Installation

For PAW-ATM 2020

All results and benchmarks were computed using the the 'distributed' folder in the master branch.
The main repository for the history of the code can be found in this repo.

40M Grid File

We have provided a 40 Million fine grid partGrid40M_unpartitioned which can be downloaded from Google Drive. Reviewers are requested to use the mfpre partitioner with --quadtree flag along with their specified number of partitions they want.

Dependencies

1. Julia 1.5.1

Julia Libraries

• ClusterManagers
• Distributed
• DistributedArrays
• StaticArrays
• Printf
• TimerOutputs

Configuration of the test case.

1. Run the partitioner on the grid file provided in Google Drive.
2. A folder called point will be generated in the directory of the partitioner.
3. Using mv shift the point to a desired location and save its path.
4. Go to the distributed folder in the master branch
5. Rename config.json.example file to config.json with the following configuration.

"cfl": 0.2
"max_iters": 1000
"type": "quadtree"
"vl_const": 50
"mach": 0.85
"aoa": 1

5. In batchscript.sh

• If using Local manager with CPU affinityinstead of SLURM, change runner_local.jl to runner.jl
• Set /path/to/julia/executable to required Julia executable
• Set x to number of processes on which the code will run.
• Set points to the total number of points, which is 39381464 for partGrid40M_unpartitioned
• Set /path/to/grid to the point folder

Execution

1. chmod +x batchscript.sh to make the execution script an executable
2. Run batchscript.sh