CODEWITHSUNDEEP
cbit-manipulation
Jason
Jason

Reputation: 23

Extracting bits using bit manipulation

I have a 32-bit unsigned int and I need to extract bits at given positions and make a new number out of those bits. For example, if I have a 0xFFFFFFFF and want bits 0,10,11 my result will be 7 (111b).

This is my attempt, it extracts the bits correctly but doesn't create the correct result. I'm shifting the result 1 place left and ANDing it with my extracted bit, apparenlty this is incorrect though?

I'm also sure there is probably a much more elegant way to do this?

#define TEST 0xFFFFFFFF

unsigned int extractBits(unsigned short positions[], unsigned short count, unsigned int bytes)
{
    unsigned int result = 0;
    unsigned int bitmask = 0;
    unsigned short bit = 0;
    int i = 0;

    for(i = 0; i < count; i++) {
        bitmask = 2 << (positions[i] -1);
        if(bitmask == 0) bitmask = 1;

        bit = bytes & bitmask;
        bit = bit >> positions[i];

        result = result << 1;
        result = result & bit;  
    }

    if(result != 31) {
        printf("FAIL");
    }

    return result;
}

int main(void)
{
    unsigned short positions[5] = {8, 6, 4, 2, 0};
    unsigned int result = extractBits(positions, 5, TEST);

    printf("Result: %d\n", result);

    return 0;
}

Upvotes: 2

Views: 1965

Answers (2)

James Waldby - jwpat7
James Waldby - jwpat7

Reputation: 8701

Since you are picking off individual bits, there's no reason to make the bit mask a variable; just shift the desired bit into the units bit, and use a mask of 1. E.g.:

...
result = (2*result) | ((bytes >> positions[i]) & 1);
...

Many compilers generate the same code for 2*result and result<<1, so use whichever you like.

Note, if you are designing the interface and don't have good reasons for using short integers for positions[] and count as you do, then don't. Be consistent and specify all the integers the same way.

Upvotes: 1

Doc Brown
Doc Brown

Reputation: 20044

Beware, untested code:

for(i = 0; i < count; i++) 
{
    bitmask = 1 << positions[i];
    bit = (bytes & bitmask)!=0;
    result = (result << 1)|bit;
}

Upvotes: 2

Related Questions