thread program errors help

[anything2]

can some one tell me where im going wrong
thanks

#include <pthread.h>  
#include <stdio.h>
#include <time.h>
#include <stdlib.h>
#include <string.h>
#include <math.h>
#include <unistd.h>
#define BUFFER_SIZE 100;
#define SIGMA 100.8
#define MU 6.6

double randno();
double normal();
double exponential();
void *producer();
void *consumer();

time_t T;


//buffer and related
double buffer[BUFFER_SIZE];
int in=0;
int out=0;
double mean;
int i=0;

//random no b/w 0~1
double randno(){
return (rand()/(double)RAND_MAX);
}

//normal distribution
double normal(){
  double x1, x2, w;                                          
  x1 = randno();
  x2 = randno();
  w = sqrt(x1*x1+x2*x2);
  return (SIGMA * x1 * sqrt( abs(2 * log(w))/w ) + MU);
}

//exponential distribution
double exponential(){
  double x;                                                                                                 
  x = randno();
  return ( -log(1-x)/MU );
}

//producer function
void *producer(){

  while(1){
    while((in+1)% BUFFER_SIZE == out)
    //do nothing as buffer is full
    {;}    

    buffer[in++] = normal();

    in %= BUFFER_SIZE;

    usleep(exponential());
  }
  pthread_exit(0);
}

//consumer function
void *consumer(){
 
  while(1){
    double sum=0;

      while( in == out ){//isempty
    usleep(exponential());
      }
   
   for(i=0; i<10; i++){
      sum+=buffer[out];
      
      out++;
      out %= BUFFER_SIZE;

    }
    mean = sum/10;
    
    printf("%d\t",in);
    T = time(NULL);
    printf(ctime(&T));
    printf("\n");
  }
  pthread_exit(0);
}

int main(){
 
  pthread_t prod,cons;
 
  pthread_create(&prod, NULL, producer, NULL);
  pthread_create(&cons, NULL, consumer, NULL);

  pthread_join(prod,NULL);
  pthread_join(cons,NULL);
 
  return 0;
}

[nadroj]

sure, you didnt include your error messages

[Salem]

> #define BUFFER_SIZE 100;
This will stop it compiling, remove the ;

You need to fix all that shared access to global data. There is no control over when one thread stops and the other takes over.

> for(i=0; i<10; i++)
Who said producer managed to queue 10 items since last time?

[anything2]

>> #define BUFFER_SIZE 100;
>This will stop it compiling, remove the ;

i removed the ;
and so it compiles

>> for(i=0; i<10; i++)
>Who said producer managed to queue 10 items since last time?

i put the while loop inside the for loop so the problem with atleast 10 in the buffer is resolved
thats where it shd have been in the first place.

about the global variables
they just share the buffer and its pointers(in/out)
there could be some conflicts but it should still give me some output

the program compiles but the output is just a lot of gibberish
even if i dont printf anything the same output comes
im using sun solaris 10

plus it gives one warning : no newline at end of file

thanks alot

[anything2]

>> #define BUFFER_SIZE 100;
>This will stop it compiling, remove the ;

i removed the ;
and so it compiles

>> for(i=0; i<10; i++)
>Who said producer managed to queue 10 items since last time?

i put the while loop inside the for loop so the problem with atleast 10 in the buffer is resolved
thats where it shd have been in the first place.

about the global variables
they just share the buffer and its pointers(in/out)
there could be some conflicts but it should still give me some output

the program compiles but the output is just a lot of gibberish
even if i dont printf anything the same output comes
im using sun solaris 10

plus it gives one warning : no newline at end of file

thanks alot

[matsp]

Be aware that the compiler may store global variables in registers, so if you have multiple threads using the same global variable(s), then you may find that they don't see the updates the other thread does (this is more likely to happen on a Sparc or x86-64 processor than an x86-32 processor, due to higher number of available registers - but there's no guarantee that it doesn't happen on ANY machine).

The solution is to use "volatile" for variables like in and out. This forces the compiler to read the variable from memory EVERY time it's used, and write back any result to the memory, without "holding it".

I'm not guaranteeing that this is the right and the whole answer, but I strongly suspect it's part of the problem.

--
Mats

[Salem]

> there could be some conflicts but it should still give me some output
M'kay...

> the program compiles but the output is just a lot of gibberish
OK, you've got some output, and some conflicts - what else do you want?


First off, I would suggest you use only a single global variable.
- the producer tries to increment it up to some limit value, and prints each change it makes
- the consumer tries to decrement it down to zero, and prints each change it makes.

Thread programming is about 1000&#37; harder than regular single threaded programs. If you create a race condition, then you can be sure this will show up when you're least able to investigate it.

[Codeplug]

Under POSIX, you don't use volatile in multi-threaded programming. Here are the only gaurentees your have with respect to memory synchronization under POSIX: http://www.opengroup.org/onlinepubs/...html#tag_04_10

The MS compiler (since version 14) is the only compiler that gives multi-threading semantics to the volatile keyword. See Multiprocessor Considerations for Kernel-Mode Drivers (warning: word document)

gg