Andy Porter

I think the preference is to have the name at (or somewhere near) the start.

We've had a discussion about this and are uneasy about the proposed solution: essentially, it changes the one special entry in the application namespace (`invoke`) to an infinite number of...

Created `gocean_doc_update` branch.

#2470 has now updated the user guide so that part is done. However, the documentation of dl_esm_inf still needs moving over to that repo.

Code builds and runs (OpenACC loops) but is very slow. One of the main culprits is `icethd_dh` which consists of loads of tiny kernels. This is because we haven't parallelised...

Attempt to use `924_module_inline` branch: Transforming ice_thd_dh with acc kernels Module inline of 'ice_var_snwblow' failed: Transformation Error: KernelModuleInlineTrans failed to retrieve PSyIR for routine 'ice_var_snwblow' due to: RoutineSymbol 'ice_var_snwblow' exists...

The first problem is simply because I'm working with the file *after* PSyclone has hoisted some automatic arrays to module scope. If I do it *before* this then `ice_var_vremap` is...

However, cannot then 'inline' either routine: Transformation Error: Routine 'snw_ent' cannot be inlined because it contains a Symbol 'zeh_cum0' with an UnknownInterface: 'REAL(KIND = wp), DIMENSION(0 : nlay_s + 1)...

Now that I look at this again, there don't seem to be any tests of the new directive and keyword for the Fortran psyir backend (it's all gen_code in LFRic).

I see from https://github.com/stfc/PSyclone/pull/2070#discussion_r1319986879 that this PR was started before we had ACCClause (#2157). However, we do now have clauses and therefore this code needs updating (to remove the use...