Newer
Older
1) PR #2477 for #2463. Add support for Fortran Namelist statements.
2) PR #2367 for #2296. Extends ModuleInfo so that it can be queried
for the Routines contained in a Module.
3) PR #2435 for #2422. Add PSyIR GenericInterface to capture the
information about which routines implement a specific interface symbol.
4) PR #2457 for #2455. Splitting and renaming DynDofMaps from
dynamo0p3.py.
5) PR #2506. Update test and integration test dependencies.
6) PR #2508 for #2507. Add a symbols_to_ignore argument to some
symbol_table operations and use it to fix inline transformation issues.
7) PR #2505 towards #1010. Use the PSyIR Fortran backend for LFRic
loops and anything inside them.
8) PR #2516 for #2513. Resolve flake8 issues and make it the only CI
linter.
9) PR #2512 towards #11. Adds comments giving the reason for any
CodeBlocks in the generated code.
10) PR #2474 for #2409. Adds comments to the generated LFRic PSy
layer describing any built-in kernels.
11) PR #2511 for #2510. Array2Loop transformations avoid character arrays
unless requested in the transformation options.
1) PR #2199 for #2189. Fix bugs with missing maps in enter data
2) PR #2269 for #2268. Rename DynInvokes to LFRicInvokes and move
to own module.
3) PR #2255 for #2237. The derived type intial values are not lost
anymore, and the "extends" or "abstract" keywords cause a
UnknownFortranType.
4) PR #1822 towards #1338. Adds support for the OMP tasking
directive and transformation.
5) PR #2346 for #2345. Make the integration tests use a parameterised
version of python, currently 3.11.4. Pin pylint to < 3.0
6) PR #2325 for #1947. Adds support for OMP/ACC Atomics and OMP Simd
directives.
7) PR #2355 for #2347. Adds testing to demonstrate that the new LFRic
metadata handling works for multi-precision kernels.
8) PR #2342 for #2340. Fix issues with VERIFY, ISHFTC and PRECISION
intrinsics and introduce a NEMO passthrough test.
9) PR #2350 for #2349. Update to HEAD of fparser master to get
fparser1 fix for parsing WHERE.
10) PR #2285 for #2284. Split LFRicRunTimeChecks from dynamo0p3.py.
11) PR #2313 towards #2197. LFRic PSyLayer uses array pointers instead
of derived types accessors in the kernel/builtin arguments.
12) PR #2321 for #2319. Refactors DynStencil into a dataclass named
LFRicArgStencil.
13) PR #2361 for #2360. Fix LFRic Loop.independent_iterations with domain
information and allow sequential directives on loops with dependencies.
14) PR #2365 for #2351. Fix ref2arrayrange allocated bug.
15) PR #2363 for #1254. Fortran FORMAT, ENTRY and IMPLICIT statements
produce CodeBlocks instead of losing the statements.
16) PR #2357 for #2353. Adds the ACC-loops NEMO transformation script
and includes its use in the integration testing.
17) PR #2380 and #2378 for #2369. Add PSyIR node.sibilings property.
18) PR #2375 for #2374. Bug fix for LFRic kernel-extraction driver
creation for netcdf.
19) PR #2385. Update tests to include python 3.12.
20) PR #2377 for #2368. Adds optional 'depth' argument to Node.walk().
21) PR #2362 for #2243 and #2312. Add support for Fortran save statements.
22) PR #2379 for #2370. Adds a Node.get_sibling_lists() method.
23) PR #2366 towards #342. Improve symbols and tighten-up ACCRoutineTrans
24) PR #2383 towards #2256. Adds support for 'DO CONCURRENT' in the
fparser2 frontend.
25) PR #2100 for #2091. Renames DynKern to LFRicKern and moves its
implementation to domain/lfric.
26) PR #2391 for #2390. Fix bug in generator.py error message.
27) PR #2364 towards #1941. Moves map of LFRic precisions to
config. file.
28) PR #2376 for #1525. Add lowering methods to all LFRic Built-ins.
29) PR #2406 for #2404. Fixes bug in setop_random kernel (was only
filling the lowest vertical level).
30) PR #2389 toward #928. Adds support for field_bcs_kernel when
parsing kernels using kernel_interface.py. This is required when
validating LFRic kernel arguments.
31) PR #2388 for #2323. Rename the array reduction 2code transformations
to use the 2loop postfix and improve their implementation.
32) PR #2400 for #1508. Add flag/config option to disable backend checks.
33) PR #2428 for #2427. Fixes bug in MaxVal2LoopTrans.
