What's new in ChaNGa 3.3 (Since 3.2)

===============================================================================

General changes:

- Move to Charm version 6.8.0

Physics Changes: (Many from Wadsley, Keller and Quinn, 2017)

- Adaptive artificial visoscity

- Smagorinsky diffusion

- Thermal diffusion

- Saitoh timestepping

- New SPH Kernels: Wendland, M6

- Geometric density averaged SPH forces.

- External gravitational potentials: constant g, patch of a disk, central body

- AGORA (2016) feedback model

- Damping force for glass creation

Parallel performance changes:

- Working GPU implementation

- Single precision floating point gravity

- SPH neighbor search improvements

===============================================================================

What's new in ChaNGa 3.2 (Since 3.1)

===============================================================================

General changes:

- Move to Charm version 6.7.1

- Restart from "tipsy binary" files.

- More flexible "dumpframe" movie making.

- Complete handling of > 2 billion particles, including 64 bit iOrders in
  the starlog file.

Physics Changes:

- Grackle cooling module.

- Jeans length pressure floor.

- Improved "beta cooling" for disks.

Parallel performance changes:

- More efficient smoothing cache.

- Better scaling for tree build.

- Better bookkeeping for multistep load balancing.

===============================================================================

What's new in ChaNGa 3.1 (Since 3.0)

===============================================================================

General changes:

- Support for greater than 2 billion particles.

- Support for NChilada file format. 

Hardware support changes:

- Remove support for CELL processor (Playstation).

- AVX instruction support.

Parallel performance changes:

- SMP load balancing using the Charm++ Ck_Loop Library

- More efficient parallel I/O using the Charm++ CkIO library.

- Faster domain decomposition.

- New load balancers
  - Better multistep load bookkeeping
  - HierarchicalLB for large core counts
  - MultistepNodeLB_notopo for node-based load balancing. 

===============================================================================

What's new in ChaNGa 3.0 (Since 2.0)

===============================================================================

Physics Changes:

- Gas Cooling: primordial hydrogen and helium, planetary disk cooling with
  cooling time proportional to orbital time.

- More efficient thermal energy equation solver.

- Star formation and feedback

- Timestepping fixes

-------------------------------------------------------------------------------

Hardware support changes:

- SMP aware: cores on an SMP node share tree data

- Use of 128 key type for deeper trees: enable with -DBIGKEYS

-------------------------------------------------------------------------------

Parallel Infrastructure Changes:

- Improved parallel treebuild

- Improved neighbor search algorithms

- Improved spatially aware load balancers

===============================================================================

What's new in ChaNGa 2.0 (Since 1.0)

===============================================================================

Physics Changes:

- Smooth Particle Hydrodynamics is now implemented.

- Hexadecapole expansions for the gravity calculation.

- Physical gravitational softening lengths for comoving simulations.

- Gravitational timestepping option.

- Non-cubical periodic boundary conditions.

- Adiabatically grow particle masses for equilibrium initial conditions.

-------------------------------------------------------------------------------

Output Changes:

- Output of movie files is available.

- Realtime visualization via liveViz is implemented.

- Output at specific redshifts or times.

-------------------------------------------------------------------------------

Hardware support changes:

- Support for CUDA.

- SSE2 implementation of force calculations.

- Better detection of 64 bit integer type.

-------------------------------------------------------------------------------

Parallel Infrastructure Changes:

- Multistep-aware load balancer.

- Rewrite of inter node caching to use the Charm++ Cache framework.

- Refactor of treewalking for easier support for new types of treewalks.

- Checkpointing using the Charm run time system.

- Overlapping treewalks.