-
Notifications
You must be signed in to change notification settings - Fork 0
Benchmarks & Validation
LiranOG edited this page May 9, 2026
·
9 revisions
GRANITE v0.6.8 | ← Architecture Overview | CHANGELOG Summary →
Full data from
docs/BENCHMARKS.md
All results from: Intel i5-8400 (6 cores), 16 GB DDR4, WSL2 Ubuntu 22.04, GCC 11.4 (-O3 -march=native), OpenMPI 4.1.2, 6 OMP threads.
Tests CCZ4 moving-puncture evolution. BL initial data. Copy BCs. ko_sigma=0.1. kappa1=0.02.
| Quantity | Value |
|---|---|
| ‖H‖₂ at t = 0.75M | 1.083 × 10⁻² |
| ‖H‖₂ at t = 500M | 1.277 × 10⁻⁴ |
| Reduction | ×84.8 |
| NaN events | 0 |
| Wall time | ~497 min |
| Throughput | 0.017 M/s |
| Quantity | Value |
|---|---|
| ‖H‖₂ at t = 0.375M | 1.855 × 10⁻² |
| ‖H‖₂ at t = 120M | 1.039 × 10⁻³ |
| Reduction | ×17.9 |
| NaN events | 0 |
Equal-mass BBH. Two-Punctures. Sommerfeld BCs. p_t = ±0.0840 (d=10M). ko_sigma=0.1.
| step | t [M] | α_center | ‖H‖₂ | Blocks |
|---|---|---|---|---|
| 2 | 3.125 | 0.81207 | 4.929e-04 | 3 |
| 4 | 6.250 | 0.86547 | 8.226e-04 | 4 |
| 32 | 50.00 | 0.96214 | 1.571e-04 | 4 |
| 64 | 100.0 | 0.96668 | 7.540e-05 | 4 |
| 128 | 200.0 | 0.96656 | 4.095e-05 | 4 |
| 256 | 400.0 | 0.97264 | 1.960e-05 | 4 |
| 320 | 500.0 | 0.96748 | 1.341e-05 | 4 |
‖H‖₂ reduction: ×61.3 | NaN events: 0 | Wall time: 98.9 min | Throughput: 0.0843 M/s
| step | t [M] | ‖H‖₂ | Blocks |
|---|---|---|---|
| 4 | 4.167 | 2.385e-03 (peak) | 4 |
| 60 | 62.50 | 1.682e-04 | 4 |
| 200 | 208.3 | 5.180e-05 | 4 |
| 480 | 500.0 | 3.538e-05 | 4 |
‖H‖₂ reduction: ×67.4 | NaN events: 0 | Wall time: 496 min
git clone https://github.com/LiranOG/Granite-NR.git && cd Granite-NR
python3 scripts/run_granite.py build
python3 scripts/health_check.py
# Single puncture — 64³, t=500M
python3 scripts/run_granite.py run --benchmark single_puncture
# BBH — 64³, t=500M (~99 min)
python3 scripts/run_granite.py run --benchmark B2_eq| Platform | Cores | Config | Throughput | t=500M wall time |
|---|---|---|---|---|
| i5-8400 desktop | 6 | BBH 64³ | 0.084 M/s | ~99 min |
| i5-8400 desktop | 6 | BBH 96³ | 0.017 M/s | ~496 min |
| 2× Xeon Gold (128 cores) | 128 | BBH 128³ | ~1.4 M/s (proj.) | ~6 min (proj.) |
| H100 SXM (post GPU) | — | BBH 256³ | ~50 M/s (proj.) | <10 sec (proj.) |
Defined in benchmarks/validation_tests.yaml:
| Test | Pass Criterion | Status |
|---|---|---|
| gauge_wave | 4th-order convergence; amplitude drift < 1e-6 | Config exists; no results yet |
| single_puncture | horizon mass drift < 1e-4 | ✅ Validated |
| binary_equal_mass | phase error < 0.01 rad vs. SXS:BBH:0001 | 🔶 Planned v0.7 |
| balsara_1 shock tube | L₁ error < 1e-3 | ✅ Unit test passes |
| magnetic_rotor | ∇·B < 1e-14 | ✅ Unit test passes |
| radiative_shock_tube | M1 vs. analytic L₂ error < 2% | ⚪ M1 not active yet |
See also: Scientific Context | HPC Deployment
v0.6.8 · Repository · Issues
- 🩺 Simulation Health & Debugging
- 📊 Benchmarks & Validation
- 🗂️ AMR Design
- 🖥️ HPC Deployment
- 🌀 VORTEX Engine
"Simulate the unimaginable."