Lz77: Help Needed Optimizing & Accessing Buffers Dos16 & Win32?

Hi! I am working on a Win32 compression technique even though the technique is actually 16-bit. Right now, I am almost ready to start debugging. But first, I want to optimize my version of LZ77. I know of a few flaws of my LZ77: an inefficient inner loop, extra unnecessary work passing the length of a match to the LZ77 main loop and partial use of C in the main loop. In order to fix this, I want to use assembler for the LZ77 code. How do I access variables and arrays from C in an assembler module? I am currently using Digital Mars C and NASM and plan to port this code to Borland C 3.0 and Open Watcom. Also, what other ways can I optimize LZ77?