SIMBI: 3D Relativistic Gas Dynamics Code

2D Relativistic Jet Simulation

jet_movie.mp4

2D Relativistic Shock Tube

shocktube2d.mp4

2D Rayleigh-Taylor in Newtonian Fluid

rt_creamer.mp4

1D Moving mesh techniques

moving_shocktube.mp4

Requirements

1. gcc >= gcc5 or clang >= clang5 (for c++17 support at the very least)
2. Cython
3. meson, pip install meson will usually suffice
4. A build system like make or ninja
5. HDF5 libraries
6. mypy, a static type checker
7. Python >= 3.10

For GPU capability

8. HIP/ROCm if wanting to run on NVIDIA or AMD GPUs, or just CUDA if running purely NVIDIA

Extras

a) CMasher for richer set of colormaps, blends into matplotlib

b) rich-argparse for pretty-print argparse

c) rich for pretty-printing console outputs

Quick setup guide

Installing

Easy Install

1. Run

   $ CC=<your_c_compiler> CXX=<your_cpp_compiler> python dev.py install [options]

Manual Install

1. Run

   $ CC=<your_c_compiler> CXX=<your_cpp_compiler> meson setup <build_dir> -D<some_option>

   from project root. It is important that this directory is not named build because the install call with create build dir for the pip installation part. For the -D<build_option> part, check the meson_options.txt file for available build options.

2. Run

   $ ninja -v -C <build_dir> install
   

   or

   $ meson install -C <build_dir>

3. If meson detected hip or cuda, the install script will install both the cpu and gpu extensions into the simbi/libs directory.

Important (!)

When compiling on a GPU, you must provide your GPU's respective architecture identifier. That is to say, if I am compiling on an NVIDIA V100 device with compute capability 7.0, I would build with:

# note the lack of a decimal
$ CC=<your_c_compiler> CXX=<your_cpp_compiler> python dev.py install --gpu-compilation --dev-arch 70 [options]
# or if manually installing
$ CC=<your_c_compiler> CXX=<your_cpp_compiler> meson setup <build_dir> -Dgpu_arch=70 -Dgpu_compilation=enabled [options]

Also, the meson.build script assumes c++20 by default. If your compiler does not support this, one must do a clean build - i.e., delete the original build directory generated by meson(!) - and run

$ CC=<your_c_compiler> CXX=<your_cpp_compiler> python dev.py install --cpp17
# or if manually installing
$ CC=<your_c_compiler> CXX=<your_cpp_compiler> meson setup <build_dir> -Dcpp_std=c++17 [options]

Running

Running a Configuration

4. If all is well, we can test. To test, try running the configuration scripts provided. For example:

   $ simbi run simbi_configs/examples/marti_muller.py --mode gpu --nzones 100 --ad-gamma 1.4 
   # or one could do 
   $ simbi run marti_muller --mode gpu --nzones 100 --ad-gamma 1.4
   # or 
   $ simbi run marti-muller --mode gpu --nzones 100 --ad-gamma 1.4
   # since the entry point is built to recursively search the simbi_configs/ folder for valid .py scripts
   # and dash-cased searches for file matches with underscores

   where --mode is a global command line option available for every config script, and --nzones and --ad-gamma are problem-specific options that are dynamically parsed based on whatever DynamicArg variables exist in the config script you create. Check out how to create one of these configuration scripts in the simbi_configs/examples/ folder! When creating your own configuration file, you must place it in a directory entitled simbi_configs/ and run simbi from your simbi_configs/ parent directory and simbi should auto detect your configuration and run the simulation.

   You can plot the above output by running

   $ simbi plot data/1000.chkpt.000_100.h5 "Marti \& Muller Problem 1" --field rho v p --tex

   The usual formula for plotting a checkpoint file is like so:

   $ simbi plot <checkpoint_file> "<name_of_physics_setup>" --field <field_string> [options]

   One can also do simbi clone --name <name_of_setup>, and a new skeleton configuration script will appear in the simbi_configs directory named <name_of_setup> that you can build off of.

5. ???

6. Profit

Physics Features (so far)

1. Special Relativistic and Newtonian Hydro up to 3D (set the regime property to classical or relativistic)
2. Supports user-defined mesh expansion / contraction (scale_factor & scale_factor_derivative methods)
3. Supports user-defined density, momentum, and energy density terms outside of grid (Implementing the dens_outer, mom_outer, AND edens_outer methods sets this)
4. Supports source terms in the Euler equations (implementing the sources property sets this)
5. Support source terms at the boundaries (implementing the boundary_sources property sets this)
6. Boundary conditions given as array of strings like so [bc_x1min, bc_x1max, bc_x2min, bc_x2max, bc_x3min, bc_x3max] where the supported boundary conditions are periodic, reflecting, outflow, inflow. If an inflow boundary condition is set, but no inflow boundary source terms are given, the code will switch to outflow boundary conditions to prevent crashes.
7. Can track a single passive scalar (implementing the passive_scalars property sets this)
8. Can insert an immersed boundary (Peskin 2002). It is impermeable by default. (Implementing the object_cells property sets this)
9. Gravity source terms (Implementing the gravity_sources property sets this)

TODO:

• Explore general IB in greater detail for sources and sinks!
• multi-gpu support
• MPI support