peijin94 / fastqsl

A module to calculate Q Factor with GPU

FastQSL

This module is licensed under a CC BY-NC-SA 4.0 License.

A module to calculate Q Factor with GPU, FastQSL

The idea is to do the most computational intensive work in GPU with compiled code (TraceBline.cu), and do the rest of the complex but not computational intensive work in Python.

A CPU version is also provided https://github.com/el2718/FastQSL.

Dependencies

• CUDA >= 11
• PyCuda
• cupy

module load for cluster computing (CUDA/11.7.0 Xvfb/1.20.13-GCCcore-11.2.0 GCCcore/11.3.0 xorg-macros/1.19.3-GCCcore-11.3.0 X11/20220504-GCCcore-11.3.0 LLVM/14.0.3-GCCcore-11.3.0 Mesa/22.0.3-GCCcore-11.3.0)

Step by step Tutorial

Preparation

Hardware

Buy a computer with Nvidia GPU

Software

• Install GPU driver, https://www.nvidia.com/Download/index.aspx, for 'apt' based Linux, try sudo ubuntu-drivers autoinstall.
• https://developer.nvidia.com/Cuda-downloads Download cuda v10.2! (don't install the newest)
• Install anaconda https://docs.anaconda.com/anaconda/install/windows/
• Install git
• Install c/c++ compiler, for Windows https://visualstudio.microsoft.com/visual-cpp-build-tools, for linux sudo apt install gcc

Build Enviroment

Test conda

Start a 'Anaconda Powershell prompt' (in Windows), a terminal (in Linux). Type command:

conda --version

You should get "conda x.x.x", which means conda is successfully installed.

New conda-env

Create a new env named 'fastqsl' and use 3.8.2 version of python:

conda create -n fastqsl python=3.9
conda activate fastqsl

after this you should be able to see a '(fastqsl)' in your command line.

Install dependencies

conda install numpy matplotlib scipy jupyterlab tqdm
python -m pip install cupy-cuda117 ipython-autotime pyvista

Run

Download code

git clone https://github.com/Pjer-zhang/FastQSL.git

Run in jupyterlab

In the env created last section, type command

cd FastQSL
python -m jupyterlab --port 9999

Normally it will pop up a browser automatically, if it didn't, you need to it manually by copy the url in the output of the command to a browser (Chrome recomended)

Enjoy.

---

Cite as

• Zhang, P., Chen, J.*, Liu, R. and Wang, C., FastQSL: A Fast Computation Method for Quasi-separatrix Layers. 2022, The Astrophysical Journal, 937, 26

@ARTICLE{2022ApJ...937...26Z,
       author = {{Zhang}, PeiJin and {Chen}, Jun and {Liu}, Rui and {Wang}, ChuanBing},
        title = "{FastQSL: A Fast Computation Method for Quasi-separatrix Layers}",
      journal = {\apj},
     keywords = {Solar magnetic fields, GPU computing, 1503, 1969},
         year = 2022,
        month = sep,
       volume = {937},
       number = {1},
          eid = {26},
        pages = {26},
          doi = {10.3847/1538-4357/ac8d61},
       adsurl = {https://ui.adsabs.harvard.edu/abs/2022ApJ...937...26Z},
      adsnote = {Provided by the SAO/NASA Astrophysics Data System}
}

fastqsl's Issues

Missing Data

Hey! Thanks for sharing this project! I wanted run your demos but failed because they all depend on sample data, e.g. "B0.sav". Is there any way to access these data sets?

Best regards, Max