Performance Optimisation and Productivity
A Centre of Excellence in HPC
OpenMP nested loop collapse
Version's name: OpenMP nested loop collapse ; a version of the OMP Collapse program.Repository: [home] and version downloads: [.zip] [.tar.gz] [.tar.bz2] [.tar]
Implemented best practices: Collapsing OpenMP parallel loops ·
The following pseudo-code shows the proposed best-practice, namely, to use the COLLAPSE
clause to fuse the outer loops
!$OMP PARALLEL DO COLLAPSE(2) DEFAULT(NONE) SHARED(...)
DO K = 1, Nk
DO J = 1, Nj
DO I = 1, Ni
!! work to do
END DO
END DO
END DO
The COLLAPSE(2) clause allows the parallelization of the perfectly nested first two loops where the compiler forms a
single loop with size Nk*Nj and then parallelizes it. In this way the work from the outer two loops
is better balanced among the threads.
There are specific conditions to be fulfilled to use the COLLAPSE clause (see, for more details on the conditions, this link for an exhaustive list of rules). The most important ones are:
- The loops must form a rectangular iteration space;
- The loops must be perfectly nested;
- The associated do-loops must be structured blocks. Their execution must not be terminated by an EXIT statement.
