comp.lang.ada
 help / color / mirror / Atom feed
From: Georg Bauhaus <rm.dash-bauhaus@futureapps.de>
Subject: Re: Performance of element access in Vector
Date: Fri, 23 Jan 2009 18:30:22 +0100
Date: 2009-01-23T18:30:23+01:00	[thread overview]
Message-ID: <4979feaf$0$31870$9b4e6d93@newsspool3.arcor-online.net> (raw)
In-Reply-To: <6tu43dFcibgkU1@mid.individual.net>

Alex R. Mosteo schrieb:
> Georg Bauhaus wrote:
> 
>> Maciej Sobczak schrieb:
>>
>>> The problem here is that the compiler does not inline Update_Element/
>>> Query_Element callbacks. Safety and performance are not in conflict
>>> here, it is just a quality of implementation issue.
>>> You can have both.
>> FWIW, if Iterate/Process is made a generic procedure as
>> in the original AI-302, and Increment is made a library
>> level procedure,  then GNAT confirms, in a sense,
>> that there might be a quality of implementation issue:
>>
>> raw array                : 00:00:29.78
>> raw array process        : 00:01:06.74
>> raw array generic        : 00:00:29.78
> 
> Do you mean, IIUC, that if 'process' where properly inlined, the timing 
> would be same for the three instances? That would be nice.

The third line seems to indicate the possibility
of inlining, provided nothing else gets in the
way---in the case of 'Access to procedure; can't say.
Complete example with Increment made a library level
procedure for all three methods:

-- 8<--

procedure Increment (I : in out Integer);
pragma Inline(Increment);


with Ada.Calendar.Formatting;
with Ada.Text_IO;
with Increment;

procedure Test_P is
   Iterations: constant := 10_000;

   type Int_Array is array(Positive range <>) of Integer;

   Start : Ada.Calendar.Time;
   Stop : Ada.Calendar.Time;

   use type Ada.Calendar.Time;

   procedure Test_0A (Container: in out Int_Array;
                      Process: not null access procedure (Element: in
out Integer))  is
   begin
      Start := Ada.Calendar.Clock;

      for J in 1 .. Iterations loop
         for Index in Container'Range loop
            Process(Container(Index));
         end loop;
      end loop;

      Stop := Ada.Calendar.Clock;
      Ada.Text_IO.Put_Line
        ("raw array process        : " &
           Ada.Calendar.Formatting.Image (Stop - Start, True));
   end Test_0A;

   generic
      with procedure Process (Element: in out Integer);
   procedure Test_0B (Container: in out Int_Array);


   procedure Test_0B (Container: in out Int_Array)  is
   begin
      Start := Ada.Calendar.Clock;

      for J in 1 .. Iterations loop
         for Index in Container'Range loop
            Process(Container(Index));
         end loop;
      end loop;

      Stop := Ada.Calendar.Clock;
      Ada.Text_IO.Put_Line
        ("raw array generic        : " &
           Ada.Calendar.Formatting.Image (Stop - Start, True));
   end Test_0B;


   procedure Test_0(Container: in out Int_Array) is
   begin
      Start := Ada.Calendar.Clock;

      for J in 1 .. Iterations loop
         for I in Container'range loop
            Container(I) := Container(I) + 1;
         end loop;
      end loop;

      Stop := Ada.Calendar.Clock;
      Ada.Text_IO.Put_Line
        ("raw array                : " &
           Ada.Calendar.Formatting.Image (Stop - Start, True));
   end Test_0;

   Z: Int_Array(Positive range 1 .. 1_000_000);
begin
   Test_0(Z);
   Test_0A(Z, Increment'Access);
   declare
      procedure Test_G is new Test_0B(Process => Increment);
   begin
      Test_G(Z);
   end;
end Test_P;


procedure Increment (I : in out Integer) is
begin
   I := I + 1;
end Increment;



      reply	other threads:[~2009-01-23 17:30 UTC|newest]

Thread overview: 26+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2009-01-18 21:42 Performance of element access in Vector Maciej Sobczak
2009-01-19  0:03 ` george.priv
2009-01-19  3:23 ` george.priv
2009-01-19 11:25   ` Georg Bauhaus
2009-01-19 16:32   ` (see below)
2009-01-20  2:17 ` Randy Brukardt
2009-01-20  8:03   ` Maciej Sobczak
2009-01-20  8:26     ` Dmitry A. Kazakov
2009-01-20 22:07       ` george.priv
2009-01-21  8:52         ` Maciej Sobczak
2009-01-21 19:25           ` george.priv
2009-01-22 10:01             ` Georg Bauhaus
2009-01-22 12:43               ` Maciej Sobczak
2009-01-22 13:52               ` george.priv
2009-01-21  8:59         ` Dmitry A. Kazakov
2009-01-21  9:19           ` Maciej Sobczak
2009-01-21 10:19             ` Dmitry A. Kazakov
2009-01-21 13:14               ` Maciej Sobczak
2009-01-21 19:00                 ` Dmitry A. Kazakov
2009-01-21 13:22             ` Georg Bauhaus
2009-01-23 14:56         ` Alex R. Mosteo
2009-01-20 23:01   ` Randy Brukardt
2009-01-21  9:15     ` Maciej Sobczak
2009-01-21 18:08       ` Georg Bauhaus
2009-01-23 14:55         ` Alex R. Mosteo
2009-01-23 17:30           ` Georg Bauhaus [this message]
replies disabled

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox