- defined ordering of headers
- lbm split kernels produced different ordering of assignments due to
  usage of a set

- new time loop caches all kernel functions with their argument dict
  -> inner loop just calls C functions with cached kwargs

- not generic enough - does not work if there a multiple blocks

- better latex display for indirect accesses
- new field type 'custom': only custom fields can be accessed indirectly
  no static bounds check possible for custom fields

- support aligned load/stores
- nontemporal stores
- aligned memory allocation for arrays and temporary buffers

- vectorization for loops with ranges that are not a multiple of vector width
- vectorization for variable sized loops if special transformation
  replace_inner_stride_with_one is run

- when running multiple pystencils instances, sometimes errors happened
  because one process might have partially written a cached file, which
  is already read before writing was finished
-> switched to "atomic write" (only on linux yet) that uses os.rename
   which is guaranteed to be atomic

- removed warnings
- added flake8 as CI target

- test run again
- notebooks not yet

- previously all objects where cached by id()
- for waLBerla simulations in each time step a new np.array view
  is created from the waLBerla field. Each of these views has a
  different id -> caching did not work for waLBerla setups
- changed hash for numpy arrays: instead of id, a tuple of
  (dataPtr, strides, shapes) is used as hash input

- scaling interface width eta instead of surface tensions tau to correct
  interface profile & surface tensions

João Victor Tozatti Risso authored 7 years ago and Martin Bauer committed 7 years ago

Concept: Generate code involving the (un)packing of fields (from)to linear
(1D) arrays, i.e. (de)serialization of the field values for buffered
communication.

A linear index is generated for the buffer, by inferring the strides and
variables of the loops over fields in the AST. In the CPU, this information is
obtained through the makeLoopOverDomain function, in
pystencils/transformations/transformations.py. On CUDA, the strides of
the fields (excluding buffers) are combined with the indexing variables to infer
the indexing of the buffer.

What is supported:
    - code generation for both CPU and GPU
    - (un)packing of fields with all the memory layouts supported by
    pystencils
    - (un)packing slices of fields (from)into the buffer
    - (un)packing subsets of cell values from the fields (from)into the buffer

Limitations:

- assumes that only one buffer and one field are being operated within
each kernel, however multiple equations involving the buffer and the
field are supported.

- (un)packing multiple cell values (from)into the buffer is supported,
however it is limited to the fields with indexDimensions=1. The same
applies to (un)packing subset of cell values of each cell.

Changes in this commit:

- add the FieldType enumeration to pystencils/field.py, to mark fields
of various types. This is replaces and is a generalization of the
isIndexedField boolean flag of the Field class. For now, the types
supported are: generic, indexed and buffer fields.

- add the fieldType property to the Field class, which indicates the
type of the field. Modifications were also performed to the member
functions of the Field class to add this property.

- add resolveBufferAccesses function, which replaces the fields marked
as buffers with the actual field access in the AST traversal.

Miscelaneous changes:

- add blockDim and gridDim variables as CUDA indexing variables.

- loop cutting
- simplification of conditionals inside loop

Fixed a severe bug. Renamed makePythonFunction of the llvm backend.
Deleted code duplicity.

- renaming because of clashes with types.py from other packages

- sympys piecewise defined functions are mapped to blend instructions
- cast function is now a class
- several bugfixes

- first vectorization tests are running
- type system: use memoized getTypeOfExpression
- casts are done using sp.Function('cast')
- C backend adapted for vectorization support
- AST nodes can required optional headers

- cache relied on uniqueness of  python id()
- id may be reused if object is freed
-> object must be held alive
-> kernel keeps all it arguments it was ever called with, alive (problematic in terms of memory consumption)

- VC 15 (2017) does not set an environment variable any more
- fallback to filesystem search

- this extension is necessary for more generic boundary treatment
- cells can now be structs, i.e. contain different data types
- instead of having numeric index dimensions, one can use the index per cell to adress struct elements

- use data type class consistently instead of strings (in TypedSymbol, Field and jit module)
- new datatype class is based on numpy types with additional specifier information (const and restrict)
- translation between data type class and other modules (numpy, ctypes)

- getShearRelaxationRate is a free function now -> works also with cumulant methods
- better error message when calling kernels with wrong or too few parameters
- entropic & incompressible is not working by default due to pdf shift -> added NotImplemented exception
- new creation function for 'raw_mrt' where all relaxation rates can be independently specified
- enhanced entropic creation funtion, supports omega output field now