Files
mercury/compiler/cse_detection.m
Mark Brown 38d9ef3479 Package the type_info_varmap and typeclass_info_varmap types into an ADT
Estimated hours taken: 25
Branches: main

Package the type_info_varmap and typeclass_info_varmap types into an ADT
called rtti_varmaps.  There are two main purposes for this:

	- We wish to extend this set of maps with new maps.  Doing this
	will be a lot easier and less error prone if all of the maps are
	packaged in a single data structure.

	- Any new maps that we add may contain redundant information that
	just makes searching the maps more efficient.  Therefore they must
	be kept consistent with the existing maps.  Having all the maps
	inside an ADT makes it easier to ensure this.

This change also includes two extensions to the maps.  First, the
typeclass_info_map is made reversible so that it is possible to efficiently
look up the constraint for a given typeclass_info variable.  Second, a new
map from prog_vars to types makes it possible to efficiently look up the
type that a given type_info variable is for.  These two changes mean that
it is no longer necessary to consult the argument of type_info/1 or
typeclass_info/1 to find this information.  (We still do put that information
there, though; changing the RTTI is left for a separate change.)

compiler/hlds_pred.m:
	Move items relating to type_infos and typeclass_infos into a section
	of their own.

	Add a type `rtti_var_info' to hold information about the contents
	of a type_info or typeclass_info variable.

	Define the rtti_varmaps abstract data type.  This data structure
	consists of the type_info_varmap and the typeclass_info_varmap.
	Add a new map, type_info_type_map, which is like the inverse
	to the type_info_varmap.  The difference is that the latter can
	point to locations that are inside a typeclass_info variable,
	whereas the former only refers to type_info variables.  Note that
	the combined maps do not form a bijection, or even an injection,
	since it is possible for two different type variables to point to
	the same location (that is, if they are aliased).

	Make the typeclass_info_varmap reversible, by using the new module
	injection.m.  Unlike the type_info_varmap, this map is always
	injective since the same typeclass_info cannot be used for two
	different constraints.

	The predicates rtti_det_insert_type_info_locn and set_type_info_locn,
	which update the type_info_varmap, contain sanity checks to ensure
	that only type variables that have already been registered with
	the type_info_type_map are used, and that the information in both
	maps is consistent.

	Use the rtti_varmaps structure in proc_info and clauses_info, in
	place of type_info_varmap and typeclass_info_varmap.

compiler/polymorphism.m:
	Remove polymorphism__type_info_or_ctor_type/2 and
	polymorphism__typeclass_info_class_constraint/2, to ensure that
	nobody tries to use the information in the type argument.  Replace
	them with two similar predicates that test if a type is type_info
	or typeclass_info, but that don't return the argument.

	Ensure that the new type_info_type_map in the rtti_varmaps is kept
	up to date by threading the rtti_varmaps through a few more places.
	Some of these places are exported, so this part of the change
	affects other modules as well.

	Fix a comment that referred to a non-existent predicate.

compiler/type_util.m:
	Remove the predicates apply_substitutions_to_var_map/5 and
	apply_substitutions_to_typeclass_var_map/5.  The functionality
	is now provided by the new ADT.

compiler/cse_detection.m:
	Rewrite update_existential_data_structures/4 to use the interface
	provided by rtti_varmaps.  The algorithm for doing this has changed
	in the following ways:

		- The first pass, which builds a map from changed locations
		in the first branch to the tvars concerned, is modified
		slightly to traverse over the keys instead of over key-value
		pairs.

		- The second pass, which previously calculated the induced
		type substitution and reconstructed the type_info_varmap
		now only does the former.

		- Applying the prog_var transformation and the induced type
		substitution is done at the end, using the interface to
		rtti_varmaps.

compiler/goal_util.m:
	Rewrite goal_util__extra_nonlocal_typeinfos/6 to avoid the need
	for using map__member/3 on the typeclass_info_varmap (about which
	the existing comments say "this is probably not very efficient..."),
	and to be more efficient in general.

	Previously, we nondeterministically generated non-local type vars
	and then tested each constraint to see if it had the type var in it.
	Now, we go through each constraint one at a time and check if any of
	the type variables in it are non-local.  This is more efficient
	because we only need to find one non-local type in order to include
	the typeclass_info in the non-locals -- the remaining (duplicate)
	solutions are pruned away.

compiler/higher_order.m:
	Use the new maps instead of looking at the arguments of type_info/1
	and typeclass_info/1 types.  We plan to remove this information
	from type_info and typeclass_info types in future.

	Previously, this module used the type argument in order to update
	the varmaps when the curried arguments of a higher order call are
	added as arguments to the procedure in which the call occurs.
	We now look up this information at the point where the curried arg
	variables are known, and store this information in higher_order_arg
	alongside the types where it used to be stored.  This structure is
	threaded through to the place where the information is needed.

	Fix a cut and paste bug in higher_order_arg_depth/1.  It was
	previously calling higher_order_args_size/1 in the recursive
	call, instead of calling higher_order_args_depth/1.

compiler/inlining.m:
	In inlining__do_inline_call, apply the substitutions to the entire
	rtti_varmaps structure, not just to the type_info_varmap.  (XXX Is
	there a good reason why the substitution should _not_ be applied
	to the typeclass_info varmap?)

compiler/magic_util.m:
	Avoid using polymorphism__type_info_or_ctor_type/2 and
	polymorphism__typeclass_info_class_constraint/2, as these are
	no longer supported.

compiler/*.m:
	Straightforward changes to use the new ADT.

library/injection.m:
	New library module.  This provides an `injection' type which is
	similar to the existing `bimap' type in that it implements
	back-to-back maps, but doesn't have such stringent invariants
	imposed.  In particular, the reverse map is not required to be
	injective.

