This morning, I have released PAQCompress v0.5.14:
This version adds the new PAQ8PX v205 compressor, which was released yesterday.
PAQ8PX v205 brings the following enhancements:
- Fixed TEXT detection in multi-content files. Cleaned up (a little bit) the general detection mechanism. - Functions squash() and stretch() are now table- based and initialized by their float equivalent functions. - Small speedup for binary files: skip adding zero inputs if exe content is not detected. - Small speedup for text files: skip adding inputs for some contexts (in ChartModel, SparseModel, SparseBitModel). - Slightly decreased WordModel's memory use for binary content. - Switched from ContextMap to ContextMap2 in NestModel (without Run Stats) in IndirectModel and XMLModel (with Run Stats). - Tuned contextmap scaling factor in WordModel and TextModel based on block type. - Ported MRB detection, decoding and encoding from paq8pxd and added support for 1-bit images and uncompressed images. - Moved dBase detection from RecordModel to detect() - similarly to paq8pxd. - Improved ChartModel. - Cosmetic: fixed assert in WordModel. - Cosmetic: turned #define to function (rleOutputRun) and enum 'state' to scoped enum in RleFilter.hpp - Cosmetic: removed #ifdef VERBOSE. - Cosmetic: Utils.hpp -> SystemDefines.hpp + BlockType.hpp + CharacterNames.hpp + SIMDType.hpp. - Updated and tidied up the help screen.
You can download this new version on GitHub by clicking here.