I_Initializer

概览

I_Initializer 放置 AlgoPlasma 的粒子初始化工具。当前分成两类入口：一类在 Fortran 内部直接生成规则粒子分布，另一类由 Python 离线生成全局二进制粒子文件，再由 MPI 程序按本地子域筛选载入。

子模块

模块

数据来源

主要职责

I01_par_distribute

调用时给定的局部网格范围和每格粒子数

在每个网格单元内生成严格均匀的位置，并用 Maxwellian 分布初始化速度。

I02_par_init_and_load

output_init_particles_bin/*.bin 二进制文件

生成 T 形区域粒子初值，并在 MPI 并行计算中将粒子分配到对应子域。

坐标与数组约定

  • 粒子数组采用 par(1:6,...)，其中 1:3 为 x,y,z，4:6 为 vx,vy,vz。

  • I01 使用归一化网格间距 dx=dy=dz=1，位置坐标与网格单元索引直接对应。

  • I02 的二进制文件每个粒子保存 6 个 float64 值，Fortran 端按局部 il/iu 范围筛选。

  • I02 当前默认粒子数很大，适合作为离线初始化数据生成流程，不是轻量级回归测试入口。

测试状态

当前 tests/006_initializer 已包含 I01_par_distribute 测试， 用于验证规则粒子分布初始化。I02 更接近离线初始化数据生成和载入流程， init_particles_bin.py 页面已给出参考输出图， 因此当前不单独作为轻量回归测试项。

Overview

I_Initializer contains AlgoPlasma particle-initialization utilities. The current module family has two entry styles: Fortran-side procedural initialization for regular particle distributions, and an offline Python binary generator followed by an MPI-aware Fortran loader.

Submodules

Module

Data source

Main responsibility

I01_par_distribute

Local mesh bounds and particles per cell passed by the caller

Generate exactly uniform in-cell particle positions and Maxwellian velocities.

I02_par_init_and_load

Binary files under output_init_particles_bin/*.bin

Generate T-shaped-domain particle initial conditions and load them into MPI subdomains.

Coordinate and Array Conventions

  • Particle arrays use par(1:6,...): 1:3 are x,y,z and 4:6 are vx,vy,vz.

  • I01 assumes normalized mesh spacing dx=dy=dz=1 so positions map directly to cell indices.

  • I02 binary records store six float64 values per particle; the Fortran loader filters them by local il/iu bounds.

  • The I02 generator currently uses a very large default particle count, so it is an offline data-preparation workflow rather than a lightweight regression test.

Test Status

tests/006_initializer currently includes the I01_par_distribute test for regular particle-distribution initialization. I02 is closer to an offline initialization-data generation and loading workflow. The init_particles_bin.py page already includes reference output figures, so it is not listed as a separate lightweight regression test item for now.