Constant-time word-size string matching

We present a novel string-matching algorithm that requires constant time for text scanning in an unusual model where (a) the input pattern and text are each packed into a single word, (b) the output is a one word bit-mask identifying the pattern occurrences in the text, and (c) there are constant-time arithmetic, bitwise, and shift instructions that operate on words whose size is proportional to the arbitrarily long input length. Our bit-parallelism techniques build upon and also greatly simplify existing parallel random access machine algorithms by using two "simple structure" rather than "small size" deterministic samples, i.e., one deterministic sample is very small (size two), while the other is a potentially very long prefix of the pattern. Pattern preprocessing takes time proportional to the word size. Our results also establish, by recent reductions, new bounds for the packed string matching problem.