Performance Optimisation and Productivity
A Centre of Excellence in HPC
Pils (DLB) version
Version's name: Pils (DLB) version ; a version of the Pils program.Repository: [home] and version downloads: [.zip] [.tar.gz] [.tar.bz2] [.tar]
Patterns and behaviours: Implemented best practices: Compensate load imbalance with DLB ·
This is the dlb-adapted version of the pils program. As the baseline program,
it iterates on different phases, splitted by MPI_barrier()’s. Each phase
executes a given number of parallel regions (determined by the grain
parameter), which main body is a loop executing iterations of task-duration.
Changes in the source code:
- Include the DLB header file.
- Include a call to
DLB_Borrow()just before each parallel region.
Changes in the execution (at run.sh):
- Preloading the DLB library.
