forest-dynamic-vortex¶
英語 | 日本語
First dynamic-AMR validation for the leaf forest: a Lamb/Gaussian vortex advected at a constant background velocity U0 across a PERIODIC quasi-2D box. The refinement sensor flags the vortex core; the fine patch must TRACK the vortex (refine ahead, coarsen behind, with a buffer so it does not outrun the patch between regrids). docs/design/dynamic-amr-design.md D1/D2/D4.
概要: Lamb 渦を動的 AMR で追従させる。細分パッチが渦を取り逃さないかを見る。
Geometry¶

Animation¶
Fields¶
この計算が最後の出力ステップで書き出した場をそのまま描いたもの。3D ケースは中央断面。


Build & run¶
cmake -S native -B native/build && cmake --build native/build -j
./native/build/orheo-forest-dynamic-vortex
このページの図を作った計算のコスト(Miyabi-G Grace CPU (72 cores, -stdpar=multicore)): 11.0 s wall。
Validation¶
| Tier | T2 — 単点比較 |
| QoI | 渦の保持・追従 |
| Reference | Lamb 渦(定性) |
| Gate | — ゲートなし |
Source & run details¶
| Item | Value |
|---|---|
| Target | orheo-forest-dynamic-vortex |
| Source | native/benchmarks/forest_dynamic_vortex.cpp |
| Lattice | — |
| GPU twin | yes (orheo-forest-dynamic-vortex-gpu) |
| Modules | dynamic_amr.hpp, forest_regrid.hpp, forest_step.hpp |
| Gallery run | 2026-08-30 @ 38ad1eaa |
| Args | out |
| Wall time | 11.01 s |
| CPU time | 540.11 s |
| Device | Miyabi-G Grace CPU (72 cores, -stdpar=multicore) |
Computation log¶
この計算の実際の出力(先頭と末尾)。
dynamic-vortex: N=64 (Nb=8 Nbz=2 kNb=8) periodic, tau=0.505 U0=0.0500
steps=600 N_regrid=20 C=0.060 R=2.50 b_buf=1 H_coarsen=1 max_level=1
[D1] max|dynamic - static| = 0.000e+00 over 300 steps -> PASS (bit-identical)
[dynamic1] final peakvort=2.3212e-02 regrids=30 cum_mass_drift=2.336e-09 cum_mom_drift=6.324e-11 max_fresh_rough=2.935e-02 leaves=233(fine=120)
[dynamic3] final peakvort=2.3492e-02 regrids=30 cum_mass_drift=2.306e-09 cum_mom_drift=1.133e-10 max_fresh_rough=2.142e-02 leaves=275(fine=168)
[fine] final peakvort=2.5244e-02 regrids=0 cum_mass_drift=0.000e+00 cum_mom_drift=0.000e+00 max_fresh_rough=0.000e+00 leaves=1024(fine=1024)
[coarse] final peakvort=2.0026e-02 regrids=0 cum_mass_drift=0.000e+00 cum_mom_drift=0.000e+00 max_fresh_rough=0.000e+00 leaves=128(fine=0)
CSVs written to out