Compiling / Python


Julia normally makes heavy use of just-in-time (jit) compilation to compile efficient versions of functions for any given input types. This has a side effect that the second time you run a given function with certain variable types is much faster than the first time, because the efficient compiled machine code is cached.

It is possible to store a system image file that contains compiled versions of functions from a Julia package, using the PackageCompiler package. This will greatly reduce the time to load and run the first instance of the already compiled functions.

To help automate this process, you can use the compile() function. It is recommended that you precompile Plots at the same time.

using ThreeBodyTB; using Plots; ThreeBodyTB.compile()

This will run the examples in order to compile the appropriate functions into the memory, and then create a file called in ~/.julia/sysimages/. To load the sysimage, run Julia as

julia –sysimage ~/.julia/sysimages/

Note that the compilation takes several minutes and significant disk space. If you want to update the code, you will have to delete and remake the sys image.


While running ThreeBodyTB using Julia is the easiest option, I understand that many users have existing codes in Python. Therefore, we have created a separate github package in python, TB3PY, that hosts a wrapper that allows the user to call the Julia functions directly like they were in python. It also has a more pythonic interface and integration with jarvis-tools.

This works by using the PyJulia interface. The wrapper will, if necessary, a) download & install Julia b) download & install ThreeBodyTB.jl, and c) create a system image for fast loading in python.