Developer’s guide

• Definition of new Xtrack-compatible beam elements
  • Introduction
  • Definition and management of the data structure
    • Allocation of beam elements on CPU or GPU
    • Python access to beam-element data
    • Underlying xobject
    • Custom __init__ method
  • Definition of the tracking function
    • Accessing beam-element data from C
    • Writing the tracking code
  • Data recording in beam elements
    • Introduction
    • Recording of an individual table
    • Recording of multiple tables
    • Internal record for elements used in standalone mode
  • Lost particles state codes
• Multiplatform programming with xobjects
  • Data management
    • Definition of a simple data structure class
    • Allocation of a data object on CPU or GPU
    • Access to the data
    • Numpy-like access
  • Kernel functions in C
    • Autogenerated C API
    • Writing a C kernel
    • Compiling the kernel
    • Calling the kernel
    • Inspecting the source code
• Code autogeneration
  • vectorize_over block
  • only_for_context directive
  • gpukern directive
  • gpuglmem directive
• Numerical reproducibility
  • Using conda to create a numerically reproducible environment
• Continuous testing of xsuite
  • Setup of the test runner machine
    • An OpenStack GPU-capable machine
    • Configure the test machine
    • Install the GitHub Actions runner