Author: Dieter Buerssner
Date: 03:06:32 02/20/04
Go up one level in this thread
On February 19, 2004 at 17:41:59, James Robertson wrote:
>... puzzling for me at least. A short time ago the following code compiled
>nicely on the server machines using linux/g++ at my school:
>
>inline int FirstSetBit(unsigned long long a)
>{
> asm ("
> bsf 4(%0), %%eax
> add $32, %%eax
> bsf (%0), %%eax
> ":: "r"(&a): "%eax");
>}
untested (did not even try to compile it):
inline int FirstSetBit(unsigned long long a)
{
int res;
__asm__ __volatile__
("bsfl 4(%1), %0 \n"
"addl $32, %0 \n"
"bsfl (%1), %0 \n"
: "=r&" (res) : "r"(&a): "cc");
return res;
}
Or perhaps:
inline int FirstSetBit(unsigned long long a)
{
int res;
__asm__ __volatile__
("bsfl %1, %0 \n"
"addl $32, %0 \n"
"bsfl %2, %0 \n"
: "=r&" (res)
: "g"((unsigned long)a), "g" ((unsigned long)(a>>32))
: "cc");
return res;
}
Might be faster (no need to calculate the adress of a in general).
You are probably aware, that you depend on undocumented behaviour of bsf (that
it leaves the output unchanged for 0)
Regards,
Dieter
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.