[m-rev.] for reviews: fix optimization in unify_proc.m

[Fergus Henderson]

On 31-Mar-2004, Zoltan Somogyi <zs at cs.mu.OZ.AU> wrote:
> For review by Fergus.
> 
> Zoltan.
> 
> compiler/unify_proc.m:
> 	Reenable an optimization of unify procedures for general du types
> 	after fixing the problem that caused it to be disabled. This required
> 	placing a feature on casts from constants to integers that ask the mode
> 	checker to preserve any binding information it has about the constant
> 	as the equivalent binding information about the integer.
> 
> compiler/hlds_goal.m:
> 	Add this feature to the list of goal features.
> 
> compiler/modecheck_call.m:
> 	When processing unsafe_cast goals, respect this feature.
> 
> tests/hard_coded/det_complicated_unify2.m:
> 	Make this test case, which was the test case for the original problem
> 	with the optimization, even tougher, to make sure the new version
> 	works. Update the test case documentation.
> 
> cvs diff: Diffing .
> cvs diff: Diffing analysis
> cvs diff: Diffing bindist
> cvs diff: Diffing boehm_gc
> cvs diff: Diffing boehm_gc/Mac_files
> cvs diff: Diffing boehm_gc/cord
> cvs diff: Diffing boehm_gc/cord/private
> cvs diff: Diffing boehm_gc/doc
> cvs diff: Diffing boehm_gc/include
> cvs diff: Diffing boehm_gc/include/private
> cvs diff: Diffing boehm_gc/tests
> cvs diff: Diffing browser
> cvs diff: Diffing bytecode
> cvs diff: Diffing compiler
> Index: compiler/hlds_goal.m
> ===================================================================
> RCS file: /home/mercury/mercury1/repository/mercury/compiler/hlds_goal.m,v
> retrieving revision 1.113
> diff -u -b -r1.113 hlds_goal.m
> --- compiler/hlds_goal.m	31 Mar 2004 08:52:22 -0000	1.113
> +++ compiler/hlds_goal.m	31 Mar 2004 09:00:10 -0000
> @@ -742,8 +742,16 @@
>  				% be hidden. This is used e.g. by the tabling
>  				% transformation to preserve the set of events
>  				% generated by a tabled procedure.
> -	;	tailcall.	% This goal represents a tail call. This marker
> +	;	tailcall	% This goal represents a tail call. This marker
>  				% is used by deep profiling.
> +	;	keep_constant_binding.
> +				% This feature should only be attached to
> +				% unsafe_cast goals that cast a value of an
> +				% user-defined type to an integer. It tells
> +				% the mode checker that if the first variable
> +				% is known to be bound to a given constant,
> +				% then the second variable should be set
> +				% to the corresponding local tag value.
>  
>  	% We can think of the goal that defines a procedure to be a tree,
>  	% whose leaves are primitive goals and whose interior nodes are
> Index: compiler/modes.m
> ===================================================================
> RCS file: /home/mercury/mercury1/repository/mercury/compiler/modes.m,v
> retrieving revision 1.276
> diff -u -b -r1.276 modes.m
> --- compiler/modes.m	31 Mar 2004 08:52:23 -0000	1.276
> +++ compiler/modes.m	31 Mar 2004 09:06:36 -0000
> @@ -190,7 +190,8 @@
>  
>  % The following predicates are used by unique_modes.m.
>  
> -:- import_module check_hlds__mode_info, hlds__hlds_data.
> +:- import_module check_hlds__mode_info.
> +:- import_module hlds__hlds_data.
>  
>  	% Modecheck a unification.
>  
> @@ -1193,9 +1194,43 @@
>  		error("modecheck_goal_expr: class_method_call")
>  	;
>  		GenericCall = unsafe_cast,
> -		modecheck_builtin_cast(Modes0, Args0, Args, Det, ExtraGoals,
> +		(
> +			goal_info_has_feature(GoalInfo0,
> +				keep_constant_binding),
> +			mode_info_get_instmap(!.ModeInfo, InstMap),
> +			(
> +				Args0 = [Arg1Prime, _Arg2Prime],
> +				Modes0 = [Mode1Prime, Mode2Prime]
> +			->
> +				Arg1 = Arg1Prime,
> +				Mode1 = Mode1Prime,
> +				Mode2 = Mode2Prime
> +			;
> +				error("modecheck_goal_expr: bad unsafe_cast")
> +			),
> +			Mode1 = in_mode,
> +			Mode2 = out_mode,
> +			instmap__lookup_var(InstMap, Arg1, Inst1),
> +			Inst1 = bound(Unique, [functor(ConsId, [])]),
> +			mode_info_get_module_info(!.ModeInfo, ModuleInfo),
> +			module_info_types(ModuleInfo, TypeTable),
> +			mode_info_get_var_types(!.ModeInfo, VarTypes),
> +			map__lookup(VarTypes, Arg1, ArgType1),
> +			type_to_ctor_and_args(ArgType1, ArgTypeCtor1, _),
> +			map__lookup(TypeTable, ArgTypeCtor1, CtorDefn),
> +			get_type_defn_body(CtorDefn, Body),
> +			ConsTagValues = Body ^ du_type_cons_tag_values,
> +			map__lookup(ConsTagValues, ConsId, ConsTag),
> +			ConsTag = shared_local_tag(_, LocalTag)
> +		->
> +			BoundInst = functor(int_const(LocalTag), []),

Why do you ignore the first argument of shared_local_tag here?
Won't that code give an incorrect result when the first argument of
shared_local_tag is non-zero?

Otherwise that change looks fine.

-- 
Fergus Henderson                    |  "I have always known that the pursuit
                                    |  of excellence is a lethal habit"
WWW: <http://www.cs.mu.oz.au/~fjh>  |     -- the last words of T. S. Garp.
--------------------------------------------------------------------------
mercury-reviews mailing list
post:  mercury-reviews at cs.mu.oz.au
administrative address: owner-mercury-reviews at cs.mu.oz.au
unsubscribe: Address: mercury-reviews-request at cs.mu.oz.au Message: unsubscribe
subscribe:   Address: mercury-reviews-request at cs.mu.oz.au Message: subscribe
--------------------------------------------------------------------------