11.10 2* Iterative cluster fit is now considered to be a new compression mode 3* The core cluster fit is now 4x faster using contributions by Ignacio 4Castano from NVIDIA 5* The single colour lookup table has been halved by exploiting symmetry 6 71.9 8* Added contributed SSE1 truncate implementation 9* Changed use of SQUISH_USE_SSE to be 1 for SSE and 2 for SSE2 instructions 10* Cluster fit is now iterative to further reduce image error 11 121.8 13* Switched from using floor to trunc for much better SSE performance (again) 14* Xcode build now expects libpng in /usr/local for extra/squishpng 15 161.7 17* Fixed floating-point equality issue in clusterfit sort (x86 affected only) 18* Implemented proper SSE(2) floor function for 50% speedup on SSE builds 19* The range fit implementation now uses the correct colour metric 20 211.6 22* Fixed bug in CompressImage where masked pixels were not skipped over 23* DXT3 and DXT5 alpha compression now properly use the mask to ignore pixels 24* Fixed major DXT1 bug that can generate unexpected transparent pixels 25 261.5 27* Added CompressMasked function to handle incomplete DXT blocks more cleanly 28* Added kWeightColourByAlpha flag for better quality images when alpha blending 29 301.4 31* Fixed stack overflow in rangefit 32 331.3 34* Worked around SSE floor implementation bug, proper fix needed! 35* This release has visual studio and makefile builds that work 36 371.2 38* Added provably optimal single colour compressor 39* Added extra/squishgen.cpp that generates single colour lookup tables 40 411.1 42* Fixed a DXT1 colour output bug 43* Changed argument order for Decompress function to match Compress 44* Added GetStorageRequirements function 45* Added CompressImage function 46* Added DecompressImage function 47* Moved squishtool.cpp to extra/squishpng.cpp 48* Added extra/squishtest.cpp 49 501.0 51* Initial release 52 53