34) PR #2430 towards #2421. Ensures local symbol table is merged into
routine scope following an inline transformation. (Works around
#2424.)
35) PR #2412 for #2411. Remove NemoKern and add the 'routine' profiling
36) PR #2401 for #2261. Split DynLoop out of dynamo0p3 and rename.
37) PR #2387 towards #1799. Add support for querying the datatype of
operations.
38) PR #2434 for #2433. Bug fix for process_nemo.py script to
aggregate -I arguments.
39) PR #2426 for #2254. Make LFRic PSyAD test harness use LFRic
logging API.
40) PR #2273 for #2272. Split LFRicFields into domain/lfric module.
41) PR #2393 for #2392. Split and rename DynKernMetadata into
into domain/lfric module (LFRicKernMetadata class).
42) PR #2240 for #2132. Update Integration test software stack.
43) PR #2454 for #2418. Fix OMPTaskTrans attempts to inline IntrinsicCalls.
44) PR #2459 for #2458. Fix broken documentation links.
45) PR #2472 for #2471. Update all copyright dates to include 2024 and
add utils/update_copyright.sh script to do this task.
46) PR #2443 for #2442. Fix bug for PSy-layer generation involving CMA
operators in the LFRic API.
47) PR #2481 for #1192. Unpin jsonschema version.
48) PR #2470 for #2468 and towards #237. Fixes Sphinx link-check errors
for links to anchors in User Guide and also updates the name of the
dl_esm_inf library and updates associated links (remove references to
puma).
49) PR #2420 towards #1786. Add and update Built-ins to support
mixed precision in the LFRic API: real to real, real to integer
and integer to real field types.
50) PR #2480 for #2479. Updates .gitignore for files generated by
examples.
51) PR #2486 for #2419. Renames Unknown[Fortran]Type to
Unsupported[Fortran]Type and DeferredType to UnresolvedType.
52) PR #2482 towards #2062. Adds a new 'origin_string()' method
to Node to return information on the origin (line number and file
name) of a PSyIR node.
53) PR #2439 for #2105. Use maxval2loop and atomics to parallelise
NEMO stpctl.f90.
54) PR #2493 for #2492. Fix bug in NemoOuterArrayRange2LoopTrans
55) PR #2465 towards #2004. Move same_range functionality to
ArrayMixin and other improvements.
56) PR #2372 towards #1960. Fix bugs on Fortran WHERE canonicalisation.
57) PR #2489 for #2488. Unify generic LoopFuseTrans and allow renaming
the second loop variable for certain cases.
58) PR #2423 for #2410. Split off and rename DynStencils to
LFRicStencils in the LFRic API.
59) PR #2490 towards #2447. Update PSyclone to use fparser 0.1.4 release.
60) PR #2501 for #2447. Create PSyclone release 2.5.0.
1) PR #1758 for #1741. Splits the PSyData read functionality into a
standalone module to remove dependence on infrastructure for
driver code.
2) PR #1766 FOR #1764. Fixes various syntax errors in some of the
Fortran test files.
3) PR #1759 for #1752. Updates to GOcean examples so that F90FLAGS
is used consistently and is also used when linking.
4) PR #1700. Various improvements to the example NEMO processing
scripts.
5) PR #1762 for #1743. Ensures declaration of precision variables
in PSyAD test harness.
6) PR #1774 for #1773. Fixes syntax error in test Fortran file.
7) PR #1620 for #1555. Create Algorithm Layer for a Kernel.
8) PR #1767 for #1739. Fixes dep-analysis bug whereby CMA operator
arguments to LFRic kernels were not given correct access value.
9) PR #1763 for #1717. Fixes invalid adjoint names for subroutine
and module.
10) PR #1776 for #1775. Fixes post_region_code_gen and improves
test fixtures.
11) PR #1784 towards #1779. Removes the use of config-provided bounds
from the NemoRange2Loop Transformation.
12) PR #1765 for #1724. Remove kernel use statements from
processed algorithm layer code.
13) PR #1785 for #1771. Refactor algorithm-creation functionality.
14) PR #1769 for #1721. Fixes bugs in use of sympy simplify() and
expand().
15) PR #1791 for #1780. Add setval_random built-in in LFRic API.
16) PR #1778 for #1751. Add ReplaceInductionVariableTrans transformation.
17) PR #1652 for #1645. Symbols representing LFRic BuiltIn Functors
18) PR #1795 for #1583. Bug fix for PSyAD for expressions involving
a unary minus.
19) PR #1809. A few fixes for code style issues.
