GetDesc();
if (Desc.Width != 0)
{
    const Uint32 Width = Desc.Width;
    const Uint32 Height = Desc.Height;
    const RADIENT_TEXTURE_FORMAT Format = Desc.Format;
    const Uint32 MipLevels = Desc.MipLevels;
}

Before decoding completes, or if decoding fails, the descriptor has zero dimensions and mip levels and RADIENT_TEXTURE_FORMAT_UNKNOWN. Query GetDesc() again after loading progresses; a previously returned empty descriptor remains empty.

Once available, the description stays unchanged and its reference remains valid while the texture asset is retained. It is available without a GPU and before GPU uploads finish. Upload failure or cancellation does not discard decoded metadata. GetDesc() may be called from any thread while the caller retains the asset.

Dimensions describe mip level zero of each array layer or cube face; height is one for 1D textures. Format describes the loaded pixels after decoding and requested format interpretation, including sRGB and block compression; it is independent of how a material samples the texture. MipLevels includes loaded and generated mip levels per layer or face. These properties describe the logical asset; they do not report renderer allocation dimensions or GPU mip residency.

Encoded texture sources retain support for cube maps and texture arrays. If a loaded format has no corresponding RADIENT_TEXTURE_FORMAT value, Format is RADIENT_TEXTURE_FORMAT_UNKNOWN; the dimensions and mip count remain available.

RadientTextureData accepts all defined formats except RADIENT_TEXTURE_FORMAT_UNKNOWN. Supply consecutive levels starting at mip 0 through pMipLevels and MipLevelCount. Each RadientTextureMipData specifies a blob, byte offset, and row stride. Different levels can share a blob or use separate blobs; zero stride selects tightly packed pixel rows or compressed block rows for that level. The supplied bytes are consumed without copying or repacking.

Mip dimensions describe the logical image in pixels. Compressed storage rounds those dimensions up to complete blocks using the format's block width and height. Even a logical 1x1 mip requires one full compression block.

GenerateMips defaults to True. For uncompressed input, missing levels through 1x1 are generated from the last supplied level. Supplied levels remain unchanged, and a complete chain needs no generation. Set it to False to load only the supplied levels. Compressed formats always load the supplied chain without generation. An incomplete compressed chain with GenerateMips = True logs a warning and still loads successfully; a complete chain or GenerateMips = False produces no warning. BC mip chains can be supplied directly or loaded from an encoded texture such as DDS or KTX.

BC mip 0 width and height must be multiples of four for all sources. Invalid raw dimensions return RADIENT_STATUS_INVALID_ARGUMENT; encoded textures with these dimensions fail loading with RADIENT_STATUS_UNSUPPORTED. Smaller dimensions in subsequent mip levels remain valid. Texture reflection reports the number of loaded levels, including any generated levels.

C callers use IRadientTextureAsset_GetDesc(Texture) to obtain a pointer to the same immutable description.