I like the test harness of 2teez, it is easily adapted for a perl -p command line. So I'll reuse it for an old sed-der comment/observation:

For mass transformations of text, trade off flexible regex for static strings when possible.

while (<DATA>) {
    s/\b module \s+ OLD \b/module NEW/xms;
    print $_;
}
[download]

__DATA__
 //Verilog HDL for "tt", "hh" "functional"
// if i write the word module here the script goofs up
`timescale 1ps/10fs


module OLD(Y, A, B ); // This is a great module!
    output Y;
    input A;
    input B;

endmodule
[download]

 //Verilog HDL for "tt", "hh" "functional"
// if i write the word module here the script goofs up
`timescale 1ps/10fs


module NEW(Y, A, B ); // This is a great module!
    output Y;
    input A;
    input B;

endmodule
[download]

Of course, this means that whatever whitespace separation you used to have between the keyword "module" and the module name is gone, but whitespace niceties usually aren't a big deal with mechanically gunkulated code. On the other hand, you don't have to ignore any lines that have commenting on them.

This is a line oriented solution, so slurping in a file and transforming that long string will require using the "g" flag and listening to sages such as choroba. For line-oriented text, I believe line oriented processing gives the fewest surprises (e.g., a surprise such as having a comment change affect code manipulation). However if the two adjacent words you are keying off of can be separated by a newline, then you have to use something like the fancier regex solutions suggested.

One thing left to worry about is whether "module OLD" in a comment should be transformed to "module NEW". But if you keep in mind the old maxim, "All comments lie," then you shouldn't worry too much ;-)