Re: [PATCH 4/6] Implement a general log2 facility in the kernel

[David Howells <dhowells@redhat.com>]

Luck, Tony <tony.luck@intel.com> wrote:

> > This facility provides three entry points:
> > 
> > 	log2()		Log base 2 of u32
> > 	ll_log2()	Log base 2 of u64
> > 	long_log2()	Log base 2 of unsigned long
> 
> The names are rather counter-intuitive. "ll" sounds like "long long", so
> why does it opearte on *unsigned* 64-bit?  Ditto for "long_log2()".
> Perhaps they should be log2_u32(), log2_u64(), etc.

Well, given that you can't represent a log of a negative number, I'm not sure
it matters.

Note that long_log2() already exists and is of arch-dependent size.

> Even better if someone can come up with the right pre-processor magic
> using sizeof/typeof so that you could just use "log2(any type)"

I then end up with 32 of these per usage if I pass in a definite 32-bit value:

	warning: comparison is always false due to limited range of data type

because of the:

		n >= (1ULL << 63) ? 63 :	\
		n >= (1ULL << 62) ? 62 :	\
		n >= (1ULL << 61) ? 61 :	\
	...

in ll_log2().

David