20) PR #1803 for #1797. Adds support for logical variables to the
21) PR #1804 for #637. Extends the dependence analysis to support
LFRic built-in kernels.
22) PR #1790 for #1787. Adds HoistLoopBoundsExprTrans transformation
which hoists complex loop-bound expressions out of loops.
23) PR #1811 for #1810. Separate LFRic LMA Operator infrastructure
into its own file.
24) PR #1808 for #1681. Extends PSyAD to reject functions.
25) PR #1633 towards #1631. New implementation of kernel meta-data
parser for GOcean API.
26) PR #1821 for #1557. Extends PSyAD to support LFRic mixed-
precision kernels.
27) PR #1551 towards #1338. Add OpenMP Clauses and its referenced
variables as part of the PSyIR tree.
28) PR #1838 for #1704. Update replace_with method to take into
account named arguments.
29) PR #1847. Make a subSignature a Signature and allow
Signature.to_language w/o arguments
30) PR #1831 for #1735. Changes the directory structure in the
test LFRic infrastructure to mimic that in the full LFRic
infrastructure. Alter the testing infrastructure to use the
Makefile to build the LFRic library.
31) PR #1814 towards #924. Add new PSyIR InlineTransformation.
32) PR #1819 for #1815. Forbid OMPTarget regions from containing
CodeBlock nodes.
33) PR #1830 for #1828. Fixes to enable explicit OpenACC Loop
parallelism with UM.
34) PR #1855 for #1840. Fix bug with dependency analysis when an
expression has the same array reference with and without indices.
35) PR #1827 for #1826. Adds support for SIR intrinsics.
36) PR #1839 for #1782. Adds a -api option to the PSyAD script.
rupertford
committed
37) PR #1844 towards #1823. Makes module inlining work in LFRic
when the same kernel is used multiple times.
38) PR #1825 for #1576. Adds the Reference2ArrayRangeTrans
transformation.
39) PR #1846. Fixes to various docstrings in transformations.py
and node.py.
40) PR #1845 for #1813. Adds a standalone PSyData extraction
library (that uses only Fortran IO).
41) PR #1836 for #1834. Adds support for explicit array-range
declarations to the dotproducttrans transformation.
42) PR #1871 for #1870. Fixes typo in link in User Guide.
43) PR #1875 for #1874. Remove pytest-pep257 dependency from setup.py
44) PR #1860 for #1538. Fixes search for config file when
`python setup.py install` used instead of pip and adds yaml config
file to tell RTD to use pip and Python 3.8 (for interop. with
SymPy).
45) PR #1768 for #1711. Add initial Sum2CodeTrans transformation.
46) PR #1885 for #1878. Update FortCL submodule to head of master to
fix invalid-boz compilation errors with newer versions of gfortran.
47) PR #1890 for #1889. Fix test to avoid temporary files being
left after pytest run.
48) PR #1850 towards #1799. Initial PSyIR node.datatype implementation
to query the resulting datatype of a PSyIR subtree.
49) PR #1869 for #1868. Extend PSyAD to make use of the
50) PR #1880 for #1865. Improves support for function calls in the
PSyIR by ensuring the associated RoutineSymbol has the correct
datatype.
51) PR #1894 for #1893. Fix bug in is_upper/lower_bound for UnknownTypes.
52) PR #1853 for #1829. Add OMP teams distribute parallel do directive
and refactor OMPLoopTrans.
53) PR #1900 for #1893. Fix bug in is_upper/lower_bound for non-typed
Symbols.
54) PR #1782 for #1496. Adds initial support for PSyAD test
harness generation for LFRic kernels. Some issues remain.
55) PR #1807 towards #1806. Adds PSyIR support for LFRic kernel
metadata.
56) PR #1906 for #1896. Update HoistLocalArrayTrans to avoid hoisting
parameter arrays and, unless explicitly requested, device parallel routines.
57) PR #1901 towards #1892 and #1898. Add support for storing and
querying the index of a metadata argument.
58) PR #1884 for #1882. Extends the inlining transformation to support
inlining of functions.
59) PR #1912 for #1851. Fixes pycodestyle (2.9) errors/warnings for
with Python 3.10.
60) PR #1899 towards #1823. Refactors KernelModuleInlineTrans so that
the in-lining happens at transformation time rather than code-
61) PR #1852 for #1750. Adds an option to VariablesAccessInfo to
report/not report array arguments as first parameter to
lbound/ubound/size as read access. This fixes several issues with the
dependency tools.
62) PR #1867 for #1861. Significantly improves test coverage,
especially of the compilation-testing utilities.
