From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <gava@univ-paris12.fr>
X-Spam-Checker-Version: SpamAssassin 3.1.3 (2006-06-01) on yquem.inria.fr
X-Spam-Level: *
X-Spam-Status: No, score=1.0 required=5.0 tests=AWL,SPF_FAIL 
	autolearn=disabled version=3.1.3
X-Original-To: caml-list@yquem.inria.fr
Delivered-To: caml-list@yquem.inria.fr
Received: from discorde.inria.fr (discorde.inria.fr [192.93.2.38])
	by yquem.inria.fr (Postfix) with ESMTP id D290EBC69
	for <caml-list@yquem.inria.fr>; Thu,  8 Feb 2007 10:56:50 +0100 (CET)
Received: from smtp19.orange.fr (smtp19.orange.fr [80.12.242.17])
	by discorde.inria.fr (8.13.6/8.13.6) with ESMTP id l189uoAD026713
	for <caml-list@yquem.inria.fr>; Thu, 8 Feb 2007 10:56:50 +0100
Received: from me-wanadoo.net (localhost [127.0.0.1])
	by mwinf1909.orange.fr (SMTP Server) with ESMTP id 5792E1C00046;
	Thu,  8 Feb 2007 10:56:50 +0100 (CET)
Received: from [192.168.0.9] (ARouen-205-1-4-25.w80-11.abo.wanadoo.fr [80.11.106.25])
	by mwinf1909.orange.fr (SMTP Server) with ESMTP id BCB801C000AE;
	Thu,  8 Feb 2007 10:56:49 +0100 (CET)
X-ME-UUID: 20070208095649773.BCB801C000AE@mwinf1909.orange.fr
Message-ID: <45CAF3E2.7020807@univ-paris12.fr>
Date: Thu, 08 Feb 2007 10:56:50 +0100
From: =?ISO-8859-1?Q?Fr=E9d=E9ric_Gava?= <gava@univ-paris12.fr>
Reply-To: gava@univ-paris12.fr
Organization: =?ISO-8859-1?Q?Universit=E9_de_Paris_12=2C_LACL?=
User-Agent: Thunderbird 1.5.0.9 (X11/20070104)
MIME-Version: 1.0
To: Jacques Garrigue <garrigue@math.nagoya-u.ac.jp>
Cc: caml-list@yquem.inria.fr
Subject: Re: [Caml-list] Multiplication of matrix in C and OCaml
References: <45CA63F5.6020301@univ-paris12.fr> <20070208.111401.55511744.garrigue@math.nagoya-u.ac.jp>
In-Reply-To: <20070208.111401.55511744.garrigue@math.nagoya-u.ac.jp>
Content-Type: multipart/mixed;
 boundary="------------000707080809030905000102"
X-Miltered: at discorde with ID 45CAF3E2.000 by Joe's j-chkmail (http://j-chkmail . ensmp . fr)!
X-Spam: no; 0.00; gava:01 gava:01 ocaml:01 ocaml:01 memcpy:01 matrices:01 matrices:01 iteration:01 ocamlopt:01 -unsafe:01 mult:01 argv:01 argv:01 maxi:01 iter:01 
X-Attachments: type="text/x-csrc" name="mult.c" name="mult.c" name="mult.ml" name="mult.ml" 

This is a multi-part message in MIME format.
--------------000707080809030905000102
Content-Type: text/plain; charset=iso-8859-1; format=flowed
Content-Transfer-Encoding: 8bit

I attached the pure sequential version of my program.

In the file, I give how I compile, How I run and the result. Here C is 5 
time faster than OCaml

Many thanks to tell me why my OCaml code is so slow.

Regards,
Frédéric Gava

ps: note that in a more generic version of the C code (that is, do 
memcpy because the data could be of any size,  (for example, matrices of 
polynomials) the ocaml code is faster and faster than the C one :-) )

--------------000707080809030905000102
Content-Type: text/x-csrc;
 name="mult.c"
Content-Transfer-Encoding: 7bit
Content-Disposition: inline;
 filename="mult.c"

/* param 1 = size of the matrices at the beginning
   param 2 = incrementation
   param 3 = size of the matrices at the end
   param 4 = number of iteration 

gcc -O2 -o cmult mult.c

time ./cmult 600 2 602 1

real    0m3.800s
user    0m3.764s
sys     0m0.020s

*/

#include <stdlib.h> 
#include <stdio.h> 
#include <math.h>

typedef struct Complexe {double re; double img;} complexe;

complexe zero;

