On Thu, Oct 29, 2015 at 4:52 PM, Jan Hubicka <hubi...@ucw.cz> wrote:
>> On Thu, Oct 29, 2015 at 4:02 PM, Jan Hubicka <hubi...@ucw.cz> wrote:
>> >>
>> >> IMHO it was always wrong/fragile for backends to look at the actual
>> >> arguments to
>> >> decide on the calling convention. The backends should _solely_ rely on
>> >> gimple_call_fntype and its TYPE_ARG_TYPES here.
>> >>
>> >> Of course then there are varargs ... (not sure if we hit this here).
>> >
>> > Yep, you have varargs and K&R prototypes, so it can't work this way.
>>
>> Well, then I suppose we need to compute the ABI upfront when we gimplify
>> from the orginal args (like we preserve fntype). Having a separate fntype
>> was really meant to make us preserve the ABI throughout the GIMPLE phase...
>
> Hmm, the idea of doing some part of ABI explicitly is definitly nice (at least
> the implicit promotions and pass by reference bits), but storing the full
> lowlevel info on how to pass argument seems bit steep. You will need to
> preserve the RTL containers for parameters that may get non-trivial (PARALLEL)
> and precompute all the other information how to get data on stack.
>
> While playing with the ABi checker I was just looking into this after several
> years (when i was cleaning up calls.c) and calls.c basically works by
> computing
> arg_data that holds most of the info you would need (you need also return
> argument passing and the hidden argument for structure returns). You can
> check
> it out - it is fairly non-trivial beast plus it really holds two parallel sets
> of infos - tailcall and normal call (because these differ for targets with
> register windows). The info also depends on flags used to compile function
> body
> (such as -maccumulate-outgoing-args)
>
> To make something like this a permanent part of GIMPLE would probably need
> quite
> careful re-engineering of the APIs inventing more high-level intermediate
> representation to get out of the machine description. There is not realy
> immediate
> benefit from knowing how parameters are housed on stack for gimple
> optimizers, so
> perhaps just keeping the type information (after promotions) as the way to
> specify
> call conventions is more practical way to go.
Yeah, I suppose we'd need to either build a new function type for each
variadic call
then or somehow represent 'fntype' differently (note that function
attributes also
need to be preserved).
Richard.
> Honza
>
>> >> But yes, the VIEW_CONVERT "stripping" is a bit fragile and I don't
>> >> remember
>> >> what exactly we gain from it (when not done on registers).
>> >
>> > I guess gain is really limited to Ada - there are very few cases we do VCE
>> > otherwise.
>> > (I think we could do more of them). We can make useless_type_conversion
>> > NOP/CONVERT
>> > only. That in fact makes quite a sense because those are types with gimple
>> > operations
>> > on it. Perhaps also VCE on vectors, but not VCE in general.
>> >
>> > Honza
>> >>
>> >> But I also don't see where we do the stripping mentioned on memory
>> >> references.
>> >> The match.pd pattern doesn't apply to memory, only in the GENERIC path
>> >> which is guarded with exact type equality. So I can't see where we end up
>> >> stripping the V_C_E.
>> >>
>> >> There is one bogus case still in fold-const.c:
>> >>
>> >> case VIEW_CONVERT_EXPR:
>> >> if (TREE_CODE (op0) == MEM_REF)
>> >> /* ??? Bogus for aligned types. */
>> >> return fold_build2_loc (loc, MEM_REF, type,
>> >> TREE_OPERAND (op0, 0), TREE_OPERAND (op0,
>> >> 1));
>> >>
>> >> return NULL_TREE;
>> >>
>> >> that comment is only in my local tree ... (we lose alignment info that is
>> >> on the original MEM_REF type which may be a smaller one).
>> >>
>> >> Richard.
>> >>
>> >> > Honza
>> >> >>
>> >> >>
>> >> >> * gnat.dg/discr44.adb: New test.
>> >> >>
>> >> >> --
>> >> >> Eric Botcazou
>> >> >
>> >> >> -- { dg-do run }
>> >> >> -- { dg-options "-gnatws" }
>> >> >>
>> >> >> procedure Discr44 is
>> >> >>
>> >> >> function Ident (I : Integer) return Integer is
>> >> >> begin
>> >> >> return I;
>> >> >> end;
>> >> >>
>> >> >> type Int is range 1 .. 10;
>> >> >>
>> >> >> type Str is array (Int range <>) of Character;
>> >> >>
>> >> >> type Parent (D1, D2 : Int; B : Boolean) is record
>> >> >> S : Str (D1 .. D2);
>> >> >> end record;
>> >> >>
>> >> >> type Derived (D : Int) is new Parent (D1 => D, D2 => D, B => False);
>> >> >>
>> >> >> X1 : Derived (D => Int (Ident (7)));
>> >> >>
>> >> >> begin
>> >> >> if X1.D /= 7 then
>> >> >> raise Program_Error;
>> >> >> end if;
>> >> >> end;
>> >> >
