From mboxrd@z Thu Jan  1 00:00:00 1970
X-Spam-Checker-Version: SpamAssassin 3.4.4 (2020-01-24) on polar.synack.me
X-Spam-Level: 
X-Spam-Status: No, score=-1.3 required=5.0 tests=BAYES_00,INVALID_MSGID
	autolearn=no autolearn_force=no version=3.4.4
X-Google-Language: ENGLISH,ASCII-7-bit
X-Google-Thread: 103376,bd45e29f9dafca87
X-Google-Attributes: gid103376,public
From: Gautier <gautier.demontmollin@maths.unine.ch>
Subject: Re: bitwise comparators
Date: 2000/01/18
Message-ID: <3883B6BA.476AD55@maths.unine.ch>#1/1
X-Deja-AN: 573985350
Content-Transfer-Encoding: 7bit
References: <3880D375.7E363123@hotmail.com> <38829638.0@news.pacifier.com>
 <3882FC1C.2BA8C959@hotmail.com> <85vmn2$ki1$1@nnrp1.deja.com>
 <38836CF2.AB738B8B@hotmail.com> <388394E0.55D1E913@maths.unine.ch>
 <38838989.4F158467@hotmail.com>
X-Accept-Language: en
Content-Type: text/plain; charset=us-ascii
X-Trace: 18 Jan 2000 00:40:07 +0100, 130.125.13.32
MIME-Version: 1.0
Newsgroups: comp.lang.ada
Date: 2000-01-18T00:00:00+00:00
List-Id: <comp.lang.ada>

Alexander Van Hecke wrote:

> > > use structs and callback functions and you have perfect generic types!

> > You imagine a rather inefficient way to implement generics.

> As a matter of fact, this way is more efficient than using templates!  Templates have to
> be copied and filled in -> lots of overhead.  If you use callback functions in structs,
> all you have to do is link the correct pointers to the adresses of the functions and
> that's it.

I see - indeed for some "percent bar" feedback purposes I do so if callback
doen't need to be inlined+optimized. It saves space.
But genericity embrace more than functions but all sorts of parameters.
One usage is to produce lots of overhead, but for each instance an optimized
code for it. A trivial example of it is the PKZip `explode' method that
appears in 2 versions, almost the same, apart of 2 parameters.
As variables they would slow down the code, so it is preferable to
instanciate twice the code with these numbers built-in and let the optimizer
squeeze the shifts etc. But, only one copy of source is needed 

generic
  needed: integer;
  mask:   unsigned_32;
procedure explode_lit ( tb, tl, td : p_Table_list; bb, bl, bd : integer );
(etc.)
Then, the instanciations:
-- **************** exploding, method: 8k slide, 3 trees *****************

procedure explode_lit8 is new explode_lit( 7, 16#7F# );

-- **************** exploding, method: 4k slide, 3 trees *****************

procedure explode_lit4 is new explode_lit( 6, 16#3F# );

See http://members.xoom.com/_XMCM/gdemont/uza_html/unz_expl__adb.htm#60_11

For that a C macro would fit a best, wouldn't it ?

But you can do more abstracted things like...
generic                              -- to provide:
          type ring_elt is private;                  -- ring element type 
          zero, one: ring_elt;                       -- 0 and 1 elements

          with function "-" (a:ring_elt) return ring_elt;    -- unary oper.
          with function "+" (a,b:ring_elt) return ring_elt;  -- binary oper.
          with function "*" (a,b:ring_elt) return ring_elt;

package Frac is
  type frac_elt is record a,b:ring_elt; end record;        -- define fraction

  frac_0: constant frac_elt:= (zero,one); 
  frac_1: constant frac_elt:= (one,one); 
(...)
and use it for creating rational numbers:
  package Rationals is new Frac_Euclid(Integer, 0,1, "-","+","-","*","/");
then rational polynomials:
  package Rationals.Polynomials is new
             Polynomials( frac_elt, frac_0, frac_1, "-","+","-","*","/","=" ); 
then using Frac again to create rational functions!

(end of generic trip - the subject is so wide... )

-- 
Gautier

_____\\________________\_______\
http://members.xoom.com/gdemont/