[patch 14/44] generic hweight{64,32,16,8}()

[Akinobu Mita]

On Thu, Feb 02, 2006 at 02:26:38AM +0100, Gabriel Paubert wrote:
> 
> The first step can be implemented slightly better:
> 
> unsigned int res = w-((w>>1)&0x55555555);
> 

Yes. I've got many advices about hweight speedup.


static unsigned int hweight32(unsigned int w)
{
        unsigned int res = w - ((w >> 1) & 0x55555555);
        res = (res & 0x33333333) + ((res >> 2) & 0x33333333);
        res = (res + (res >> 4)) & 0x0F0F0F0F;
        res = res + (res >> 8);
        return (res + (res >> 16)) & 0x000000FF;
}

static unsigned int hweight16(unsigned int w)
{
        unsigned int res = w - ((w >> 1) & 0x5555);
        res = (res & 0x3333) + ((res >> 2) & 0x3333);
        res = (res + (res >> 4)) & 0x0F0F;
        return (res + (res >> 8)) & 0x00FF;
}

static unsigned int hweight8(unsigned int w)
{
        unsigned int res = w - ((w >> 1) & 0x55);
        res = (res & 0x33) + ((res >> 2) & 0x33);
        return (res + (res >> 4)) & 0x0F;
}

static unsigned long hweight64(__u64 w)
{
#if BITS_PER_LONG < 64
	return hweight32((unsigned int)(w >> 32)) +
				hweight32((unsigned int)w);
#else
	__u64 res = w - ((w >> 1) & 0x5555555555555555ul);
	res = (res & 0x3333333333333333ul) + ((res >> 2) & 0x3333333333333333ul);
	res = (res + (res >> 4)) & 0x0F0F0F0F0F0F0F0Ful;
	res = res + (res >> 8);
	res = res + (res >> 16);
	return (res + (res >> 32)) & 0x00000000000000FFul;
#endif
}