	This type is used to model the relationship between prog_constraints
	and program variables that hold typeclass_infos for them.  Namely,
	the typeclass_info for a constraint can be held in two different
	variables, but one variable can never hold the typeclass_info for
	two distinct constraints.

library/library.m:
	Add the new library module.

library/list.m:
	Add list__foldl_corresponding and list__foldl2_corresponding, which
	traverse two lists in parallel, which one or two accumulators, and
	abort if there is a length mismatch.

NEWS:
	Mention the changes to the standard library.
2005-07-22 12:32:07 +00:00

847 lines
31 KiB
Mathematica

%-----------------------------------------------------------------------------%
% Copyright (C) 1995-2005 The University of Melbourne.
% This file may only be copied under the terms of the GNU General
% Public License - see the file COPYING in the Mercury distribution.
%-----------------------------------------------------------------------------%
%
% Common subexpression detection - hoist common subexpression goals out of
% branched structures. This can enable us to find more indexing opportunities
% and hence can make the code more deterministic.
% This code is switched on/off with the `--common-goal' option.
%
% Main author: zs.
% Much of the code is based on switch_detection.m by fjh.
%
%-----------------------------------------------------------------------------%
:- module check_hlds__cse_detection.
:- interface.
:- import_module hlds__hlds_module.
:- import_module hlds__hlds_pred.
:- import_module io.
:- pred detect_cse(module_info::in, module_info::out, io::di, io::uo) is det.
:- pred detect_cse_in_proc(proc_id::in, pred_id::in,
module_info::in, module_info::out, io::di, io::uo) is det.
%-----------------------------------------------------------------------------%
%-----------------------------------------------------------------------------%
:- implementation.
:- import_module check_hlds__det_util.
:- import_module check_hlds__inst_match.
:- import_module check_hlds__mode_util.
:- import_module check_hlds__modes.
:- import_module check_hlds__switch_detection.
:- import_module check_hlds__switch_detection.
:- import_module check_hlds__type_util.
:- import_module hlds__goal_util.
:- import_module hlds__hlds_data.
:- import_module hlds__hlds_goal.
:- import_module hlds__hlds_out.
:- import_module hlds__instmap.
:- import_module hlds__quantification.
:- import_module libs__globals.
:- import_module libs__options.
:- import_module parse_tree__prog_data.
:- import_module parse_tree__prog_type.
:- import_module assoc_list.
:- import_module bool.
:- import_module int.
:- import_module list.
:- import_module map.
:- import_module multi_map.
:- import_module require.
:- import_module set.
:- import_module std_util.
:- import_module string.
:- import_module svmap.
:- import_module term.
:- import_module varset.
%-----------------------------------------------------------------------------%
% Traverse the module structure, calling `detect_cse_in_goal'
% for each procedure body.
detect_cse(!ModuleInfo, !IO) :-
module_info_predids(!.ModuleInfo, PredIds),
detect_cse_in_preds(PredIds, !ModuleInfo, !IO).
:- pred detect_cse_in_preds(list(pred_id)::in,
module_info::in, module_info::out, io::di, io::uo) is det.
detect_cse_in_preds([], !ModuleInfo, !IO).
detect_cse_in_preds([PredId | PredIds], !ModuleInfo, !IO) :-
module_info_preds(!.ModuleInfo, PredTable),
map__lookup(PredTable, PredId, PredInfo),
detect_cse_in_pred(PredId, PredInfo, !ModuleInfo, !IO),
detect_cse_in_preds(PredIds, !ModuleInfo, !IO).
:- pred detect_cse_in_pred(pred_id::in, pred_info::in,
module_info::in, module_info::out, io::di, io::uo) is det.
detect_cse_in_pred(PredId, PredInfo0, !ModuleInfo, !IO) :-
ProcIds = pred_info_non_imported_procids(PredInfo0),
detect_cse_in_procs(ProcIds, PredId, !ModuleInfo, !IO).
:- pred detect_cse_in_procs(list(proc_id)::in, pred_id::in,
module_info::in, module_info::out, io::di, io::uo) is det.
detect_cse_in_procs([], _PredId, !ModuleInfo, !IO).
detect_cse_in_procs([ProcId | ProcIds], PredId, !ModuleInfo, !IO) :-
detect_cse_in_proc(ProcId, PredId, !ModuleInfo, !IO),
detect_cse_in_procs(ProcIds, PredId, !ModuleInfo, !IO).
detect_cse_in_proc(ProcId, PredId, !ModuleInfo, !IO) :-
detect_cse_in_proc_2(ProcId, PredId, Redo, !ModuleInfo),
(
Redo = no
;
Redo = yes,
globals__io_lookup_bool_option(very_verbose, VeryVerbose, !IO),
(
VeryVerbose = yes,
