Instructions

1. Set the initial state of the universe in initial_state.txt (see next section for format).
   • You can generate a random initial state by running python3 input.py > initial_state.txt. Set the number of objects within input.py (the variable is called n).
2. Set the desired parameters at the top of gravity.cpp.
   • Set all to false to accurately evaluate runtime, which is printed in the .err file.
   • Set OUTPUT_VIDEO_FRAMES to true if you want to see the video (see Step 4).
   • Set CHECK_MOMENTUM to true to calculate momentum and print it in .err.
   • Set BENCHMARK_PAPI to true to use PAPI, and set PAPI_TYPE_DESCRIPTION. The result is printed in the .err file.
3. Run using sbatch job.sh.
   • You can change the number of nodes and ranks there.
4. If OUTPUT_VIDEO_FRAMES is true, the .out file will contain pixel data corresponding to video frames. Use ffmpeg to generate the video.
   • Download the .out file.
   • Run cat <name of .out file> | ffmpeg -y -f rawvideo -pixel_format gbrp -video_size 1024x768 -i - -c:v h264 -pix_fmt yuv420p video.mov.
   • Open video.mov with a video player.
   • The version of ffmpeg in the repository is for Windows. Visit https://ffmpeg.org/download.html for other versions.
   • You may need to add .../ffmpeg/bin/ to the PATH environment variable.

Format for initial_state.txt

Syntax:

• First line: num_time_steps num_objects
• Next num_objects lines: mass radius pos.x pos.y pos.z vel.x vel.y vel.z

Meaning:

• num_time_steps: Number of time steps for which to run the simulation
• num_objects: Number of objects in the initial state of the universe
• Each of the lines after the first line gives the initial state of one object
• mass: Mass of object
• radius: Radius of object
• pos.x, pos.y, pos.z: x-, y-, and z-components of the object's initial position
• vel.x, vel.y, vel.z: x-, y-, and z-components of the object's initial velocity