63) PR #1924 for #1923. Fix issues building Developer and Reference
guides.
64) PR #1913 for #1848. Fix issues with case-sensitivity in the
SymbolTables.resolve_imports method.
65) PR #1917 for #1916. Extends InlineTrans to support calls to
routines in different Containers.
66) PR #1940. Replace Python 3.10 with 3.11 in GHA test suite.
67) PR #1943 for #1942. Fixes incorrect path in tutorial
Makefile setup.
68) PR #1905 for #1903. OpenMP schedule can be set to "none", which skips
the schedule clause in the OpenMP directive. This is the new default in
generic PSyIR (GOcean and LFRic still default to "static").
69) PR #1892 towards #1824. get_kernel_schedule supports resolving
interfaces (e.g. for the LFRic mixed precision kernels).
70) PR #1911 for #1909. Update fparser dependency.
71) PR #1908 for #1879. Completes the new meta-data parsing
functionality for the LFRic API.
72) PR #1918 for #1883. Updates kern_call_arg_list() (for LFRic) to
use PSyIR. (Step towards using PSyIR in LFRic PSy-layer generation
and required for PSyKE driver generation.)
73) PR #1898 towards #1708. Extends PSyAD LFRic harness support to
kernels with field arguments that contain geometry information.
74) PR #1982 for #1981. Updates the GHA configuration to use
Python 3.7 instead of 3.6.
75) PR #1956 for #1955. Updates the tests/.pylintrc file to remove
unsupported options.
76) PR #1888 for #1783. Adds NEMO OpenMP transformation scripts
to the examples.
77) PR #1933 for #1902. Adds run-time checks when hoisting
local work arrays.
78) PR #1963 towards #1883. More work to improving the use of PSyIR
in LFRic (as needed for PSyKE driver creation).
79) PR #1964 for #1961. Adds lower_to_language_level to DynLoop to
ensure loop variable names preserved in PSyKE driver creation.
80) PR #1952 for #310. Adds support for ACCEnterData and ACCUpdate
directives in the NEMO API.
81) PR #1925 for #1832. Add a 'force' option to the ParallelLoopTrans.
82) PR #1988 for #1507. Drop 'six' dependency and fix many pylint issues.
83) PR #1968 towards #1823. Fixes additional issues with module-
inlining in LFRic. (Interfaces, imported global variables and some
validation problems.)
84) PR #1996. Addendum to #1952 to cover missed lines.
85) PR #1994 for #1986. Fix issues with LFRic OMP support by skipping
dependence analysis for coloured loops.
86) PR #1997 for #888. Updates the handling of the '-v'
command-line flag for requesting version information. The
psyclone script will now quit after outputting the version
information.
87) PR #1970 for #1966. Add an overwrite_datatype option to
StructureReferences for when an API knows what type a DeferedType is.
88) PR #2003. Pins the Sphinx version to < 6.1 to avoid a bug in
the link-checking functionality. Also corrects various links to
89) PR #1972 for #1967. Ensures that proper Signatures are created
for structure accesses that appear in LFRic kernel arguments.
90) PR #1907 for #1366. Introduces the IntrinsicCall node and the
IntrinsicSymbol symbol. Adds ALLOCATE, DEALLOCATE and RANDOM_NUMBER
intrinsics.
91) PR #2007 for #2000. Removes pinning of Sphinx version in setup.py
now that the linkcheck bug has been fixed in a new release.
92) PR #1976 for #1974. Fixes issues with literal constants and
array accesses in VariableAccessInfo for the LFRic API.
93) PR #2015 for #2012. Fixes crash in access-information
collection when there is an LFRic 'domain' loop.
94) PR #2014. Extends psyir_from_source() so that the caller
can specify if the Fortran is in fixed format.
95) PR #1993 towards #1953. Extends the LFRicKernelMetadata
class to support stencil specifications.
Joerg Henrichs
committed
96) PR #2020 for #2019. Fix inconsistencies in handling of precision
97) PR #2023 for #1975. Extends DynReferenceElement to ensure that
it adds related kernel arguments to KernCallArgList as PSyIR.
98) PR #2030 for #1529. Update eg19 Makefile so that compilation
works.
99) PR #2029 for #1567. Fix visibility issues with imported symbols.
100) PR #2024 for #1984. Fixes the dependence analysis for the case
where a single element of an array of fields is passed as an
argument to an LFRic built-in.
101) PR #1980 for #1978. Updated LFRic Infrastructure files in
PSyclone repository and PSyData libraries to support 32- and
64-bit mixed precision.
