- small (length < 5) arrays with shape and stride information had to be
  memcpy'd to the GPU before every kernel call
- instead of passing the information as arrays, the single elements are
  passed
- leads to more function arguments, but simplifies GPU kernel calls

-> changes in all backends required

- put all buffer related stuff into separate functions
- should be functionally equivalent

Dominik Ernst authored 6 years ago and Martin Bauer committed 6 years ago

Transformations on the Sympy equation level, that aim to reduce the amount of registers the Compiler needs. Primarily aimed at GPUs. The function livenessOptTransformation is a sequence of three sub transformations, which has been determined by a genetic optimization algorithm.
The file test_liveness_opts.py is a simple code that performs the transformation and tests whether the resulting equations are still generatable and compilable. Numerical results are not checked.

Make AssignmentCollection iteratable

See merge request software/pystencils!17

- field access are now equal if their underlying fields compare as equal
  not only if the id of their fields are equal

- bugfix in vector-typing of piecewise functions
- cast_function is now a sympy atom - fixes problems with sympy > 1.1
- replace_inner_stride_with_one is a bit faster now

- if a field access is printed in compass or index notation seems to
  depend on the sympy version
- removed doctests that rely on the printed form

- Added test for linear equation system and moved it to pystencils.utils

- added isotropic version

- new functions to easily extract stencil coefficients and visualize them
- Moved stencil functions from lbmpy to pystencils

- constants are typed - no need any more for this parameter

- hash function was not correctly implemented
- sometimes this caused a bug where latex version was not correctly displayed

-> test_numerical_1D notebook