Abstract: A method and apparatus for compressing and decompressing text and image data by forming fixed-length codewords from a variable number of data symbols. Data symbols are shifted into registers in the first half of a buffer, while an uncompressed string of data symbols are shifted into registers in the second half of the buffer. A systolic array of processors compares each data symbol in the second half of the buffer with each data symbol in the first half of the buffer. Each processor compares pairs of data symbols, and selectively passes the data symbols to an adjacent processor. A fixed-length output is provided indicating the length and the starting point of the longest substring in the first half of the buffer that matches a substring from the second half of the buffer.