This is intended for creating mipmaps as part of an pre-run asset build or as a no-fuss way to generate them that avoids involving any particular graphics API. However, it will not run as quickly as when using a graphics API. So, if updating mipmaps during a frame, or generating a substantial volume at load time, consider another approach.

For reference, a GPU-based approach usually involves setting up a source and target image and using blit commands to copy and downscale from one to the other.

In exactly one C or C++ file that includes this file, add

 #define AD_MIPMAP_IMPLEMENTATION

before the #include for this file. This will put the implementation into that file.

Fill out a AdmBitmap struct with a bitmap and pass it to adm_generate_mipmaps.

AdmBitmap bitmap;
bitmap.width = …;
bitmap.height = …;
bitmap.bytes_per_pixel = …;
bitmap.pixels = …;

int levels;
AdmBitmap* mipmaps = adm_generate_mipmaps(&levels, &bitmap);

mipmaps is an array which includes bitmap as the first element. Then, once you're done, adm_free_mipmaps deallocates the array (including the original).

adm_free_mipmaps(mipmaps, levels);

If you don't want the original to be in the returned array, use adm_generate_mipmaps_no_base.

To use another memory allocator instead of calloc and free, define both of the following before the #include that introduces the implementation.

#define ADM_ALLOCATE(bytes, allocator) …
#define ADM_DEALLOCATE(memory, allocator) …

This code is public domain under the CC0 license.