Scripts and data for studying scaling of PyClaw on KAUST's Cray XC40
To investigate:
- Dynamic loading
- Separate costs of various parallel parts
- Weak scaling
- 2D/3D
- fixed # of steps/adaptive stepping
- Classic/SharpClaw solvers (and influence of larger WENO stencils)
- Strong scaling
- All the same as above
- Single-node efficiency?