io__write_string("% Repeating mode check for ", !IO),
hlds_out__write_pred_id(!.ModuleInfo, PredId, !IO),
io__write_string("\n", !IO)
;
VeryVerbose = no
),
modecheck_proc(ProcId, PredId, !ModuleInfo, Errs, _Changed,
!IO),
( Errs > 0 ->
error("mode check fails when repeated")
;
true
),
(
VeryVerbose = yes,
io__write_string("% Repeating switch detection for ",
!IO),
hlds_out__write_pred_id(!.ModuleInfo, PredId, !IO),
io__write_string("\n", !IO)
;
VeryVerbose = no
),
detect_switches_in_proc(ProcId, PredId, !ModuleInfo),
(
VeryVerbose = yes,
io__write_string("% Repeating common " ++
"deconstruction detection for ", !IO),
hlds_out__write_pred_id(!.ModuleInfo, PredId, !IO),
io__write_string("\n", !IO)
;
VeryVerbose = no
),
detect_cse_in_proc(ProcId, PredId, !ModuleInfo, !IO)
).
:- type cse_info
---> cse_info(
varset :: prog_varset,
vartypes :: vartypes,
rtti_varmaps :: rtti_varmaps,
module_info :: module_info
).
:- pred detect_cse_in_proc_2(proc_id::in, pred_id::in, bool::out,
module_info::in, module_info::out) is det.
detect_cse_in_proc_2(ProcId, PredId, Redo, ModuleInfo0, ModuleInfo) :-
module_info_preds(ModuleInfo0, PredTable0),
map__lookup(PredTable0, PredId, PredInfo0),
pred_info_procedures(PredInfo0, ProcTable0),
map__lookup(ProcTable0, ProcId, ProcInfo0),
% To process each ProcInfo, we get the goal,
% initialize the instmap based on the modes of the head vars,
% and pass these to `detect_cse_in_goal'.
proc_info_goal(ProcInfo0, Goal0),
proc_info_get_initial_instmap(ProcInfo0, ModuleInfo0, InstMap0),
proc_info_varset(ProcInfo0, Varset0),
proc_info_vartypes(ProcInfo0, VarTypes0),
proc_info_rtti_varmaps(ProcInfo0, RttiVarMaps0),
CseInfo0 = cse_info(Varset0, VarTypes0, RttiVarMaps0, ModuleInfo0),
detect_cse_in_goal(Goal0, InstMap0, CseInfo0, CseInfo, Redo, Goal1),
(
Redo = no,
ModuleInfo = ModuleInfo0
;
Redo = yes,
% ModuleInfo should not be changed by detect_cse_in_goal
CseInfo = cse_info(VarSet1, VarTypes1, RttiVarMaps, _),
proc_info_headvars(ProcInfo0, HeadVars),
implicitly_quantify_clause_body(HeadVars, _Warnings,
Goal1, Goal, VarSet1, VarSet, VarTypes1, VarTypes),
proc_info_set_goal(Goal, ProcInfo0, ProcInfo1),
proc_info_set_varset(VarSet, ProcInfo1, ProcInfo2),
proc_info_set_vartypes(VarTypes, ProcInfo2, ProcInfo3),
proc_info_set_rtti_varmaps(RttiVarMaps, ProcInfo3, ProcInfo),
map__det_update(ProcTable0, ProcId, ProcInfo, ProcTable),
pred_info_set_procedures(ProcTable, PredInfo0, PredInfo),
map__det_update(PredTable0, PredId, PredInfo, PredTable),
module_info_set_preds(PredTable, ModuleInfo0, ModuleInfo)
).
%-----------------------------------------------------------------------------%
% Given a goal, and the instmap on entry to that goal,
% find disjunctions that contain common subexpressions
% and hoist these out of the disjunction. At the moment
% we only look for cses that are deconstruction unifications.
:- pred detect_cse_in_goal(hlds_goal::in, instmap::in, cse_info::in,
cse_info::out, bool::out, hlds_goal::out) is det.
detect_cse_in_goal(Goal0, InstMap0, !CseInfo, Redo, Goal) :-
detect_cse_in_goal_1(Goal0, InstMap0, !CseInfo, Redo, Goal, _InstMap).
% This version is the same as the above except that it returns
% the resulting instmap on exit from the goal, which is
% computed by applying the instmap delta specified in the
% goal's goalinfo.
:- pred detect_cse_in_goal_1(hlds_goal::in, instmap::in, cse_info::in,
cse_info::out, bool::out, hlds_goal::out, instmap::out) is det.
detect_cse_in_goal_1(Goal0 - GoalInfo, InstMap0, !CseInfo, Redo,
Goal - GoalInfo, InstMap) :-
detect_cse_in_goal_2(Goal0, GoalInfo, InstMap0, !CseInfo, Redo, Goal),
goal_info_get_instmap_delta(GoalInfo, InstMapDelta),
instmap__apply_instmap_delta(InstMap0, InstMapDelta, InstMap).
% Here we process each of the different sorts of goals.
:- pred detect_cse_in_goal_2(hlds_goal_expr::in, hlds_goal_info::in,
instmap::in, cse_info::in, cse_info::out, bool::out,
hlds_goal_expr::out) is det.
detect_cse_in_goal_2(Goal @ foreign_proc(_, _, _, _, _, _), _, _, !CseInfo,
no, Goal).
detect_cse_in_goal_2(Goal @ generic_call(_, _, _, _), _, _, !CseInfo,
no, Goal).
detect_cse_in_goal_2(Goal @ call(_, _, _, _, _, _), _, _, !CseInfo, no, Goal).
detect_cse_in_goal_2(unify(LHS, RHS0, Mode, Unify, UnifyContext), _, InstMap0,
!CseInfo, Redo, unify(LHS, RHS, Mode,Unify, UnifyContext)) :-
(
RHS0 = lambda_goal(Purity, PredOrFunc, EvalMethod, FixModes,
NonLocalVars, Vars, Modes, Det, Goal0)
->
