[gecode-users] another question about stats

[Christian Schulte]

Sorry, I currently have no access to linux but tried it under windows (by
using the windows variant of sleep). It works perfectly for me, even for
2.1.1. Then I googled usleep and suggest that you read the usleep manpage as
well. That might be the problem (the number might be too large).

 

Christian

 

--

Christian Schulte, www.it.kth.se/~cschulte/

 

From: users-bounces at gecode.org [mailto:users-bounces at gecode.org] On Behalf
Of David Przybilla
Sent: Wednesday, February 25, 2009 3:06 PM
To: users at gecode.org
Subject: Re: [gecode-users] another question about stats

 

Sorry I just clicked "Answer" Button and didnt notice the email was sent to
your personal email.
I know it is not the latest,however I need to work on 2.1.1..sorry about the
downgrade :) thank you for the fast answer..
here is the whole code, it is just one file:

#include "gecode/minimodel.hh"
#include "gecode/examples/support.hh"
 
using namespace Gecode;

class MiPropagador: public
Gecode::UnaryPropagator<Gecode::Int::IntView,Gecode::Int::PC_INT_DOM>
{

public:

     Actor*    copy(Space* home,bool share){
       return new (home) MiPropagador(home,share,*this);
     }
    
      PropCost  cost(void) const {
        return PC_UNARY_LO;
      }


     MiPropagador(Space* home, bool share, MiPropagador&
p):Gecode::UnaryPropagator<Gecode::Int::IntView,Gecode::Int::PC_INT_DOM>(hom
e,share,p){}
     
     MiPropagador(Space* home, Gecode::Int::IntView
x):Gecode::UnaryPropagator<Gecode::Int::IntView,Gecode::Int::PC_INT_DOM>(hom
e,x){}
     
     Gecode::ExecStatus propagate(Space *home, Gecode::ModEventDelta med){
         usleep(1000000); // works on linux, it is supposed to delay
execution for a second
         std::cout<<"The propagator has been called!"<<std::endl;
         GECODE_ME_CHECK(x0.gq(home,0));
         return Gecode::ES_NOFIX; 
     }
      
      
      
};


Gecode::ExecStatus positiva(Space* home,Gecode::Int::IntView x)
  {
       (void) new (home) MiPropagador(home,x);
       return Gecode::ES_OK;
  };

class EjemploPropagador: public Example {
        private: 
             IntVarArray a;
        public:
                
            EjemploPropagador(const Options& opt): a(this,2,-10,10){ 
                       
                       post(this,a[1]==a[0]+2);
                       positiva(this,a[0]);
                       positiva(this,a[1]);
 
Gecode::branch(this,a,Gecode::INT_VAR_NONE,Gecode::INT_VAL_MIN );
                                   
                } 
           
            EjemploPropagador(bool share, EjemploPropagador& s) :
Example(share,s){
                        a.update(this, share, s.a);
                }
              
                virtual Space*
                        copy(bool share) {
                                return new EjemploPropagador(share,*this);
                        }
                
                virtual void
                        print(std::ostream &os) {
                                std::cout << "\a = " << a << std::endl;
                        }
};



int  main(int argc, char** argv) {

  SizeOptions opt("EjemploPropagador");
  opt.icl( ICL_DOM);
  opt.solutions ( 0);
  opt.parse(argc,argv);
  Example::run<EjemploPropagador,DFS>(opt);

  return 0;
}


The result I get is:

Initial
        propagators:   3
        branchings:    1

Summary
        runtime:       0.000 (0.000000 ms)
        solutions:     9
        propagations:  54
        failures:      0
        clones:        8
        commits:       16
        peak memory:   10 KB


however it takes more than 0.0ms to complete execution.

Regards, David



2009/2/24 Christian Schulte <cschulte at kth.se>

Actually, that is weird as in the examples also setup time is measured. But
I fear you have to chip in some more detail what you do.

 

The runtime between 1* and 2* should be roughly the same (why do you compare
one old version against another old version, 2.1.1 is not the latest). But
the statistics can be trusted, we use it all the time and we know that it
does not deviate from the OS measures.

 

Christian

 

--

Christian Schulte, www.it.kth.se/~cschulte/
<http://www.it.kth.se/%7Ecschulte/> 

 

From: users-bounces at gecode.org [mailto:users-bounces at gecode.org] On Behalf
Of David Przybilla
Sent: Tuesday, February 24, 2009 5:24 PM
To: users at gecode.org
Subject: [gecode-users] another question about stats

 

I made a small testing sample, I made a simple propagator and some variables
subscribing to it.
Inside the propagate method there is a usleep line, which is supposed to
delay the execution for some seconds.
When I run the example, the delay is there, however when  stats are shown,
they say the example`s runtime was 0ms.
what does "runtime" measure? am I doing  anything wrong?
Also, Im making performance comparisons between some programs on Gecode old
versions and between the same ones on Gecode 2.1.1,
I wonder whether I should  measure the time on my own or just use the
runtime thrown by gecode`s stats

Thank you in advance!

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.ps.uni-sb.de/pipermail/users/attachments/20090225/4b8a8050/attachment.htm>