Mixing Asm W/ C In Compression?

Hi! I am working on file compression and want to finish converting the code to find an LZ77 block to assembler. Right now, it's far from optimal: parts of it are still in C, and the internal loop needs optimizing. I am using Digital Mars C and have Borland C++ 3.0, Open Watcom C 2.0 and NASM. One problem I have is accessing an array in a C module from an Assembler module. Another is passing arguments to a function in C. I'm doing this in Win32 but also want to do this in DOS16/32 and Win64.

 

My version of LZ77 is nonoptimal for other reasons: I use a brute method for the scan. I asked for help optimizing it elsewhere and got some help. I was able to make the internal compare faster and told to use hash tables. I am trying to understand it but am having a hard time. How else can I optimize LZ77?