svn commit: r209604 - head/lib/libc/gmon

[Marcel Moolenaar]

On Jun 30, 2010, at 12:17 PM, Bruce Evans wrote:
>> I like this. What about the following (white-space corrupted)
>> simplification:
> 
> Almost OK.  I orginally wrote it with a conditional operator, but decided
> that that might be harder to understand.
> 
>> Index: gmon.c
>> ===================================================================
>> --- gmon.c	(revision 209604)
>> +++ gmon.c	(working copy)
>> @@ -110,24 +110,9 @@
>> 	p->tos[0].link = 0;
>> 
>> 	o = p->highpc - p->lowpc;
>> -	if (p->kcountsize < o) {
>> -#if !defined(__powerpc__)
>> -		s_scale = ((float)p->kcountsize / o ) * SCALE_1_TO_1;
>> -#else /* avoid floating point */
>> -		int quot = o / p->kcountsize;
>> +	s_scale = (p->kcountsize < o) ?
>> +	    ((uintmax_t)p->kcountsize << SCALE_1_TO_1) / o : SCALE_1_TO_1;
> 
> Shifting by 65536 is a bit much :-).  Multiplication by 65536 might
> give the same code as shifting by 16, but I think shifting is clearer.

Doh... :-)

Ok. I'll remove SCALE_1_TO_1 and instead define SCALE_SHIFT.
Substitutions in expressions as expected.

Thanks,

-- 
Marcel Moolenaar
xcllnt at mac.com