Help Needed Optimizing Lz77 Compressor

Hi! I am working on some compression techniques, but they are very slow. One of the performance hogs is my versions of lz77. I asked about it elsewhere, and one person helped me optimize the lz77 main loop, while another pointed out hash tables. The former helped slightly, but I couldn't quite understand the latter. I tried to use an 8k array of bits to determine whether a word occurred before. It helped slightly. Can somebody here better explain the hash table technique and give me other ways to optimize lz77?