Joerg Henrichs

line_length example is failing

2

The CI test uses `make`, which works, but the instructions also state that you should be able to run `runme.py`, but that doesn't work anymore: ``` /psyclone/examples/line_length$ python ./runme.py Traceback...

[PSyIR] Add support for reduction clauses on parallel directives

4

This extends #512. Reductions and 'omp parallel do' are not needed for my ROMS work, but it would complete the omp directive support in PSyIR, so I opened a separate...

OMP do-directive accepts non-loop node

2

During #171 it was discovered that you can add a ProfileNode as child of a OMP-do-directive. This results in incorrect code (inside an omp-do loop must only be loops). In...

Incomplete variable access information for LFRic kernels, e.g. in LoopFusion

4

The `reference_accesses` methods for an LFRic kernel adds the arguments to the access list, but (esp. in the case of fields) does not provide any index information. As a result...

Using Symbols for VariableAccessInfo

3

1) The class `VariableAccessInfo` stores variable names as strings. Once available, they should become symbols (whose properties can be looked up using a SymbolTable). 2) The 'array' vs 'scalar' handling...

The stencil-based index information for a GOcean kernel is currently created out of strings, in gocean1p0._record_stencil_accesses: ``` i_expr = GOKern._format_access("i", i, current_depth) j_expr = GOKern._format_access("j", j, current_depth) var_accesses.add_access(signature, arg.access, self,...

Loop Fusion Transform can create incorrect code

1

The Fuse transformation does not take a potential stencil operation into account. Since now we have stencil information for gocean, this can be tested. Example: ``` do j do i...

Currently the focus of kernel extraction is LFRic, especially the support for non-local symbols. This ticket is about porting these changes to support a generic PSyIR and gocean. There is...

The UM physics contain e.g.: ``` real.. q_latest(tdims%i_start:tdims%i_end,tdims%j_start:tdims%j_end, & bl_levels), & ``` This turns up as an unknown Fortran type, so even when looking at the symbol table for variable...