A Hardware Design Generator of Streaming Data Compressors Based on Exponential-Golomb Coding

Exponential-Golomb code is a variable-length code used in lossless data compression. It entails encoding nonnegative integers into code words whose lengths correlate with the magnitudes of the encoded numbers. This correlation can be altered through variations in the Golomb parameter used for the encoding process. A compression algorithm utilizing Exponential-Golomb coding is proposed, which encodes a fixed number of integers together with the same value for the Golomb parameter, forming a block of encoded samples. A lossy compression variant is also proposed, which employs partial data discarding, allowing the user to set an upper bound on the total bit-length of the compressed block of input samples. An extension to Exponential-Golomb code is also proposed, which enables the encoding of two's complement signed integers while preserving the sign bit in lossy compression. The algorithms have been implemented as parameterizable hardware design generators written in Chisel hardware description language (HDL). Generated hardware instances have undergone testing through simulations and verification on a commercially available FPGA platform.