RE: [Dev-C++] speed test

[Ioannis V. <no...@ho...>]

> -----Original Message-----
> From: dev...@li... 
> [mailto:dev...@li...] On Behalf 
> Of Ioannis Vranos
> Sent: Friday, August 16, 2002 5:06 PM
> To: dev...@li...
> Subject: RE: [Dev-C++] speed test
> 
> 
> > -----Original Message-----
> > From: dev...@li... 
> > [mailto:dev...@li...] On Behalf 
> > Of mo...@ju...
> > Sent: Friday, August 16, 2002 8:02 AM
> > To: dev...@li...
> > Subject: [Dev-C++] speed test
> > 
> > 
> > Running the program below in Dev-C++, Version 4, took between 
> > 209 and 235
> > seconds on my system.  I don't know why the running times 
> > vary, but even
> > taking the fastest time of 209 seconds, it compares 
> unfavorably with a
> > similar program compiled (36 seconds) or interpreted  (55 
> seconds) in
> > QBasic 4 .  
> 
> 
> Did you printed the result after every x++ inside the for loop in QB?
> 
> 
> 
> > 
> > Does Dev-C++ really produce such leisurely code?  Or can the 
> > program be
> > speeded up?
> 
> 
> Of course it can. Printing in console is slow (the time cost of a
> function call, etc).
> 
> 
> 
> >  
> > Allan
> > 
> > // This program prints (1) integers 1 to 1,000,000
> > //                                (2) starting & ending times
> > //======================================================
> > #include <stdio.h>
> > #include <stdlib.h> 
> > #include <time.h>
> > 
> > int main(int argc, char *argv[])
> 
> Theese arguments are not needed.
> 
> 
> > {
> > struct tm *ptr;
> > time_t loc_time;
> > long int x;
> > 
> > loc_time  = time(NULL);
> 
> 
> No accurate results with this method.
> 
> 
> 
> > ptr = localtime(&loc_time);
> > 
> > for (x = 1;x <= 1000000;x++)
> >         printf("%d ",x);
> 
> 
> Here you have 100000 function calls for printing in the console.
> 
> 
> > 
> > printf("\nStarting time is ");
> > printf(asctime(ptr));
> > 
> > loc_time =  time(NULL);
> > ptr = localtime(&loc_time);
> > printf("\nEnding time is ");
> > printf(asctime(ptr), "\n");
> > 
> > system ("PAUSE");
> > 
> > return 0;
> > }
> 
> 
> A better program:
> 
> 
> #include <stdio.h>
> #include <stdlib.h> 
> #include <time.h>
> 
> int main()
> {
>  clock_t t1, t2;
> 
>  long int x;
> 
>  t1=clock();
> 
>  for (x = 1; x <= 1000000;x++)
>      ;
>      
>  printf("%ld\n", x);
> 
>  t2=clock();
> 
>  printf("\n\nProcessor time: %lu", (unsigned long)t2-t1);
>  printf("\nIn seconds: %f\n\n", (double)(t2-t1)/CLOCKS_PER_SEC);
> 
>  system ("PAUSE");
> 
>  return 0;
> }
> 
> 
> 
> The above prints the final value of x only.
> 
> 
> 
> Using MINGW GCC 3.1 with optimizations:
> 
> 
> gcc temp.c -o temp -ansi -pedantic-errors -Wall
> -fexpensive-optimizations -O3 -ffloat-store -mcpu=pentiumpro
> 
> 
> C:\c>temp
> 1000001
> 
> 
> Processor time: 0
> In seconds: 0.000000
> 
> Press any key to continue . . .
> 
> 
> 
> Here obviously the compiler performed an optimisation with 
> which he saw
> the final result of the loop and avoided the loop completely.
> 
> 
> Running the GCC 3.1 without the optimisation switches:
> 
> 
> C:\c>gcc temp.c -o temp -ansi -pedantic-errors -Wall
> 
> C:\c>temp
> 1000001
> 
> 
> Processor time: 10
> In seconds: 0.010000
> 
> Press any key to continue . . .
> 
> 
> 
> Here it performed the loop and it took it 10 msec (something similar
> will happen with Dev-C++ eith optimisations off).
> 
> 
> In summary, your time cost was imposed due to 1000000 
> function calls to
> printf().


Also even if you use printf() inside the for loop, things will go much
faster if you minimize the console (command prompt) window during the
execution. And this because windows will not have to display every
number.


Ioannis

* Ioannis Vranos
* Programming pages: http://www.noicys.cjb.net
* Alternative URL: http://run.to/noicys