Φ-Flow version 2 is the first major update since public release.
It introduces a functional style for writing simulations and provides lots of utility to go along with it. The most fundamental addition is a new Tensor class that wraps native tensors and extends them by named and typed dimensions. This makes code simulation more generic and robust, and less prone to errors.
New major features
General & UI
API documentation with pdoc
phi.verify() and verify.py
Matplotlib fallback GUI
ModuleViewer
Physics & Fields
FLIP simulation functions and demo
Improved PointCloud support
Field I/O, batched Scenes
Unified SampledFields
Math
Profiling
TensorFlow 2 with eager execution
Tensors with named and typed dimensions
Automatic conversion from NumPy
Automatic dtype casting
Automatic reshaping based on dimension type and name
Sparse matrix generation
PyTorch / TensorFlow parity
Experimental Jax support
Non-uniform tensors
Delayed allocation and stacking
Unified gradient computation
Context blocks for FP precision and backend selection
Φ-Flow version 2 is the first major update since public release.
It introduces a functional style for writing simulations and provides lots of utility to go along with it. The most fundamental addition is a new
Tensorclass that wraps native tensors and extends them by named and typed dimensions. This makes code simulation more generic and robust, and less prone to errors.New major features
General & UI
Physics & Fields
Math