Skip to content

Latest commit

 

History

History

sysimg

ReadMe.md

This does not work: yes, in terminal it does, and PythonPlot is much faster. But when using this as an IJulia kernel, the kernel does not start (Precompiling IJulia [..] ERROR: LoadError: OutOfMemoryError() and Exception: EXCEPTION_ACCESS_VIOLATION at 0x7ffedc8a89ae -- jl_table_assign_bp at C:/workdir/src\iddict.c:48 in expression starting at C:\Users\tfiers\.julia\packages\IJulia\6TIq1\src\IJulia.jl:36).

For archival purposes:

These scripts can be used to generate a custom "system image" with PackageCompiler.jl, like here.

The goal is to have to wait less long for package imports and first function calls in a fresh Julia session.

Build

To build the image, run, in this directory:

julia build.jl

This takes 10+ minutes.

Use

To use the generated system image:

  • On the command line, in the repo root:

    julia --sysimage=sysimg/mysys.dll

  • As a 'kernel' in Jupyter: see the relevant IJulia docs.
    Make sure to also add the flag --project=@..

Notes

  • If on Linux or MacOS, you can replace the .dll extension inside these scripts with .so or .dylib respectively (though leaving it as is will also just work).

  • For me, the generated Jupyter kernel definition file is located at:
    C:\Users\tfiers\AppData\Roaming\jupyter\kernels\julia-mysys-1.9\kernel.json

  • A relevant / helpful project: https://github.com/terasakisatoshi/sysimage_creator