Sergi Siso

Oncee #2527, #2424, #2271 are fixed, we need to replace many: `for node in expr.walk(Reference)` with the `reference_accesses` version because it currently misses all references that are not nodes in...

After #2527 a symbol is always call.routine.symbol and it contains all the callee information (arguments signature, if its pure, elemental, available on GPU). But for intrinsics it still lives in...

We have at least one duplicated transformation with: ``` OMP_TRANS_PAR_DO = OMPParallelLoopTrans(omp_schedule="static") ``` ``` OMP_TRANS_PAR_DO = OMPLoopTrans(omp_schedule="static", omp_directive="paralleldo") ``` But the first one does not accept the collapse option.

The Jupyter notebook tutorial on Binder has been commented out (but not removed), because it was out-of-date with current practices. We could bring it up-to-date and re-enable it. Also we...

Improve and document Reference methods to get the datatypes

1

For: ``` integer, dimension(10) :: a x = a(i) ``` `psyir[0].rhs.datatype` evaluates to integer, which makes sense in the context of psyir nodes because datatype is a property that exists...

fparser2 frontend doesn't support character declarations with any kind of length specification, therefore, both e.g. `character(len=10)` ... and `character(len=*)` end up as `UnsupportedFortranType`.

A bit of NEMO that is currently not executed on the GPU with the OpenMP transformation are expressions like: ` zmax(2) = MAXVAL(ABS(un(:, :, :)))` `zmax(3) = MAXVAL(- tsn(:, :,...