Re: [PATCH 2/5] bitops: compile time optimization for hweight_long(CONSTANT)

[Wu Fengguang <fengguang.wu@intel.com>]

On Mon, Feb 01, 2010 at 01:48:25PM -0700, Andrew Morton wrote:
> > -static inline unsigned long hweight_long(unsigned long w)
> > -{
> > -	return sizeof(w) == 4 ? hweight32(w) : hweight64(w);
> > -}
> > +#define hweight_long(x)				\
> > +(						\
> > +	__builtin_constant_p(x) ?		\
> > +		__builtin_popcountl(x) :	\
> > +		(sizeof(x) <= 4 ?		\
> > +			 hweight32(x) :		\
> > +			 hweight64(x))		\
> > +)
> >  
> >  /**
> >   * rol32 - rotate a 32-bit value left
> 
> Peter's been mucking with a compile-time HWEIGHT().  An outdated
> version of that is presently in linux-next.

Ah I find it. Thanks.

> (I wonder if it could have used __builtin_popcount)
> 
> (I wonder which gcc versions support __builtin_popcount)

This is how Jamie Lokier first recommended it to me:

        Checked GCC 3.4.3 / 4.1 / 4.4: It expands as a compile-time
        constant if the argument is a compile-time constant, so can be
        used in BUILD_BUG_ON() and even for array sizes etc.

Is there an official lowest GCC version that Linux supports?

> Anyway, I suspect you should be using that rather than tweaking
> hweight_long().

OK.

But.. if ever the GCC builtins can be used:

        __builtin_popcount(unsigned int)
        __builtin_popcountl(unsigned long)
        __builtin_popcountll(unsigned long long)

it would be more natural to use hweight_long() or even introduce
hweight_longlong() for (transparent) compile time computation.

Thanks,
Fengguang