The　performance　of　XPath　evaluation　significantly　affects　XML　applications,　while　the　popularity　of　multi-core　environment　provides　the　great　opportunity　for　performance　improvement　through　parallel　computing.　A　novel　parallel　XPath　evaluation　method　named　pM2　(parallel　Matrix　Match),　which　is　based　on　node　relation　matrix,　is　proposed　in　this　paper.　It　has　two　main　stages　includes　parallel　relation　matrix　construction　and　parallel　query.　Node　relation　matrix　of　XML　dataset　is　constructed　according　to　region　codes　beforehand,　while　parallel　query　is　realized　by　using　parallelized　query　primitives.　The　navigate　feature　of　pM2　makes　it　easy　to　implement　complete　semantics　of　XPath.　Iterative　processes　in　matrix　construction　and　query　primitives　are　optimized　in　data　parallelism　style.　Experiments　show　pM2　is　a　practical　XPath　evaluation　method　with　good　performance.　©　2013　Binary　Information　Press.