Sergi Siso

@arporter This is ready for next review

@arporter @rupertford This is ready for review. I decided to just fix one of the issues (prevent duplicated routines when inlining from different invokes if the routine is an exact...

@rupertford This is ready for another review.

We may be talking about 2 different things here: - The reference_accesses of the whole loop, which includes a read of the boundaries at the beginning of the execution. -...

I agree to exclude *all* codeblocks in the OMPTargetTrans, in the Nemo script I have it as a condition to the loop that walks the Loop nodes. But it would...

I am taking this. I will not make it a `ParallelRegionTrans` since by itself is not really a Parallel Region (it needs another directive to make the contents parallel) but...

I just realize there are a few outdated things in the documentation example you are using that we don't do anymore. These are the return schedule and memento (just return...

> How can nodes inside such a node be structured - the OMPSerialDirective will have a schedule containing many nodes - do all of my OMPTaskloopDirective nodes that I care...

Have you check the functionality in `from psyclone.psyir.tools import DependencyTools` ?

There is also the `import psyclone.core import VariablesAccessInfo`. I think the forward/backward_dependence are just for one level of PSy-layer fields but the other can deal with any variable dependency at...