EXACT Mini Apps

We collected traces for two applications from the DOE codesign center EXACT: Center for Exascale Simulation of Combustion in Turbulence. These two applications are based on BoxLib library.

EXACT CNS

The first application integrates the Compressible Navier Stokes (CNS) equations with constant viscosity and thermal conductivity. We ran the program with two dataset sizes in a strong scaling mode. The first run parameters are as follows: 

  • number of cells: 512
  • maximum grid size: 64
  • number of steps: 5

Available traces:

CNS 16 processes x 4 threads

CNS 64 processes x 1 threads

CNS 64 processes x 4 threads

CNS 256 processes x 1 threads

CNS 256 processes x 4 threads

The second run parameters are as follows: 

  • number of cells: 1024
  • maximum grid size: 64
  • number of steps: 5

Available traces:

CNS Large 64 processes x 4 threads

CNS Large 256 processes x 1 threads

CNS Large 256 processes x 4 threads

CNS Large 1024 processes x 1 threads

Graphical Presentation of CNS Boxlib Communication Pattern

The following figure shows the amount of data communicated between 1024 cores (for the large dataset). (click on the figure for a higher quality variant)

EXACT Multigrid

The second application is a multigrid solver based on BoxLib. We ran the application with two input sizes. The first run parameters are as follows:  

  • number of cells: 512
  • maximum grid size: 64
  • number of steps: 5

Available traces:

MultiGrid 16 processes x 4 threads

MultiGrid 64 processes x 1 threads

MultiGrid 64 processes x 4 threads

MultiGrid 256 processes x 1 threads

MultiGrid 256 processes x 4 threads

 

The second run parameters are as follows: 

  • number of cells: 1024
  • maximum grid size: 64
  • number of steps: 5

Available traces:

MultiGrid Large 64 processes x 4 threads

MultiGrid Large 256 processes x 1 threads

MultiGrid Large 256 processes x 4 threads

MultiGrid Large 1024 processes x 1 threads

 

Graphical Presentation of Multigrid Communication Pattern

The following figure shows the amount of data communicated between 1024 cores (Large problem size). (click on the figure for a higher quality variant)