Re: [PATCH] Remove some divide instructions

[Geert Uytterhoeven]

On Thu, 28 Oct 2004, Linus Torvalds wrote:
> On Thu, 28 Oct 2004, Zachary Amsden wrote:
> 	#define __constant_log2(y) ((y==1) ? 0 : \
> 				    (y==2) ? 1 : \
> 				    (y==4) ? 2 : \
> 				    (y==8) ? 3 : -1)	/* We could go on ... */
> 
> 	#define do_div(x,y) ({					\
> 		unsigned long __mod;				\
> 		int __log2;					\
> 		if (__builtin_constant_p(y) && 			\
> 		    !((y) & ((y)-1)) &&				\
> 		    (__log2 = __constant_log2((y))) >= 0) {	\
> 			mod = x & ((y)-1);			\
> 			(x) >>= __log2;				\
> 		} else {					\
> 			.. inline asm case ..			\
> 		}						\
> 		__mod; })
> 
> which looks like it should work, but it's getting so ugly that I suspect I 
> should be committed for even thinking about it.
> 
> (And no, I didn't test the above. It is all trivially optimizable by a 
> compiler, and I can't see how gcc could _fail_ to get it right, but hey, I 
> thought the previous thing would work too, so I'm clearly not competent to 
> make that judgement... ;)

Perhaps this is why you see in many header files code like this:

    static inline xxx_const_case(...) { ... }
    static inline xxx_nonconst_case(...) { ... }

#define xxx(...) __builtin_constant_p(...) ? xxx_const_case(...) \
					   : xxx_nonconst_case(...)

i.e. gcc is better in optimizing away calls to non-called functions?

Gr{oetje,eeting}s,

						Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
							    -- Linus Torvalds
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/