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=-1.9 required=5.0 tests=BAYES_00 autolearn=ham
	autolearn_force=no version=3.4.4
X-Google-Thread: 103376,bb7cc916bd63ab43
X-Google-Attributes: gid103376,domainid0,public,usenet
X-Google-Language: ENGLISH,ASCII-7-bit
Path: 
 g2news2.google.com!news3.google.com!goblin2!goblin1!goblin.stu.neva.ru!tiscali!newsfeed1.ip.tiscali.net!newsfeed.tiscali.ch!npeer.de.kpn-eurorings.net!npeer-ng1.de.kpn-eurorings.net!newsfeed.arcor.de!newsspool3.arcor-online.net!news.arcor.de.POSTED!not-for-mail
Date: Mon, 19 Jan 2009 12:25:42 +0100
From: Georg Bauhaus <rm.dash-bauhaus@futureapps.de>
User-Agent: Thunderbird 2.0.0.19 (Macintosh/20081209)
MIME-Version: 1.0
Newsgroups: comp.lang.ada
Subject: Re: Performance of element access in Vector
References: 
 <0a8baaf0-19f7-40c0-a041-884e93fa7020@w39g2000prb.googlegroups.com>
 <c8a97381-667c-4fda-8a1b-b4035f4281b8@o40g2000yqb.googlegroups.com>
In-Reply-To: 
 <c8a97381-667c-4fda-8a1b-b4035f4281b8@o40g2000yqb.googlegroups.com>
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
Message-ID: <49746336$0$30238$9b4e6d93@newsspool1.arcor-online.net>
Organization: Arcor
NNTP-Posting-Date: 19 Jan 2009 12:25:42 CET
NNTP-Posting-Host: e34275ff.newsspool1.arcor-online.net
X-Trace: 
 DXC=Z>K7^:oI?QXaAeROF2PWMQic==]BZ:af^4Fo<]lROoRQ^YC2XCjHcbY=3CT@19W_A[;9OJDO8_SKVNSZ1n^B98iZAFS[\Zmg8?_
X-Complaints-To: usenet-abuse@arcor.de
Xref: g2news2.google.com comp.lang.ada:4406
Date: 2009-01-19T12:25:42+01:00
List-Id: <comp.lang.ada>

george.priv@gmail.com schrieb:
> On Jan 18, 4:42 pm, Maciej Sobczak <see.my.homep...@gmail.com> wrote:
>> In a nearby thread I have asked about the performance of
>> Update_Element in Vector.

>> Each of the tests increments all vectors elements, using several
>> element access methods.

>> Just to stir the discussion a bit, a straightforward implementation of
>> the same test in C++ (with the same base compiler and a single -O2
>> option) runs in 13s with indexed access and in 9.5s with iterators -
>> both are based on the use of references, which do not exist in Ada.
>> Not very easy to neglect this difference.
>>
>> All comments are welcome.

> Very interesting.
> 
> It seems that compiler is unable to inline.  I've added extra test to
> see what will be performance with raw array:

Starting from your test I have added a fake loop.
It is made to simulate the indirection typical of
Vector access, calling Increment via 'Access .
Indeed, inlining does not seem to happen.
Is this a GNAT feature?

gnatmake -O1 -W -g -gnato test_p.adb
raw array                : 00:00:57.47
raw array process        : 00:02:13.70

gnatmake -s -O2 -gnatn -gnatN -W -gnatp test_p.adb
raw array                : 00:00:29.42
raw array process        : 00:01:04.18


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

procedure Test_P is
   Iterations: constant := 10_000;

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

   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;

   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);
end Test_P;