CaLES v2.0.0‑smartflow — First Stable Release with SmartFlow Integration

This release marks the first stable integration of CaLES with the released SmartFlow v1.0.0, a modular framework for multi-agent deep reinforcement learning (DRL) applied to turbulence modeling, flow control, and numerical algorithm development.

By coupling SmartFlow with CaLES, this release provides a turnkey platform for reinforcement learning research.

---

🚀 What's New

• SmartFlow v1.0.0 Integration
  • Seamless coupling between CaLES (Fortran) and SmartFlow (Python)
  • Support for multi-agent DRL

CaLES v2.0.0 – First Stable Release

CaLES v2.0.0

CaLES (Canonical Large‑Eddy Simulation) is a fully GPU‑accelerated, finite‑difference DNS/LES solver for incompressible wall‑bounded flows in massively parallel environments. This major release builds on the CaNS DNS codebase and introduces production‑quality LES capability.

---

🚀 Highlights

• LES support with wall‑models

  • Classical Smagorinsky & dynamic Smagorinsky models
  • Van Driest near‑wall damping
  • Log‑law wall model option

• Flexible temporal schemes

  • Low‑storage, explicit 3rd‑order Runge–Kutta
  • Implicit Crank–Nicolson for viscous terms (1D or 3D)

• High‑performance solvers

  • Fast direct Poisson/Helmholtz solver via eigenfunction expansions
  • FFTW/cuFFT (and CPU fallback) with automatic transform selection
  • Gauss‑elimination for non‑uniform Z‑grids
  • FFT‑based 3D Helmholtz option in fully‑implicit mode

• Massive parallelism & GPU support

  • MPI + cuDecomp pencil decomposition for multi‑GPU runs
  • 2DECOMP&FFT for CPU global transposes & I/O
  • OpenACC‑annotated kernel loops