Testing various BCn decoding libraries

Code for "Comparing BCn texture decoders" blog post.

Libraries Tested:

Used as git submodules:

• amd_cmp: https://github.com/GPUOpen-Tools/compressonator, 2022 Jun 20 (b2b4ee3). MIT.
• bc7enc_rdo: https://github.com/richgel999/bc7enc_rdo, 2021 Dec 27 (e6990bc). MIT or public domain.
• bcdec: https://github.com/iOrange/bcdec, 2022 Sep 15 (026acf9). MIT or Unlicense.
• convection: https://github.com/elasota/ConvectionKernels, 2022 Jun 23 (350416d). MIT.
• DirectXTex: https://github.com/microsoft/DirectXTex, 2022 Jun 6 (679538e). MIT. Does not easily build on non-Windows: requires bits of DirectX-Headers and DirectXMath github projects, as well as an empty <malloc.h> and some stub <sal.h>.
• icbc: https://github.com/castano/icbc, 2022 Jun 3 (502ec5a). MIT.

Embedded directly into the repo:

• etcpak: https://github.com/wolfpld/etcpak, 2022 Jun 4 (a77d5a3). BSD 3-clause. Just the needed source files, modified to leave only decompression bits.
• mesa: https://github.com/mesa3d/mesa, 2022 Jun 22 (ad3d6d9). MIT. Just the needed source files, modified to leave only decompression bits.
• squish: https://sourceforge.net/projects/libsquish/, 2019 Apr 25 (r110, v1.15). MIT. Upstream is not a Git repository.
• swiftshader: https://github.com/google/swiftshader, 2022 Jun 16 (2b79b2f). Apache-2.0. Minimal set of source files.

Test Images:

I produced most of the .dds files under test_images folder myself, using a simple "export DDS" script for Unity. Within Unity, used various import options to produce BCn formats out of original images.

The original images are from:

• ambientcg/: various texture & HDRI files from ambientCG.
• frymire: University of Waterloo image set.
• kodim_23: Kodak lossless image suite.
• lythwood_room: Lythwood Room HDRI from Poly Haven.
• testcard: BC2 DDS file from bcdec.
• webp_gallery_3: WebP Lossless and Alpha Gallery page.