Psyclone

1) 250. Change setup.py (used for pip installation) to only use
version 0.0.7 of fparser.

1) 91 and PR 111 Remove unnecessary __init__.py files.

2) PR 107 Fix errors in Fortran test files revealed by update
to fparser 0.0.6. Update to catch new AnalyzeError raised by
fparser.

3) 90 and PR 102 Add support for integer, read-only, scalar
arguments to built-in kernels. Add new inc_X_powint_n Built-in
that raises a field to an integer power.

4) 67 Bug fix - printing Invoke object caused crash.

5) 112 and PR 113 - add travis support for codecov coverage tool

6) 118 and PR 122 - add .pylintrc file in tests directory
to disable some warnings that we don't care about (for testing
code).

7) 117 and PR 125 - bug fix for passing kernel arguments by value
in the GOcean API.

8) 105 and PR 115 - add support for coloured loops when
performing redundant computation.

9) 84 and PR 128 - make w2v and wtheta discontinuous function
spaces

10) PR 93 - add support for a swap-loop transform for the GOcean
API.

11) PR 152 - add pylintrc so that pylint's line-length limit
matches that of pep8.

12) PR 143 - bug fix for DAG generation in the GOcean API.

13) PR 153 - added -v option to the psyclone script to show the
version

14) PR 136 - add support for intergrid kernels (required for
multigrid)

15) PR 149 - add support for GH_READWRITE access for
discontinuous fields

16) PR 163 - updating PSyclone to work with the latest version of
the parser (0.0.7) as the API has changed

17) PR 169 - add support for specifying stencils in the GOcean
kernel meta-data.

1) 53 and PR 88 Add a version file so that PSyclone does not
need to be installed to generate documentation.

2) 94, 96, 98 and PR 97 Make name="name" support in invoke
calls more robust e.g. names must be valid fortran and match
should be insensitive to case.

3) 76 and PR 89 Add support for the new LFRic quadrature
api. Currently only gh_quadrature_xyoz is supported.

4) 99 and PR 100 Make the author list for the user manual
strictly alphabetical.

5) 50 and PR 78 Implement a redundant-computation transformation.
This enables a user to request that a kernel (or kernels) is
computed out into the halo region and any affected halo swaps
be updated (and/or removed) as required.

6) 103 and PR 104. Bug fix to ensure that a halo swap is not
inserted if read annexed dofs are already clean due to a previous
write to the L1 halo.

1) 20 and PR 28 Add dependence analysis (and a Move
transformation) to PSyclone. This will subsequently allow for
the safe implementation of more complex schedule transformations
and optimisations.

2) 16 and PR 29 Update and add new PSyclone-generated builtins
to support the functionality required by the current LFRic code.

3) 43 and PR 44 Update of Met Office-specific install script to
symlink the generator.py file rather than copy it.

4) 40 and PR 47 Update example lfric/eg3 to use builtins and to
merge invokes together as much as possible

5) 45 and PR 49 Change of kernel metadata names from
evaluator_shape to gh_shape and modifications to its values.

6) 3 and PR 51 Restructure repository to allow installation via
pip. Due to this change we can now use travis and coveralls within
github for automated testing. Also, new releases are now
automatically uploaded to pypi (via travis) and status badges can
now be used so have been added to the README.md file (which is
shown on the main github page).

7) 55 and PR 66 Update the documentation to cover installation
on systems running OpenSUSE.

8) 61 and PR 71 Re-name existing Dynamo0.3 builtins to follow
a consistent scheme and add support for some new ones (including
setval_{c,X}, X_minus_bY etc.). See psyclone.pdf for the full list.

9) 32 and PR 39 Adds the ability to build the generated code
when performing Python tests. Generated code can now be checked
that it compiles as part of the development process helping avoid
making releases with errors in them. Note, existing tests still
need to be updated.

10) 54 and PR 63 The view() method (used to look at the internal
representation of a schedule) now outputs coloured text (for
easier viewing) if the termcolor package is installed.

11) 69 and PR 73 setup.py now uses an absolute path to avoid
failing tests (with file not found) in some environments when
running py.test from a different directory to the tests
themselves.

12) 59 and PR 75 added metadata support for inter-grid kernels
which will be used for the multi-grid code.

13) 82 and PR 83 modified a test to work with different versions
of graphviz which output files with different ammounts of white
space

14) 52 and PR 60 added support for evaluators. An evaluator is
specified by setting the gh_shape metadata in kernels
appropriately.

15) 85 and PR 86 - minor correction to documentation (section
on Inter-Grid kernel rules was in the wrong place).

1) 22 Updated PSyclone to support enforce_operator_bc_kernel (in
addition to the existing enforce_bc_kernel) for boundary
conditions. PSyclone no longer adds in boundary condition calls
after a call to matrix_vector_kernel, it is up to the user to add
them in the algorithm layer in all cases. Also fixes a bug
introduced in 1.4.0 issue 12.

1) 2 Add support for kernel meta-data changes required to support
Column-wise operators (Column Matrix Assembly).

2) 6 Implement support for Column Matrix Assembly (CMA) in
PSyclone.

3) 12 Add support for the any_w2 function space descriptor

4) 33 Update documentation referring to fparser.