The Polaris- transformed code is shown below: do parallel i = 1 . n do j = 1, cpus fxO(i.j) = 0 enddo enddo do parallel i = 1 . n do j = 1 , n ... algorithm for recognizing reductions searches for assignment statements within a given loop of the form: A(ai , 012, ac ac ac) = -4(qi.o2. ... In our example the reduction statements could simply be enclosed by lock/unlock pairs, allowing the loop to be executed in parallel.
|Title||:||Conference proceedings of the 1995 International Conference on Supercomputing|
|Publisher||:||Assn for Computing Machinery - 1995|