Port over loop handling utils from PSyACC

[jwallwork23]

Partially addresses #2717.

This PR merges functions from the loop module of PSyACC into PSyclone:

• is_outermost checks that the Loop doesn't have any ancestors which are Loops.
• is_perfectly_nested checks that the Loop nest starting at a given Loop is "loops all the way down" until the last.
• is_simple checks that a Loop nest is perfect and that the deepest level only contains literal assignments (i.e., the loop nest is trivially parallelisable).
• nest_variable_names extracts the tuple of variable names for the Loop nest starting at a given Loop.

[sergisiso]

@jwallwork23 Isn't this equivalent to the Loop.independent_iterations method? (I think we considered this name for that method but decided the independent iterations could be more generic)

[jwallwork23]

@jwallwork23 Isn't this equivalent to the Loop.independent_iterations method? (I think we considered this name for that method but decided the independent iterations could be more generic)

Oh good, if there is already an equivalent method then that's one less thing to merge in :) Thanks!

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 99.86%. Comparing base (8367de1) to head (90bc743).

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #2754   +/-   ##
=======================================
  Coverage   99.86%   99.86%           
=======================================
  Files         354      354           
  Lines       49010    49044   +34     
=======================================
+ Hits        48946    48980   +34     
  Misses         64       64           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[arporter]

Thanks @jwallwork23. I have a couple of questions about two of the new methods so I haven't gone on to look at the tests yet.
Ref. guide builds cleanly though.

[arporter]

Please put these imports in alpha-order with the other ones (ordered by module name).

[arporter]

This looks pretty complicated. I think you'd be better off using the loop_body property of Loop rather than doing a walk. Could you use/adapt the implementation at:

PSyclone/src/psyclone/psyir/transformations/parallel_loop_trans.py

Lines 271 to 325 in 8367de1

	num_collapsable_loops = 0
	next_loop = node
	previous_iteration_variables = []
	while isinstance(next_loop, Loop):
	previous_iteration_variables.append(next_loop.variable)
	num_collapsable_loops += 1
	if not isinstance(collapse, bool):
	if num_collapsable_loops >= collapse:
	break
	# If it has more than one child, the next loop will not be
	# perfectly nested, so stop searching. If there is no child,
	# we have an empty loop and we also stop here.
	if len(next_loop.loop_body.children) != 1:
	if (next_loop.loop_body.children and
	isinstance(next_loop.loop_body[0], Loop)):
	next_loop.loop_body[0].append_preceding_comment(
	"Loop cannot be collapsed because it has siblings")
	break
	next_loop = next_loop.loop_body[0]
	if not isinstance(next_loop, Loop):
	break
	# If it is a loop dependent on a previous iteration variable
	# (e.g. a triangular iteration space), it can not be collapsed
	dependent_on_previous_variable = False
	for bound in (next_loop.start_expr, next_loop.stop_expr,
	next_loop.step_expr):
	for ref in bound.walk(Reference):
	if ref.symbol in previous_iteration_variables:
	dependent_on_previous_variable = ref.symbol
	break
	if dependent_on_previous_variable:
	if verbose:
	next_loop.append_preceding_comment(
	f"Loop cannot be collapsed because one of the "
	f"bounds depends on the previous iteration variab"
	f"le '{dependent_on_previous_variable.name}'")
	break
	# Check that the next loop has no loop-carried dependencies
	dtools = DependencyTools()
	if not ignore_dep_analysis:
	if not next_loop.independent_iterations(
	dep_tools=dtools,
	signatures_to_ignore=list_of_signatures):
	if verbose:
	msgs = dtools.get_all_messages()
	next_loop.preceding_comment = (
	"\n".join([str(m) for m in msgs]) +
	" Consider using the \"ignore_dependencies_"
	"for\" transformation option if this is a "
	"false dependency.")
	break

(and then replace/update that code to use this new method)?

[arporter]

I'm not sure about this one. It seems very niche (why assignments only with Literals?). How does the use case compare with that for Loop.independent_iterations?

[arporter]

tuple[str] please.

[arporter]

Not yours but would you mind putting all of these in alpha order in fact?