strstr challenge

create the fastest C strstr function implementation.
Rules:
1.Must be portable between different architectures(ARM,x86/x86-64,MIPS)
2.Must handle strings of any length that C standard library strstr can handle.
3.Cannot rely on undefined behavior or platform specific code.
strstr reference: http://www.cplusplus.com/reference/cstring/strstr/

With respect to size of data to search through yes, you won't find better algorithmic time complexity than O(n). That doesn't mean you can't make genuine speed improvements.

It's high-time somebody wrote a C implementation of the Knuth-Morris-Pratt algorithm and got this over and done with, that's all the author wanted.