[m-rev.] diff: delete a bad assumption in structure sharing

Peter Wang novalazy at gmail.com
Tue Jun 10 16:20:19 AEST 2008


On 2008-06-10, Julien Fischer <juliensf at csse.unimelb.edu.au> wrote:
>
> On Tue, 10 Jun 2008, Peter Wang wrote:
>
>> Branches: main
>>
>> compiler/structure_sharing.domain.m:
>> 	Delete a bad assumption that no procedures in builtin modules
>> 	introduce sharing between arguments.  For example,
>> 	`private_builtin.typed_unify(in, out)' does.
>>
>> Index: compiler/structure_sharing.domain.m
>> ===================================================================
>> RCS file: /home/mercury/mercury1/repository/mercury/compiler/structure_sharing.domain.m,v
>> retrieving revision 1.35
>> diff -u -p -r1.35 structure_sharing.domain.m
>> --- compiler/structure_sharing.domain.m	5 Jun 2008 06:25:18 -0000	1.35
>> +++ compiler/structure_sharing.domain.m	10 Jun 2008 05:38:05 -0000
>> @@ -892,11 +892,6 @@ predict_called_pred_is_bottom(ModuleInfo
>>         % "unify", "index", "compare" or "initialise".
>>         pred_info_get_origin(PredInfo, Origin),
>>         Origin = origin_special_pred(_)
>> -    ;
>> -        % 4. (XXX UNSAFE!! To verify) any call to private_builtin and builtin
>> -        % procedures.
>> -        PredModule = pred_info_module(PredInfo),
>> -        any_mercury_builtin_module(PredModule)
>>     ).
>>
>> :- func top_sharing_not_found(pred_proc_id) = sharing_as.
>
> You could be more specific about this since most of the procedures in
> builtin and private_builtin will be bottom.

The other disjuncts (before the one that was deleted) catch those
procedures.  I did check the results for builtin and private_builtin
individually.

Peter

--------------------------------------------------------------------------
mercury-reviews mailing list
Post messages to:       mercury-reviews at csse.unimelb.edu.au
Administrative Queries: owner-mercury-reviews at csse.unimelb.edu.au
Subscriptions:          mercury-reviews-request at csse.unimelb.edu.au
--------------------------------------------------------------------------



More information about the reviews mailing list