ModuleInfo = !.CseInfo ^ module_info,
instmap__pre_lambda_update(ModuleInfo,
Vars, Modes, InstMap0, InstMap),
detect_cse_in_goal(Goal0, InstMap, !CseInfo, Redo, Goal),
RHS = lambda_goal(Purity, PredOrFunc, EvalMethod, FixModes,
NonLocalVars, Vars, Modes, Det, Goal)
;
RHS = RHS0,
Redo = no
).
detect_cse_in_goal_2(not(Goal0), _GoalInfo, InstMap, !CseInfo, Redo,
not(Goal)) :-
detect_cse_in_goal(Goal0, InstMap, !CseInfo, Redo, Goal).
detect_cse_in_goal_2(scope(Reason, Goal0), _GoalInfo, InstMap,
!CseInfo, Redo, scope(Reason, Goal)) :-
detect_cse_in_goal(Goal0, InstMap, !CseInfo, Redo, Goal).
detect_cse_in_goal_2(conj(Goals0), _GoalInfo, InstMap, !CseInfo, Redo,
conj(Goals)) :-
detect_cse_in_conj(Goals0, InstMap, !CseInfo, Redo, Goals).
detect_cse_in_goal_2(par_conj(Goals0), _, InstMap, !CseInfo, Redo,
par_conj(Goals)) :-
detect_cse_in_par_conj(Goals0, InstMap, !CseInfo, Redo, Goals).
detect_cse_in_goal_2(disj(Goals0), GoalInfo, InstMap, !CseInfo, Redo, Goal) :-
( Goals0 = [] ->
Redo = no,
Goal = disj([])
;
goal_info_get_nonlocals(GoalInfo, NonLocals),
set__to_sorted_list(NonLocals, NonLocalsList),
detect_cse_in_disj(NonLocalsList, Goals0, GoalInfo,
InstMap, !CseInfo, Redo, Goal)
).
detect_cse_in_goal_2(switch(Var, CanFail, Cases0), GoalInfo, InstMap,
!CseInfo, Redo, Goal) :-
goal_info_get_nonlocals(GoalInfo, NonLocals),
set__to_sorted_list(NonLocals, NonLocalsList),
detect_cse_in_cases(NonLocalsList, Var, CanFail, Cases0, GoalInfo,
InstMap, !CseInfo, Redo, Goal).
detect_cse_in_goal_2(if_then_else(Vars, Cond0, Then0, Else0), GoalInfo,
InstMap, !CseInfo, Redo, Goal) :-
goal_info_get_nonlocals(GoalInfo, NonLocals),
set__to_sorted_list(NonLocals, NonLocalsList),
detect_cse_in_ite(NonLocalsList, Vars, Cond0, Then0, Else0, GoalInfo,
InstMap, !CseInfo, Redo, Goal).
detect_cse_in_goal_2(shorthand(_), _, _, _, _, _, _) :-
% these should have been expanded out by now
error("detect_cse_in_goal_2: unexpected shorthand").
%-----------------------------------------------------------------------------%
:- pred detect_cse_in_conj(list(hlds_goal)::in, instmap::in, cse_info::in,
cse_info::out, bool::out, list(hlds_goal)::out) is det.
detect_cse_in_conj([], _InstMap, !CseInfo, no, []).
detect_cse_in_conj([Goal0 | Goals0], InstMap0, !CseInfo,
Redo, Goals) :-
detect_cse_in_goal_1(Goal0, InstMap0, !CseInfo, Redo1, Goal1,
InstMap1),
detect_cse_in_conj(Goals0, InstMap1, !CseInfo, Redo2, Goals1),
( Goal1 = conj(ConjGoals) - _ ->
list__append(ConjGoals, Goals1, Goals)
;
Goals = [Goal1 | Goals1]
),
bool__or(Redo1, Redo2, Redo).
%-----------------------------------------------------------------------------%
:- pred detect_cse_in_par_conj(list(hlds_goal)::in, instmap::in, cse_info::in,
cse_info::out, bool::out, list(hlds_goal)::out) is det.
detect_cse_in_par_conj([], _InstMap, !CseInfo, no, []).
detect_cse_in_par_conj([Goal0 | Goals0], InstMap0, !CseInfo, Redo,
[Goal | Goals]) :-
detect_cse_in_goal(Goal0, InstMap0, !CseInfo, Redo1, Goal),
detect_cse_in_par_conj(Goals0, InstMap0, !CseInfo, Redo2, Goals),
bool__or(Redo1, Redo2, Redo).
%-----------------------------------------------------------------------------%
% These are the interesting bits - we've found a non-empty branched
% structure, and we've got a list of the non-local variables of that
% structure. Now for each non-local variable, we check whether each
% branch matches that variable against the same functor.
:- pred detect_cse_in_disj(list(prog_var)::in, list(hlds_goal)::in,
hlds_goal_info::in, instmap::in, cse_info::in,
cse_info::out, bool::out, hlds_goal_expr::out) is det.
detect_cse_in_disj([], Goals0, _, InstMap, !CseInfo, Redo, disj(Goals)) :-
detect_cse_in_disj_2(Goals0, InstMap, !CseInfo, Redo, Goals).
detect_cse_in_disj([Var | Vars], Goals0, GoalInfo0, InstMap,
!CseInfo, Redo, Goal) :-
(
instmap__lookup_var(InstMap, Var, VarInst0),
ModuleInfo = !.CseInfo ^ module_info,
% XXX we only need inst_is_bound, but leave this as it is
% until mode analysis can handle aliasing between free
% variables.
inst_is_ground_or_any(ModuleInfo, VarInst0),
common_deconstruct(Goals0, Var, !CseInfo, Unify,
FirstOldNew, LaterOldNew, Goals)
->
maybe_update_existential_data_structures(Unify,
FirstOldNew, LaterOldNew, !CseInfo),
Goal = conj([Unify, disj(Goals) - GoalInfo0]),
Redo = yes
;
detect_cse_in_disj(Vars, Goals0, GoalInfo0, InstMap,
!CseInfo, Redo, Goal)
).
:- pred detect_cse_in_disj_2(list(hlds_goal)::in, instmap::in, cse_info::in,
