lcodepy-team / lcodepy

Experimental quasistatic plasma wakefield simulation code, 2D and 3D geometry, for CPUs and GPUs

lcodePy

Overview

LCODE is a free software for numerical simulation of particle beam-driven plasma wakefield acceleration. LCODE is based on the quasistatic approximation, capable of simulation in 2D and 3D geometry, and can use GPUs and CPUs.

For now, this under active development. This is a complete overhaul of the old C version in Python.

• Website: https://lcode.info
• Documentation: https://lcode.info/docs

Instalation

We use Anaconda and we recommend installing lcode in a separate environment. Any other python installation should work fine, but has not been tested.

• Create a new environment and install the dependencies:

conda create -n lcode-env -c conda-forge numba numpy scipy matplotlib mpi4py

or

conda env create -f conda-env.yml  

where conda-env.yml is avalible in sources.

• Acivate the new environment:

conda activate lcode-env

• Optional: in order to run simulations on GPU, add cupy and other necessary packages to the line of dependencies when creating a new eviroment. Check in advance if you have the drivers for your GPU installed. It is not necessary to install CUDA Toolkit in advance. For any other questions about cupy, please check https://docs.cupy.dev/en/stable/install.html

conda create -n lcode-env -c conda-forge numba numpy scipy matplotlib cupy

• Install lcode:

pip install lcode

or download sources from GitHub and run the forlowing command in downloaded directory:

pip install .

---

You can also find a more mature 2D version of LCODE at http://lcode.info/.

Fill free to contact [email protected] for assistance with it.

lcodepy's Issues

Change I/O of the code to hdf5 (h5py) according to OpenPMD standart

Revise and merge Simulation class

Implement the laser scheme in Python for CPU

We need to immediately think about the subsequent transfer of the numerical scheme to GPU. To do this, we need to write the scheme in vector form. Ivan is supervising.

Fix performance issue in beam generator.

Merge beam generator

Test simulation quality comparing with lcode2d.C

Test tasks:
-Run all standard tests like test1, test2 and so on in both lcode2d.py and lcode2d.C and compare the fields and particle distributions.
-Identify the sources of possible errors in Py version.

Formulate a numerical scheme for laser calculation in 3D

Should ask Roman and Irina and discuss and write the numerical scheme. Roman and Ivan are supervising. I propose to simultaneously write the scheme in latex here for quick transfer to any necessary documents later.

Deadline: I suggest striving to complete this task in 4 weeks, so the deadline is October 10th.

Revise the Diagnostics class.

Add a dual plasma approach for the LCODE 3D (GPU)

Implement the laser scheme in Python for GPU

I haven't decided yet who will do it.

Change numba to cupy for kernel functions

Implement mpi parallelization (Lcode 2d)