Re: [PATCH 1/2] Consolidate condition check for return value in mp asin and acos

[Andreas Jaeger <aj at suse dot com>]

On 10/28/2013 07:47 AM, Siddhesh Poyarekar wrote:
> On Fri, Oct 25, 2013 at 06:48:21PM +0530, Siddhesh Poyarekar wrote:
>> On 25 October 2013 18:16, Allan McRae <allan@archlinux.org> wrote:
>>> You are doing more than add the probe here.  I would have thought that
>>> clean-up would need to be a separate patch, or at least have a ChangeLog
>>> entry.
>>
>> OK, I'll split it up in my next iteration.
> 
> 
> So this is part one of the patch which consolidates the conditionals
> in the return of __sin32 and __cos32 to prepare the code for sstemtap
> probes.
> 
> Siddhesh
> 
> 	* sysdeps/ieee754/dbl-64/sincos32.c (__sin32): Consolidate
> 	conditional check for return value.
> 	(__cos32): Likewise.
> 
> diff --git a/sysdeps/ieee754/dbl-64/sincos32.c b/sysdeps/ieee754/dbl-64/sincos32.c
> index f253b8c..11654d6 100644
> --- a/sysdeps/ieee754/dbl-64/sincos32.c
> +++ b/sysdeps/ieee754/dbl-64/sincos32.c
> @@ -147,10 +147,9 @@ __sin32 (double x, double res, double res1)
>    __dbl_mp (x, &c, p);		/* c = x  */
>    __sub (&b, &c, &a, p);
>    /* if a > 0 return min (res, res1), otherwise return max (res, res1).  */
> -  if (a.d[0] > 0)
> -    return (res < res1) ? res : res1;

This part should give:
a.d[0] > 0 && !(res < res1)

> -  else
> -    return (res > res1) ? res : res1;

And this condition:
!(a.d[0] > 0) && !(res > res1)

> +  if ((a.d[0] > 0 && res > res1) || (a.d[0] <= 0 && res < res1))

In that case it should be res >= res1 for the first paren and "res <=
res1" for the second.

Andreas


> +    res = res1;
> +  return res;
>  }
>  
>  /* Receive double x and two double results of cos(x) and return result which is
> @@ -181,10 +180,9 @@ __cos32 (double x, double res, double res1)
>    __dbl_mp (x, &c, p);		/* c = x                  */
>    __sub (&b, &c, &a, p);
>    /* if a > 0 return max (res, res1), otherwise return min (res, res1).  */
> -  if (a.d[0] > 0)
> -    return (res > res1) ? res : res1;
> -  else
> -    return (res < res1) ? res : res1;
> +  if ((a.d[0] > 0 && res < res1) || (a.d[0] <= 0 && res > res1))
> +    res = res1;
> +  return res;
>  }
>  
>  /* Compute sin() of double-length number (X + DX) as Multi Precision number and
> 


-- 
 Andreas Jaeger aj@{suse.com,opensuse.org} Twitter/Identica: jaegerandi
  SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg, Germany
   GF: Jeff Hawn,Jennifer Guild,Felix Imendörffer,HRB16746 (AG Nürnberg)
    GPG fingerprint = 93A3 365E CE47 B889 DF7F  FED1 389A 563C C272 A126