DaniWeb Forum Index > Software Development > C

C RSS

Bit shifting and masking.

•

Join Date: Mar 2009

Posts: 11

Reputation: boydale1 is an unknown quantity at this point

Solved Threads: 0

boydale1

Offline

Newbie Poster

Bit shifting and masking.

Oct 19th, 2009

I was given homework to write a program that finds if a double (64 bit) is negative. The double uses twos complement, so the first bit will tell if it is positive (0) or negative (1), but you all probably know that.

The problem I am having is that there can be no floating point operations, just bit masking and shifting. The function can be written with no if statements. It should return 1 if it is negative and 0 if it is not. I thought about masking just the sign bit. I also thought about shifting the sign bit all the way through. Is there any way to just access the first bit and use that in the return statement.

Thanks for any help!

•

Join Date: May 2006

Posts: 3,114

Reputation: WaltP has much to be proud of

Solved Threads: 281

WaltP

Offline

Posting Sensei

Oct 20th, 2009

0x8000000000000000 is the 'sign bit' of a 64 bit value. AND it with the value you want to check. If TRUE, the sign bit is set.

The 3 Laws of the Procrastination Society:
1) Never do today that which can be put off until tomorrow
2) Tomorrow never comes

•

Join Date: Sep 2009

Posts: 76

Reputation: Gaiety is an unknown quantity at this point

Solved Threads: 2

Gaiety

Offline

Junior Poster in Training

-1

Oct 20th, 2009

decalre a union with one double value and char array of 8 bytes.
in Little Endian format , always the sign will be MSB.
so take 7th array, and perform AND operation with 1(char) left shifted 7 times.
and return the resultant value.

NOTE: the returned value is not 1 here but is 128 in case of -Ve.
not sure its efficient.

AND ing with 0X8000000000000000 is good idea but bit wise operations on float value is not valid.

Please correct me if i am wrong.
i'm too eager to know, other efficient methods.

below code is just an example .

 Help with Code Tags 
 C Syntax (Toggle Plain Text) 
int getsign( );
union sign {
        double d1;
        char d[8];
  }u;
 
int main()
{
 u.d1 = -16.32;
 
 if(getsign())
        printf("\n %lf is -Ve\n",u.d1);
 else
        printf("\n %lf is +Ve\n",u.d1);
 
 u.d1 = 16.32;
 
 if(getsign())
        printf("\n %lf is -Ve\n",u.d1);
 else
        printf("\n %lf is +Ve\n",u.d1);
 
 return 0;
}
 
int getsign(double d1)
{
  char m=1;
  return m<<7 & u.d[7];
}
int getsign( );
union sign {
        double d1;
        char d[8];
  }u;

int main()
{
 u.d1 = -16.32;

 if(getsign())
        printf("\n %lf is -Ve\n",u.d1);
 else
        printf("\n %lf is +Ve\n",u.d1);

 u.d1 = 16.32;

 if(getsign())
        printf("\n %lf is -Ve\n",u.d1);
 else
        printf("\n %lf is +Ve\n",u.d1);

 return 0;
}

int getsign(double d1)
{
  char m=1;
  return m<<7 & u.d[7];
}

Last edited by Gaiety; Oct 20th, 2009 at 1:20 am. Reason: return 0;

Minds are like parachutes - they only work when they are open
Gaiety

•

Join Date: May 2006

Posts: 3,114

Reputation: WaltP has much to be proud of

Solved Threads: 281

WaltP

Offline

Posting Sensei

Oct 20th, 2009

As requested, Gaiety:

•

Originally Posted by Gaiety

decalre a union with one double value and char array of 8 bytes.
in Little Endian format , always the sign will be MSB.
so take 7th array, and perform AND operation with 1(char) left shifted 7 times.
and return the resultant value.

NOTE: the returned value is not 1 here but is 128 in case of -Ve.
not sure its efficient.

Way too complex for a new programmer...

•

Originally Posted by Gaiety

AND ing with 0X8000000000000000 is good idea but bit wise operations on float value is not valid.

Not true. What is true is ANDing with a float's maintissa or exponent and getting a useful result is difficult to impossible, but the sign bit is a defined bit in a defined location so it's perfectly valid. If I'm wrong, please point (link) me to the rules & explanation.

