Fast Search Algorithm

which article ?

also, you dont quantify 'too slow' or how many files you are dealing with

>> First. Find files matched search filters.
>> Second. Read matched files one by one and search string which is inputted from read file.

I'd assume you have some timing checks in your program, and that the first step doesn't take a lot of time in comparison to the second, yes ? if you still think this time is significant, then maybe you need an 'observable list', so that putting entries into it causes 1 or more threads to be fired off as per 'Second', which Im covering next

So, instead of 'one by one', for the file search for the 'Second' step, have you thought about splitting the list into a number of sub-lists, using a thread for (lets say n) sub-lists, and putting the information on some short of shared memory structure ?

The trick will be, past a certain number of threads and you'll start finding bottle-necks elsewhere - IO for example, unless you're using memory mapping to read the files (hint, you should be), so you'll need to work out a 'balance'

Im assuming you actually have a decent file search algorithm, but, its worthwhile looking at other implementations and profiling them, as well

[edit] unless you're using memory mapping - well, I guess even with memory mapping, you'll eventually find the amount of memory as a constraint [/edit]