lz : Lossless & Lossy Floating Point Compression

Sources : http://github.com/leobago/lz

Description:

This compression library (lz) should be use to compress large floating-point datasets. It allows for both, lossless and lossy compression. It aims to provide high compression throughput and even faster decompression. It receives a block of floating-point numbers (one per row) and decomposes each floating-point number into an array of bytes, forming a matrix of bytes. The matrix has 4 or 8 columns (for single and double precision respectively) and as many rows as floating-point numbers in the block. Then, it transposes the matrix resulting in a new matrix where the first rows show low entropy (exponent) and the last rows have high entropy (last mantissa bits). Then, lz compresses the first rows, keep the high entropy rows uncompressed and if desired by the user, discard the last rows.