عèööðððð Óñôùøøøøóò Óó Äó Blockin Blockinð Ëëññððööøý Ë Blockinóöö

Abstra t. Lo al similarity omputation between two sequen es permits dete ting all the relevant alignments present between subsequen es thereof. A well-known dynami programming algorithm works in time O(mn), m and n being the lengths of the subsequen es. The algorithm is rather slow when applied over many sequen e pairs. In this paper we present the rst bit-parallel omputation of the s ore matrix, for a simpli ed hoi e of s ores. If the omputer word has w bits, then the resulting algorithm works in O(mn logmin(m;n;w)=w) time, a hieving up to 8-fold speedups in pra ti e. Some DNA omparison appli ations use pre isely the simpli ed s ores we handle, and thus our algorithm is dire tly appli able. In others, our method an be used as a raw lter to dis ard most of the strings, so the lassi al algorithm an be fo used only on the substring pairs that an yield relevant results.