Средства и методы разработки

Мало что понятно и денег хотят за чтение. Зачем так сложно, если можно пойти на веб-страничку автора: http://th.if.uj.edu.pl/~dudaj/
Asymmetric Numeral Systems (ANS, Wikipedia article, slides, PCS article) family of entropy coders (heart of data compressors). Previously, a compromise was required: Huffman coding allowed for fast but suboptimal compression, arithmetic coding for nearly optimal but slow (costly). ANS offers compression ratio as arithmetic coding, at similar speed as Huffman coding. Here is a list of implementations and compressors switched to ANS. For example Facebook ZSTD and Apple LZFSE use Finite State Entropy implementation of tANS variant, CRAM 3.0 genetic data compressor of European Bioinformatics Institute and experimental branch of Google VP10 video codec use rANS variant. Additionally, chaotic behavior of tANS makes it also perfect for simultaneous encryption... Интерес представляет fpaqb кодер, для наших применений. Но что-то по объему занимаемой памяти для упаковки я не понял... сходу он кажется чудовищным (сотни кБайт). В fpaqa такого нет, но деление ничего хорошего не предвещает, тем более на не-константу. The coder is implemented in the order-0 compressors fpaqa, fpaqb, and fpaqc and the context mixing compressor lpaq1a from the PAQ series. fpaqa uses lookup tables for both compression and decompression. It uses a 10 bit state and the probability is quantized to 7 bits on a nonlinear scale (finer near 0 and 1 and coarse near 1/2). The stack size is 500,000. Increasing these numbers would result in better compression at the expense of a larger table. fpaqb uses direct calculations except for division, which uses a table of inverses because division is slow. fpaqc is fpaqb with some speed optimizations. The coder for fpaqc is used in lpaq1a (a context mixing program), replacing the arithmetic coder in lpaq1.