•

Originally Posted by Gaiety

Please correct me if i am wrong.
i'm too eager to know, other efficient methods.

Done...

The 3 Laws of the Procrastination Society:
1) Never do today that which can be put off until tomorrow
2) Tomorrow never comes

•

Join Date: Sep 2009

Posts: 76

Reputation: Gaiety is an unknown quantity at this point

Solved Threads: 2

Gaiety

Offline

Junior Poster in Training

Oct 20th, 2009

•

Originally Posted by WaltP

As requested, Gaiety:

Way too complex for a new programmer...

Not true. What is true is ANDing with a float's maintissa or exponent and getting a useful result is difficult to impossible, but the sign bit is a defined bit in a defined location so it's perfectly valid. If I'm wrong, please point (link) me to the rules & explanation.

Done...

could you please provide me a small code snippet on how that can be done?
do we need to use pointer stuff..!

 Help with Code Tags 
 C Syntax (Toggle Plain Text) 
 
int main()
{
        double nums[] ={16.32,-16.32};
        int i=0;
        while(i<2)
        {
        if(getsign(&nums[i]))
                printf("%lf is -ve\n",nums[i]);
        else
                printf("%lf is +ve\n",nums[i]);
        i++;
        }
 
return 0;
}
 
int getsign(double *d1)
{
        char *p;
        p=(char *)d1;
        p=p+7;
        return *p & 1<<7;
}
int main()
{
        double nums[] ={16.32,-16.32};
        int i=0;
        while(i<2)
        {
        if(getsign(&nums[i]))
                printf("%lf is -ve\n",nums[i]);
        else
                printf("%lf is +ve\n",nums[i]);
        i++;
        }

return 0;
}

int getsign(double *d1)
{
        char *p;
        p=(char *)d1;
        p=p+7;
        return *p & 1<<7;
}

Last edited by Gaiety; Oct 20th, 2009 at 6:16 am. Reason: code added

Minds are like parachutes - they only work when they are open
Gaiety

•

Join Date: Jul 2009

Posts: 201

Reputation: yellowSnow is a splendid one to behold

Solved Threads: 35

yellowSnow yellowSnow is offline

Offline

Posting Whiz in Training

Oct 20th, 2009

Mmm ... interesting. I always thought that bitwise operators were only valid with integral operands. The only way I can think of using bitwise operators with floats/doubles is to use casting pointers. Perhaps something like:

 Help with Code Tags 
 C Syntax (Toggle Plain Text) 
int isNegative(float f) {
    int *ip = (int*)(&f);
    return (*ip & 0x80000000 ? 1 : 0);
}
int isNegative(float f) {
    int *ip = (int*)(&f);
    return (*ip & 0x80000000 ? 1 : 0);
}

But then, this may also be too complex for a new programmer.

Last edited by yellowSnow; Oct 20th, 2009 at 10:11 am.

Manic twiddler of bits

•

Join Date: Mar 2009

Posts: 11

Reputation: boydale1 is an unknown quantity at this point

Solved Threads: 0

boydale1

Offline

Newbie Poster

Oct 20th, 2009

Can you explain this last line. I understand the bit wise 'and' to isolate the bit, but I am lost on the last line. (return statement)

return (*ip & 0x80000000 ? 1 : 0);

Thanks for your help

•

Join Date: Jul 2009

Posts: 201

Reputation: yellowSnow is a splendid one to behold

Solved Threads: 35

yellowSnow yellowSnow is offline

Offline

Posting Whiz in Training

Oct 21st, 2009

•

Originally Posted by boydale1

Can you explain this last line. I understand the bit wise 'and' to isolate the bit, but I am lost on the last line. (return statement)

return (*ip & 0x80000000 ? 1 : 0);

Thanks for your help

The example I posted demonstrates the use of a ternary operator. Here's a brief summary on how the ternary operator works:

http://cprogramminglanguage.net/c-ternary-operator.aspx

Essentially it is a short form equivalent of the if-else construct. I posted it this way because your original post stated that you weren't permitted to use any 'if' type statements.

Have a read and post back if you still have questions.

Last edited by yellowSnow; Oct 21st, 2009 at 6:01 am.

Manic twiddler of bits