102) PR #2034 for #1864: Add support for adjoint test-harness
generation for kernels with operator arguments.
103) PR #2038 for #1998. Adds support for r_tran_operator_type to
LFRic (dynamo0.3).
104) PR #2036 for #2033. Splits core/access_info into its constiutent
classes and moves utility functions out of the DependencyTools
module.
105) PR #2056 for #2055. Extends the Reference2ArrayRange
transformation such that it rejects any References that are args to
the DEALLOCATE intrinsic.
106) PR #2047 for #2046. Adds a 'return' statement to the
lower_to_language_level() methods.
107) PR #2032 for #92. Forbid include statements (which should have been
processed by the preprocessor or fparser before PSyclone).
108) PR #1973 for #1904. Fix inlining issues with complex access
expressions (e.g. array offsets and partial derived types accesses).
109) PR #2044 for #1295. Changes LFRic psyir to avoid use of exec
and ensures that LFRicConstants is not created *before* the config.
file has been read.
110) PR #2066 for #1587. Introduces new DebugWriter visitor for
fast generation of Fortran-like content in error messages.
111) PR #1914 for #1891. Improves error handling when there are
undeclared arguments.
112) PR #2067 towards #1618. Fixes bugs relating to gh_sum for
scalars and use of any_space_* spaces in the new metadata
handling. Also corrects a (tested) docstring so that it sets
up a Config object first.
113) PR #2063. Improve search for optimisation script and handling
of any errors when it is imported.
114) PR #2071 for #2042. Adds support for the Fortran 'parameter'
statement to the fparser2 frontend.
115) PR #2068 for #352 and #1422. Adds a 'DoWhile' node to represent
Fortran DO WHILE and unbounded DO loops.
116) PR #2072 tidying of IfBlock following #2068.
117) PR #2061 towards #598. Improves the use of dependence analysis
in the OMP transformations to infer which variables are firstprivate
in some cases.
118) PR #2059 for #2057. Adds kernel-extraction support for the
NEMO API.
119) PR #2073. Fixes typos across the repo.
120) PR #2025 for #1483. Adds driver generation for LFRic kernel
extraction (PSyKE).
121) PR #2048 towards #1772. Implements a new MetadataToArgumentsRules
class using the new LFRic metadata functionality.
122) PR #2085 for #840. Add support for minval, maxval and sum Fortran
intrinsics.
123) PR #2077 for #2075. Specify LFRic built-in metadata descriptions
within LFRic built-in classes.
124) PR #2104 towards #1618. Fix invalid use of subroutine instead of
function in an LFRic example algorithm.
125) PR #2098 towards #2091. Split off and rename DynKernCallFactory
to LFRicKernCallFactory.
126) PR #2103 towards #1618. Support Fortran interfaces whose symbols
127) PR #2099. Update of tutorials.
128) PR #1937 for #1935. Adds all LFRic TL and adjoint kernels under
examples.
129) PR #2112 towards #1618. Splits LFRic eg11 source file into
driver and algorithm.
130) PR #2118 for #2117. Ensure build flags passed to LFRic
infrastructure
131) PR #2026 for #1991. Improves the support for building a
standalone driver for an extracted kernel. All required modules are
now inlined into the driver source file.
132) PR #2086 for #2013. Adds support for logical scalar arguments
in PSyAD test harness generation.
133) PR #2060 towards #1618. Adds an optional code path for LFRic
Algorithm-layer handling using the new metadata processing
functionality.
134) PR #2123. Updates GOcean project link in UG.
135) PR #2121 for #2109. Updates PSyclone to fparser 0.1.1 release.
136) PR #2128 towards #1618. Bug fix in LFRicAlgInvoke2PSyCallTrans
for kernels requiring more than 1 quadrature object.
137) PR #967 towards #2132. Add GH workflows for self-hosted runner.
138) PR #2084 for #1801. Improve handling of assume-shape arrays with
specified bounds.
139) PR #2129 towards #1618. Bug fix for -ve Literal args to kernels
in PSyIR handling of LFRic Alg Layer.
140) PR #2133 for #2126. Add RTD YAML config files to allow the
documentation (user guide, developer guide etc) to be managed
separately and to allow software environments other than the
default to be specified.
141) PR #2092 for #2008. Refactors Symbol interfaces and adds support
for Static/SAVE variables.
142) PR #2089 for #2005. Add a SymbolTable merge method and improve symbol
renaming capabilities.
143) PR #2141 towards #1618. Add support for Symbols imported from
Containers with renaming.
144) PR #2136 for #2134. Fix compilation tests bug with assumed shape arrays.
