SIMD MIMD

Vector computers need only indexed memory and operations. MIMD requires new primitives like fork/join processes and shared/private memory

Statement mutual independence

All variables read is its Input Set I(S), all written to Output Set O(S), two statements are independent if intersection of their output sets are the empty set, I(S_j) intersection O (S_i) is empty set, and I(S_i) intersection O(S_j) is empty set... these three properties 'output independence', 'flow independence', and 'anti independence' these are known as Bernsteins Conditions

Date dependency graphs

a way to characterize the parallelism of a algorithm, and how to reorganize the computation to improve that paralelization. From linear additions, to binary tree of additions, same # of computations but the binary tree form is better on SIMD/MIMD architectures