TLNS3D SOLVE FILTFX Parallel Loops

Orphaned Parallel DO achieves Coarse Grain parallelism

Use Static scheduling

Tune by avoiding Barriers at the loop level

Assure verified again

. . .

do 11 nv=1,5

!$OMP DO

do 10 k=1,kmp2

do 10 j=1,jmp2

do 10 i=1,imp2

fw(i,j,k,nv) = ...

10 continue

!$OMP END DO NOWAIT

11 continue

!$OMP BARRIER

. . .

Previous slide Next slide Back to first slide View graphic version