cse_info::out, bool::out, list(hlds_goal)::out) is det.
detect_cse_in_disj_2([], _InstMap, !CseInfo, no, []).
detect_cse_in_disj_2([Goal0 | Goals0], InstMap0, !CseInfo, Redo,
[Goal | Goals]) :-
detect_cse_in_goal(Goal0, InstMap0, !CseInfo, Redo1, Goal),
detect_cse_in_disj_2(Goals0, InstMap0, !CseInfo, Redo2, Goals),
bool__or(Redo1, Redo2, Redo).
:- pred detect_cse_in_cases(list(prog_var)::in, prog_var::in, can_fail::in,
list(case)::in, hlds_goal_info::in, instmap::in,
cse_info::in, cse_info::out, bool::out, hlds_goal_expr::out) is det.
detect_cse_in_cases([], SwitchVar, CanFail, Cases0, _GoalInfo, InstMap,
!CseInfo, Redo, switch(SwitchVar, CanFail, Cases)) :-
detect_cse_in_cases_2(Cases0, InstMap, !CseInfo, Redo, Cases).
detect_cse_in_cases([Var | Vars], SwitchVar, CanFail, Cases0, GoalInfo,
InstMap, !CseInfo, Redo, Goal) :-
(
Var \= SwitchVar,
instmap__lookup_var(InstMap, Var, VarInst0),
ModuleInfo = !.CseInfo ^ module_info,
% XXX we only need inst_is_bound, but leave this as it is
% until mode analysis can handle aliasing between free
% variables.
inst_is_ground_or_any(ModuleInfo, VarInst0),
common_deconstruct_cases(Cases0, Var, !CseInfo,
Unify, FirstOldNew, LaterOldNew, Cases)
->
maybe_update_existential_data_structures(Unify,
FirstOldNew, LaterOldNew, !CseInfo),
Goal = conj([Unify, switch(SwitchVar, CanFail, Cases)
- GoalInfo]),
Redo = yes
;
detect_cse_in_cases(Vars, SwitchVar, CanFail, Cases0, GoalInfo,
InstMap, !CseInfo, Redo, Goal)
).
:- pred detect_cse_in_cases_2(list(case)::in, instmap::in, cse_info::in,
cse_info::out, bool::out, list(case)::out) is det.
detect_cse_in_cases_2([], _, !CseInfo, no, []).
detect_cse_in_cases_2([Case0 | Cases0], InstMap, !CseInfo, Redo,
[Case | Cases]) :-
Case0 = case(Functor, Goal0),
detect_cse_in_goal(Goal0, InstMap, !CseInfo, Redo1, Goal),
Case = case(Functor, Goal),
detect_cse_in_cases_2(Cases0, InstMap, !CseInfo, Redo2, Cases),
bool__or(Redo1, Redo2, Redo).
:- pred detect_cse_in_ite(list(prog_var)::in, list(prog_var)::in,
hlds_goal::in, hlds_goal::in, hlds_goal::in, hlds_goal_info::in,
instmap::in, cse_info::in, cse_info::out, bool::out,
hlds_goal_expr::out) is det.
detect_cse_in_ite([], IfVars, Cond0, Then0, Else0, _, InstMap, !CseInfo,
Redo, if_then_else(IfVars, Cond, Then, Else)) :-
detect_cse_in_ite_2(Cond0, Then0, Else0, InstMap, !CseInfo, Redo,
Cond, Then, Else).
detect_cse_in_ite([Var | Vars], IfVars, Cond0, Then0, Else0, GoalInfo,
InstMap, !CseInfo, Redo, Goal) :-
(
ModuleInfo = !.CseInfo ^ module_info,
instmap__lookup_var(InstMap, Var, VarInst0),
% XXX we only need inst_is_bound, but leave this as it is
% until mode analysis can handle aliasing between free
% variables.
inst_is_ground_or_any(ModuleInfo, VarInst0),
common_deconstruct([Then0, Else0], Var, !CseInfo,
Unify, FirstOldNew, LaterOldNew, Goals),
Goals = [Then, Else]
->
maybe_update_existential_data_structures(Unify,
FirstOldNew, LaterOldNew, !CseInfo),
Goal = conj([Unify, if_then_else(IfVars, Cond0, Then, Else)
- GoalInfo]),
Redo = yes
;
detect_cse_in_ite(Vars, IfVars, Cond0, Then0, Else0, GoalInfo,
InstMap, !CseInfo, Redo, Goal)
).
:- pred detect_cse_in_ite_2(hlds_goal::in, hlds_goal::in, hlds_goal::in,
instmap::in, cse_info::in, cse_info::out, bool::out,
hlds_goal::out, hlds_goal::out, hlds_goal::out) is det.
detect_cse_in_ite_2(Cond0, Then0, Else0, InstMap0, !CseInfo, Redo,
Cond, Then, Else) :-
detect_cse_in_goal_1(Cond0, InstMap0, !CseInfo, Redo1, Cond, InstMap1),
detect_cse_in_goal(Then0, InstMap1, !CseInfo, Redo2, Then),
detect_cse_in_goal(Else0, InstMap0, !CseInfo, Redo3, Else),
bool__or(Redo1, Redo2, Redo12),
bool__or(Redo12, Redo3, Redo).
%-----------------------------------------------------------------------------%
% common_deconstruct(Goals0, Var, !CseInfo, Unify, Goals):
% input vars:
% Goals0 is a list of parallel goals in a branched structure
% (disjunction, if-then-else, or switch).
% Var is the variable we are looking for a common deconstruction on.
% !.CseInfo contains the original varset and type map.
% output vars:
% !:CseInfo has a varset and a type map reflecting the new variables
% we have introduced.
% Goals is the modified version of Goals0 after the common deconstruction
% has been hoisted out, with the new variables as the functor arguments.
% Unify is the unification that was hoisted out.
:- pred common_deconstruct(list(hlds_goal)::in, prog_var::in, cse_info::in,
cse_info::out, hlds_goal::out, assoc_list(prog_var)::out,
