[Next] [Previous] [Top]
Early Parallel Programming Efforts
- Control parallel (PFP code on BBN TC2000)
- Shared interleaved storage of grid arrays
- Lock-protected parallel charge accumulation
- Data parallel (CM Fortran code on CM200)
- Gather communication is 30-40% of execution time
- Scatter-add communication is 20-30% of execution time
Gyrokinetic PIC Simulation on Massively Parallel Processors - (lecture given 6/23/95)
[Next] [Previous] [Top]
Generated with CERN WebMaker