Optimization of projector application
The application of projectors (currently in nlefsm, and soon in the forthcoming PAW implementation) might cause inefficiencies in parallelization. Several strategies can be used to ameliorate the problem.
Blueprint information
- Status:
- Complete
- Approver:
- Alberto Garcia
- Priority:
- Medium
- Drafter:
- Alberto Garcia
- Direction:
- Approved
- Assignee:
- Alberto Garcia
- Definition:
- Approved
- Series goal:
- Accepted for 4.1
- Implementation:
- Implemented
- Milestone target:
- 4.1-b2
- Started by
- Alberto Garcia
- Completed by
- Alberto Garcia
Whiteboard
Initial efforts are addressed at nlefsm in the 4.1 series. Branch: lp:~albertog/siesta/4.1-nl.
The main problem is that the loop over KB projectors is "global", and a substantial amount of work, proportional to the size of the system, has to be done by all MPI processes.
The initial steps in nlefsm optimization have reduced the cpu weight of that loop by:
- Pre-computation of the atoms really involved in KB overlaps with the orbitals handled by the process.
- Early exit from the loop after simpler tests
- Other minor optimiizations
Further progress is now associated to blueprint "new-distributi