list(assoc_list(prog_var))::out, list(hlds_goal)::out) is semidet.
common_deconstruct(Goals0, Var, !CseInfo, Unify, FirstOldNew, LaterOldNew,
Goals) :-
common_deconstruct_2(Goals0, Var, before_candidate,
have_candidate(Unify, FirstOldNew, LaterOldNew),
!CseInfo, Goals),
LaterOldNew = [_ | _].
:- pred common_deconstruct_2(list(hlds_goal)::in, prog_var::in,
cse_state::in, cse_state::out, cse_info::in, cse_info::out,
list(hlds_goal)::out) is semidet.
common_deconstruct_2([], _Var, !CseState, !CseInfo, []).
common_deconstruct_2([Goal0 | Goals0], Var, !CseState, !CseInfo,
[Goal | Goals]) :-
find_bind_var(Var, find_bind_var_for_cse_in_deconstruct, Goal0, Goal,
!CseState, !CseInfo, yes),
!.CseState = have_candidate(_, _, _),
common_deconstruct_2(Goals0, Var, !CseState, !CseInfo, Goals).
%-----------------------------------------------------------------------------%
:- pred common_deconstruct_cases(list(case)::in, prog_var::in,
cse_info::in, cse_info::out, hlds_goal::out, assoc_list(prog_var)::out,
list(assoc_list(prog_var))::out, list(case)::out) is semidet.
common_deconstruct_cases(Cases0, Var, !CseInfo, Unify,
FirstOldNew, LaterOldNew, Cases) :-
common_deconstruct_cases_2(Cases0, Var, before_candidate,
have_candidate(Unify, FirstOldNew, LaterOldNew),
!CseInfo, Cases),
LaterOldNew = [_ | _].
:- pred common_deconstruct_cases_2(list(case)::in, prog_var::in,
cse_state::in, cse_state::out, cse_info::in, cse_info::out,
list(case)::out) is semidet.
common_deconstruct_cases_2([], _Var, !CseState, !CseInfo, []).
common_deconstruct_cases_2([case(ConsId, Goal0) | Cases0], Var,
!CseState, !CseInfo, [case(ConsId, Goal) | Cases]) :-
find_bind_var(Var, find_bind_var_for_cse_in_deconstruct, Goal0, Goal,
!CseState, !CseInfo, yes),
!.CseState = have_candidate(_, _, _),
common_deconstruct_cases_2(Cases0, Var, !CseState,
!CseInfo, Cases).
%-----------------------------------------------------------------------------%
% This data structure represents the state of the search for
% deconstructions in all the branches of a branched control structure
% that deconstruct a given variable with the same functor.
% Initially, we don't know what unification we will hoist out, so the
% state is before_candidate. When we find a unification we want to
% hoist out, this fixes the functor, and the state is have_candidate.
% If we find that some branches unify that variable with some other
% functor, we have multiple_candidates, which means that we don't hoist
% out any of them. (Although our caller may try again with another
% variable.)
%
% The goal field contains the unification we are proposing to put
% before the branched control structure. The first_old_new field
% gives the mapping from argument variables in the old unification
% in the first branch to the freshly created variables in the goal
% being hoisted before the branched control structure. The
% later_old_new field contains the same information for the second
% and later branches.
:- type cse_state
---> before_candidate
; have_candidate(
goal :: hlds_goal,
first_old_new :: assoc_list(prog_var),
later_old_new :: list(assoc_list(prog_var))
)
; multiple_candidates.
:- pred find_bind_var_for_cse_in_deconstruct(prog_var::in, hlds_goal::in,
list(hlds_goal)::out, cse_state::in, cse_state::out,
cse_info::in, cse_info::out) is det.
find_bind_var_for_cse_in_deconstruct(Var, Goal0, Goals,
!CseState, !CseInfo) :-
(
!.CseState = before_candidate,
construct_common_unify(Var, Goal0, !CseInfo,
OldNewVars, HoistedGoal, Goals),
!:CseState = have_candidate(HoistedGoal, OldNewVars, [])
;
!.CseState = have_candidate(HoistedGoal,
FirstOldNewVars, LaterOldNewVars0),
Goal0 = _ - GoalInfo,
goal_info_get_context(GoalInfo, Context),
(
find_similar_deconstruct(HoistedGoal,
Goal0, Context, OldNewVars, Goals0)
->
Goals = Goals0,
LaterOldNewVars = [OldNewVars | LaterOldNewVars0],
!:CseState = have_candidate(HoistedGoal,
FirstOldNewVars, LaterOldNewVars)
;
Goals = [Goal0],
!:CseState = multiple_candidates
)
;
!.CseState = multiple_candidates,
Goals = [Goal0],
!:CseState = multiple_candidates
).
:- pred construct_common_unify(prog_var::in, hlds_goal::in,
cse_info::in, cse_info::out, assoc_list(prog_var)::out,
hlds_goal::out, list(hlds_goal)::out) is det.
construct_common_unify(Var, GoalExpr0 - GoalInfo, !CseInfo, OldNewVars,
HoistedGoal, Replacements) :-
(
GoalExpr0 = unify(_, Term, Umode, Unif0, Ucontext),
Unif0 = deconstruct(_, Consid, Args, Submodes, CanFail, CanCGC)
->