145) PR #2115 for #2114. Fortran allocate with type-spec becomes a CodeBlock.
146) PR #2130 towards #1161. Adds support for HUGE and TINY intrinsics.
147) PR #2095 towards #2091. Split off and rename DynInvoke to LFRicInvoke.
148) PR #2146 for #2145. Skip CI link checks for mybinder as they
occasionally fail.
149) PR #2135 towards #1618. Adds (limited) PSyIR support for Fortran
interfaces without a name.
150) PR #2152 for #2149. Adds checks in Literal constructor that
supplied string is a valid format.
151) PR #2143 for #2043. Adds basic PSyIR support for Fortran COMMON
blocks.
152) PR #2150 towards #2132. Improve self-hosted CI workflow.
153) PR #2090 for #1545. Split off and rename DynCollection to
LFRicCollection.
154) PR #2140 for #2139. Encapsulates the result of the dependency-
analysis' get_in_out_parameters() in a single class.
155) PR #2160 for #2159. Adds support for LFRic kernel arguments with two
new precision names (r_bl and r_phys).
156) PR #2170 for #2169. Correct import of fparser2
Declaration_Construct.
157) PR #2131 towards #1618. Improves the error handling when we find
a Routine with the same name as an LFRic Builtin.
158) PR #2147 for #2119. Add elemental and pure routine attributes.
159) PR #2181 for #2175. Bug fixes for processing of Fortran
declarations and integer literals.
160) PR #2107 for #2106. Improvements to NEMO processing scripts and
inclusion in Integration Testing.
161) PR #2198 for #2194. Fixes handling of a function where the
precision of the result is only defined within the function.
162) PR #2187 towards #1618. Remove builtin use statements when
lowering.
163) PR #2191 for #2185. Fix element order used in PSyAD test
harness.
164) PR #2208 towards #2205. Ensures that functions or subroutines
that CONTAIN other functions/subroutines are put into a CodeBlock.
165) PR #2124 fir #2101. Refactors OpenMP inference of private/first-
private variables and those requiring synchronisation.
166) PR #2210 for #2206. Ensure symbol renaming included when deep-
copying a symbol table.
Sergi Siso
committed
167) PR #2209 for #2204. Add support for EQV and NEQV and fix select
case bug with logical comparisons.
168) PR #2211 for #2183. Change PSyAD test-harness generation to use
the same variable names as the TL kernel being tested.
169) PR #2212 for #2203. Fix bug in PSyIR fparser2 frontend when
DO WHILE contains normal DO loops.
170) PR #2220 for #2219. Adds 'shared_with' argument to the
node.ancestor() method.
171) PR #2174 for #2171. Add support for obtaining the partial datatype
of an UnknownFortranType.
172) PR #2163 for #1537. Support array-slicing notation in SymPy
comparisons.
173) PR #2221 for #2214. Fix names of invokes containing a single
kernel in the PSyIR of the LFRic Alg layer.
174) PR #2225 for #2224. Fix erroneous CI link check error by
omitting test for the particular link.
175) PR #2172 for #1833. Fixes the LFRic tutorials so that they
build correctly.
176) PR #2227 for #2226. Adds new Node.path_from(ancestor) method.
177) PR #2158 for #81. Ensures all example/test Fortran code for
LFRic follows the LFRic naming convention.
178) PR #2218 towards #2132. Adds LFRic integration tests and
new, associated optimisation scripts. Includes bug fixes for
matmul inlining transformation.
179) PR #2230 for #2228. Improve reporting of errors due to kernel
functors not explicitly included in algorithm use statements.
180) PR #2207 for #1419. Split the DataSymbol constant_value attribute
into is_constant (bool) and initial_value (PSyIR Node). This allows
parsing Fortran declarations with initial_value that are not constant.
181) PR #2231 for #2213. The Fortran frontend and backend now also
consider that UnresolvedInterface symbols can come from the body of
a Fortran Interface.
182) PR #2240 for #2234. Fix WhileLoop issue in the NEMO OpenACC kernels
script.
183) PR #2182 for #2179. Fix LFRic eg14 nvfortran compilation and add
nvfortran examples and tutorials steps in the Integration tests.
184) PR #2250 for #2248 and #1575. Fixes bug in line-length limiter
to prevent it breaking a line before the first non-whitepsace
character.
185) PR #2262 towards #446. Implement reference_access for Call nodes.
186) PR #2184 for #2166. Adds Sympy support for user-defined types.
187) PR #2270 for #1575. Fixes the line-length limiter for code that
has an indent greater than the max. line length.
