[v2,1/4] iee754: provide gcc builtins based generic sqrt functions

  Reviewed-by: Adhemerval Zanella  <adhemerval.zanella@linaro.org>
---
 sysdeps/generic/math-use-builtins.h |  3 +++
 sysdeps/ieee754/dbl-64/e_sqrt.c     |  6 ++++++
 sysdeps/ieee754/flt-32/e_sqrtf.c    | 16 ++++++++++------
 3 files changed, 19 insertions(+), 6 deletions(-)
  

On 6/2/20 2:35 AM, Vineet Gupta via Libc-alpha wrote:
> Reviewed-by: Adhemerval Zanella  <adhemerval.zanella@linaro.org>
> ---
>  sysdeps/generic/math-use-builtins.h |  3 +++
>  sysdeps/ieee754/dbl-64/e_sqrt.c     |  6 ++++++
>  sysdeps/ieee754/flt-32/e_sqrtf.c    | 16 ++++++++++------
>  3 files changed, 19 insertions(+), 6 deletions(-)
> 
> diff --git a/sysdeps/generic/math-use-builtins.h b/sysdeps/generic/math-use-builtins.h
> index 8a39ef58bc95..fc724c824a17 100644
> --- a/sysdeps/generic/math-use-builtins.h
> +++ b/sysdeps/generic/math-use-builtins.h
> @@ -60,4 +60,7 @@
>  # define USE_COPYSIGNF128_BUILTIN 0
>  #endif
> 
> +#define USE_SQRT_BUILTIN 0
> +#define USE_SQRTF_BUILTIN 0
> +
>  #endif /* math-use-builtins.h */
Please also update the current architecture specific math-use-builtins.h
file: sysdeps/s390/fpu/math-use-builtins.h
  

On 6/2/20 5:51 AM, Stefan Liebler via Libc-alpha wrote:
> On 6/2/20 2:35 AM, Vineet Gupta via Libc-alpha wrote:
>> Reviewed-by: Adhemerval Zanella  <adhemerval.zanella@linaro.org>
>> ---
>>  sysdeps/generic/math-use-builtins.h |  3 +++
>>  sysdeps/ieee754/dbl-64/e_sqrt.c     |  6 ++++++
>>  sysdeps/ieee754/flt-32/e_sqrtf.c    | 16 ++++++++++------
>>  3 files changed, 19 insertions(+), 6 deletions(-)
>>
>> diff --git a/sysdeps/generic/math-use-builtins.h b/sysdeps/generic/math-use-builtins.h
>> index 8a39ef58bc95..fc724c824a17 100644
>> --- a/sysdeps/generic/math-use-builtins.h
>> +++ b/sysdeps/generic/math-use-builtins.h
>> @@ -60,4 +60,7 @@
>>  # define USE_COPYSIGNF128_BUILTIN 0
>>  #endif
>>
>> +#define USE_SQRT_BUILTIN 0
>> +#define USE_SQRTF_BUILTIN 0
>> +
>>  #endif /* math-use-builtins.h */
> Please also update the current architecture specific math-use-builtins.h
> file: sysdeps/s390/fpu/math-use-builtins.h

Fixed. I should have added s390 to my list of arches to test.

Thx for taking a look.

-Vineet
  

On 01/06/2020 21:35, Vineet Gupta wrote:
> Reviewed-by: Adhemerval Zanella  <adhemerval.zanella@linaro.org>

LGTM with the small nit below and the s390 fix pointed by Stefan.

Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>

> ---
>  sysdeps/generic/math-use-builtins.h |  3 +++
>  sysdeps/ieee754/dbl-64/e_sqrt.c     |  6 ++++++
>  sysdeps/ieee754/flt-32/e_sqrtf.c    | 16 ++++++++++------
>  3 files changed, 19 insertions(+), 6 deletions(-)
> 
> diff --git a/sysdeps/generic/math-use-builtins.h b/sysdeps/generic/math-use-builtins.h
> index 8a39ef58bc95..fc724c824a17 100644
> --- a/sysdeps/generic/math-use-builtins.h
> +++ b/sysdeps/generic/math-use-builtins.h
> @@ -60,4 +60,7 @@
>  # define USE_COPYSIGNF128_BUILTIN 0
>  #endif
>  
> +#define USE_SQRT_BUILTIN 0
> +#define USE_SQRTF_BUILTIN 0
> +
>  #endif /* math-use-builtins.h */

Ok.

