Sailfish logo

Table Of Contents

Previous topic

About

Next topic

Media files

This Page

External links

Release notes

2013.1

  • 4 new outflow boundary conditions: Grad’s approximation (NTGradFreeflow), Yu’s method (NTYuOutflow), node copying method (NTCopy), Neumann’s BC (NTNeumann).
  • New boundary conditions based on the regularized LB dynamics (NTRegularizedDensity, NTRegularizedVelocity).
  • New no-slip boundary condition based on the Tamm-Mott-Smith approximation (NTWallTMS).
  • NTHalfBBWall is now supported when the AA memory layout is used.
  • Exact difference method (EDM) of applying body forces.
  • Model minimizing round-off errors (–minimize_roundoff)
  • New examples: four rolls mill, 2D lid-driven cavity with the entropic model, external geometry from a npy file, Taylor-Green flow, duct flow, channel flow, capillary wave, two-phase Poiseuille flow.
  • Restored the cutplane visualization engine for 3D simulations to a working state.
  • Mechanism of easy inclusion of additional code fragments (commit ae7bdff4)
  • Regularized LB dynamics (–regularized).
  • Entropic LB model fixes.
  • Compatiblity fixes for sympy-0.7.2 and numpy-1.7.0.
  • Support for distributed simulations on LSF clusters.
  • D3Q27 lattice.
  • Support for drawing walls in the pygame visualization interface.
  • New on-line, remote visualization tool (utils/visualizer.py).
  • LinearlyInterpolatedTimeSeries data source for boundary conditions.
  • Indirect node addressing mode for sparse geometries.
  • Possibility to use neighbor link tagging instead of orientation (NTHalfBBWall in more complex geometries).
  • SIGHUP handling (creates a checkpoint dump).
  • Shuffle propagation (speeds up execution on some lower-end GPUs).
  • Force objects (on-GPU calculations of forces on solid objects).
  • On-GPU flow statistics (kinetic energy, Reynolds stresses, etc).

2012.2

  • Added support for dynamic forces (time-dependent, or dependent on the macroscopic fields).

2012.1

  • Added support for the AA access pattern, which provides substantial memory savings (~50%).
  • Non-fluid nodes are masked in the output fields.