188) PR #2259 for #1932. Adds support for the 'gang' and 'vector'
clauses on OpenACC Loop directives.
189) PR #2276 for #2274. Small fixes for various complaints
produced by latest version of pycodestyle.
190) PR #2291 for #2290. Bug fix to KernelModuleInlineTrans -
ensures that the new RoutineSymbol added to the Container is given
a DefaultModuleInterface.
191) PR #2238 for #2234. Adds support for a generic interface to
perform variable comparison when the datatype is unknown. (Required
for canonicalising SELECT CASE constructs.)
192) PR #2292 for #2288. Workaround for Sphinx-doctest bug by
building html dev guide before running the doctests.
193) PR #2260 for #2245. Splits LFRicLoopBounds class from
dynamo0.3.py.
194 PR #2241 for #2215. Adds support for Fortran names being the
same as Python keywords when using sympy within PSyclone, e.g. in
comparisons between expressions.
195) PR #2295 for #2294. Extends the PSyData extraction library
to support character variables
196) PR #2157 for #1396. Add tree-update mechanism and new ACCClause
nodes for data movement.
197) PR #2281 for #2258. Adds a PSyData wrapper for the TAU profiler.
198) PR #2283 for #2282 and #2286. Adds a get_ubound_expression
method to ArrayMixin (analagous to get_lbound_expression).
199) PR #2293 for #2301. Add 'is_independent' property/query to Loop.
200) PR #2309 for #1558. Add support for out-of-order parameter statements.
201) PR #2148 towards #2105. Adds {Min,Max}Val2Code transformations.
202) PR #2298 for #1987. Change all intrinsic Operations to IntrinsicCalls.
203) PR #2320 for #2242. Pin Sphinx dependency to <7 to fix RTD building step.
204) PR #2264 for #2263. Split LFRicScalarArgs from src/psyclone/dynamo0p3.py.
205) PR #2328 for #2316. Update PSyclone to use 0.1.3 of fparser.
206) PR #2329 towards #2204 and closes #2315. Bug fix to fparser2
frontend to ensure that a PROGRAM with a CONTAINS is captured as
a CodeBlock.
207) PR #2327 towards #2132. Updates the LFRic integration tests
to use a Spack environment and the latest HEAD of trunk of LFRic.
Also extends the integration tests to upload performance numbers
to a GitHub gist.
208) PR #2154 for #95 and #994. Enforce valid Fortran names for
invoke names in the LFRic API.
209) PR #2332 for #2318. Small bug fix to the LFRic extraction
driver creation. eg17/full_example_extract now builds and runs OK.
210) PR #2287 for #2253. Allow for GH_WRITE access to continuous
field arguments to inter-grid kernels in the LFRic API.
211) PR #2299 for #2297. Add option to get original symbol name
(for an imported symbol) in VariablesAccessInfo.
212) PR #2233 for #1772. Correct kernel metadata for LFRic adjoint code.
213) PR #2334. Fix missing backtick in docstrings.
214) PR #2306 for #2305. Rename DynHaloExchange to LFRicHaloExchange.
release 2.3.1 17th of June 2022
1) PR #1747 for #1720. Adds support for If blocks to PSyAD.
2) PR #1669 for #450. Remove set_dirty/clean from ACC regions
3) PR #1745 towards #1731. Split PSyLoop class from Loop node.
4) PR #1761 for #1760. Create release 2.3.1
1) PR #1672. Fixes BOZ literal standard incompatibility in the
nvidia profiling wrapper code.
2) PR #1670 for #1650. Fixes various inconsistencies in the symbol
table implementation.
3) PR #1674 for #1644. Fixes bug in matmul transformation.
4) PR #1673 for #1671. Fixes an error in the Makefile for the
'time evolution' LFRic tutorial.
5) PR #1628 for #1491. Extends PSyAD pre-processing phase to
expand any expressions involving active variables.
6) PR #1666 towards #1618. Modified prototype PSyIR-based
algorithm generation to use a transformation to convert from
invoke's to processed algorithm code, rather than it being done as
part of lowering in the backend. This allows algorithm code with
invoke's to be output by the backend.
7) PR #1613 for #1238 and #1528. Improves the view() method so that
it returns a str rather than printing to stdout. Also allows colour
to be disabled.
8) PR #1656 for #1649 and #1645. Add equality operator to all core
PSyIR nodes in order to compare nodes by value.
9) PR #1680. Refactors some of the code in various ACC directives to
minimise duplication and switch to f-strings.
10) PR #1639 for #1630. Add support for named arguments in Call and
Operation PSyIR nodes, including Fortran front-end and backend support.
