关于GCU、沐曦GPGPU、MLU、0卡V100资源4月7日恢复上架的公告>>> 关于共建具身智能开源数据集的倡议>>> 关于云脑任务中统一路径访问方式的公告>>> 关于将启智集群GPU资源迁移至智算集群的公告>>>

lxu 85ab466f4f initial commit		1 year ago
code	initial commit	1 year ago

data	initial commit	1 year ago

LICENSE	initial commit	1 year ago

README.md	initial commit	1 year ago

README.md

nlfff_gpu
- Usage
- Algorithm
- Contact

nlfff_gpu

A Large-Scale Nonlinear Force-Free Field Three-Dimensional Magnetic Fields Data of Solar Active Regions

We deploy a cuda version for nlfff. Then we do experiments with RTX 3090, which has achieved 800x times faster speed than CPU. The gpu-memory should be 4x times larger than the size of output 3D data.

Usage

run code/use.ipynb script, which include:

prepare input magnetic vector data (we provide a input sample in data/raw)
compile code
prepare boundary data
run grid1, grid2, grid3
run quality checking
check output files (we provide a output sample in data/pro)

This repository is a development repository, see code/use.ipynb for a usage example

Subsequent major releases will also be merged into the toolkit pynlfff.

Algorithm

The original CPU version code was Wiegelmann(wiegelmann@mps.mpg.de) NLFFF extrapolation algorithm

@article{Wiegelmann2012Nov,
    author = {Wiegelmann, T. and Thalmann, J. K. and Inhester, B. and Tadesse, T. and Sun, X. and Hoeksema, J. T.},
    title = {{How Should One Optimize Nonlinear Force-Free Coronal Magnetic Field Extrapolations from SDO/HMI Vector Magnetograms?}},
    journal = {Sol. Phys.},
    volume = {281},
    number = {1},
    pages = {37--51},
    year = {2012},
    month = nov,
    issn = {1573-093X},
    publisher = {Springer Netherlands},
    doi = {10.1007/s11207-012-9966-z}
}

The GPU version of NLFFF was here

@misc{Zhang2022,
  author = {Xinze, Zhang},
  title = {The cuda version for nlfff},
  year = {2022},
  publisher = {GitHub},
  journal = {GitHub repository},
  howpublished = {\url{https://github.com/deepsolar/nlfff_gpu}},
  commit = {83d83f6ade2bc0eed429e1f90ee39636460bbedc}
}