Re: [fixed-point] Patch to parse _Fract, _Accum, _Sat

["Chao-ying Fu" <fu at mips dot com>]

Hello,

  Minor changes are made as follows.  Thanks!

Regards,
Chao-ying

2006-11-08  Chao-ying Fu  <fu@mips.com>

    * tree.h (struct tree_common): Add a comment about that saturating_flag
    increases the size of tree_common by a full word.
    (make_fract_type): Add a comment.
    * tree.c (build_common_tree_nodes_2): Test fixed_point_supported_p to
    build fixed-point tree nodes conditionally.
    * c-parser.c (c_parser_declspecs): Add a comment for the link to
    the DTR defining _Fract, _Accum, and _Sat.


----- Original Message ----- 
From: "Steven Bosscher" <stevenb.gcc@gmail.com>
To: "Chao-ying Fu" <fu@mips.com>
Cc: "Thekkath, Radhika" <radhika@mips.com>; "Stephens, Nigel"
<nigel@mips.com>; <gcc-patches@gcc.gnu.org>
Sent: Wednesday, November 08, 2006 3:32 PM
Subject: Re: [fixed-point] Patch to parse _Fract, _Accum, _Sat


> On 11/8/06, Chao-ying Fu <fu@mips.com> wrote:
> > Hello,
> >
> >   Thanks for your review!
> >
> > > > +  unsigned saturating_flag : 1;
> > >
> > > You probably have noticed that the flags are grouped in sets of 8
bits,
> > > and that there are four groups of 8 bits.  You've added another bit,
> > > and increased the size of tree_common by a full word.  Is that change
> > > on purpose?
> >
> >   About saturating_flag, if I cannot add a new flag in either
tree_common
> > or tree_type, I will need to use lang_flag_2  in c-tree.h
> > Ex:
> > #define C_TYPE_SATURATING(TYPE) TYPE_LANG_FLAG_2 (TYPE)
>
> Neh, you're implementing a set of types in the middle-end, so using a
> C only flag is Badness.
>
> I'd keep it as you have it with saturing_flag for now, but add a FIXME
> there so that it doesn't slip in if/when you merge to the trunk many
> months from now.  We're probably going to have to shuffle with those
> flags anyway when all the 'tree' cleanup projects go in.
>
> Gr.
> Steven
>

Attachment: gcc.diff
Description: Binary data