How to Write Fast Code SIMD Vectorization CMU 2008 Slides (slides 18 645

Carnegie MellonHow to Write Fast Code SIMD Vectorization 18-645, spring 2008 13th and 14th LectureInstructor: Markus Püschel Guest Instructor: Franz Franchetti TAs: Srinivas Chellappa (Vas) and Frédéric de Mesmay (Fred)Carnegie MellonOrganization OverviewIdea, benefits, reasons, restrictions History and state-of-the-art floating-point SIMD extensions How to use it: compiler vectorization, class library, intrinsics, inline assemblyWriting code for Intel’s SSE Compiler vectorization Intrinsics: in