11) PR #1682 for #1632. Extend the LFRic API to support transport
field type. Required for mixed-precision support.
12) PR #1678 for #1667. Work towards getting the LFRic OpenACC
example (eg14) to compile.
13) PR #1676 for #1644 Extend matmul2code transformation to
support matrix matrix multiplication. Previously it was limited to
matrix vector.
14) PR #1697 for #1695. Small update to LFRic stub infrastructure to
fix compilation failure when building tutorial code.
15) PR #1699. Fix failing dynamopsy test by specifying dm=true
explicitly
16) PR #1684 for #1683. Adds support for MIN and MAX LFRic Built-ins.
17) PR #1702 for #1695. Correction to LFRic infrastructure ('emdi') to
bring up-to-date with current LFRic usage.
18) PR #1705. Corrections for the NEMO tutorial.
19) PR #1692 for #1688. Prevents scoping regions from sharing symbol
tables. Adds attach() and detach() methods to SymbolTable for
managing this connection.
20) PR #1703 for #1689. Fix stencil depth bug in LFRic halo exchange
when a discontinuous writer follows a discontinuous stencil reader.
21) PR #1608 towards #1542. Allow GH_WRITE access for fields on
continuous function spaces in LFRic API.
22) PR #1722. Changes all PSyIR code over to using f-strings.
23) PR #1725. Allows parsing an empty Fortran file.
24) PR #1716 for #1713. Fix bug in matmul transformation.
25) PR #1693 for #1661. Improve dependency analysis tools.
26) PR #1347 for #1730. The Fortran backend now generates declarations with
dependencies in the appropriate order.
27) PR #1729. Minor fixes for MarkDown link checker and path to
test file.
28) PR #1733 for #1707. Update PSyclone master to the latest version of
fparser to enable running of PSyclone tests.
29) PR #1679. Add a limit parameter to the PSyIR node ancestor method.
30) PR #1749 for #1748. Updates version of fparser specified in setup.py
to 0.0.15.
31) PR #1740 for #1738. Update the operator implementation in the
cut-down LFRic infrastructure and change the USE statements generated
for CMA operators.
release 2.2.0 17th March 2022
1) PR #1439 for #1074. Adds a GOcean example for the use of the NaN-
checking functionality (in PSyData).
2) PR #1463 for #1461. VALID_LOOP_TYPES constant made consistent
across APIs to permit check on loop type in dependency tools.
3) PR #1473 for #1423. Adds the ChunkLoopTrans PSyIR transformation.
4) PR #1475 for #595. Remove memento and transformation return values.
5) PR #1470 for #1459. Refactor OMPLoopTrans to support OMPLoop
6) PR #1452 for #1451. Add doctest for the Dev Guide to GitHub Actions
and tidy the examples so that they work.
7) PR #1503 for #1502. Fix incorrect path in tutorial README.
8) PR #1489 for #1488. Configure MD link checker to skip links to MO wiki.
9) PR #1501 for #1484. Remove Python2 validation, update
documentation, start removing Python2-specific code and start
using any pylint-suggested Python3-specific code constructs.
10) PR #1476 for #1456. Fix precision problems in the PSyAD harness
code generation and add harness documentation.
11) PR #1482 for #1481. Renames symbol_from_tag->find_or_create_tag
12) PR #1521 for #1193. Update Python 3.5 to 3.6 in the GitHubActions CI.
13) PR #1512 for #1511 and #1516. Fix silent compilation test failures,
and make sure all files are compiled in tmpdir.
14) PR #1531. Fixes incorrect imports of OMPTaskwaitTrans in examples
in class docstrings.
15) PR #1498 for #533. Adds the use of SymPy for comparing expressions
for equality. Removes math_equal() and
ArrayRange2LoopTrans.string_compare() methods.
16) PR #1492 for #1457. Adds support for the transformation of Loops
17) PR #1520 towards #1504. The Fortran frontend marks class type
definitions as UnknownFortranType in the PSyIR.
18) PR #1522 for #1519. Make the LoopSwapTransformation generic and
move it to the psyir.transformations module.
19) PR #1539 for #1524. Re-enable a previously deleted compilation test
due to a compilation error now fixed.
20) PR #1535 for #1534. Adds testing of Python 3.10 in GHA and
updates PSyclone to work with 3.10. Also makes GHA build examples
with 2 threads (as 2 cores are available) and runs tests with as
many threads as there are cores available (again 2 at this time).
21) PR #1510 for #1499. Add SymbolTable method to resolve the properties