/* version generic */

__inline__ void* add(void *x, void *y)
{
complexe a=((complexe*) x)[0];
complexe b=((complexe*) y)[0];
return (void *) &((complexe){a.re+b.re, a.img+b.img});
}

__inline__ void* mult(void *x, void *y)
{
complexe a=((complexe*) x)[0];
complexe b=((complexe*) y)[0];
return (void *) &((complexe){a.re*b.re-a.img*b.img, a.re*b.img+a.img*b.re});
}

complexe random_complexe(void)
{
 complexe c = {c.re=(double)(rand()%1000), c.img=(double)(rand()%1000)};
 return c;
}


void multiply_generic(int n, void *zero, void* (*add)(void *a, void *b), void* (*mult)(void *a, void *b), void *a, void *b, void *c)
{
 register int i,j,k;
 void *tmp;
 void *cc;
 
  for (i=0; i<n; i++) 
   {
    for (j=0;j<n;j++)
     {
      tmp=zero;
      for (k=0;k<n;k++)
       {
        tmp=add(tmp,(mult(b+(k*n+j),a+(i*n+k))));
       }
      cc=c+(i*n+j);
      cc=tmp;
     }
  }
}


int main(int argc, char* argv[])
{
 complexe *a, *b, *c;
 int i,j,k,t1,pas,maxi,iter;

 zero.re=(double)0;
 zero.img=(double)0;

 i=atoi(argv[1]);
 pas=atoi(argv[2]);
 maxi=atoi(argv[3]);
 iter=atoi(argv[4]);

 srand(1000);
 while (i<maxi)
  {  
   /* initialisation des matrices */
   a=(complexe *)malloc(i*i*sizeof(complexe));
   b=(complexe *)malloc(i*i*sizeof(complexe));
   c=(complexe *)malloc(i*i*sizeof(complexe));
      
   for (j=0;j<i;j++)
    for (k=0,t1=j*i;k<i;k++)
     {
      a[t1+k]=random_complexe();
      b[t1+k]=random_complexe();
     }

  for (k=1;k<=iter;k++) multiply_generic(i,&zero,add,mult,a,b,c);

  /* suppression des vieilles matrices */
  free((complexe *)a);
  free((complexe *)b);
  free((complexe *)c);

  i=i+pas;
  }
 return 0;
}

--------------000707080809030905000102
Content-Type: text/plain;
 name="mult.ml"
Content-Transfer-Encoding: 7bit
Content-Disposition: inline;
 filename="mult.ml"

(* param 1 = size of the matrices at the beginning
   param 2 = incrementation
   param 3 = size of the matrices at the end
   param 4 = number of iteration 

(Objective Caml version 3.09.1)

ocamlopt -unsafe -o ocamlmult mult.ml

time ./ocamlmult 600 2 602 1

real    0m17.043s
user    0m16.937s
sys     0m0.052s

*)


let multiplication n zero add mul a b c =
  let tmp=ref zero in
  for i=0 to n-1 do
   for j=0 to n-1 do
    tmp:=zero;
    (* let rec loop tmp k =
     if k=n then tmp else loop (add tmp (mul b.(k*n+j) a.(i*n+k))) (k+1)
    in 
     c.(i*n+j)<-loop zero 0 *)
    for k=0 to n-1 do
     tmp:= add (!tmp) (mul b.(k*n+j) a.(i*n+k)) 
    done; 
    c.(i*n+j)<-(!tmp)    
   done
  done

let random_complex n = {Complex.re=Random.float n; im=Random.float n}

let matrixRandom n = Array.init (n*n) (fun _ -> random_complex 1000.)

let _ = 
	Random.self_init();
        let i = ref (int_of_string (Sys.argv).(1))
	and pas=(int_of_string (Sys.argv).(2))
	and maxi=(int_of_string (Sys.argv).(3)) 
	and iter=(int_of_string (Sys.argv).(4)) in
	while !i<maxi do
         let a=matrixRandom (!i)
         and b=matrixRandom (!i)
         and c=Array.make ((!i)*(!i)) Complex.zero in
	  for j=1 to iter do
           ignore(multiplication (!i) Complex.zero Complex.add 
		                                      Complex.mul a b c);
          done;
	 i:=!i+pas;
	done

--------------000707080809030905000102--