Unif = deconstruct(Var, Consid, Args, Submodes, CanFail,
CanCGC),
( Term = functor(_, _, _) ->
GoalExpr1 = unify(Var, Term, Umode, Unif, Ucontext)
;
error("non-functor unify in construct_common_unify")
),
goal_info_get_context(GoalInfo, Context),
create_parallel_subterms(Args, Context, Ucontext,
!CseInfo, OldNewVars, Replacements),
map__from_assoc_list(OldNewVars, Sub),
goal_util__rename_vars_in_goal(GoalExpr1 - GoalInfo, Sub,
HoistedGoal)
;
error("non-unify goal in construct_common_unify")
).
:- pred create_parallel_subterms(list(prog_var)::in, prog_context::in,
unify_context::in, cse_info::in, cse_info::out,
assoc_list(prog_var)::out, list(hlds_goal)::out) is det.
create_parallel_subterms([], _, _, !CseInfo, [], []).
create_parallel_subterms([OFV | OFV0], Context, UnifyContext, !CseInfo,
OldNewVars, Replacements) :-
create_parallel_subterms(OFV0, Context, UnifyContext, !CseInfo,
OldNewVars1, Replacements1),
create_parallel_subterm(OFV, Context, UnifyContext, !CseInfo,
OldNewVars1, OldNewVars, Goal),
Replacements = [Goal | Replacements1].
:- pred create_parallel_subterm(prog_var::in, prog_context::in,
unify_context::in, cse_info::in, cse_info::out,
assoc_list(prog_var)::in, assoc_list(prog_var)::out,
hlds_goal::out) is det.
create_parallel_subterm(OFV, Context, UnifyContext, !CseInfo, !OldNewVar,
Goal) :-
VarSet0 = !.CseInfo ^ varset,
VarTypes0 = !.CseInfo ^ vartypes,
varset__new_var(VarSet0, NFV, VarSet),
map__lookup(VarTypes0, OFV, Type),
map__det_insert(VarTypes0, NFV, Type, VarTypes),
!:OldNewVar = [OFV - NFV | !.OldNewVar],
UnifyContext = unify_context(MainCtxt, SubCtxt),
create_atomic_unification(OFV, var(NFV),
Context, MainCtxt, SubCtxt, Goal),
!:CseInfo = (!.CseInfo ^ varset := VarSet) ^ vartypes := VarTypes.
%-----------------------------------------------------------------------------%
:- pred find_similar_deconstruct(hlds_goal::in, hlds_goal::in,
prog_context::in, assoc_list(prog_var)::out, list(hlds_goal)::out)
is semidet.
find_similar_deconstruct(HoistedUnifyGoal, OldUnifyGoal, Context,
OldHoistedVars, Replacements) :-
(
HoistedUnifyGoal = unify(_, _, _, HoistedUnifyInfo, OC) - _,
HoistedUnifyInfo = deconstruct(_, HoistedFunctor,
HoistedVars, _, _, _),
OldUnifyGoal = unify(_, _, _, OldUnifyInfo, _NC) - _,
OldUnifyInfo = deconstruct(_, OldFunctor, OldVars, _, _, _)
->
HoistedFunctor = OldFunctor,
list__length(HoistedVars, HoistedVarsCount),
list__length(OldVars, OldVarsCount),
HoistedVarsCount = OldVarsCount,
assoc_list__from_corresponding_lists(OldVars, HoistedVars,
OldHoistedVars),
pair_subterms(OldHoistedVars, Context, OC, Replacements)
;
error("find_similar_deconstruct: non-deconstruct unify")
).
:- pred pair_subterms(assoc_list(prog_var)::in, prog_context::in,
unify_context::in, list(hlds_goal)::out) is det.
pair_subterms([], _Context, _UnifyContext, []).
pair_subterms([OldVar - HoistedVar | OldHoistedVars], Context, UnifyContext,
Replacements) :-
pair_subterms(OldHoistedVars, Context, UnifyContext, Replacements1),
( OldVar = HoistedVar ->
Replacements = Replacements1
;
UnifyContext = unify_context(MainCtxt, SubCtxt),
create_atomic_unification(HoistedVar, var(OldVar),
Context, MainCtxt, SubCtxt, Goal),
Replacements = [Goal | Replacements1]
).
%-----------------------------------------------------------------------------%
% This section handles the case where the functor involved in the
% common subexpression contains existentially typed arguments,
% whether or not they are constrained to belong to a typeclass.
% In such cases, what the compiler used to consider several distinct
% types (the types of say the first the existentially typed argument
% in the deconstructions in the different branches) become one (in this
% case, the type of the first existentially typed argument in the
% hoisted out deconstruction). The prog_vars describing the types
% of the existentially typed arguments (i.e. containing their
% typeinfos) change as well, from being some of the variables in
% in the original deconstructions to being the corresponding variables
% in the hoisted out deconstruction.
%
% As an example, consider a disjunction such as
%
% (
% HeadVar__2_2 = x:u(TypeClassInfo_for_v_8, V_4),
% ...
% ;
% HeadVar__2_2 = x:u(TypeClassInfo_for_v_14, V_6)
% ...
% )
%
% The main part of cse_detection will replace this with
%
% HeadVar__2_2 = x:u(V_17, V_16)
% (
% TypeClassInfo_for_v_8 = V_17,
% V_4 = V_16,
% ...
% ;
% TypeClassInfo_for_v_14 = V_17,
% V_6 = V_16,
% ...
% )
