noporpoise / bitarray Goto Github PK

Hi,

It seems that there is an issue in the function

static word_t __inline windows_popcount(word_t w)
{
  w = w - ((w >> 1) & (word_t)~(word_t)0/3);
  w = (w & (word_t)~(word_t)0/15_3) + ((w >> 2) & (word_t)~(word_t)0/15_3);
  w = (w + (w >> 4)) & (word_t)~(word_t)0/255*15;
  c = (word_t)(w \* ((word_t)~(word_t)0/255)) >> (sizeof(word_t) - 1) \* 8;
}

line 170 in bit_array.c

Best Regards,
Long Yuan

Add bit_array_find_next_set_bit(), a generalised version of bit_array_find_first_set_bit().

If I add 8 bytes to a bit array:
55 31 5A 12 02 31 C3 1E
And then print the bit array's words[0] with printf( "%016" PRIx64"", val ):
1ec33102125a3155
And then I loop through the bits with:
for ( int idx = 0; idx < 64; idx++ )
{
if ( bit_array_get_bit( ba, idx ) == 0 )
printf( "0" );
else
printf( "1" );
}
then I get (I added spaces to make it easier to see):
10101010 10001100 01011010 01001000 01000000 10001100 11000011 01111000
Which is wrong for a big-endian processor. It should be:
01010101 00110001 01011010 00010010 00000010 00110001 11000011 00011110
You can see that the order of the bits within each byte is backwards. You're assuming little-endian ordering.

Any chance you can tag a new release? Been 57 commits since 2013.

Need a comparison with std::bitset, so it's more attractive