> diff --git a/sysdeps/ieee754/dbl-64/e_sqrt.c b/sysdeps/ieee754/dbl-64/e_sqrt.c
> index d42a1a4eb6e9..518a8ae5cdaf 100644
> --- a/sysdeps/ieee754/dbl-64/e_sqrt.c
> +++ b/sysdeps/ieee754/dbl-64/e_sqrt.c
> @@ -41,6 +41,7 @@
>  #include <math_private.h>
>  #include <fenv_private.h>
>  #include <libm-alias-finite.h>
> +#include <math-use-builtins.h>
>  
>  /*********************************************************************/
>  /* An ultimate sqrt routine. Given an IEEE double machine number x   */
> @@ -50,6 +51,10 @@
>  double
>  __ieee754_sqrt (double x)
>  {
> +#if USE_SQRT_BUILTIN
> +  return __builtin_sqrt (x);
> +#else
> +  /* Use generic implementation.  */
>    static const double
>      rt0 = 9.99999999859990725855365213134618E-01,
>      rt1 = 4.99999999495955425917856814202739E-01,
> @@ -138,6 +143,7 @@ __ieee754_sqrt (double x)
>  	return (x - x) / (x - x); /* sqrt(-ve)=sNaN */
>        return 0x1p-256 * __ieee754_sqrt (x * 0x1p512);
>      }
> +#endif /* ! USE_SQRT_BUILTIN  */
>  }
>  #ifndef __ieee754_sqrt
>  libm_alias_finite (__ieee754_sqrt, __sqrt)

Ok.

> diff --git a/sysdeps/ieee754/flt-32/e_sqrtf.c b/sysdeps/ieee754/flt-32/e_sqrtf.c
> index b339444301aa..d85a04162983 100644
> --- a/sysdeps/ieee754/flt-32/e_sqrtf.c
> +++ b/sysdeps/ieee754/flt-32/e_sqrtf.c
> @@ -16,12 +16,15 @@
>  #include <math.h>
>  #include <math_private.h>
>  #include <libm-alias-finite.h>
> -
> -static	const float	one	= 1.0, tiny=1.0e-30;
> +#include <math-use-builtins.h>
>  
>  float
>  __ieee754_sqrtf(float x)
>  {
> +#if USE_SQRTF_BUILTIN
> +	return __builtin_sqrtf (x);
> +#else
> +	/* Use generic implementation.  */
>  	float z;
>  	int32_t sign = (int)0x80000000;
>  	int32_t ix,s,q,m,t,i;
> @@ -70,10 +73,10 @@ __ieee754_sqrtf(float x)
>  
>      /* use floating add to find out rounding direction */
>  	if(ix!=0) {
> -	    z = one-tiny; /* trigger inexact flag */
> -	    if (z>=one) {
> -		z = one+tiny;
> -		if (z>one)
> +	    z = 0x1p0 - 0x1.4484cp-100; /* trigger inexact flag */

Period and double space before '*/'.

> +	    if (z >= 0x1p0) {
> +		z = 0x1p0 + 0x1.4484cp-100;
> +		if (z > 0x1p0)
>  		    q += 2;
>  		else
>  		    q += (q&1);
> @@ -83,6 +86,7 @@ __ieee754_sqrtf(float x)
>  	ix += (m <<23);
>  	SET_FLOAT_WORD(z,ix);
>  	return z;
> +#endif /* ! USE_SQRTF_BUILTIN  */
>  }
>  #ifndef __ieee754_sqrtf
>  libm_alias_finite (__ieee754_sqrtf, __sqrtf)
> 

Ok.