%
% However, this is not enough. Since TypeClassInfo_for_v_8 and
% TypeClassInfo_for_v_14 may (and probably will) be eliminated later,
% it is imperative that the data structures in the proc_info that refer
% to them be updated to eliminate references to those variables.
% Those data structures may originally contain something like this:
%
% type_info varmap:
% T_1 (number 1) -> typeclass_info(TypeClassInfo_for_v_8, 1)
% T_3 (number 3) -> typeclass_info(TypeClassInfo_for_v_14, 1)
% typeclass_info varmap:
% x:v(T_1) -> TypeClassInfo_for_v_8
% x:v(T_3) -> TypeClassInfo_for_v_14
% variable types map:
% V_4 (number 4) :: T_1
% V_6 (number 6) :: T_3
%
% They must be updated like this:
%
% type_info varmap:
% T_1 (number 1) -> typeclass_info(V_17, 1)
% typeclass_info varmap:
% x:v(T_1) -> V_17
% variable types map:
% V_4 (number 4) :: T_1
% V_6 (number 6) :: T_1
:- pred maybe_update_existential_data_structures(hlds_goal::in,
assoc_list(prog_var)::in, list(assoc_list(prog_var))::in,
cse_info::in, cse_info::out) is det.
maybe_update_existential_data_structures(Unify, FirstOldNew, LaterOldNew,
!CseInfo) :-
(
Unify = unify(_, _, _, UnifyInfo, _) - _,
UnifyInfo = deconstruct(Var, ConsId, _, _, _, _),
ModuleInfo = !.CseInfo ^ module_info,
VarTypes = !.CseInfo ^ vartypes,
map__lookup(VarTypes, Var, Type),
type_util__is_existq_cons(ModuleInfo, Type, ConsId)
->
update_existential_data_structures(FirstOldNew, LaterOldNew,
!CseInfo)
;
true
).
:- pred update_existential_data_structures(
assoc_list(prog_var)::in, list(assoc_list(prog_var))::in,
cse_info::in, cse_info::out) is det.
update_existential_data_structures(FirstOldNew, LaterOldNews, !CseInfo) :-
list__condense(LaterOldNews, LaterOldNew),
map__from_assoc_list(FirstOldNew, FirstOldNewMap),
map__from_assoc_list(LaterOldNew, LaterOldNewMap),
RttiVarMaps0 = !.CseInfo ^ rtti_varmaps,
VarTypes0 = !.CseInfo ^ vartypes,
% Build a map for all locations in the rtti_varmaps that are
% changed by the application of FirstOldNewMap. The keys
% of this map are the new locations, and the values are
% the tvars (from the first branch) that have had their
% locations moved.
%
rtti_varmaps_tvars(RttiVarMaps0, TvarsList),
list__foldl(find_type_info_locn_tvar_map(RttiVarMaps0, FirstOldNewMap),
TvarsList, map__init, NewTvarMap),
% Traverse TVarsList again, this time looking for locations
% in later branches that merge with locations in the first
% branch. When we find one, add a type substitution which
% represents the type variables that were merged.
%
list__foldl(find_merged_tvars(RttiVarMaps0, LaterOldNewMap, NewTvarMap),
TvarsList, map__init, TSubst),
% Apply the full old->new map and the type substitution
% to the rtti_varmaps, and apply the type substitution to the
% vartypes.
%
list__append(FirstOldNew, LaterOldNew, OldNew),
map__from_assoc_list(OldNew, OldNewMap),
apply_substitutions_to_rtti_varmaps(TSubst, map__init, OldNewMap,
RttiVarMaps0, RttiVarMaps),
map__map_values(apply_tvar_rename(TSubst), VarTypes0, VarTypes),
!:CseInfo = !.CseInfo ^ rtti_varmaps := RttiVarMaps,
!:CseInfo = !.CseInfo ^ vartypes := VarTypes.
:- pred apply_tvar_rename(tsubst::in, prog_var::in, (type)::in, (type)::out)
is det.
apply_tvar_rename(TSubst, _Var, Type0, Type) :-
Type = term__apply_substitution(Type0, TSubst).
:- pred find_type_info_locn_tvar_map(rtti_varmaps::in,
map(prog_var, prog_var)::in, tvar::in,
map(type_info_locn, tvar)::in, map(type_info_locn, tvar)::out) is det.
find_type_info_locn_tvar_map(RttiVarMaps, FirstOldNewMap, Tvar, !NewTvarMap) :-
rtti_lookup_type_info_locn(RttiVarMaps, Tvar, TypeInfoLocn0),
type_info_locn_var(TypeInfoLocn0, Old),
( map__search(FirstOldNewMap, Old, New) ->
type_info_locn_set_var(New, TypeInfoLocn0, TypeInfoLocn),
svmap__det_insert(TypeInfoLocn, Tvar, !NewTvarMap)
;
true
).
:- pred find_merged_tvars(rtti_varmaps::in, map(prog_var, prog_var)::in,
map(type_info_locn, tvar)::in, tvar::in, tsubst::in, tsubst::out)
is det.
find_merged_tvars(RttiVarMaps, LaterOldNewMap, NewTvarMap, Tvar, !TSubst) :-
rtti_lookup_type_info_locn(RttiVarMaps, Tvar, TypeInfoLocn0),
type_info_locn_var(TypeInfoLocn0, Old),
( map__search(LaterOldNewMap, Old, New) ->
type_info_locn_set_var(New, TypeInfoLocn0, TypeInfoLocn),
map__lookup(NewTvarMap, TypeInfoLocn, NewTvar),
( NewTvar = Tvar ->
true
;
svmap__det_insert(Tvar, term__variable(NewTvar),
!TSubst)
)
;
true
).
%-----------------------------------------------------------------------------%