Unfortunately itвЂ™s quite complex to decompress using two Huffman codes that are themselves encoded with a canonical Huffman code. DEFLATEДEFLATE offers good compression, and decompression runs in a few kilobytes of RAM. We can discount plain binary encoding as it does not meet the вЂњgood compressionвЂќ requirement. Plain binary codingДЇixed-width binary coding of the distance and length does not produce an outstanding result, but it is fast and requires no additional state memory when decompressing. LZMA uses a single range coderВ to encode literals, distances, and lengths, dynamically adapting its probability model as it encodes.LZ4 uses a byte-oriented prefix encoding for distance and length and leaves literals uncompressed.LZJU90 uses two bit-oriented prefix encodings for distance and length and leaves literals uncompressed.DEFLATE uses two separate dynamic Huffman codes, one to encode literals and lengths and the other to encode distances.Plain binary encoding does nothing to reduce literals, distance, and length, relying entirely on the string parser to detect redundancies.

LZSS-based encoders try to reduce the coding of literal, distance, and length using an additional compression scheme: Encoding with standard algorithmsвЂ¦explained

The main differences between all these algorithms is the way that the literal and (distance, length) pairs are encoded and constraints placed on the distance and length of the match because of this encoding. Typical coding is a leading flag bit, identifying a literal or a match, followed by an 8-bit literal or a (distance, length) pair referencing previous content behind the cursor. Once literals and matches are identified, the encoder takes each and emits them to the bitstream. Effective compression relies on being able to encode the literals and references in fewer bits than the string being replaced. Standard compression schemes, such as DEFLATE, LZMA, LZJU90, and LZ4, use an LZSS-style kernel to parse the input string into a stream of literals and matches, a match being a back reference to a substring that can be subsituted at the cursor position. When we improve emCompress, we also improve the SEGGER Linker, which is a win all round! LZSS, the ubiquitous algorithm The company ethos is that we develop and use our own software, and therefore it is natural to use emCompress components as a way of compressing data sections held in flash that must be copied to RAM on startup. In discussing the linker and the future direction of our emCompress product, we examined the possibility of improving emCompress and, using that improvement, fold it into the SEGGER Linker. SMASH compresses as efficiently as DEFLATE, uses only a few words of decompressor state, and has an ultra-small, ultra-fast, ultra-simple decompressor. With some experimentation, the end result is a coding technique we named SMASH, the Small Microcontroller Advanced Super-High compression scheme. Limited state in RAM when decompressing.High compression ratio (uncompressed size divided by compressed size).One of the things that is driven from the top in SEGGER is that we can always do better.В Not satisfied with standard schemes, we wanted to optimize emCompress, SEGGERвЂ™s compression library, for:

File minimizer algorithm