The Evolutionary Model of Physics Large-Scale Simulation on Parallel Dataflow Architecture

1
The Evolutionary Model of Physics Large-Scale
Simulation on Parallel Dataflow Architecture

• Dr. Andrey V. Nikitin
• Dr. Ludmila I. Nikitina
• Lomonosov Moscow State University

2
Physics Large-Scale Simulation

• Nonlinear 3D simulation of experimentally
  observed magnetohydrodynamic (MHD) burst in the
  DIII-D tokamak
• Fast (10-6 sec) processes
• Slow (10-2 sec) processes
• 3D calculations
• Different scales for time and space

3
Physics Large-Scale Simulation

• Model
• V velocity
• B magnetic field
• P - pressure

4
Physics Large-Scale Simulation

• Numerical model
• straight field line coordinate system

5
Physics Large-Scale Simulation

• Requirements to the numerical simulation process
• Inversion of matrixes 106
• Near real time
• Evolution of simulation process

6
Dataflow model of calculations

• Dataflow graph decomposition to layers

7
Dataflow model of calculations
8
Dataflow model of calculations
Write time
Search time
Total time
Effectiveness
9
Dataflow model of calculations

• Goal

10
Genetic algorithm
Algorithm
Partition matrix
Chromosome x  
11
MHD dataflow graph
12
Genetic algorithm
Convergence of GA (mutation)
13
Genetic algorithm vs. MK
Convergence of GA and MK
14
Conclusion

• GA was used to organize evolutionary computations
  of physics large-scale simulation on parallel
  dataflow architecture
• GA allows to find optimal distribution of data by
  modules with high effectiveness
• Code NFTC (1 iteration) SUN Ultra 400 Mhz/2 FPU
  50s, for dataflow 410-2s.
• Total time of typical computations
• Current (SUN Ultra Enterprise) 30 hrs
• Dataflow (modules - 102,module capacity - 106,PE
  104/10 GFlops) real time

15

• Questions?